&EPA
United States Industrial Environmental Research EPA-600/7-78-110a
Environmental Protection Laboratory June 1978
Agency Research Triangle Park NC 27711
Electrostatic
Precipitator
Technology
Assessment:
Visits in Japan,
November 1977 -
Appendices
Interagency
Energy/Environmei
R&D Program Report
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
ft. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the INTERAGENCY ENERGY-ENVIRONMENT
RESEARCH AND DEVELOPMENT series. Reports in this series result from the
effort funded under the 17-agency Federal Energy/Environment Research and
Development Program. These studies relate to EPA's mission to protect the public
health and welfare from adverse effects of pollutants associated with energy sys-
tems. The goal of the Program is to assure the rapid development of domestic
energy supplies in an environmentally-compatible manner by providing the nec-
essary environmental data and control technology. Investigations include analy-
ses of the transport of energy-related pollutants and their health and ecological
effects; assessments of, and development of, control technologies for energy
systems; and integrated assessments of a wide range of energy-related environ-
mental issues.
EPA REVIEW NOTICE
This report has been reviewed by the participating Federal Agencies, and approved
for pubiication. Approval does not signify that the contents necessarily reflect
the views and policies of the Government, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/7-78-110a
June 1978
Electrostatic Precipitator
Technology Assessment:
Visits in Japan, November 1977
Appendices
by
Grady B. Nichols
Southern Research Institute
2000 Ninth Avenue, South
Birmingham, Alabama 35205
Contract No. 68-02-2610
W.A. 5
Program Element No. EHE624
EPA Project Officer: James H. Abbott
Industrial Environmental Research Laboratory
Office of Energy, Minerals, and Industry
Research Triangle Park, NC 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
-------
ABSTRACT
This volume of the two-volume report consists of appendices representing
technical papers supplied to a team of U.S. investigators during a parti-
culate control technology assessment visit to Japan. The visit included
discussions with personnel from universities, industries, and major
installations involved with partlculate control, Significant research
activities were noted in both the academic and industrial sectors related
to partlculate control and measurements.
The report proper, EPA-600/7-78-110, summarizes the results of the
individual discussions, observations during the tours, and discussions
of technical papers.
ii
-------
CONTENTS
Appendix A — University of Tokyo 1
Initiation Condition and Mode of Back Discharge 2
Masuda and Mi zvino
Back Discharge Phenomena in Bias Controlled
Pulse Charging System 37
Masuda et al.
Light Measurement of Back Discharge 40
Masuda and Mizuno
Initiation Condition and Mode of Back Discharge
for Extremely High Resistivity Powders 63
Masuda et al.
Utility Limit and Mode of Back Discharge in
Bias-Controlled Pulse Charging System 73
Masuda et al.
The Analysis of Electric Wind in Electrostatic
Precipitator (by Laser Doppler Velocimeter) 81
Masuda et al.
Fundamental Analysis of Electron Beam Gas
Elimination 91
Masuda et al.
Motion of a Microcharge Particle Within
Electrohydrodynamic Field 100
Masuda and Matsumoto
111
-------
CONTENTS (continued)
Appendix A (Continued)
A Preliminary Study of Re-entrainment in an
Electrostatic Precipitator 108
Bassett et al.
Recent Progress in Electrostatic Precipitation 123
Masuda
Flashover Measurements of Back Discharge 142
Masuda and Mizuno
Basic Studies on Back Discharge Mode and
Streamer Propagation 176
Masuda and Mizuno
Present Status of Electrostatic Precipitator Technology 179
Masuda
Appendix B 228
IHI's New Precipitation Techniques 229
Ishikawajima-Harima Heavy Industries Co., Ltd.
Tokyo, Japan
Appendix C — Hitachi Ltd. 235
High Temperature Electrostatic Precipitator
for Coal Fired Boiler 236
Imanishi et al.
High Temperature Electrostatic Precipitator
for Coal Fired Boiler 270
Oataki et al.
Electric Field Distribution in Wide Plate
Spacing Electrostatic Precipitator 280
Misaka et al.
Elimination of SO2 and NO in a Corona Discharge Field 284
Ootsuka et al.
Hitachi EP-SB Type Electrostatic Precipitator 289
Measurement of Suspended Particulates 291
Ootsuka et al.
Hitachi Dust Collection Equipment and Systems 294
List of Hitachi Installations 320
iv
-------
CONTENTS (continued)
Appendix D — Sumitomo Heavy Industries, Ltd. 322
Roof-Mounted Electrostatic Precipitator 323
Sumitomo Heavy Industries, Ltd., Oct. 1977
Appendix E — Nippon Steel-Kimitsu Works 365
(No papers)
Appendix F — Shinwa Trading and Engineering Co. 366
(No papers)
Appendix G — Isogo Power Station 367
(No papers)
Appendix H — Kyoto University 368
Dynamics of Naturally Cooled Hot Gas Duct 369
Hotta et al.
Particle Size Classification by Deposition Angle
in a Gas Centrifuge at Reduced Pressure 377
Tanaka et al.
Electrification of Gas—Solid Suspensions Flowing
in Steel and Insulating-Coated Pipes 383
Masuda et al.
Experiments on the Electrical Dislodging of a
Dust Layer 393
Makino et al.
Comparison of Dust Cleaning Performance of
Collapse and Mechanical Shaking Types of Fabric Filters 400
linoya et al.
On the Economically Optimal Design of Bag Filter 408
linoya et al.
Performance of a Micro-Cyclone 424
linoya and Nakai
International Seminar on Dust Collection 432
Concept of Research in Particle-Gas-Separation 434
Lbffler
-------
CONTENTS (Continued)
Appendix H (Continued)
International Seminar on Dust Collection (Continued)
Current Research on Particle Removal at the
Harvard Air Cleaning Laboratory 452
First
Research Activities on Dust Collection 453
Directory of Foreign Attendants 471
Roster of Attendants, Japan 473
Error in Measurement of Gas Flow Rate in
Gas-Solids Two-Phase Flow by Use of a
Horizontal Diffuser 477
Masuda et al.
Dust Cleaning Dynamics in Reverse Collapse
Type Bag Filter 483
Makino et al.
A Method of Measuring Pressure Drop Parameters
for Multi-Compartment Bag Filter—Mechanical
Shaking Type and Reverse Collapse Type 490
linoya et al.
Performance of Fibrous Powder Bed Filter 504
linoya et al.
Appendix I -- University of Osaka 511
Growth of Aerosol Particles by Condensation 512
Yoshida et al.
Stability of Fine Water Droplet Clouds 518
Kousaka et al.
Behavior of Aerosols Undergoing Brownian
Coagulation, Brownian Diffusion and Gravitational
Settling in a Closed Chamber 552
Ikuyama et al.
A New Technique of Particle Size Analysis of
Aerosols and Fine Powders Using an Ultramicroscope 560
Yoshida et al.
Turbulent Coagulation of Aerosols in a Pipe Flow 565
Okuyama et al.
vi
-------
CONTENTS (continued)
Appendix I (continued)
Effects of Brownian Coagulation and Brownian
Diffusion on Fine Particle Size Analysis by
Sedimentation Method 59°
Oku*ana et al.
Constant Pressure Filtration of Power-Law
Non-Newtonian Fluids (incomplete) 598
Shirato et al.
Experimental Study of Thermophoresis of
Aerosol Particles 599
Kousaka et al.
Turbulent Coagulation of Aerosols in a
Stirred Tank 604
Okuyama et al.
The Effect of Neighbouring Fibers on the
Single Fiber Inertia-Interception Efficiency
of Aerosols 611
End et al.
Pressure Drop and Collection Efficiency
of Irrigated Bag Filter 612
Yoshida et al.
Growth of Aerosol Particles by Steam Injection 618
Yoshida et al.
Effect of Brownian Coagulation and Brownian
Diffusion on Gravitational Settling of
Polydisperse Aerosols 634
Yoshida et al.
Change in Particle Size Distributions of
Polydisperse Aerosols Undergoing Brownian
Coagulation 641
Yoshida et al.
Application of Particle Enlargement by
Condensation to Industrial Dust Collection 648
Yoshida et al.
Appendix J — Mitsubishi Heavy Industries, Ltd. 677
The Latest Dust Collecting Technique 678
Saito et al.
vli
-------
APPENDIX A
UNIVERSITY OF TOKYO
-------
INITIATION CONDITION AND MODE
OF
BACK DISCHARGE
Senlchl MASUDA, PhD. Department of Electrical Engineering
University of Tokyo
Aklra MIZUNO, MSc. "
7-3-1, Kongo, Bunkyo-Ku, Tokyo,
Japan
-------
Summary
Modes of back discharge occurring in the electrostatic precipitator
were studied using, instead of a dust layer, the model samples of glass
and mica plates with a pin-hole, and tissue papers. It was confirmed
that back discharge started to occur when the apparent field strength
in the sample layers exceeded its breakdown field strength. Back discharge
became to be a streamer corona under atmospheric condition. It could be
classified into space streamer mode, surface streamerjnode and mixed_
streamer mode, depending upon the field distribution around the breakdown
point in the sample layers. The first and the third modes occurred when
the field strength in the air gap, Ea, exceeded about 5 kV/cm, and positive
ions were generated in the whole gas space. The second mode appeared when
Ea was lower than about 5 kV/cm, and ion generation was limited to the
near surface region. Among the factors affective on the back discharge,
the dust resistivity was the most essential. For low dust resistivity,
space streamers tended to develop from the breakdown points when the
applied voltage was raised. For high dust resistivity, on the other
hand, the number of breakdown points increased and surface discharge was
pronounced. Remarkable difference in modes was seen when using positive
corona. Neither space streamer nor surface discharge occurred and the
flashover voltage was higher than that with negative corona.
-------
1. Introduction
Back discharge is one of the most difficult problems in electrostatic
precipitators impairing their performance in many industrial plants. '
This is an abnormal kind of discharge which is triggered by breakdown in
a high resistivity dust layer deposited on the collecting electrode and
which lowers the flashover voltage, reducing particle charge and causing
a severe drop in collection efficiency. The nature of back discharge
depends on many factors such as the electrical properties of the dust
layer and the chemical properties of the particles themselves, and its
form is very complicated. Therefore more intensive and basic investigations
are required to solve the back discharge problem, and also to assess
precipitator performance, when back discharge occurs.
Back discharge occurring when using negative corona can be classified
into two major discharge modes. One is the streamer mode, occurring
with high gas density, the other the glow corona mode, occurring with
low gas density.
Normally streamers, formed at the breakdown point on the layer,
proceed Into the gas space towards the discharge electrode or to the
accumulated charges on the dust surface, or In both directions, depending
upon the field distribution around the starting point. It is appropriate
to classify the streamer mode into three sub-modes; space streamer mode,
surface streamer mode and mixed streamer mode. The third one appears
In most of the practical cases.
In this paper, an experimental study on the back discharge of streamer
mode carried out under atmospheric pressure and room temperature is
reported. Studies on the back discharge of a glow corona mode will be
reported separately.
-------
2. Initiation condition and initial mode of back discharge
At first the initiation condition and the initial mode of back
discharge was studied using a needle to plane electrode system, located
inside a thermostat where humidity could also be controlled. It was
tested whether the initiation of back discharge was governed by the
breakdown field strength of a layer, Eds, measured separately using
-parallel plane electrodes. In order to change Eds of a sample, two
glass plates, each having a pinhole, were used on top of one another as
the layer sample, as indicated in Fig.l. By altering the position of
one plate and thus changing the distance between the holes, the value
of Eds could be changed. The resistivity of the glass plate, Pd, was
6 x 1011 ohm-cm and the diameter of the pinhole was 0.5 mm. The thickness
of one plate was 2.0 mm. An image intenslfier tub£ (EMI, type 9912) was
used at its maximum gain (about 106) to observe a back discharge glow
at Its Initiation. Current pulse was observed at the same time by a
cathode ray oscilloscope with a band width of 10 MHz.
The breakdown field strength of the layer In corona field at the
Initiation of back discharge, Eds', was estimated as follows/
Voltage-current density (V-J) curves with the layer for various values
of Eds were measured, where J represents average current density at the
measuring electrode. They are shown by solid curves 1, 2 and 3 in Fig.2
when the electrode gap was 60 mm. In the same figure, the air load V-J
curve (without glass plates) was plotted by a dotted line, the plane
electrode being raised to the position of the surface when the glass
plates were present* i.e. a gap of 56 mm. The voltage across the glass
plates, &V, is given by AV = V - V, where V and V are the electrode
-------
voltages corresponding to the same value of current density, J, with and
without the glass plates, respectively. With the increase in applied
voltage from zero, a Trichel pulse current appeared at the measuring
electrode when corona started at the needle electrode. With the voltage
further increased, repetition frequency of the Trichel pulse current
increased and D.C. current component appeared as shown in Fig.3-a. This
D.C. current component also increased when voltage was increased. When
the point Ai (i = 1 - 3) (Fig.2) was reached, feeble but continuous glow
appeared at the pinhole (Fig.3-b-2), leading to a slight non-linear increase
in current. Breakdown of the layer at the pinhole occurred at this point.
The current wave form at this point is shown in Fig.3-b-l. We took this
point as the initiation point of back discharge. From the value of
AV at this point, (AV)0, and the layer thickness, t, we get Eds' = (AV)0 / t.
The values of Eds' thus obtained at the points AI - Aj in Fig.2 were
compared with those of Pd x Jo, where pd is the layer resistivity and
Jo the current density at the corresponding points. A good agreement
could be seen between Eds' and pd x Jo as indicated in Table 1. The
values of Eds measured using parallel plane electrodes are also given.
The values of Eds' agreed well with these values. This suggests that
the breakdown of the layer occurred at a layer field strength nearly
equal to Eds measured by a parallel plane electrode system. This continuous
glow mode of back discharge at its initial stage should be considered as
a kind of glow discharge. Hence this should be refered to as "onset-glow
mode". It should be distinguished from the more intense "steady-glow mode"
(2)
appearing under the conditions to be reported separately.
With the further increase in voltage, a very small surface streamers
-------
randomly appeared at the limited region around the upper edge of the
pinhole (Fig.3-c-2) at point B1 (1 = 1 • 3) 1n F1g.2, corresponding to
72 2
0 = 0.5 to 1.0 x 10 A/cm . (The expression of current density 1n A/cm
lost Us sense here, since most of the current flew through the pinhole
hereafter.) F1g.3-c-l Indicates current pulse of this streamer, which
should be refered to as an "onset-streamer". The Trlchel pulse current
was still observed to exist. When the voltage was slightly raised above
points Bi, space streamers and large surface streamers occurred from
the pinhole at point |Ci (1=1- 3) in Fig.2 (see F1g.3-d-2), accompanied
by large current pulses (Fig.3-d-l). A more Intense rise 1n current
occurred beyond these streamer starting voltages. It should therefore
be noted that the criterion for occurence of the layer breakdown
should be clearly distinguished from that for occurence of streamers
which are the essential cause for rapid current Increase. In the field
measurements of V-J curves, only the initiation points of streamers
could be detected because of much higher signal to noise ratio expected.
For glass plate with a pinhole described above, the onset-glow appeared
at the layer breakdown voltage. However, when a sufficiently high
resistivity layer, such as a mica plate with a pinhole (pd > 10 ohm-cm),
was used, a random breakdown occurred at first. With a slight increase in
voltage, it was followed by a repetitive breakdown, as Indicated in
F1g.3-e. Then this was followed by a stable onset-glow. Hence, the layer
breakdown voltage, Vb, was different from the starting voltage of the
onset-glow, Vo, in this case. The streamer discharge in gas space 1s
followed by a flashover occurring at a voltage much lower than that without
back discharge. Thus there are four major critical voltages for back
-------
discharge under atmospheric condition; layer breakdown voltage Vb,
onset-glow starting voltage Vo, streamer starting voltage^st and finally
flashover voltage Vs. The random breakdown, onset-glow and onset-streamer
constitute an initial stage of back discharge where the current rise
remains still comparatively low. This stage should be refered to as
"onset-stage".
3. Back discharge in streamer mode
With the increase in voltage beyond the point C1 in Fig.2, streamers
are emitted either into gas space towards discharge electrode or along the
surface of the layer, or in both directions. Hence, back discharge in this
mode should be refered to as "streamer mode", more specifically space
streamer mode, surface streamer mode and mixed streamer mode as a combination
of the former two. When the voltage was further raised, space streamer
proceeded towards the discharge electrode and it bridged across the
electrode gap until it finally turned into flashover. It could be expected
that the most essential factors deciding the respective mode of streamers
would be the strengths of tangential and vertical field around the breakdown
point of the layer as well as corona current. Thus, these effects were
studied separately. The detailed mechanism of propagation for these streamers
(4)
will be discussed in another paper.
3.1 Effects of vertical field and current
Along with the study of the effect of vertical field, Ea, that of the
corona current density, J, was also investigated. These two factors,
Ea and J, are closely coupled to each other in an actual precipitator,
8
-------
while their magnitudes at back discharge initiation largely differ from
case to case, depending upon the dust layer resistivity as described in
section 3.3. To investigate the effects of these two factors separately,
a grid electrode was inserted between the needle and plane electrodes as
shown in Fig.4. A transient fluctuation in the grid electrode potential
was eliminated by using a condenser of 0.5 yF capacity connected paralled
to its H.V.source. By the change in needle electrode voltage Va and
grid electrode potential Vg, vertical field strength Ea and current
density J could be varied independently. The value of Ea was calculated
from the ratio Vg / (grid-to-plane spacing). In these experiments two
glass plates each having a pinhole (0.5 mm in diameter) were used as
before. The resistivity was 6 x 10 ohm-cm and the breakdown field
strength 20.7 kV/cm. Fig.5 shows curves of current density 0 plotted
against voltage of needle electrode Va with the grid potential Vg being
kept constant. The mode and current wave form of back discharge under
atmospheric condition were observed with the aids of an image intensifier
tube and a cathode ray oscilloscope. From these observations and the
curves in Fig.5, the mode diagram of back discharge was depicted on
Ea-J domain as shown in Fig.6-a. No back discharge occurred in region I
because of low current density. When the current exceeded a certain value
at which the layer breakdown condition described before was fulfilled,
back discharge in the onset-glow mode occurred in region II (see Fig.3-b-2).
The further increase in current resulted in the onset-streamers occurring
around the edge of the pinhole in region III (see Fig.3-c-2). It should
be noted that the critical current densities for the transitions between
regions I, II and III were nearly constant respectively independent of
-------
Ea, as is shown by the flat curves A and B. The two regions II and III
should be refered to as "onset-stage" region, With the further Increase
In current beyond the other critical curves C and D, back discharge 1n
the streamer modes (surface and mixed streamer modes) took place in region
IV and V. The region IV, for lower value of Ea, 1s the surface streamer
region where the surface streamer mode was predominant and space streamers
were few (see F1g.6-b). In this region current density J saturated at
curve E because of space charge limitation (see Fig.5), and no flashover
could be resulted between the grid and the plane electrodes. Whereas In
region V, when Ea exceeded 5 kV/cm, both the surface and space streamers
occurred to form the mixed streamer mode (see F1g.7-b). Again the critical
current density for the transition from the region III to IV and V was
nearly constant, except for a corner area G. When J 1n the region V
exceeded curve F, the streamer turned Into a flashover. The critical value
of the field strength between regions IV and V (curve H) was about 5 kV/cm
under the atmospheric pressure and room temperature, which had been taken
as a criterion for the occurence of streamer under these conditions. It
should, however, be noted that the Initiation and growth of space and
surface streamers is'also governed by current density «3.
3.2 Effect of tangential field
In the present case where the surface resistivity of the layer is
extremely high, the surface charge would be firmly bound to its original
position. In this case the tangential field around the breakdown point
will become a function of the surface charge density on the layer, 00» at
the Instant ttf breakdown at which the potential of the breakdown point
10
-------
becomes almost zero. The value of o0 in turn is given by eEds where e is
the dielectric constant of the layer. If o0 is sufficiently high, the
breakdown of the point will directly trigger the surface streamer. In the
opposite case onset-glow appears prior to the occurence of surface streamer,
so far as the vertical field strength in gas space is not sufficiently high
for the space streamer to be triggered. Such a high vertical field
strength does not normally exist at the initiation of back discharge, unless
the layer resistivity is in the range of 5 x 10 to 10 ohm-cm as
discussed later. Thus, the effect of o0 on back discharge in streamer mode
was studied. Two glass plates were used as before so that Eds and, hence,
o0 could be changed. Photographs of the back discharge for two values
of breakdown strength are shown in Fig.7. When Eds was 13.8 kV/cm, a space
streamer was dominant proceeding to the discharge electrode (Fig.7-a).
When Eds was 33.8 kV/cm, the mixed streamer mode appeared where a remarkable
surface streamer in the vicinity of the pinhole could be observed (Fig.7-b).
This was because the tangential field strength became larger as o0 increased.
The surface discharge became especially dominant when the value of o0
-9 2
exceeded about 5 x 10 C/cm .
3.3 Effect of dust resistivity
A tissue paper was used as a sample in this experiment. This was
g
because its apparent resistivity p. could easily be changed from 10 to
14
10 ohm-cm by adjusting the ambient humidity. Thus the effect of p . on
the back discharge mode under normal temperature was studied. Voltage-
g
current density curves for different values of p^, ranging from 6 x 10
to 2 x 10 ohm-cm, are shown in Fig.8 where the electrode gap was kept
at 60 mm. Photographs of the back discharge for three different values
11
-------
of pd are shown in Fig.9, where the values of J were 1n the same order.
g
When the resistivity was 6 x 10 ohm-cm (curve 1 in Fig.8), no back
2
discharge occurred until flashover took place at V = 65 kV and J = 7.6 yA/cm .
For the case of needle to plane electrode system and experimental conditions
Investigated, the initiation condition of back discharge, Eds - Pd x Jo,
did not become to be fulfilled prior to the occurence of flashover when
Pd was lower than about 5 x 10 ohm-cm. In other words, the initiation
voltage of back discharge was higher than the flashover voltage of the.
gap because of too low value of pd- When the value of pd slightly exceeded
this critical value, space streamers occurred as soon as the layer broke
down, owing to the large voltage drop across the gas space. For instance,
when the resistivity was 0.9 x 10 ohm-cm (curve 2), the streamer starting
voltage Vst was about 27 kV. The number of breakdown points was less
and streamers proceeded into space towards the discharge electrode, as
shown in Fig.9-a. The occurence of space streamers lowered the flashover
voltage Vs to a great extent. It was observed that, when pd was between
about 5 x 10 and 0.9 x 10 ohm-cm, excessive sparking tended to occur.
In this range of pd, Vst would be lowered with the increase in pd, so that
1t finally becomes lower than Vs as in the case of curve 2 in Fig.8.
A slight Increase in voltage beyond Vst would cause flashover because
12
Vst remained still close to Vs. For pd higher than 10 ohm-cm (curve
3 and 4), the back discharge streamers started to occur at a much lower
voltage and current density. There was a larger interval between Vst and
Vs so that the excessive sparking disappeared. There were more breakdown
points with a general glow surrounding each point. In this case a surface
glow dominated and space streamers were few. This tendency became
12
-------
pronounced with the increase in prf (Fig.9-c).
The different discharge modes were caused by the difference in the
ratio of the voltage drop across the dust layer to that across the gas
space. If the resistivity was high, the voltage drop across the dust
layer was high even at a low current density on the initial stage of back
discharge, whereas the voltage drop across the gas space was low. As a
result, the development of a space streamer was suppressed, and a surface
discharge occurred. In this case many weak points broke down and the
current increased readily without excessive sparking. However, when
voltage was raised, the space streamer occurred also in this case, taking
the form of a general glow bridging across the gap. A more severe
increase in current occurred at this later stage. It can be seen that
flashover occurred almost at the same voltage inspite of a large difference
in p., once back discharge occurred. This agrees well with the results
of G.W. Penney, ' i.e. the flashover voltage of back discharge was not
affected by the value of resistivity. This flashover voltage was almost
half the value of that under non-back discharge condition.
3.4 Charging efficiency in different regions
For negative corona, back discharge is a source of positive ions
to produce a bipolar ion atmosphere in gas space. The effect on particle
charging, however, is different depending on the mode of back discharge.
In the surface streamer mode, the ion source 1s considered to be surface-
like, but in the space or mixed streamer modes, ion generation in space
may occur. These were confirmed by measuring particle charge using the
electrode system as shown in Fig.10.' ' This system enabled the change
13
-------
in back discharge mode by changing the field strength between grid and plane
electrodes, Ea (see Fig.6). A steel ball, 3.0 mm in diameter, was dropped
between plane and grid electrodes and its saturation charge was measured
by a Faraday cage. Distance between plane and grid electrodes was 50 mm,
grid to discharge electrodes 30 mm, and a mica plate having many pinholes
was used as a layer. Fig.11 is an example of the results obtained, showing
the saturation charge of a steel ball as a function of its position d from
the plane electrode. The values indicated in the bracket represent the
theoretical saturation charge due to monopolar ions, calculated from
Pauthenier's equation. ' In the surface streamer region (curve 1), the
value of charge was about 90 X less than the theoretical value but the
sign of particle charge remained the same as that of the discharge electrode.
The value of charge decreased as the particle crossed nearer the plane
electrode. This result indicates that the back discharge of this mode
can be considered as a surface-like ion source so that the density of
positive ions decreased into the space. In the mixed streamer region where
space streamer was pronounced (curve 2), particle charge scattered largely
around its average value which was a fairly high positive value and almost
the same regardless of position. This result might indicate that positive
ions were generated abundantly inside the whole space and took dominant
role in particle charging. The effect of streamer tip to collide with
a particle might also be a factor. The curve 3 represents the transition
region between the foregoing two. In this case particle charges were
also positive but as low as in the case of curve 1.
14
-------
3.5 Back discharge with positive corona
It was observed that the mode of back discharge with positive
corona at the needle was completely different, as shown 1n F1g.l2.
Tissue paper was used and the electrode gap was 60 mm. Voltage-current
density curves are shown in Fig.13 for various values of pd- In this
case breakdown points were distributed uniformly on the surface, no space
,— \._._
or surface streamers could be observed and the discharge mode Was only
glow mode Independent on resistivity. The abnormal Increase 1n current
was small and the flashover voltage when back discharge occurred was
approximately 1.5 times higher than that for back discharge with negative
corona at the needle. The relationship between the flashover voltage of
back discharge, Vs, and gap distance d 1s shown 1n F1g.l4 for the positive
and negative coronas. Vs of the positive corona was higher than that of
the negative corona for gap distance range of 1.0 to 10.0 cm. The flashover
voltage of the positive corona under back discharge condition was also
higher than that without back discharge when the layer was removed (air
load). The mechanism for this behavior is considered to be due to a stable
nature of negative glow corona at the breakdown point and to the positive
(Q\
corona at the needle tip being converted to Mentistein's glow corona.1 '
The latter may be resulted by copious negative Ions fed to the needle
electrode, from which electrons would be shedded to form a continuous
and stable positive glow discharge at the needle tip.
-------
4. Conclusions
From the experimental studies described above, using the model samples
of tissue papers, glass and mica plate, the following results on the effects
of apparent resistivity and breakdown field strength on back discharge were
obtained.
(1) With the negative corona at the needle, the layer breakdown started
to occur when Eds = prf x J is fulfilled. It was followed by the onset-
glow mode occurring with a slight increase in voltage. A rapid increase
in current, however, occurred only when the streamers started to occur
at a critical voltage Vst. Thus there are four major critical voltages
for back discharge under atmospheric condition; layer breakdown voltage Vb,
onset-glow starting voltage Vo, streamer starting voltage Vst and flashover
voltage Vs. For the case of electrode system investigated, the initiation
condition of back discharge may not be fulfilled prior to the occurence
of flashover when p. does not exceed about 5 x 10 ohm-cm. When p. is
in the range of 5 x 10 to 10 ohm-cm, it becomes Vst * Vs, so that
12
excessive sparking occurs. When pd > 10 ohm-cm, Vst becomes sufficiently
lower than Vs so that excessive sparking disappears, but abnormal increase
in current occurs.
(2) There are three sub-modes in the streamer mode; space streamer mode,
surface streamer mode and mixed streamer mode, depending upon the field
distribution around the breakdown point in the sample layers. This in turn
is a function of p ., Eds, Ea and J. In the space streamer mode, positive
ion generation in space occurs and particle aquires a fairly high positive
charge. Whereas in the surface streamer mode, positive ion generation is
limited to the surface region and the sign of particle charge is the same
16
-------
as that of the needle electrode. In most of the actual cases, however,
the mixed streamer mode appears.
(3) With positive corona at the needle, back discharge mode is completely
different. The flashover voltage is higher than that under back discharge
condition with the negative corona.
Acknowledgements
This research was sponsored by the Ministry of Education, Japan,
as its Special Research Project (I) (Project No.011914). The authors
are gratefully indebted for its support. Thanks are also due to
Mr. Masao Kuroda for his help given to a part of the experiments.
Nomenclature
p. apparent dust resistivity
Eds breakdown field strength of the layer measured by parallel plane electrodes
Eds'breakdown field strength of the layer in corona field
a0 surface charge density at the instant of breakdown
e dielectric constant of the layer
Ea vertical field strength in the gap
Va discharge electrode voltage
Vg grid electrode potential
Vb layer breakdown voltage
Vo onset-glow starting voltage
Vst streamer starting voltage
Vs flashover voltage
J current density at the measuring electrode
Jo current density at the initiation of back discharge
17
-------
References
•
1 S. Masuda, Recent Progress 1n Electrostatic Precipitation,
Static Electrification 1975, Institute of Physics Conference
Series, No.27, p. 154 (1975)
2 S. Masuda and A. Mlzuno, Flashover Measurements of Back Discharge,
J. Electrostatics (to be published)
3 S. Masuda, Reverse Ion1sat1on Phenomena In Electrostatic Predpltator,
J.I.E.E. Japan. 35-102 (1960) p. 1482
4 S. Masuda and A. Mlzuno, Light Measurement of Back Discharge,
J. Electrostatics (to be published)
5 6.W. Penney and S.E. Craig, Sparkover as Influenced by Surface
Conditions 1n D.C. Corona, A,I.E.E. Trans, pt.1, vol. 79, May 1960,
pp.112-18.
6 M. Kuroda, 1975 Graduation Thesis, Department of Electrical Engineering,
University of Tokyo
7 H. Pauthenier, Moreau-Hanot, Rev. Gen. Elect., Tome XIV, No. 18
p. 583 (1932)
8 L.B. Loeb, Electrical Coronas, p.95 (1965) Univ. of California
Press
18
-------
Curve
1
2
3
Initiation
point
Al
A2
A3
AV (kV)
5.6
9.2
14.0
Jo (A/cm2)
2.5X10~8
4.1X10" 8
6.9X10"8
Eds'= AV/tj
(kV/cm)
14.0
23.0
35.0
/>dxj0|
(kV/cm)
15.0
24.6
41.4
Eds (kV/cm)
15.7
25.1
39.0
Table 1 Comparison of Eds, Eds' and
19
-------
D.C.H.V.
NEEDLE ELECTRODE
PINHOLE
TEST SAMPLE
(DOUBLE PLATES)
PLANE ELECTRODE
MEASURING ELECTRODE
Figure 1
-------
20 -
GAP = 60 (mm)
Eds » 15.7 (kV/cm)
Eds =25.1 (kV/cm)
3: Eds = 39.0 (kV/cm)
10 15 20
V ( KV )
Figure 2
30
21
-------
rrrxr
->r~Kt>'rr*T**'~~*niryt*ir*mr-w~,- ,r -.. ™
._• ?r ,2^^^^:
'. cT COMp""E"jTT1>^3:J
i-o
TS^I 1S^^
L -^fc.A... .j^.yji^v-u^m^ic^M.-'
0.7 x 10~8 A/cm2
(a) Trichel pulse ( 12 kV, 2.0 x 10~8 A/cm )
2 x 10~8 A/cm2
(1) Current
(b) Onset-glow ( 16 kV, 4.0 x 10"8 A/cm2 )
(2) Photograph obtained
with intensifier
3 x 10~8 A/cm2
ONSET-STREAMER PULSE
(1) Current (2) Photograph obtained
(c) Onset-streamer (18 kV, 1.1 x 10'7 A/cm2 ) w1th 1ntens1f1er
Figure 3a
22
-------
1.5 x 10
-•'
A/cm'
(1) Current
(d) Streamer ( 21 kV, 4.0 x 10"7 A/cm2 )
(2) Photograph obtained
with 1ntens1f1er
' ''
^h^v.-^r •
- -=- -''- ' -*'•
3 x 10"8 A/cm2
(e) Repetitive breakdown (10 kV, 4.0 x 10"8 A/on2 )
(Mica plate)
Figure 3b
23
-------
Va
NEEDLE ELECTRODE
GRID ELECTRODE
Vg
PINHOLE
GLASS PLATES
R = 100
mm
-- **
PLANE
ELECTRODE
Figure 4 (modified)
24
-------
< in -
10
-8
10 15 20 30 40
Va (kV)
Figure 5
25
-------
10
-5
FLASHOVER
10
-6
CM
E
u
-7
10
-8
0
v
MIXED
H STREAMER
REGION
IV
SURFACE
STREAMER
REGION
III ONSET-STREAMER REGION
X X x X X j
II ONSET-GLOW REGION
I NO BACK DISCHARGE
216
Ea (kV/cm)
8
10
Figure 6a
26
-------
Figure 6b .
Photograph of back discharge in the surface streamer region JJVJ
(Ea : 4.0kV/cm, J : 5. 0 x 10-7 A/cm2)
27
-------
(a) Eds LOU
Eds = 13.8 (IcV/cm)
V = 30 (kV)
1 = 29
(b) Eds HIGH
Eds = 33.8 (kV/cm)
V =40 (kV)
= 23
Figure 7
28
-------
10
u
0,1
0,01
FLASHOVER
65 (kV)
7.6 (/iA/cm2)
= 6.0 x 10 (ohm-cm)
2: /»d = 0.9 x lO^ohm-cm)
3: /»d = 1,6 x 1012 (ohm-cm)
4: /»d = 2.0 x 1013(ohm-cm)
I
10 15 20
V ( KV )
Figure 8
30 40
29
-------
(a) />d - 0.9 x 10
J - 3.2
II
(ohm-cm)
OlA/cm2)
(b) /»d • 1.6 x 10
J = 5.5
12
(ohm-cm)
Figure 9
(c) /»d - 2.0 x 10"
J - 2.2
(ohm-cm)
30
-------
fi
p _
••
•
i
^
i
^
i
1
X
f
. •
JL
10 mm
T
d
f'
-oV9
1
a
j_JL
)Va
30 mm-*
F D: DISCHARGE ELECTRODE
(SAW TOOTH ELECTRODE)
G: GRID ELECTRODE
P: PLANE ELECTRODE
M: MEASURING ELECTRODE
T: MICA PLATE WITH PINHOLES
B: STEEL BALL WITH 3.0 mm
IN DIAMETER
F: FARADAY CAGE
Figure 10 (modified)
31
-------
+10
+ 8
+ 6
L
o
P"
X
a
0
- 2
-1
0
(2) MIXED STREAMER REGION
Ea * 6.0 (kV/cm)
J = 2.0 '•'—*'
(44.7 X)
(41.7%) ;C(43.4X)
(3) TRANSITION REGION Ea = 4.6 (kV/cm)
J = 1.7
I r(5.4 56)
(2.2 35) IT——-^
(1) SURFACE STREAMER REGION
Ea = 3.6 (kV/cm)
.9 X)
J = 1.6 (AiA/cro)
10
20
30
50
d (mm)
Figure 11
32
-------
Figure 12
33
-------
10
CM
<
n.
0,1
0,01
GAP = 60 (ran)
1: AIR LOAD
l
2: Ai
3: Al
10" (ohm-cm)
1012 (ohm-cm)
4: Al = 1013 (ohm-cm)
FLASHOVER
I
7 10 15 20
V (*kV)
30
40 50
Figure 13
34
-------
60 r
CO
tu
o
CO
50
40
30
20
10
0
0
50
ELECTRODE GAP d (mm)
100
Figure 14
35
-------
Figure caption
Flg.l Electrode system for studying back discharge
F1g.2 Voltage-current curves under back discharge condition for
different values of breakdown field strength
(A pair of glass plates, each having a single plnhole; sample
resistivity p^ « 6 x 10 ohm-cm)
Fig.3 Current wave forms and modes of back discharge
Fig.4 Electrode system for studying the effects of vertical field and current
F1g.5 Voltage-current density curves for different values of grid
potential Vg (ref. F1g.4)
Fig.6 Effect of tangential field and current density on mode of back discharge
(a) Mode diagram 1n field-current domain
(b) Photograph of back discharge in surface streamer region (IV)
(Ea = 4.0 kV/cm, J = 5.0 x 10~7 A/cm2)
F1g.7 Effect of tangential field on back discharge in the mixed streamer mode
(A pair of glass plates, each having a single pinhole. Electrode
gap = 50 mm)
F1g.8 Effect of dust resistivity pd on voltage-current density curves
under back discharge condition when negative corona is used
(Tissue paper, 1.0 mm in thickness)
Fig.9 Effect of dust resistivity on back discharge mode
(Tissue paper)
Fig.10 Electrode system for measuring particle charging
F1g.ll Saturation charge v.s. position d for different back discharge modes
Fig.12 Back discharge under positive corona point
(Tissue paper, V = +40 kV, 0 = 2.8 x 10"6 A/cm2, P£J • 1013 ohm-cm)
Fig.13 Voltage-current density curves under back discharge condition
when positive corona is used
(Tissue paper, 1.0 mm 1n thickness)
F1g.l4 Flashover voltage v.s. gap distance under back discharge condition
with positive and negative coronas
(Tissue paper, pd » 1.2 x 10 ohm-cm, 1.0 mm in thickness)
36
-------
Proe. 4th Int. Clean Air Congress
Paper No. V-52 (May 1977, Tokyo)
BACK DISCHARGE PHENOMENA
IN BIAS CONTROLLED PULSE CHARGING SYSTEM
Ph6nomene de d&charge de Back dans le systfcme de chargement pr6ventif
a pulsations controlees
MASUDA.S.andDOI.I.
Department of Electrical Engineering. University of Tokyo
Tokyo. Japan
HATTORI, I. and SHIBUYA, A.
Ishikawajima-Harima Heavy Industries Company Limited
Tokyo. Japan
INTRODUCTION
Recently an emphasis is given in the field of electrostatic
precipitation to the solution of back discharge. As is
known, this phenomena occurs when the apparent dust
resistivity of the deposited layer upon the collecting
electrode exceeds about 5X10'° ohm-cm. When it occurs,
many troublesome problems arise in the precipitation
process, such as an excessive sparking hindering the increase
of voltage and the decrease in particle charge owing to the
ions of opposite polarity emitted from the back discharge
points, hence reducing largely the collecting performance.
In principle, a technical solution lies in meeting the
conditions to avoid the occurrence of back discharge
phenomena, such that the apparent field strength Ed inside
the dust layer given as the product of current density id and
its dust layer resistivity pa does not exceed its breakdown
strength £,),'>.
=id xpd
-------
THE FOURTH INTERNATIONAL CLEAN AIR CONGRESS
II. SPECIALITIES IN BEHAVIOR OF BACK
DISCHARGE
pulse height was lowest among them.
The start and mode of visible back discharge are
governed by sample resistivity p,j, pulse width r, pulse
repetition frequency f, pulse height Vp, and main dc
voltage Vc. The effect of E and i are contained implicitly in
these results. It is found that behaviors of back discharge in
the pulse charging system is largely different from those
under- a conventional electrode system. Hence, they are
studied in detail.
2.1 Effect of Dust Resistivity
Among the effects of various factors described above.
that of Pd is the most essential. In this pulse charging
system, the visible back discharge does not occur normally
unless Pd exceeds about 1013 ohm-cm, when Vc 10 Hz, duty cycle = 0.1. However,
even in this range, random back discharge exceptionally
occurs very rarely. It starts with the appearance of an
unstable glow on the sample surface from which space
streamers suddently develop and turn to a flashover. When
the frequency is high or pulse width large, the occurrence
of back discharge becomes more frequent, but the starting
condition becomes more obscure. The situation changes
completely when the sample resistivity exceeds 10''
ohm-cm. In this case, with the initiation of back discharge,
a very stable glow appears at first on the sample surface.
With the increase in the dc voltage, a spot like glow points
a'ppear in the sample layer, their number increases, and
finally, streamers occur toward the discharge electrode,
bridging across the electrode gap. But these streamers are
very hard fo turn flashover in contrast to those under
Pd<1013 ohm-cm. An interesting phenomena occurring in
the resistivity range higher than 10'3 ohm-cm is that
streams jump from points to points over the whole sample
surface. When the sample resistivity exceeds 1014 ohm-cm,
a very noticeable phenomena starts to occur, such that back
discharge also appears on the third electrode if it is covered
by the layer. This back discharge appears only when the
field strength E is sufficiently high so that breakdown of
the layer on the third electrode can take place owing to the
oncoming positive ions from the counter electrode. Also in
the resistivity range beyond this order, if the main field
strength E is sufficient enough, a feeble glow-like back
discharge becomes possible to occur on the sample layer on
counter electrode, even by the dark current, even without
supply of pulse current.
2.2 Effects of Pulse Width and its Repetition Frequency
Pulse width and its repetition frequency have also an
appreciable effect on the initiation and mode of back
discharge. Back discharge becomes more active with the
increase in pulse widths. The increase in pulse repetition
frequency and pulse height have also the same effect, and
these tendencies are observed in the whole range of sample
resistivity investigated. For larger pulse widths or higher
repetition frequencies more streamers develop toward the
discharge electrode. These effects are most pronouncedly
observable when pulse width is raised, followed by the
increase in pulse repetition frequency, whereas the effect of
III. CRITICAL VOLTAGE AND CURRENT
FOR BACK DISCHARGE INITIATION
The visible back discharge starts, depending upon the
sample resistivity pd at a certain critical current density i
which, however, varies as the function of the main field
strength E. The magnitude of i determines the charging
rate, whereas that of E governs the particle saturation
charge and Coulomb force. Hence, the utility limit of this
charging system is to be judged from the critical values of i
and E at which the visible back discharge initiates.
Therefore the critical values of the main dc voltage and
pulse current, Vcc and lc, are measured for respective value
of pd, where the values of T, f, and Vp are changed as
parameters.
As described already, back discharge occurs very rarely
when Pd<1013 ohm-cm. When the resistivity Pd is in the
order of 10" ohm-cm no appreciable difference is
observed with the change of these parameters. In the Figs.
2(a) and (b) are shown the relationship between voltage
Vcc and lc forpd = (1.13~1.80)X10!7 ohm-cm where f =
100 Hz, and Vp = 10 and 25 kv. A critical voltage Vcc as
high as 50—80 kv can be attained for 20 cm spacing
between discharge and counter electrodes, and zero bias
voltage. It should be noted that the difference in the
magnitude of Vp results in a large changes in Vcc—Ic
characteristics. This change occurs only when f>100 Hz. In
general, the increase in Pd results in a decrease in Vcc. So
( - U.U * l.M)iUU!k»
t • 1000
»0
tfO.lt *• I.M)t»'*On
— i • 1000 M
Me "
MO •
100 -
M •
JO •
It •
f - M
T - 10 B
M
»
« • M l
t - H
Fig. 2
O
Relationship
current 1
between critical voltage V^, and critical
38
-------
THE FOURTH INTERNATIONAL CLEAN AIR CONGRESS
far as f<100 Hz, the characteristic of Fig. 2(a) remains
unchanged until Pd becomes about 10** ohm-cm, although
the magnitudes of e.ach plot changes. If Pd exceeds about
10'4 ohm-cm, back discharge becomes extremely easier to
occur, and the drooping characteristics always appears
independent of V and f, as shown in Figs. 2(c) and (d).
The magnitudes of both Vcc and Ic become extremely low.
Generally the use of' a narrow pulse width r and a low
pulse height Vp is preferable, because a high critical voltage
Vec is obtainable in this case. Under the conditions of this
experiment, the highest value of Vcc is obtained when r is
reduced to its minimum of 10—20 ps. In this case Vce = 50
kv is obtainable even when pa is as high as 10'* ohm-cm.
It can be concluded that, with the use of the pulse
charging system, back discharge can be very effectively
avoided when a sufficiently narrow pulse width and low
pulse height are used.
IV. RESULTS IN FIELD TEST
The above results were confirmed in a pilot plant test
performed at the exit of an electrostatic precipitator
located at an iron ore sintering furnace. The pilot plant
consists of a combination of a pulse charging zone identical
to that shown in Fig. 1 and a collecting zone out of the
electrostatic screen, the details of which were reported
elsewhere3). There were two stages of this combination in
series, and the gas transit time through a single charging
zone was about 0.4 sec. The dust consisted mainly of iron
oxide (Fe}O3) particles and contained small amounts of
salts of alkaline metals and alkaline earth metals, by several
percents in total. The resistivity of the dust layer was very
high, in the range of 10'*-10M ohm-cm under operating
conditions. The particle size of dust was extremely small,
and more than 70 percents were in the range less than 1 /im
in diameter. This was because most of the coaser particles
had been collected in the preceding conventional pre-
cipitator. Instead of a pulse voltage a sinusoidal voltage was
applied between the discharge and third electrodes, so that
its equivalent pulse width re was taken as a half period.
Fig. 3 shows the relationship between the peak voltage
Vp and the current I flowing into the counter electrodes of
one charging zone, where the main dc voltage Vc between
the third and the counter electrodes was kept at 40 kv and
the frequency of the ac voltage was either 62.5 or 50 Hz.
The curves (I) and (2) show the Vp-I characteristics when
the thickness of dust layer deposited on the counter
electrodes was large. The equivalent pulse width r, for
curve (1) is 10 ms, whereas that for curve (2) 1 ms. The
sharp rise in curve (1) clearly shows the occurrence of a
severe back discharge. This could, however, be amended to
a great extent by reducing the equivalent pulse width Te to
1 ms as indicated by curve (2). The curves (3) and (4)
indicate the characteristics when the dust layer thickness
was kept very small. These two curves show a remarkable
effect of decreasing the layer thickness as a counter
measure to back discharge. The effect of decrease in pulse
width can also be observed here. The characteristics of
(»
Mia dc velum ' *>
notation
O
•
A
A
l.(M>
1
1
' 10
10
((Hi)
»2.J
62.1
SO
JO
l*y«r
cMckniM
•Mil
Ur««
•Mil
!«»•
"' V
\J
1 M
O) I -IOM
\
<» 1
L
— PEAK VOLTAGE V^(KV) "
Fig. 3 Effect of peak voltage V. and equivalent pulse width
re on current I
curve (4) is a normal one without back discharge.
The collection performance of 65-80 percent could be
achieved, which exceeded the required level for an after
collector to be installed. This performance level corre-
sponds to a very satisfactory figure, if the extremely high
resistivity and small size of dust as well as the very short
treating time of about I s in total within active zones be
taken into account.
ACKNOWLEDGEMENT
Fellowship of the Fundac, ao de Amparo a Pesquisa do
Estado de Sao Paulo (FAPESP), Brazil, given to one of the
authors (loshiaki Doi) is highly appreciated.
REFERENCES
1) Masuda, S., Mizuno, A. & Akutsu. K., "Initiation and
Mode of Back Discharge," J. Electrostatics (to be
published).
2) Luthi, J. E., Grundlagen Zur Elektrostatischen
Abscheidung von hochohmigen Stauben, Dissertation
ETH-Zurkh.No. 3924(1967).
3) Masuda, S., Doi, I.. Aoyama, M. & Shibuya, A., "Bias-
Controlled Pulse Charging System for Electrostatic
Precipitator," Staub-Reihalt.Lu/rM,1,19(1976).
39
-------
Reprinted from
Journal of Electrostatics, 2 (1976/1977) 375—396
& Elsevier Scientific Publishing Company, Amsterdam — Printed in The Netherlands
(tteprinted with permission)
LIGHT MEASUREMENT OF BACK DISCHARGE
SENICHI MASUDA and AKIRA MIZUNO
Department of Electrical Engineering, University of Tokyo, 7—3—1, Hongo, Bunkyo-Ku,
Tokyo (Japan)
(Received January 19, 1977; in revised form April 7,1977)
ELSEVIER SCIENTIFIC PUBLISHING COMPANY. AMSTERDAM
40
-------
Journal of Electrostatics, 2(1976/1977)375—396
© Elsevier Scientific Publishing Company, Amsterdam — Printed in The Netherlands
LIGHT MEASUREMENT OF BACK DISCHARGE
SENICHI MASUDA and AKIRA MIZUNO
Department of Electrical Engineering, University of Tokyo, 7—3—1, Hongo, Bunkyo-Ku,
Tokyo (Japan)
(Received January 19, 1977; in revised form April 7, 1977)
Summary
Light measurements of back discharge are made under negative d.c. high voltage and
negative pulse high voltage application, with the aid of a photomultiplier tube and an
image converter camera connected to an image intensifier. The light signal of back dis-
charge in the mixed streamer mode indicates that it consists of two parts, the primary
light wave rising very rapidly, and the secondary light wave rising more slowly. The
former proceeds into space while the latter proceeds along the layer surface. In the space
streamer mode, the primary light wave is dominant and the secondary light wave is very
weak. When a sufficiently high pulse voltage is applied under lower pressure, back dis-
charge is triggered by free electrons supplied from the needle electrode. As the pressure
is increased, an abrupt change in the triggering carriers occurs from electrons to negative ions.
1. Introduction
Back discharge is a phenomenon which impairs the collection performance
of electrostatic precipitators. It is an abnormal discharge starting from the
dust layer deposited on the collecting electrode and triggered by its break-
down. The authors studied the initiation condition and mode of back dis-
charge as well as its flashover voltage, and clarified the effects of various
parameters affecting them, as reported separately [1,2]. Studies were further
made on the propagation of back discharge in the streamer mode and its
triggering process by the use of a light measurement technique. As reported
separately [3], particles are ejected at back discharge points from the layer,
so that, after a short transient period of time, pinholes are formed in the
layer. Hence, through these studies, the precipitated layer was modelled by
the insulating discs having a 0.5 mm pinhole backed by a metal electrode.
2. Experimental apparatus
2.1 Electrode system
A needle-to-plane electrode system is used with a gap d of 50 mm or 20
mm, as shown in Fig. 1. A glass plate, having resistivity Pd of 6 X 10n ohm.cm,
2.0-mm thickness, and a pinhole of 0.5-mm diameter, is located on the plane
electrode as a test layer. This is to improve reproducibility of the phenomena.
41
-------
NEEDLE
ELECTRODE
-D.C.H.V.
o
in
MIRROR „
PLANE
ELECTRODE I *
r i
>• i
••-•J
P.M
P.M.: PHOTOMULTIPLICR TUBE
CRO : CATHODE RAY
OSCILLOSCOPE
(a) MEASURING SYSTEM FOR STUDYING
BACK DISCHARGE STREAMER PROPAGATION
1
-D.C.H.V.
o
in
\
E^^^gd
DELAY CABLE 4- i
DELAY
UNIT
I.C.: IMAGE CONVERTOR
CAMERA
1.1.: IMAGE INTENSIFIER
S : SMALL SPARK GAP
TRIGGERING PULSE
(b) TRIGGERING CIRCUIT FOR IMAGE
CONVERTOR CAMERA
Fig.l. Experimental apparatus.
P.B.: PULSE GENERATOR
(c) TRIGGERING CIRCUIT FOR IMAGE
42
-------
A mica plate, having Pd greater than 1015 ohm.cm, 0.20-mm thickness, and
a pinhole of 0.5-mm diameter, is also used for the measurement with pulse
high voltage. As reported separately, difference in sample resistivity does
not affect the discharge in the gas space to be studied [2]. The plane elec-
trode consists of measuring and guard electrodes.
2.2 Photomultiplier measurements
The light signal from a point inside the back discharge streamer is mea-
sured simultaneously with current. The light is f ocussed onto a slit in front
of a photomultiplier tube, using a concave mirror as shown in Fig. l(a). An
area of 0.5-mm square at the measured point can be resolved. The measured
point can be traversed by altering the position of the photomultiplier. The
output resistance of the photomultiplier tube and the resistance for current
waveform detection are 50 ohm, equal to the characteristic impedance of the
cable used, so that distortion in wave-form is avoided. The input resistance
of the dual beam oscilloscope (Tektronics 7844) is 50 ohm, and rise time of
the measuring circuit is about 2 ns. The needle electrode is applied with a
negative d.c. high voltage in this measurement. The effect of surface resistiv-
ity of the sample is studied by locating the electrode system inside a thermostat
in which the humidity can also be controlled.
2.3 Streak photograph measurement
Propagation of the back discharge streamer is also measured by a streak
photograph method using an image converter camera (John Hadland, Ima—
Con) combined with an image intensifier tube (EMI, type 9912) having a
maximum gain of 106. Fluctuation in the period between successive back dis-
charge pulses, especially when d.c. high voltage is used, is very large (of the
order of 100 /LIS) compared to the duration of the phenomenon (shorter than
1 jus). Therefore, two different triggering circuits are used to synchronize
the image converter camera accurately to the start of the back discharge to
be observed. Figure 1 (b) shows the one which is used when d.c. high voltage
is applied to the needle electrode. The measuring electrode is connected to
one end of a delay cable, and its other end is grounded through a small spark
gap. Then, the spark in the gap, caused by the increase in voltage at the mea-
suring electrode, results in an earth potential appearing at the electrode after
a certain delay time, so that back discharge is triggered. This spark signal is
also fed to the image converter camera so that it is completely synchronized
to the phenomenon. In this measurement, the electrode system is located in-
side the thermostat. Figure l(c) shows the second triggering circuit. A nega-
tive d.c. high voltage is applied so that a negative corona occurs at the needle
electrode and a faint onset-glow [1] occurs at the sample. Then, a single or
periodic negative pulse high voltage is applied superimposed to the negative
d.c. voltage. After a certain formation time, the back discharge is triggered.
The pulse signal is also fed to the image converter camera. In order to study
the triggering mechanism in more detail, the measurement is performed under
43
-------
vacuum, using a pulsed high voltage alone. The circuit shown in Fig.l(c) is
used also in this case, and the electrode system is located inside a vacuum
chamber.
3. Results obtained
3.1 Negative d.c. high voltage application
Measurements are performed using a negative d.c. high voltage to be ap-
plied to the needle electrode, with the glass plate sample on the plane elec-
trode.
3.1.1 Mixed streamer mode
The authors reported that back discharge in the form of surface streamer
becomes dominant when the sample surface has a charge density a0 higher
than about 5 X 10~9 C/cm2 at the instant of breakdown, and a sufficiently
high surface resistivity [1]. In the present experiment, the value of o0 =
of the sample used is about 5 X 10"9 C/cm2, where e is the dielectric con-
100
(1) Mixed Streamer Mode
( R, « 3.5 x 10U 11 )
(If) Space Streamer node
( R, » 2.0 x 101' u )
(1U) A1r Load
13
-5
£1.0
B
3-
0.1
0 100200
time (ns)
(a) z-AXIS
0 100200
time (ns)
(b) r-AXIS
10 15 20
VOLTAGE ( kV )
Fig.2. Voltage—current curves of back discharge for different modes. Sample: glass plate
with a pinhole.
Fig. 3. Change in light signal of back discharge in space (see Fig. l(a)).
44
-------
(a) LIGHT SIGNAL AT THE BREAKDOWN POINT
(z = 0, r = 0) AND CURRENT WAVEFORM
( V = -28 kV, I = 13
DISCHARGE PULSE
(b) LIGHT SIGNAL AT THE TIP OF THE
NEEDLE ELECTRODE (z = 48 mm)
Fig.4. Light and current waveform of back discharge in the mixed streamer mode.
45
-------
stant of the sample. When the surface resistivity Rt is set to 3.5 X 1014 ohm
by reducing ambient humidity (R.H. = 20%, T = 40°C), both surface and
space streamers occur to form the mixed streamer mode [1]. The following
measurements are performed under these conditions of humidity, tempera-
ture and atmospheric pressure inside the thermostat.
A voltage—current characteristic in this case is shown by curve (i) in Fig.2.
The light measurements described below are performed for the back discharge
occurring when the applied voltage is —26 kV. The change in light signal in
the normal direction z and in the tangential direction r are shown in Fig.3
(see Fig. l(a)>. The light signal at the breakdown point (z = 0, r = 0) shows
that the back discharge pulse in the mixed streamer mode consists of two
parts. There is a "primary light wave" which rises very rapidly and lasts
about 20 ns, followed by a "secondary light wave" which rises more slowly
and lasts about 200 ns. The former proceeds hi the 2-direction and the latter
in the r-direction. The interval between the primary and secondary light
waves, however, varies widely, depending upon the value of applied voltage
and surface resistivity. It can be seen that the primary light wave advances
towards the needle electrode with a speed of about 4 X 107 cm/s, while the
secondary light wave propagates along the sample surface with a speed of
about 2.5 X 107 cm/s. Figure 4(a) shows an example of the typical light and
current waveforms of back discharge in the mixed streamer mode. The
primary light wave corresponds to the first rise in current waveforms, as indi-
cated by P in Fig.4(a), having a small pulse height and a charge content of
1—2 X 10~9 C/pulse, as described in Section 3.1.2. The secondary light wave
corresponds to the second rise in current, as indicated by 8 in Fig.4(a) which
has a much larger pulse height and a charge content of 2—4 X 10"* C/pulse.
Table 1 shows the charge content per pulse in this mode for various applied
voltages. The charge content per pulse remains almost the same even when
the applied voltage is changed, so that the change in current results from
the change in the average repetition frequency of the back discharge pulse.
The light signal at the tip of the discharge electrode is shown in Fig.4 (b).
The first pulse, indicated by A, occurs at the same instant as that for a back
TABLE 1
Charge per single back discharge pulse (mixed streamer mode)
Voltage (kV) Q(C) Ta (ms) /-Q/T,(A) /measured (A)
16
18
22
25
28
2.0 X 10'1
2.0 X 10'*
2.8 X 10'8
3.0 X 10'*
2.8-3.6 X 10-»
60
34-36
12
6.5
3.2-3.5
0.34 X 10'*
0.6 X 10'*
2.4 X 10'«
4.6 X 10"
1.08 X 10"*
1.3 X 10"
2.5 X 10-
4.4 X 10-
8.0 X 10-
1.3 X 10-
Ta: average pulse repetition period
46
-------
•o
1
0
*~ 17 mm —
.L.
-
0 TOO 200 300
time (ns)
(a) STREAK PHOTOGRAPH OF PRIMARY AND SECONDARY LIGHT WAVES
(P:760 Torr, V:-28 kV, 1:13 yA) (see Fig.l-b)
v/N
•
0
in
•o
1
0
" 45 mm ~"
0
1
time
(b) BACK DISCHARGE TRIGGERED BY APPLYING PULSE HIGH
VOLTAGE IN ADDITION TO -D.C. VOLTAGE
(P:410 Torr, Vp:-10 kV , T :10 us, D.C.:-15 kV)
(see F1g.l-c)
Fig.5. Streak photograph of back discharge in the mixed streamer mode. N: needle elec-
trode, O: breakdown point, P: primary light wave, S: secondary light wave.
discharge pulse and has the same shape as the light signal shown in Fig.3 at
z * 48 mm when the time scale is magnified. The other pulses in Fig.4(b)
have an entirely different waveform which corresponds to that of the Trichel
pulse.
Streak photographs taken by the method described in Section 2.3 are
shown in Fig. 5. Figure 5(a) is a side view obtained by the method in Fig.l(b).
This photograph clearly shows the development of primary and secondary
47
-------
light waves. Figure 5(b) is also a side view obtained by the method in Fig.l(c)
where pulsed high voltage is applied to the d.c. high voltage. In this case, the
measurement is performed under P = 410 torr, because the triggering proves
to be difficult under atmospheric pressure. Dry air is used inside a vacuum
chamber so that the surface resistivity can be kept sufficiently high and the
mixed streamer mode appears. These photographs indicate that, once the
onset-glow mode has turned into the streamer mode, the continuous glow
disappears and the streamers are triggered each time by the layer breakdown.
The reproducibility can be much improved by the method of Fig. 1 (c). As a
result, it is found that secondary light wave appears at the instant when the
primary light wave nears the discharge electrode. It should also be pointed
out that the pulse repetition period in the mixed streamer mode is about
two orders of magnitude larger than that in the space streamer mode to be
described later.
3.1.2 Space streamer mode
As reported separately, back discharge in the space streamer mode appears
when the vertical field strength exceeds about 5 kV/cm and the value of o0
= e£ds is comparatively small [1]. In this experiment, however, it is found
that, as long as the surface resistivity R3 is low, the space streamer mode oc-
curs even though the value of a0 exceeds about 5 X 10~9 C/cm2. The voltage-
current characteristic in this case when Rs = 2 X 1011 ohm is shown by
curve (ii) in Fig.2 and the light signal from the breakdown point, and current
waveform, are shown in Fig. 6 (a).
In this mode, the secondary light emitting spot is very weak and occurs
randomly so that it cannot clearly be observed by the photomultiplier. The
charge content per single current pulse is about 10"' C/pulse. The repetition
period, however, is much smaller (50—300 jus) than that in the mixed streamer
mode, and hence the current is several times higher than that in the latter
mode (see Fig.2). Figure 6(b) shows a streak photograph of back discharge
in side view in the space streamer mode, which is taken under P = 410 torr.
The light under atmospheric pressure is too weak to be seen even with the
aid of the image intensifier, unless the applied d.c. high voltage is excessively
high. This high d.c. voltage causes an instability in the streamer which results
in random sparking. In addition, its repetition period under atmospheric
pressure becomes larger than the time frame of the streak camera. All these
problems can be solved when the pressure is lowered to the value used of
410 torr. In this case, room air (R.H. = 76%, T = 20°C) is also used in the
vacuum chamber so that a sufficiently low value of Rs is obtained. Three
successive back discharge glows are seen in the photograph. It should be
noted that glow at the needle electrode does not appear at the time when
back discharge glow disappears. It can be seen on the sample side that the
space streamers occur at first, followed by a faint glow at the breakdown point
which corresponds to the weak secondary light wave described above.
48
-------
(a) LIGHT SIGNAL FROM THE BREAKDOWN POINT
AND CURRENT WAVEFORM
(P: 760 Torr, V: -22 kV, I: 12 UA)
1
o
£
0
- 45 mm-
10
50 60
20 30 40
time (us)
(b) STREAK PHOTOGRAPH FROM SIDE VIEW
(P: 410 Torr, V: -10 kV)
Fig.6. Light and current pulse and streak photograph of back discharge in the space
•treamer mode.
3.1.3 Flashover
The flashover from back discharge in the streamer mode is studied by the
streak photograph method described in Fig.l(c), where pulsed high voltage
(Vp - —10 kV, TP = 1.0 ms) is applied to the d.c. high voltage of —15 kV.
Room air (R.H. = 76%, T = 20°C) is used inside th«- vacuum chamber as be-
fore. The light signal at the breakdown point and current waveform are
shown in Fig. 7(a). There are two stages, A and B, the former corresponding
to the back discharge pulses at its initial stape, the latter to those at its final
49
-------
•PULSE WIDTH: 1 ms
FLASHOVER
URRENT
(a) LIGHT AND CURRENT SIGNALS AT FLASHOVER
(P: 410 Torr, Vp: -10 kV, T : 1 ms, D.C,
-15 kV)
N
1
i
0
•o
1
>- 45 mm -
10
40
50
20 30
time (us)
(b) BACK DISCHARGE TRIGGERED BY APPLYING PULSE HIGH VOLTAGE
N
i
o
0
-45 mn —
10
4C
50
20 30
time (us)
(c) STREAK PHOTOGRAPH OF FLASHOVER CAUSED DY BACK DISCHARGE
FROM SIDE VIEW
Fig.7. Flaihover caused by back discharge (see Fig.l(c)).
50
-------
stage. The large and continuous light signal in stage A is due to a saturation
in the photomultiplier tube used. The streak photographs in stages A and B
are shown in Fig.7(b) and (c), respectively, where the light intensity in the
latter is reduced to one half of the former by an iris. When the pulsed voltage
is applied, the back discharge in the space streamer mode is triggered after a
certain formation time of about 50—100 /-is. It should be noted that the first
streamer is highly luminous owing to the full voltage appearing between two
electrodes. This streamer disappears when the voltage stored in the capacitance
between the electrodes falls. The following streamers are much weaker because
the triggering of the streamers — the sample breakdown — could happen be-
fore the electrodes have been charged up to the source voltage. However,
during the course of repeated streamer discharge, the streamer channel could
be sufficiently heated up and localized to form a "leader" [4]. The leader
proceeds along the streamer channel towards the discharge electrode, and
finally turns into a flashover at point C, which should be taken as the high
voltage arc in this case.
3.2 Negative pulse high voltage application
3.2.1 Streak photograph of back discharge
Streak photographs of back discharge are taken when a periodic negative
pulsed high voltage having a square waveform is applied. The method de-
scribed in Fig. l(c) is used where the d.c. high voltage source is removed.
Figure 8(a)(i) indicates the waveform of the pulse voltage used, and Fig.
8(a) (ii) its initial part. The pulse rise time is 0.5 /us, its height Vp is —25 kV,
and its width rp (10 /is) is much longer than the time-scale of the phenomena
observed. The electrode gap is 50 mm. A mica plate with a pinhole of 0.5-
mm diameter and 0.2-mm thickness is used as the test layer. Figures 8 (b)(i)
and (ii) indicate the results obtained where the time-scale is changed. A
corona glow appears at first at the needle electrode at least 500 ns after the
application of the pulse voltage. Hence, the needle is applied with the full
pulse voltage when the corona glow occurs. This glow lasts for about 100 ns,
emitting electrons which will be attached to electro-negative gas molecules
to form a dense negative space charge around the needle tip. The negative
corona is thus quickly choked. After a delay time of 400-800 ns, the
primary light wave of back discharge starts to occur, triggered by the accu-
mulated negative charge on the sample layer. This delay time, which may be
the transit time of the carriers from the needle to the sample layer, is too
short to be explained by ion transit time, suggesting the role of electrons
for triggering carriers in this case of pulse voltage application, as described
in the next section. The propagation speed of the primary light wave is
about 5 X 107 cm/s, and the secondary light wave appears again when the
primary light wave approaches the needle electrode. Once the primary light
wave reaches the needle electrode, a continuous second glow corona appears
at its tip. Figure 8 (b) (ii) shows the photograph taken with a much lower
-------
(i)
(11)
(a) WAVE FORM OF APPLIED PULSE HIGH VOLTAGE
(i)
-
100 200 300 400 500
time (ns)
2 3
time (us)
(b) STREAK PHOTOGRAPHS OF BACK DISCHARGE l/ITH
PULSE HIGH VOLTAGE (sec Fig.l-c)
(d: 50 rm, P: 170 Torr, Vp: -25 kV, i : 10 us,
f: 10 Hz, nica plate with a pinhcle)
Fig.8. Back discharge with pulse high voltage application. P: primary light wave, S:
secondary light wave, R: return light wave.
52
-------
streak speed. The second glow appearing at the needle electrode moves
gradually towards the plane electrode with a speed of about 0.7 X 107 cm/s.
This glow should be referred to as the "return light wave" [5]. When the
voltage is further raised, both the secondary light wave and return light wave
proceed into space to form an intensive glow in the middle of the gap and
cause flashover as described later. The propagation velocity of the primary
light wave and the intensity of the secondary light wave vary with the change
in the pulse voltage as shown in Fig.9. The velocity of the primary Hght wave
is about 5 X 101 cm/s for V'p = - 25 kV, 3 X 10' cm/s for Vp = -20 kV, and
2 X 107 cm/s for Vp = —15 kV. The velocity of the primary light wave, as
well as the intensity of the secondary light wave, increase as the voltage is
raised. It should also bo noted that the secondary glow at the needle tip dis-
appears when the voltage is lowered
(i)
Vp: -25 kV
(ii)
Vp: -20 kV
(iii)
Vp: -15 kV
I
0 100 200 300 400 500
time (ns)
Fij.9. Streak photographs of back discharge with pulse. /': 160 torr, rp: 10 MS, /: 10 Hz,
sample: mica plate with a pinhole
53
-------
0 100 200 300 tine (ni)
(i) P: 36C Torr
.
0 100 200 300 time (ns)
(ii) P: 410 Torr
.
100 200 300 tine (ns!
(lii) P: 510 Torr
(a) STREAK PHOTOGRAPHS
i) P: 360 Torr
(ii) P: •
(in) P:
(b) LIGHT NT SIGNALS
Fig.10. Change in trig^crin^ delay tim«- \'p: -20 kV, -p lit MS, / in H/, Dimple: mica
plate with a pinli
54
-------
3.2.2 Triggering delay of back discharge and triggering carriers
There exists a triggering delay time r& from the first glow at the needle
electrode to the initiation of back discharge. During this delay time, carriers
are considered to migrate across the gap. Carriers are considered to be elec-
trons because of an extremely high velocity estimated from d/T&. This value,
estimated from Fig,8(b), is 0.6—1,2 X 107 cm/s, which is about three orders
of magnitude higher than that for negative ions [6]. The velocity of elec-
trons is given in [7] as a function of E/P. Taking the average field intensity
E - Vp/d, we get E/P • 31.0 V/cm.torr in this case, giving an approximate
value of electron velocity of about 1.2 X 107cm/s.
In order to confirm the carriers to be electrons, the triggering delay time
is measured using streak photographs as well as oscillograms of the current
waveform, and is compared with the carrier transit time measured separate-
ly. Figure 10 (a) shows the streak photographs obtained under different
pressures, where the electrode gap is 20 mm, and the mica plate with a pin-
hole is used. It is checked every time that the corona glow appears at the
needle tip after the pulse has reached its peak voltage. In this measurement,
room air (R.H. • 61%, T » 17° C) is used inside the vacuum chamber as be-
fore. Figure 10(b) shows current waveforms and light signals from the whole
gap, measured under the same conditions but not simultaneously with the
streak photograph. The first rise in current corresponds to the glow at the
needle tip, while the second rise corresponds to the occurrence of back dis-
charge streamer. These waveforms enable more accurate evaluations of r&
which agree very well with those estimated from the streak photographs.
The values of r-
g 100 ns
*
I
10 ns
I I I I
310 360 460 560 660
PRESSURE P (Torr)
Fig.ll. Triggering delay time and carrier transit time as a function of pressure.
55
-------
|-o PULSE GENERATOR
Q.C. -2 kV
H: NEEDLE ELECTRODE
G: GRID ELCCTRODE
P: PLANE ELECTRODE
CRO: CATHO&E RAY OSCILLOSCOPE
(Vp: -22 kV. i : 10 ws, f: 10 Hz)
(a) MEASURING SYSTEM
(ii) P: 410 Torr
v; v.
[NT SIGNALS
Fig. 12. Carrier transit time measurement.
56
-------
as a function of pressure P. An abrupt increase in rd, more than two orders
of magnitude, occurs at P - 560 torr, and suggests a sudden decrease in
carrier mobility. The width of the pulse voltage is raised to 100 MS above
this pressure. It is expected that, with the decrease in gas mean-free-path, the
range of free electrons emitted from the needle tip is lowered, and finally
becomes shorter than the electrode gap at P > 510 torr. Above this critical
pressure, the mobility of ions produced by electron attachment may govern
the value of TJ.
Figure 12(a) indicates the electrode system for measuring the carrier
transit time Tf. In order to suppress the disturbing effect of the displacement
current caused by the movement of carrier space-charge, a grid electrode
with 0.5-mm-square mesh is located near the plane electrode to cover its
whole surface. The distance between the plane and the grid electrodes is
2.0 mm. The gap between the needle and the grid electrodes is 20 mm, equal
to that in Fig.10. A d.c. voltage of —2.0 kV is applied to the grid electrode
to drive the incoming carriers to the plane electrode and a capacitor with
4.0 pF capacitance is connected in parallel to eliminate the effect of tran-
sient fluctuation in grid potential. In this measurement, the mica plate is re-
moved, and a pulse voltage with Vp = —22 kV, rp = 10 MS, f = 10 Hz, is ap-
plied to the needle electrode. The pulse voltage appearing between the
needle and grid electrodes is —20 kV, equal to that between the needle and
the plane electrodes in Fig.10.
Figure 12(b) shows the current signal obtained at the plane electrode under
different pressures. These signals should correspond to the arrival of the
carriers. Light signals from the needle electrode are also indicated as the time
origin. The time elapsed from this origin to the peak of the current signal
may be taken as the carrier transit time, 7>. The values of rt obtained are
plotted against pressure P in Fig.ll. A fairly good coincidence can be seen
between r& and rt in the pressure range lower than 510 torr. At P = 610 torr,
a very large discrepancy appeared between TA and r^. It should, however, be
noted in Fig.l2(b) (iv) that the peak value of current in this case becomes
very small. This suggests that the number of incoming fast carriers becomes
too small to trigger the back discharge in the case of TA measurement. It is
expected that the second peak may appear in Fig.l2(b)(iv) at a much longer
time delay beyond the frame, corresponding to the ion transit time.
Figure 13 indicates the values of djr^ for P < 510 torr and dfrt obtained
from Fig. 12, the approximate values of the carrier velocity, as functions of
E/P. The electron velocity taken from [7] is also indicated. The very good
agreement between these values confirms that the fast carriers triggering back
discharge in the low pressure range are electrons. The pressure, above which
electrons cannot trigger back discharge, should be referred to as the critical
pressure Pc. In this case, Pc is 510 torr. It should also be added that free
electrons, smaller in number, also arrive at the plane electrode in the pressure
range higher than Pc, although the triggering may be effected by ions in this
range.
57
-------
1,0
i
10 20 30
E/P
Fig. 13. Carrier velocity as a function of £/P.
3.2.3 Back discharge in N2 and SF6 gases
The results described so far clearly indicate the importance of electron
affinity of gas molecules encountered to determine the critical pressure Pc.
In addition, it is known that streamer propagation is also strongly affected
by the gas electron affinity. These effects are studied in more detail using
Nj having no electron affinity and SF6 gas with a very high electron affinity.
A streak photograph of back discharge in N2 gas, when the pulse voltage is
applied, is shown in Fig. 14 (a). In this case, the chamber is evacuated at first
to about 1 torr, and, thereafter, N3 gas (99.99%) is introduced up to P =
310 tonr. It is observed that the triggering delay in this case is 200 ns, much
higher than that for air, but still indicating the carriers to be electrons. A
faint glow continues to exist at the needle tip after the initial strong glow.
There exists a certain delay from the pinhole breakdown to the initiation of
the back discharge streamer, whereas no delay time exists between the prim-
ary and the secondary light waves in contrast to the case in air.
The streak photograph of back discharge occurring in SF6 gas is shown
in Fig.14 (b). SF6 gas is introduced into the vacuum chamber in the same
way as in the case of N3 gas. Even at a low pressure of P - 160 torr, the
triggering delay amounts to 4 /us, much larger than the expected value for
electrons. It is expected that the critical pressure Pc lies at a much lower
value than 160 torr in this gas with a very strong electron affinity, The small
value of 4 /us triggering delay may be attributed to the increase in ion mobil-
58
-------
N
200 400 600 800 time (ns)
(a) N2 gas
(P: 360 Torr, Vp: -20 kV, T : 10 us)
0 1 2
(b) SF6 gas
time (us)
(P: 160 Torr, Vp: -24 kV, T : 10 us)
Fig. 14. Back discharge with pulse under Na and SF, gas. Sample: mica plate with a pin-
hole.
ity at the reduced pressure of P = 160 torr. It should be noted that the
streamer propagation into space is highly suppressed, and the secondary
light wave completely disappears.
3.2.4 Flashover
The streak photographs of flashover occurring in air and N2, when a single
pulse voltage is applied, are shown, respectively, in Fig.l5(a) and (b). There
is no remarkable difference in character of the two pictures, although the
former is blurred by a strong halation. After the glow at the needle tip, the
back discharge streamer is triggered from the pinhole. At the instant the
streamer nears the tip, the second streamers are launched from both sides
to meet at the middle point and finally to turn into a highly luminous chan-
nel of a flashover.
59
-------
0 1
(a) AIR
(P: 310 Torr, Vp:
3 4 5
time (ps)
-20 kV, T : 10 us)
0 10 20 30 time (us)
(b) N2 gas
(P: 260 Torr, Vp: -25 kV, T : 50 us)
Fig. 15 Flashover caused by back discharge with pulse. Sample: mica plate with a pinhole.
4. Conclusions
The following conclusions are obtained from the light measurements of
back discharge:
(1) Ir the mixed streamer mode occurring in air under atmospheric pres-
sure, the light signal consists of primary and secondary waves. The primary
light wave corresponds to a space streamer and the secondary light wave to
a surface streamer, each resulting in a different current pulse. The former
has a charge content of 1—2 X 10"9 C/pulse, while that of the latter is 2—4
X 10'8 C/pulse, both remaining almost constant, independent of applied
voltage. The period of successive back discharge pulses, however, becomes
smaller as the voltage is raised. This mode of back discharge occurring under
atmospheric air is considered to be triggered by negative ions supplied from
the discharge electrode.
(2) In the space streamer mode, the primary light wave is dominant, while
60
-------
the secondary light wave is extremely weak. The charge per single current
pulse is about 1—2 X 10~9 C/pulse. The pulse repetition frequency in this
mode, however, is one or two orders of magnitude larger than that in the
mixed streamer mode, so that total current in the former mode becomes
larger than that in the latter under equal applied voltage.
(3) Flashover seems to be caused by the leader developing along the channel
left by the preceding streamers when d.c. voltage is applied.
(4) When a sufficiently high pulse voltage is applied under lower pressure,
back discharge is triggered by free electrons supplied from the needle elec-
trode. As the pressure is increased, an abrupt change in the triggering car-
riers occurs from electrons to negative ions. The electron affinity of gas
molecules is a major factor in this process.
Acknowledgement
This research was sponsored by the Ministry of Education, Japan, as its
Special Research Project (I) (Project No. 011914). The authors are grate-
fully indebted for its support.
Nomenclature
d electrode gap
£& breakdown field strength of the layer
£ average field strength of the gap (Vp/d)
f frequency of pulse
P pressure
Pe critical pressure for the change in the triggering carrier
R.H. relative humidity
R, surface resistivity of the layer
T temperature
Ta average pulse repetition period
Vp peak value of pulse voltage
e dielectric constant of the layer
Pd apparent resistivity of the layer
GO surface charge density at the instant of layer breakdown
rd triggering delay time
TP width of pulse voltage
rt carrier transit time
N needle electrode
0 breakdown point
P primary light wave
R return light wave
S secondary light wave
References
1 S. Masuda and A. Mizuno, Initiation condition and mode of back discharge,
J. Electrostatics, to be published.
2 S. Masuda and A. Mizuno, Flashover measurements of back discharge, J. Electro-
statics, to be published.
61
-------
3 8. Masuda, A. Mizuno and K. Akutsu, Initiation condition and mode of back discharge
for extremely high-resistivity powders, to be presented at the 1977 Annual Meeting of
IEEE IAS, Los Angeles, October 1977.
4 I. Tsuneyasu, Observations of air breakdown in positive point to plane gaps under
impulse voltage and its mechanisms, J. Inst. Electr. Eng. Jpn., 96B (1976) 63.
5 N. Ikuta, T. Ushita and Y. Ishiguro, Positive streamer corona and its propagation mech-
anism, J. Inst. Electr. Eng. Jpn., 90 (1970) 1816.
6 E.W. McDaniel and M.R.C. McDowell, Low-field mobilities of the negative ions in
oxygen, sulfur hexafluoride, sulfur dioxide, and hydrogen chloride, Phys. Rev., 114
(1959) 1028.
7 H. Ryzko, Drift velocity of electrons and ions in dry and humid air and in water
vapour, Proc. Phys. Soc., 85 (1965) 1283.
62
-------
INITIATION CONDITION AND MODE OF BACK DISCHARGE FOR EXTREMELY HIGH RESISTIVITY POWQERS
Stnicht MASUDA, Aklra MIZUNO and Kenauke AXUTSU
Department of Electrical Engineering, University
of Tokyo, 7-3-1, Hongo, Bunkyo-Ku, Tokyo, JAPAN
Summary
Initiation condition of back discharge occurring
la easa of extremely high resistivity powders, Includ-
ing tht Cflco. However, it was considered chat the
Initiation condition might become entirely different
because of the layer space charge when the pj- value
becomes extremely high and exceeds the level of 10 14 -
1015 Ocm. The appearence of limiting thickness itself
means that, in spite of a very high pd-value, the back
discharge does not occur until a certain layer thick-
ness is exceeded, suggesting th« thickness to be
another factor. The authors observed on the other
hand that in case of very high o^-value the back
discharge could be resulted even in the absence of lj
supplied by ionic current, ^6^ which is to be described
in the next section. According to the authors' obser-
vation, the mode of back discharge in this case was
also somewhat different froa those In the case of
electrostatic precipitation. Hence, the initiation
condition and moda of back di?cha'k;e ware studied for
the case of extremely high reuiscivity powders. W
2. PRILIHINALY OBSERVATIONS
It is studied at first whether back discharge can
taki place without the aid of ionic current when tha
Pj-value becomes extremely high. Fig.l shows tha
experimental apparatus."' In ordsr to rajaet all tha
possibility of ionic currant to be supplied, two sphere
electrodes with 14 cm diameter are used inside a shield-
ing chamber (1m x IB » 1m). The total stray currant
can ba kept below 0.1 • 0.3 nA under the applied volt-
'age of 50 kV. Polyethylene powder with PJ « 1015 (1cm
is negatively charged by tribo-electrlficatlon and fed
with the aid of air flow from a plastics nozzle onto
the surface of the lower electrode B, where both elect-
rodes are grounded. After the powder supply is stopped,
the electrode B is connected to either a negative or a
positive high voltage source. When it is connected to
a positive H.V.source, the powder adheres to the olect-
rode surface. At a certain voltage, back discharge is
detected with an image intenslfier tube (EMI, type 991"),.
and it turns into a spark at +40 kV. When a negative
voltage is applied to the electrode B, all the powder
violently jumps up to deposit on the surface of the
opposite electrode A. These phenomena can occur after
several hours of powder deposition, but completely
disappear after 40 hours.
..11
":*-?••; •'• •••.-.•^. 10
Flg.l Experimental Apparatus (I) for Prellalnary
Observation
Another experiment is performed to examine these
phenomena in more detail, using the apparatus shown
in Fig.2. The powder, negatively charged by tribo-
electrification, Is fed into a space between two para-
llel electrodes applied with a dc high voltage. The
powder Is separated in the afield to deposit on the
surfaces of both electrodes. This is because a irtaorlty
portion of the powder is charged positively, altho-ixh
the net powder charge is negative. After several 10
seconds back discharge starts to occur at the upper
area of the positive electrode at which the nxount of
deposition is maximum. It triggers a new back, disobar;;i
to occur at the opposite area on the negative c lev ".roc!..,
owing to the copious ions supplied fron the initial
back discharge point. These back discharges pro1.:•,•.-!.a
gradually downwards on both electrodoi towards t'-f
lower areas. This phenomenon, an •.•<* "j_..,, disrharyu
propagation" by the authors, also occurs in (•'.'Aero-
static precipitators whsn the pj-valu* axc
It is observed that the lay»r chickens at t'.:
initiation of back dischai-ce is much larger in t'. • ,'.i > •
IAS 77 ANNUAL
63
-------
Fig. 2
Experimental Apparatus (II) for Preliminary
Observation
two cases compared with the usual case of powder
coating with ionic current supplied.
These two experiments clearly indicate that back
discharge can take place without the aid of ionic
current when the pd-value exceeds the level of 1015 flea
and a sufficiently high external field Is applied in
a favourable direction. This suggests the necessity
of formulating its initiation condition on a »ore
general basis.
3. THEORETICAL CONDITION OF BACK DISCHARGE INITIATION
W« assume the continuous and uniform flows of
charged powder particles and ions coming from a gas
space perpendicular onto a grounded plane electrode to
form a uniform powder layer on It (Fig. 3). The layer
thickness is growing with a constant rate b (m/s).
We calculate die field distribution Inside the layer,
starting from the following fundamental equations in
one-dimensional case:
POWER LAYER
U
I
•— x — J
. • i
'.'•:'. ' •'.'
.1.
i
«!
illhlllll
'I'l'l'l'l
U
'REED PARTICU SOURCI
§
5
One-Dimensional Model of Powder Deposition
Process
Fig. 3
1) Continuity equation:
-div Td - 3idx/3x - 3q
ii) Polsson's equation:
^^d^d -
ill) Ohm's law:
- -td9Edx/3x
(2)
(3)
Edx/Bd
where q, - charge density inside the layer, ed - appa-
rent dielectric constant of the layer, and l,jx and E^
arc the x-compor ents of current density Td and field
intensity ?d insi.de the layer respectively, taken
positive to the left direction.
He take the following boundary conditions;
(I) Boundary between the layer and gas space:
a) Position of the boundary:
X - bt
(m)
(5)
b) Current density at the boundary:
(outside the layer)
1 « i.
o i
• constant
(inside the layer)
i,,(X) - E. /p.
dx dx d
(A/m2)
(A/m2)
c) Field intensity at the boundary:
Edx ' Eext
constant
(V/m)
(6)
(7)
(8)
(II) Boundary between the layer and the plane electrode;
d) Potential at the boundary:
Ud(0) - 0
(V) (9)
• ionic current density, 1 • parti-
where t - time, it - ionic current density, 1 •
cle current density, lo - total current density,
q_ - charge of a single particle, a - particle radius,
t - packing ratio of the layer, E-xt - externally
applied field, U
-------
Fig.4 Field Dlstrlbuclon inside a Deposited
Layer (general case; XQ - Ld, tQ - Td)
Pig.4 depicts the field distribution inside the layer
•t the instant t - Td, 2*d and 3rd respectively. It
can be seen that the field distribution is an exponen-
tial function of time and depth, propagating with a
velocity b In the x-direction. The maximum field
strength occurs at x - 0 (the boundary between the
layer and the plane electrode). It Is therefore
expected that back discharge may start from this bound-
ary when this maximum field strength Emax exceeds the
breakdown threshold value Eds. Hence the general form
of initiation condition of back discharge should be
' E
ext
or
max " ext
*Eds
<
'xt
t- t
_^
t- ZT
—
«-*d
X
Fig. 5
X • bt
Field Distribution'inside a Deposited
layer (low resistivity case - Case(I);
Two different cases may be of special interest.
One Is the Case (I) where the pd-value is comparatively
low as In the case of electrostatic precipitation.
the other is the Case(II) where the p. -value is
extremely high as In the case of electrostatic powder
coating.
Fig.6 Field Distribution Inside a Deposited
Layer (high resistivity case - Case(II);
Case (I): In this case, the value of Ld becomes very
small and conditions X. »Lj and to » Td are fulfilled,
so that Eq.(23) and (24) reduce to Eq. (1) where the
effect of layer thickness disappears. The field distri-
bution In this case is shown in Fig. 5, which indicates
the effects of the layer space charge and external
field to be negligibly small.
Case (II) : In this case, the condition X0 « Ld and
t0 « TJ are fulfilled, so that we get from Eq.(23)
and (24)
E»a
ext
E
VLd • *o/Td
(26)
' (tds - Eext>'
-------
electrode and the ground.
It Hhould finally be pointer! out that Eq. (25) and
(27) alao suggest the back discharge to occur even
without the supply of ionic current It when the layer
thickness X beeooec sufficiently large. It la expected
that the layer apace charge qo - lQ/b in thia case
•ay be fairly low so that, unless the particle charge
4p is kept sufficiently high, the adhesion of the layer
•ay be lower.
The field distribution in this case is depicted
in Fig.6. The higher the total source current density
10 compared with the powder deposition rate b. the
steeper becomes the slops of the curves, this being
I0/«db wh«n *o°d * Eexf
4. EXPERIMENTAL VERIFICATION
In the case of TJ » t0, an experimental verifica-
tion of Eq. (23) and (24) are made by comparing the
experimental value of £„._ froa Eq. (30) at the Initia-
tion of back discharge with the value of Ej, measured
separately using parallel plane electrodea. This
verification la made Intentionally for the extreme
case where the ionic current ij. is kept negligibly
rail! compared with the particle current i_ so that
we can asaume ig • i.. Pig. 7 shows the experimental
apparatus which meets this requirement and allows an
independent control of Eaxc and i. (« i_). The value
of t|Xt can be changed by the grid potential V., while
the value of 10 by changing the discharge electrode
voltage V* of powder gun as well as the powder feeding
rate V.
record
age * o power gun as we as e power eeng
V.. The value of /5°U0(O/ej}dt is measorsd by
rding the capacitor voltage Vc, where
where C • capacity • 10"' V and S • surface area of the
measuring electrode. The spoxy-resin powder, having a
pj-value of 2 M 10™ Ocm and a specific dislectrie
conitant c( • 2.9 ia used, The site of the powder
particle is la the range of 0.01 - 0.09 n. The
experiment is conducted under a constant temperature
and relative humidity (20 C* and 40 I) . for the powder
•ample ussd, we get i^ • 2 • 10 ' a, so that the condi-
tion for Case (XX) (14 » t0i !• iatlsfled. The (tart
of back discharge ia detected by it* accompanying light
•aiiaion wins th« imags inteniifier tuba (SKI, type
9912)
lucTwec
El ::
Kg. 7
MIMR (CURATOR
Cl i COmiltOR
6 ! MMURINI CORDHHR
V i VIIMTIM HID lUCTRMTn
« . RieoniR
N i MIRROR
II, INK INTIRimiR TIM
BlMtraititia Fewdor tepoiition Apparatui
for IxpKlMatal VHtfipatim of lack
DUeharge Initiation Condition
The values of the capacitor voltage Vc and the
layer thickness X at the initiation of back discharge,
denoted by (Vc)0 and X<, respectively, are measured for
various values of 10, thereby changing the layer space
charge density qo. From the value of (Vc)0 and Eext,
the value of EMX at the initiation of back discharge,
(Enax)0, is calculated using Eq.(30). The measurement
of Xj, is made by cutting the layer and observing its
cross-section with a microscope.
The relationship between (EBax)o *nd *o measured
in the wide renge of io - lO"10 - 10"8 A/cm2 and
powder feeding rate W. - 10 - 80 g/mln la plotted in
Fig.8, The value of Eds measured by parallel plane
electrodes, denoted by Ejg", is also given, which
decreasss with layer thickness. It can bs seen that.
In aplte of such a wide variation in qe made in this
experiment, the value of (Eg^o remains almost constant
in the thickness range between 0.3 and 1.0 mm investi-
gated, and further that its value agrees very well with
the value of E]7 at a small thickness of about 0.1 mm.
Thia supports the theoretical conclusion that back
discharge should initiate from the limited srea adjacont
to the plane electrode.
The experiment in the smaller thickness range la
oaltted because of its difficulty. Mien the ionic
current is mads negligibly small, no back discharge can
be reaulted with the powder sample used in the thickness
range lower than 0.2S mm. If the Ionic current ie
supplied, a fairly large error cannot be avoided because
of by-pass current to flow directly to the plane elect-
rode at the start of powder deposition and the Uniting
time to to become very short.
However, the results so far obtsinsd seem to
provide a sufficient support for the validity of the
theoretical initiation condition of back discharge
described in the prscsding section.
ISOk-
100
L«
1
'dt
I
o o.i
. LAYER THICKNISi I, (««•)
Fig.I Ceapariion between (Bflajf)e *&*
1.0
66
IAJ 77 ANNUM
-------
S. HOPE OF BACK DISCHARGE
The code of back dlscharage (or extreaely high
resistivity powders is observed using an electrostatic
powder deposition apparatus shown In Fig.7. The
negative corona Is primarily used. In order to observe
the light emission at th« boundary between the plane
electrode and the powder layer, the measuring electrode
Is nade of a conductive glass plate as is reported by
Ting and Hughes.(?) The Image Intenslfler tube Is also
used with Its oaxleua gain of about 10*.
5.1 Back discharge at very low Ionic current
Under the condition of negligibly small Ionic
current (lover than about 10"" A/ca2) as described
before, the thickness of the layer can grow sufficient-
ly large. But, finally the back discharge takes place,
resulting In craters. Flg.9-a show* a photograph of
light emissions during the course of powder deposition,
taken froa both the front and the back side at the
begglnlng stage, of back discharge (Image Intenslfler,
exposure tloe - 5.0 s) . The layer thickness at this
tlae is about 0.5 ca. The back discharge occurs at
discrete points, and fairly large craters are produced.
The light emissions are pulslve and oove randomly
around the layer so far as the deposition Is continued.
The Intensity of light ealsslon Is stronger at the
boundary than on the surface. This does not change
even If the positive corona is used. Uhen the powder
feed Is stopped, the light ealsslon disappears.
However, when the grid voltage is sufficiently raised
to enable a spontaneous back discharge, or a sufficient
aaount of Ionic current is supplied froa an external
source, the light emission appears again in a fora of
fixed and stable glows. Flg.9-b shows the craters
appeared on the layer surface.
•-.-£«• ;v.••»-.-
• *-
«A-V
• .
(a) light emission
(b) craters
Fig.9 Back Discharge and Craters at Very Low
Ionic Current (li • 10~12 A/cm*)
S.2 Back discharge at higher ionic current
When a sufficiently large ionic current in suppli-
ed, back discharge can occur at a lower thickness, less
than 0.25 na, but Its Intensity is so low that it can
hardly be detected visually. When observed with the
luge Intenslfler tube, it can be seen that the back
discharge takes a fora of general glow, the whole
surface glowing uniformly and nc glow spot being
detected. Flg.lO-a shows a photograph of such glow
taken froa the front and the back side simultaneously
with the powder being fed, when lo - 5 « 10~7 A/ca2
and X • 0.15 art (image Intenslfler tube, exposure tloe
2.0 s). The intensity of the general glow Is stronger
at the boundary than on the surface, when the negative
corona is used. This reverses when the posltiv*
corona Is used. The craters cannot be detected
visually.
I 1 I I 1 1 1
(a) normal photograph
o 910; oioi
JCM («•]
(b) oicroscoplc photograph
Fig. 10 B.I -k Discharge at. High Current Density
(J0 -
ICT7 A/c=:)
S. 3 Microscopic observation of general glow
In order i > study the structure of the general
glow In more d. t.ill, an observation is aade using a
microscope cou; led with the Image Intenslfler tube.
When the electrode becomes alaost covered by powder,
light emission In the fora of general glow starts to
occur. Flg.lO-b is a photograph of this light ealsslon
taken froa the back side (exposure tlae • 3.0 s). This
shows the existence of nany discrete glow points,
scattered with 3 distance In the order of particle size.
As in the case cf vi-ry low ionic current, the light
eolsslons occur r.n.dcnly at many points, moving around
the layer when the powder Is being fed. A number of
seal! voids arc fon-cd at the saa« time, and they also
move around along the boundary. It should be noted
that, in spite of this general glow appearing, the
powder can continue to deposit up to a certain value of
thickness, about 0.25 en. The reason for this powder
penetration will be discussed later. When the powder
feed is stopped with Ionic current being supplied, the
fixed glow points appear also In this case, so far as
the layer thickness Is larger than 0.1 cm. Fig. 11-a
and b are the photographs of the layer back side with
back discharge which clearly indicate these fixed glows
to appear at the voids.
(a) fixed glow points (b) voids
Fig.11 Microscopic Observation of Central Clow
IAS 77 ANNUAL
67
-------
-H.V
KUDU (iCCTtOI
ttMS PUTC
•
) "•
mCKW>U>»
0±)
SWtfll UYI«
[AP.TH (ItCTROOC
Fig.12 Experimental Apparatus for Observation
of Breakdown Channel* Inside the Powder
Layer
- f »*--v
L • .*' ,"'
L . '- V. .V
;-^
(a) before general
glow
I , I , I i I i I. I
i eiiitiiiot
Uttl (—)
(b) after general
glow
Fig.14 Microscopic Observation of Micro-Craters
(a) breakdown channel
(b) cross-aectlon
Fig. 13 Microscopic Observation of Breakdown
Channels Inside the Powder Layer
Breakdown channel* inside the povder layer i* *lio
observed using an apparatus shown in Fig.12, vhere the
luge intenslfler tube 1* also used. Flg.l3-a and b
•re the photographs of the layer cross-section with and
without back discharge. These Indicate the breakdown
channels cleaning froa the bright glow point* at the
voids on the boundary and penetrating into the inside
of layer.
5.4 Microscopic observation of craters
No change is visually detected In the smoothne**
of the layer surface after the occurence of general glow.
Hence, • acre detailed exaeination la oad* by Micro-
scopic observation.
Fig.l4-a and b indicate the microscopic photo-
graph* of the layer surface with and without the
general glow. The thickness of the layer is 0.2 ma.
It it clearly shewn in Flg.l4-b that a nuaber of very
mall crater* are formed by the general glow, which 1*
naaed "•icro-craters". The else of the micro—crater 1*
alaost equal to particle alze. These nicro-craters can
be observed within the thickness range between 0.1 and
0.25 BB for the powder used. In the rang* Its* than
0.1 aa, they become indistinguishable from the Irregu-
larities of the deposited surface. In the thickness
range larger than 0.2S ma, they turn into the much
larger craters usually observed.
Fig. IS is a photograph of the uaual craters which
occur aider a fairly low Ionic current at a larger
layer thlcknes* (X - 0.5 aa). The crater else la much
larger coopered with the Blcro-crater. In most of such
craters, several large particles are observed to be
regaining on the bottoo vrlth • loose packing, which may
allow the plasma coluan of a continuous glow to be
maintained therebetween.
1. 111111111
I 11 i > 11 g i e i
KMI -
Fig.IS Microacoplc Observation of Normal Crater*
S.S Current wave form of back discharge and
formation of craters
Fig.l6-a show* a current wave form of back die-
charge In the case of very low ionic current when the
larger crater* are being forned at the thickness of
O.S ma. In this case. lj and 1- are so sasll (it -
10~12 A/on2, lp • 10"' A/cm?) that the current consists
mainly of back discharge pulse* at the crater*.
Flg.l6-b Indicates a current wave fora of back discharge
of general glow type when the powder supply 1* stopped
or i
68
and many fixed glow points are existing (lo - 7 « 10~
A/cm2). In this case, current consists of dc component
alone. This suggest* that the general glow 1* the back
discharge in the glow node already reported,'*' occurr-
ing at many breakdown polnta (micro-craters).
An experiment is Bade to study the formation
procea* of craters in core detail. A mica plate having
a pinhola with 1.0 SB diameter and l.S ma thickness i*
used on • plane electrode of a needle-plane electrode
syatea aa a layer aaaple (Fig.17). -The epoxy-reain
powder 1* filled in the plnhole. A negative high
voltage 1* applied to the needle electrode, and back
discharge 1* produced at the plnhole. At this inatant,
the powder ejection la resulted, as shown In Fig.18-a.
When a crater i* completed Inside the plnhole the powder
ejection stops, and back discharge turns from a pulsive
breakdown into a stable glow of the onset-glow node.'*'
Flg.l8~b Indicates the current wave fora during and
after this process. At the Initial stage when the
powder ejection 1* being aade, back discharge la pulsive
consisting of repetitive breakdowns. Then, It turns
into a non-pulslve glow when the crater is completed.
IAS 77 ANNUAL
-------
2.5 . 10~7
A/du
(•) very 1 - Sonic
rrent (crater occurring)
j^m
m. —
(b) higher Ionic current (general glow)
Fig. 16 ( irr.n; ' of Back Discharge
Fig.17 Experimental Apparatus for Observation
of Powder Ejection
(a) powder ej*
(b) currrnt w.ivr .1: r.itor formation
Fig.18 • •tlon
•
It - ^.8 that an appreciable
difference exists In tl.e thickness characteristics
^•aan^o ""'1 "da- ^h* leason for this may lie
In tl.e illf' ,;> mustier of the ueak points to
' (Eaax'o. tne
•l£gerlng will be
the plane surface
1. The equivalent
• area can be tjaken ••
O.I no as K
-------
Xo, can be estimated from the theory using the value
of Eds.
(2) In case the ionic current is negligibly snail, the
Halting thickness, XQ, can become high and the large
craters can be formed. In this case, pulsive light
emissions take place during the course of powder
deposition when back discharge started.
(3) In case the higher ionic current is supplied, the
value of Xo becomes smaller and back discharge takes
a form of general glow. This consists of a number of
very small glow spots. In this case, micro-craters
are formed.
(4) So far as the negative corona is used, the light
emission Is stronger at the powder-electrode boundary
than the powder surface.
ACKNOW.EDCEHEHT
The authors are very grateful to Prof. A.W.Bright
of the Southampton University for die very valuable
discussions which stimulated and helped this work.
Ihey also thanks Mr. Toshlyuki Salto for his help in
doing a part of the experimental works.
REFERENCES
1. H.J.White, Industrial Electrostatic Precipitation,
Addison Wesley, 1963
2. S.Masuda, Recent Progress in Electrostatic
Precipitation, Static Electrification: 1975,
Institute of Physics Conference Series Ho.27,
p.154 (1975)
3. J.D.Bassett, R.P.Corbett and J. Cross, Institute
of Physics Conference Series., No.27, pv221 (1975)
4. S.Masuda, A.Mizuno and K.Akxtsu, Initiation Condi-
tion and Mode of Back Discharge, J. of Electro-
statics (to be published)
5. S.Kasuda, A.Hieuno, Flashover Measurements of
Back Discharge, J. of Electrostatics (to be
published)
6. S.Masuda, A.Mizuno and K.Akutsu, Proc. of 3rd Int.
Conf. on Static Electricity, 24-a (Grenoble,
April, 1977)
7. S.Masuda, K.Akutsu and T. Salto, Proc. 1976 Annual
Conference of I.E.E.Japan, No.470 (1976)
8. S.Hasuda, I.Doi, I.Hattori and A.Shibuya,
Utility Limit and Mode of Back Discharge in Bias-
Controlled Pulse Charging System (to be presented
to 1977 Annual Conf. of IAS, IEEE, Los Angels)
9. Yui-Cheong Ting, J.F.Hughes, Proc. of 3rd Inter-
national Congress on Static Electricity. 27-a,
(Grenoble, April, 1977)
7Q IAS TT ANNOAl
-------
Errata
(A) Boundary condition c) [Eq.(8>] should be corrected as
c) Field Intensity at the boundary:
(ed/eo)Edx(X) " Eext " constant (v/m)
(H)
According to this correction, [E ] should be replaced with [E* ]
6X1
in the following equations (20) - (30), where
- (eo/ed>Eext
(B) Eqs.(ll) and (15) should be corrected as
qd(X) -
Edx(X)/Pd)/b
" Eext/bpd " ''do
(11)
at
Thus,
qd(x,t) - qdo«xp(-(t-x/b)/Td)
" qdo«xp(-(X-x)/Ld)
Then, Eq. (19) should be
'eoEext ddx
cE(x'° • «doLd(1 •
(C) Eq.(31) should b«
S/ °Ua(t»dt/C
(15)
(16)
(17)
(19)
(31)
(D) Figures (4), (5), (6) and (8) ihould bi modifltd •• followi,
according to th« correction (A),
3T.
Pig,4 Field Diaerlbution inside a Deposited
Layer (general e*ae| X0 « L,j, e0 • TJ)
71
-------
Eext
('o/(d)Eext
0
fEdx
i
r<
i
i
:*<
i
— E
ext
x o d ext
Kd- c
—** E.
N. ds
. t-i, \t-2i "/Vt-3i .
V d ^v d \7 d
\l \l \1
• » J x
i i >
0 L. 2L. JL. w 0 I ft =•-»•
add o i.. JL. J|. .
, add
* ' bt . - bl
Fig. 5 Field Distribution inside a Deposited Fig.6 Field Distribution inside a Depoaiu-.l
Layer (low resistivity case - Case(I); Layer (high resistivity cane - Case (II);
Xo >>Ld» co *" Td^ ^o ^kji' to <
-------
UTILITY LIMIT AMD MODE OF BACK DISCHARGE IN BIAS-CONTROLLED PULSE CHARGING SYSTEM
Senichl MASUDA, PhD.
loshlakl DOI, PhD.
Ichiro HATTORI, PhD.
Akira SHIBUYA, BSc.
Department of Electrical Engineering,
University of Tokyo
7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan.
ii
Ishlkawajlma-Harlma Heavy Industries,
3-2-16, Toyosu. Koto-Ku, Tokyo, Japan.
Back discharge Is investigated in the bias-contro-
lled pulse charging system to be used in an electro-
ttatic precipltator, and Its practical utility limit is
established In terns of critical main dc voltage V
and critical current Ic at which the visible back cc
discharge Initiates. Among the factors affecting this
utility limit, the apparent resistivity of sample layer,
pj, has the most essential effect. The width and height
of pulse voltage are the second important factors.
then P£ < lO*^ ohm-cm, the occurence of visible back
discharge is very seldom in this pulse charging system.
Mitt Pj exceeds 1013 ohm-cm, it becomes necessary to
decrease the width and height of pulse voltage to impr-
ove the loss of the utility limit. When pd > 101* ohm-
f.m, a large reduction appears in both Vcc and Ic, which,
hootver, can be amended to some extent by using a very
narrow pulse width and a low pulse height.
1. INTRODUCTION
Back discharge, one of the najior troubles in elec-
trostatic precipitators, is an abnormal corona discharge
occurring on the surface of dust layer deposited on the
collecting electrode when the apparent resistivity of
the layer exceeds the threshold of about 5 x 1010 ohm-
cm. In this case the potential drop across the layer
becomes so high that breakdown takes place In the layer
and triggers back discharge. Hence, Its initiation
condition Is given by the layer breakdown condition,
tfolch, In case of electrostatic precipitators, takes
the fora*1):
Pd Id * Eds <»
where ij - apparent current density in dust layer, and
«ds - breakdown field strength of the dust layer. It is
evident from Eq. (1) that the solution of back discharge
trouble will be enabled by reducing either pd or ij
without decreasing the main field strength in corona
•pace. Luthi proposed a method to reduce ij Independent
of the main field strength, which is Indicated In
Mg.l (a)'2'. The third electrodes are arranged to the
vicinity of the discharge electrode, and a periodical
pulse discharge la applied therebetween. A dc high
voltage is applied between the third and collecting
electrodes to maintain the main field. This method
proved to have an excellent performance for suppression
of back discharge in laboratory tests, but in its prac-
tical application the following essential difficulties
raaalaed to be solved:
1) The difficulty in suppressing dc corona to occur in
the pulseless period when the distance between the
third and discharge electrodes is Increased to the
level of 10 cm, necessary for scale-up, and the gas
and dust conditions fluctuate. This dc corona
deteriorates the control performance of this method.
2) The prohibitively high initial and running costs of
the pulse voltage source to be used.
A practical solution to these difficulties was
provided by the authors by inserting a dc bias-voltage
In aeries to the pulse voltage, as shown in Fig.l (b),
which is called the "bias-controlled pulse charging
system"**'. Through the control of maximum field stre-
ngth at the discharge electrode, this bias-voltage can
COLLBCTX1IG ELECTRODE
COUNTER ELECTRODE
THIRD ELECTRODE
DISCHARGE
ELECTRODE
THIRD ELZCTRODZ-
rr
U) PULSE CHARGING SYSTEM
(b) UAS-COHTROUZD PUISB
CHARGING STST2M
YQ: DO HIGH VOLTAGE SOMCE
V i PULSE VOLTAGE SOURCE (-)
ri DC BIAS-VOLTAGE SOURCE {+)
Fig.l Pulse charging system
not only Insure the choking of the dc corona to occur
In the pulseless' period, but also enables the use of
ac or halfwave voltages Instead of the very expensive
sharp pulse voltage.
A problem arised, however, in the course of its
development that the initiation of visible back discha-
rge, which better represents the actual performance
drop than the Initiation of dust layer breakdown accor-
ding to Eq.(l), Is not only affected by PJ and ld, but
also by the strength of the main corona field. This .
means that the practical utility limit of this system
for each value of PJ should be judged from the critical
values of main dc voltage Vce and current Ie at which
the visible back discharge Initiates. As is known,
these two quantities are the most essential parameters
determining the collection performence, since the
saturation charge imparted to dust particles in corona
field is proportional to the main field strength while
the charging time constant is inversely proportional to
the ratio of ionic current density to field strength.
Furthermore, it also became clear that behaviours of
back discharge In the pulse charging system are largely
different from those observed In the conventional twin-
electrode system.
Thus, the utility limit of this systea and the
mode of back discharge were studied in the pd-range of
1011 - 1014 ohm-cm.
2. EXPERIMENTAL APPARATUS
The electrode system used in the present laborato-
ry tests has an Identical construction to those used in
the pilot-plant tests described later, and is shown In
Fig.l (b) and Fig.2. The total surface area o? the two
counter (collecting) electrodes in In2, and the dista-
nce between the third and counter electrodes is 20 en.
The whole electrode system is placed inside a tumidity
controlled chamber, in which air humidity can be chang-
ed in a wide range under normal temperature ar.t! press •
US 77 ANNUAL
73
-------
ELtCTMDC
COIHTO ELECnODI
DISCHMCB
ELECTIOM
needle electrodes
Cewter t« thit* electrode dltUac* - iOca
ditchers* to third electrode dbtmee - So
Fig.2 Ileetrode system used la laboratory and
pilot-plant tests
ure. Instead of dust laysr, paper towele ara attached
onto the surfaces of the counter and third electrodes.
because their Pd-yalua can easily be changed within a
broad rang* of 1011 - 10" ohm-cm by controlling the
air relative humidity.
The pj-valua of the paper towel la very sensitive
to pressure so that it cannot successfully be measured
by the parallel electrodes eimonly used. The probe
method, ehowa in Fig. 3, is used throughout the present
experiments, since it proved to provide very satisfact-
ory results for the soft materials such es a paper
towel1 . The potential of the very light probe elect-
rode, Vj, le Meowed by a tero Method such that Vj ia
adjusted to give If • 0. At this point, we geti
V2 * Vd • idPdd (2)
where d - sample thicknen. Hence, the pj-velue can
be obtained from the relation
pd * V f xl4 <3>
where S • area of measuring electrode, and X, • current
fro* Measuring electrode. A protecting circuit consis-
ting of r, L, K and C is provided to the high sensibil-
ity current meter Ij for the purpose of preventing a
damage to occur when a sparking takes place either
between the plate and probe electrodes or between the
needle and probe electrodes.
The pulse voltage aouree ussd in the prseent
laboratory testa provides a periodical square-wave
pulse with e rise time of 1 yl and a minimum duty cycle
of 0.1, where the parameters can be changed ia the
range' pulee height • 0 - SO kV, pulse width • lOul -
10 me, and pules repetition frequency • 0 - 1 kls.
Throughout the laboratory testa the do bias-voltage ia
omitted.
S. OlinVATIOH 07 iA« PIIOUM1
The initiation and mode of the visible baek disch-
arge art affsctsd by the apparent raaiativlty of the
sampls layer pd, pulse width T, pulse repetition
frequency f, pulaa height V-, and the main de voltage
Vc. The effects of field strength and current density
in the gas **>* aaaple layer are contained implicitly
in the tffeeta of the above parameters.
3.1 Iffeet of >d
Among the effeeta of the parameters deaeribed
above, that of 04 ia the most essential. In this
system the visible baek discharge occurs only scarcely
guard
electrode
tin
r-2Kn , R-3KJI
L-10H , OlOpr
Tig.3 Probe method for measuring aample reeistivity
in the resistivity range of pd<1013 ohm-cm, when Ve <
100 kV, V. <30 kV (negative polality), f > 10 HE, and
duty cycle<0.1. Once it appears, it takee the form of
an unetable glow, from which a streamer suddenly devel-
ops toward the discharge electrode and turns into a
flashovar;' When the frequency, f, or pulee width, T,
ip raised, the oecurence of the vieible back discharge
becomes more frequtnt, while its starting condition
becomes more obeeure.
The situation changes completely when the p^-valua
exceeds 10" ohm-cm. The visible back discharge eppea-
rs in the form of a etable dlffuss glow. With the inc-
rease in the main dc voltage Vc, the spot-like glow
points appeer on the sample surface with increasing
number, as shown in Fig.4 (a) and (b). When Ve la
further Increased, the streamers develop from the glow
points toward the diecharga electrode, and finally
bridge aerosa the gap between the counter and diecharge
electrodes. A remarkable difference of these etreamars
from those occurring under pj
-------
parameters
p. (ohm-cm)
vc (kv)
Vp (kV)
f (Hz)
T (/Js)
It <;,A)
ic (,,A)
(a)
(1.2-4.4)
x lO1^
*5.
- 20
100
100
100
55
(b)
It
.58
- 20
100
100
}OO
175
(c)
II
60
- 20
1OO
l.OOO
25*
164
(d)
.t
35
- 20
L.OOO
100
100
60
(e)
..
5*
.=_iO
100
100
100
100
Flg.4
Effects of Vc, Vp.
f and T on activity of back discharge
of pd, the tendencies described above are observed In
tht whole rang* of od. It Is felt that the effect of
pulse width, T. Is the oost essential, and next to It
COM those of Vc and Vp. and finally that of f.
4. UTILITY LIMIT
The utility Holt of this system expressed In
ttras of Che critical dc naln voltage Vcc and critical
average pulse current Ic Is measured under various
values of parameters pd, Vp. and T, with th« valu* of
f being kept at 100 or 200 Hi. The results obtained
for the pd-values of about 1QH, 1012, 10*3 and lO^4
oh»-CB are shown In Fig.5 - 8 respectively. The utili-
ty Halt giving the highest possible valued of Vcc and
Ic corresponds to the best collection performance.
In conformity with Che fact that visible back
discharge hardly occurs In the Pd-value range < 10
ohm-co, the utility Halt as high as (Vcc - 80 - 60 kV.
Ic • 50 - 30 uA) Is attainable In this range, as shown
in Flg.S and 6. The effects of T and Vp are saall
except for cases of a very high Vp-value, as In Flg.S
(d) and Fig.6 (c), (d), where drooping characteristics
with the increase in T appear. A detailed examination
shows that a slight loss in the utility Halt occurs
with the Increase In the oj-value from about 10*1 to
1012 ohB-cm.
A remarkable loss In the utility limit appears
when pd exceeds 10 oh:a-cn, as shown In Fig. 7. The
decrease In Ic occurs In cooraon. and that In Vcc also
takes place for larger values of Vp and T.
When pd finally exceeds the threshold of 10*4 oh»-
c», a very large reduction both In Vcc and Ic appears,
as shown In Flg.S. However, It should be noted that
the use of a very narrow pulse wlJth of Bb~ut 10 wS and
a sufficiently low pulse height of about -10 kV provides
a substantial laprovenent. The merit of a narrow pulse
width and a low pulse height la also clearly observed
in Fig.7.
Froa tht results obtained it cay b* concluded that
the most favourable pd-rang* for thia syati-a to solve.
back discharge trouble will be that lower than lO1^ oha-
ca, while a sllgh' difficulty Day appear In the rang*
of pd - 10^ - 10 oho-ce and the use of a narrow
pulse width will becoa* necessary. The rang* of Pd -
1014 sceaa to provide the upper limit for this systea,
where only the use of a very narrow pulse width could
save th* performance drop to some extent.
5. PILOT-PLANT TEST
The predictions obtained froa th* laboratory teats
are examined at a pilot-plant located at an iron or*
sintering furnace In connection to the exit of a conven-
tional type electrostatic preclpltator. This pilot-
plant consists of a charging lone Identical to that
shown In Fig.1 (b) and Fig.2 and a collecting zone out
of zig-zag arranged negative and positive channel elect-
rodes, as shown In Fig.9. Th* gas transit time through
a single charging zone is about 0.4 a. The dunt consi-
sts nalnly of ^03 particles and contains small
araounts of salts of alkaline metals as well as alkaline-
earth netals by several percents ID total. Its
particle size Is extremely small, and oore than 70 X
are in the ran.
-------
0
parameters
Pd ( ohm-cm )
Vc (KV)
Vp (KV)
f (Hz )
7 ( jJS )
It (pA )
Ic (JJA)
(a) -
( 1.2-4.4)
x 10 l4
45
-20
100
100
100
55
(b)
r
58
-20
100
100
300
175
(C)
'
60
-20
100
1,000
«+m
164
(d)
"
33
-20
1,000
100 '
100
60
(e)
-
54
-10
100
100
100
t -f\_^^ f* "^
4-Ov> >x
Fig.4 Effects of Vc ,Vp,f and T on activity of back discharge
-------
^ (a) V —10W _ (c) V_— 20KV
ws
>
bl
<
g
s
I
M
jj
g
p >
Sc
}«> -„
Af * »
p -(0.94 i. 1.17)xl011oh»-c« g
i - 100 Hx >
1
3
S
. . . . , . a
p
100
» 0*
1 . *• t
0 SO " 0 SO
PUtSI CUWBII Ic(liA) f T . lOOOii* PULSE CORMMT IC8
§
A
g
s
g
g
tl
(b) V — 1SCT 7 " IOOV* £
•l°° 0I zoi!! J8
t§V "* B
i
5
i
g
C
M
- - - , 1 , «
(J) V --2SCT
P
100
"4f •
1 r
SO
PULSE CURKEMT I-(vA)
so
PULSE CWWEKT
Fig.5 Utility limit - 1 (p.- 10" ohm-cm)
a
|
^
I
5
S
^
c
0
10
JS
r
^
i
M
8
(a) V — 10W I2 ~
P P.-(1.13 *• 1.80)xlO oh— e« S
100 d *•
f - 100 •« S
>
T • 0 u
«'" I
>
|
0 50 - I00um
KOLSR CUMUWT I_{pA) . «,„.
(b) V —I5CT
p a
100 C
t
t
^* 1
T £
i
B
, , , , r , S
(c) V --ZOCT
100
°*/« i
T
i
0 SO
PULSE CUUEBT I.(liA)
C
(d) Vp--2SCT
100
» ft
1
f
t
PULSE COMENT Ic(vA)
PULSE CUUtENT I,(|iA)
C
Fig.6 Utility limit - 2 (pd a 101Z ohm-cm)
77
IAS 77 ANNUM
-------
c
5
100
*•".
so
FHLSE CUMEHT IC(HA)
I
I
s
s
i
Ck) V — 1SXV
100
if*
•
*r
t
cntiare
IC(WA)
2.0Q)xlO'
,D
f - 100 te
r T -lOOOu*
A J00v*
• _ 200IM
y loon*
O 20IM
y——>--- 10u*
100
(e) V^—IOCT
0 SO
FOLSC CUUBIT I.(llA)
(d) V^— JSCT
100
s
g
8
,13
so
PULSE COUOIT Ic
i
i
s
£
3
1
(•) V — 10CT Pj-O-11 v 1.28)xlO otaM« g
SO " f-Wfc -g
w • 0 ^
|
<
* 4 T • 500 na g
Tj^ ^ .^^ ^
. „„ a
* »0ll« „
I.I. +*. «n D
(c)
»0 V -20W
p
J A_
• ^ A
0 10 ..0 10
"•* i i -i ID uji
rout CUUWT ic(tiA) x 10 ** »ms« COU«T ic(gA)
|
>
i
i
i
H
^*
g
I
» VISCT ^
g
o 9
AT 2
* 3
c
8
W)
5° T — 2JW
*•
0 •!
. . 1 - *
10
roue amuyr IC(I>A)
10
TOUE cuuuurr IC(VA)
IAS 77 ANNUAL
'Fig.8 Utility limit - 4 (pd » 1014 oh»-c«)
78
-------
2nd CHARGING ZONE
'
CO-
• (+>
r-
\
2nd COLLECTING ZONE
\
1st CHARGING ZONE
•t+)
CO
-
1st COLLECTING ZONK
Fig.9 Conitruction of pilot-plant preclpitator (PAC-ES type)
ae voltage, la uaed Instead of • sharp pulse voltage.
Hence, ite equivalent pulse width T, can be aeeumed to
be approximately 10 ma or 1 ma.
Tig. 10 ahowa tha relationship between the peak
voltage Vp and the current I flowing into the counter
electrodes of the first charging aone, where Ve la kept
at 40 kV. Tha curve (4) indicatea tha normal V. - X
characteristics without back dlaeharga. It can ba
seen from tha curves (1) and (2) that back discharge,
repreeented by abnormal Increase in current I, la
enhanced by the growth of thickness of duet layer. A
reurkabla improvement is observed to occur when an
electrode rapping with aufficient atrength and frequ-
ency is provided so that tha layer thickness becomes
small. The curves also clearly indicate the merit of
tiling a narrow pulae width.
The collection performance of 65 - 65 X can ba
achieved under the optimum condition of curve (4),
which exceed! the requirement set forth to the after-
collector to be Installed. This performance level
repreaente a very aatiafactory figure, considering the
extremely high oa-value, small particle else, and a
vary short treatment time,
It ie concluded from thia test that thie ay a tern
can provide a eolutipn to back diacharge occurring
under pd • 10" - 10" ohm-cm when a half-wave voltaga
with T,< 1ms is uaed in combination to an effective
electrode rapping (aee Fig,7 (a)),
6. CONCLUSION
The following conclusions are obtained from the
foregoing studissi
1) The bias-controlled pulse charging system end pulse
charging system can provide an effective technical
solution to the back discharge trouble up to the
PJ -value of about 10" ohm-cm,
2) Toe meet favourable range of pd for these systems ie
up to 10" ohm-em.
J) A.slight difficulty appeara in the range pd • 10" -
10** ohB-em, where the uee of a narrow pulse wldft
beeomee necessary for improving the performance lose*
4) The PJ-value of 1014 seems to provide an upper limit
for tneae systems, where beck discharge starts to
oeeur alao en the third electrode when V. is raised
sufficiently high. At thie resiitiviey level back
discharge ean eceui en the counter electrode even
by a dark current when Ve ia raised beyond e certain
value.
S) The use of an extremely nerrow pulee width in the
order of 10 ul and a low pulie height may provide a
possibility of improving a large performance drep
to occur at 04 > 10l* ohm-em.
I
(1) T. - 10 H
ic volUj* fe- 40KV
notation
0
e
A
A
t.(M>
1
1
10
10
f(Hl)
its
IIS
so
so
layer
thieknti*
•Mil
l»rg«
•Mil
large
(1) T • 1 ••
\
10
FMK VOkTACI V (KV) '
Tig.10 Relationship between counter electrode
current X and peak voltage Vp
(first charging aonei Vc • 40 kV,
bias-volt age Vt • 0)
Aeknowlidiaaent
the authors wieh to express their gratitude to the
ee-votken for their supports and aaiistaness gived te
this work, Mse me of the authors (Xoshiakl Dei)
wishes to acknowledge the fellowship grants given by
Japanese Ministry of Education and by fundaelo de tapare
a Peequisa do latado de lie Paulo (fAMIP), Iraeil
(fli. 75/699).
79
JAI 77 ANNUAL
-------
References
(1) S.Hasuda and A.Mizuno: Proc. 3rd Int. Conf. Static
Electricity, 24-a (April, 1977 in Grenoble in
France)
(2) J.E.Liithi: Dissertation ETH-Zurich, No.3924 (1967)
(3) S.Kaauda, I.Doi, H.Aoyama and A.Shibuya: Staub-
Reinhalt. tuft, Bd.36, S.19 (1976)
(4) S.Masuda and S.Obata: Proc. 1975-Cen. Conf. of Inst.
Elect. Engrs. Japan, Eaper No.906 (1975)
IAS 77 ANNUM $0
-------
THE ANALYSIS OF ELECTRIC WIND IN ELECTROSTATIC PRECIPITATOR
(BY LASER DOPPLER VELOCIMETER)
S. Masuda, K. Akutsu, K. Aihara
Department of Engineering
University of Tokyo
Introduction
The electrostatic precipitator charges dust particles to
remove them electrostatically, but the effect of electrical wind
(ion wind) in precipitating process of particles cannot be
ignored. In the electrostatic precipitator, there is a negative
electrical wind from a discharging electrode and a reverse elec-
trical wind due to inverse ionization phenomena. The velocities
of these winds are at least several m/sec. On the other hand,
the electrical moving velocity of the charged dust particle in
the electrical field of 5 kV/cm is about 0.1 ^ 1.0 m/sec for the
particle with diameters ranging from 1 to 10 ym and at most 10
cm/sec for the submicron particle with diameter of about 0.1 ^
1.0 ym, Therefore, you have to treat the precipitating process
of particles with diameter of less than 10 ym as, so called,
EHD process which considers hydrodynamic field as well as elec-
trical field.
Here, in this experiment, we have analyzed the electrical
wind using the Laser Doppler Velocimeter. First, we have analyzed
the negative electrical wind from the needle-point in point-to-
plane electrodes. Secondly, we have generated inverse ionization
by placing a mica plate with a hole on the plate electrode instead
of dust particles and have analyzed the negative ionization
electrical wind. Finally, we have investigated the behavior of
submicron particles in the vicinity of a boundary layer near the
plate electrode when there is a current parallel to the plate.
Measurement Method
The Laser Doppler Velocimeter is designed to measure the
velocity of particles, which move at the same speed as current,
by measuring the Doppler shift of scattered light from the par-
ticles. As scattering particles, we have used the D.O.P. particles
IDOP Di-Octil-Phtalic Acid, average particle diameter of 0.3 ym].
An experimental apparatus is shown in Figure 1. in order to dis-
tinguish Vn and V'n in Figure 1, we have used the frequency shift
system which shift the frequency of the other beam by 40 MHz. The
size of measuring region is 116 ym wide, 1138.8 ym long and 0.008
nun3 volume.
81
-------
Experimental Results
Negative Electrical Wind
Velocity distributions of negative electrical wind (ion wind)
in a perpendicular direction and in a parallel direction to the
plate are shown in Figures 2 and 3. Figure 4 shows the velocity
changes of the negative electrical wind which is perpendicular to
the plate by applying various amounts of negative voltage to the
needle-point. In the immediate vicinity of the plate, the negative
wind velocity is proportional to the square of the voltage.
Reverse lonization Electrical Wind
We have used the mica plate which has a pin-hole with a diam-
eter of 0.5 mm instead of a dust particle. Figure 5 shows the
measurement of velocity distributions of inverse ionization elec-
trical wind in a perpendicular direction to the plate in the area
above pin-hole. Figure 6 shows velocity changes in r direction.
In the range of the voltage (15 kV ^ 20 kV) and the electrical
current 1 yA *v 10 pA), the increase in voltage and electrical
current will increase the velocity of inverse ionization electrical
wind in a perpendicular direction to the plate, however when the
voltage and the electrical current exceed above range and when the
streamers can be visually identified, the velocity will not be
dependent on the voltage and electrical current and will be
around 10 m/sec. A similar amount of inverse ionization electrical
wind can be obtained by increasing the number of pin-holes and
by superposing the influencing ranges of wrecking points. Also
the negative electrical wind from the needle point was increased
in about 0.5 m/sec (however, voltage = 25 kV).
Behavior of D.O.P. Particles in Vicinity of Bounday Layer
Figure 7 presents the behavior of D.O.P. particles near the
plate when there is a parallel current r* direction in Figure 1)
to the plate.
Conclusion
Since a negative electrical wind maintains a fairly high
velocity [4 m/sec in 5 kV/cm and 6.4xlO-*A/m2] as far as immediately
before the plate, we think that most of the particles will be
carried to the vicinity of precipitating point. The vicinity of
precipitating point with a boundary layer and an eddy of about
30 cm/sec will be created just outside of this boundary layer.
Also, since the inverse ionization electrical wind will exist in
the area of several millimeters located a 1 cm in front of a
wrecking point and have a high velocity of maximum 12 ^ 13 m/sec,
most of the particles in this area will probably be blown away.
82
-------
Acknowledgment
We would express our sincere appreciation to beneficial sug-
gestions given by Professor Toshimitsu Asakura, Hokkaido University
concerning the Laser Doppler Velociraeter.
References
1. Mishina, Asakura: Application to the Measurements of Light
Heterodine—Doppler Velocimeter—Applied Physics, Volume 42,
Edition 6 (1973), 560.
83
-------
H.V.
NEEDLE ELECTRODE
D.O.P. SMOKE
GENERATOR
PLATE
ELECTRODE
Figure 1. Schematic diagram of experimental apparatus.
ft A
-------
AIR FLOW FOR D.O.P. SUPPLY
H.V. - -30. I - 5 A»A, r - 0
10
I
Is
oo—o—o
0
NEEDLE
10
20
30
40
Z, mm-
so
-PLATE
Figure 2. Negative I.W. distribution.
85
-------
I I I I I I I I I
I
E
-H.V. - 25. Kv
I -3,/iA
r = 0, mm
>
__ 9
I '
10 20 30 40
Z, mm
Figure 3. Distribution of negative I.W (Z direction).
50
86
-------
UJ
1U.U
5.0
1.0
0.5
0.1
j
: I
—
•
—
—
•
r
• • • | . .. ., ••£"/ 1 ""I1
•«?/
M
• /?»*rr
«/V
7 r °
/ J* A
/ tj a
/ f •
OC - H.V « / •
4 v
OC - H.V2 ^
i iiliiiil i i i 1 ii i ill
:Z
:Z
:Z
:Z
:Z
:Z
:Z
49.2
49.0
33.0
15.0
6.0
1.0
5.0
mm
mm
mm
mm
mm
mm
mm
:
—
-
•M
-
(B.D. occur)
• • t\
r • O
^™l
10
50 100
Kv
Figure 4. Negative I. W. (Z direction) vs. -H. V.
87
-------
f,0
o
3
UJ
> 5
BD: PLATE—"-NEEDLE
N: NEEDLE—»PLATE
AIR FLOW FOR D.O.P. SUPPLY
I
10
20 30
Z, mm
40
50
Figure 5. I.W. distribution (Back Discharge Occur).
68
-------
t
!
kU
10
I t
T
T
T
T
I.W. MAX. BD PLATE
VELOCITY N NEEDLE
Z » 48.0, mm
H.V. - 25. K v, - 30 AiA
NEEDLE
PLATE
1 I
6
BD: 2.25
N: 3.59 •
8 mm
Figure 6. I.W. distribution of r direction (Z - 48.0 mm)
89
-------
1.0
0.5
I
s
a
•
<
-
—
I I
I
•
-H.V. • 2
1 - 3.6, /L
i <
Z, mm
•
5, Kv
A r • 0, mm
I I
i
m
\
<
•
•
(
•
«
•
1
»
4
•
»
-
«
' (\
-
i>
i
•1
1
46
47
48
49
50
7. Submicron particles velocity nearby boundary layer.
90
-------
FUNDAMENTAL ANALYSIS OF ELECTRON BEAM GAS ELIMINATION
S. Masuda, K. Akutsu, and M. Hirano
Introduction
Presently, the most effective method considered for elimi-
nating NOx (mainly NO) in the burned exhaust gas is the electron
beam denitrate method. This is the radiological chemical method
in which N2 or HaO in exhaust gas is activated by a high energy
electron beam which was accelerated to 1~2 MeV and it will oxidize
NO and then nitrate it. Therefore, this method requires neither
special catalytic agent and oxidation agent nor reheating or
pressurizing exhaust gases. Also, by added NH3 in the electron
beam injecting space, it will have solid nitrate aerosol and
will prevent corrosion of equipment by nitric acid as well as
possible precipitation. Most of all, this process can desulphurate
as well as denitrate and the resulting product of this process
(3NIUNO3- (NHi,) zSOi,) can be utilized as a fertilizer.
As explained above, the electron beam denitrate method is
a very epoch-making process, but initial and operating costs of
the electron beam generating equipment is very high. In order to
overcome this, it is planned to combine the electron beam injec-
tion into the electrostatic precipitator, namely to install the
precipitating electrode in the beam injection space. This change
can make an overall system compact, reduce the installing space
and cost, but it is necessary to analyze how the electrical field
generated in the electron beam injection space will affect deni-
trate reaction. This paper here has analyzed it experimentally.
Experiment Apparatus and Method
Gas Flow System (see Figure 1)
The electro-magnetic valves are located at the entrance and
the exit of reaction cell to control a flow of sample. As a
sample, a constant flow rate of (1 1/30 sec) N2 is used as a
carrier and small amounts of NO and NH3 are supplied through
capillary tubes. The concentration of sample is controlled by
measuring pressure with manometer. Valves Vi and V2 are closed
during experiment to have a stand-still condition.
Optical Measurement System (see Figure 2)
The measurement of NO was done by Infrared-Ray Molecule
Extinction Method which uses a luminescence source of NO itself.
This is based on the fact that only NO can absorb the luminescence
(5.3 urn) which is produced when NO in a vibration excitation
state returns to a stable state. The removal of NO under the
electron beam application can be measured with high sensitivity-
91
-------
Fi gure 1
EXPERIMENT APPARATUS
V2
ml*"*
J
l|
I- 'Ukr—
r^ L
1
^
^
-^
s
b
c 1
^r
?
II
II
I
^
V1
A
1
q
ns
i
c.
J7\ /t
NH2 NO
SC: DYNAMITRON SCANNING HORN
S: SHUTTER
L: NO-LAMP
D: InSb-DETECTOR
A: AMPLIFIER
LO: LOCK-IN AMPLIFIER
R: CHART RECORDER
OSC: RC-OSCILLATOR
M: MANOMETER
F: FLOWMETER
V1.V2: ELECTRO-MAGNETIC VALVES
92
-------
Figure 2
DIMENSIONS OF REACTION CELL [mm]
GE
W: CaF2 WINDOW
GE: GUARD ELECTRODE
ME: MAIN ELECTRODE
SW: STAINLESS STEEL WINDOW
93
-------
high-speed-sequential-measurement in a real time. This technique
has negligible interfernece by coexisting materials (H2O, N2,
aerosols, etc.). Also, in order to have a better signal to noise
ratio, the RC oscillator is used as a power source to modulate
at Hz and the frequency composition of 120 Hz is extracted by the
lock-in-amplifier.
Beam Application System
The dynamitron in University of Tokyo-Nuclear Research General
Center was used as an electron beam application source. The
application of beam is done by opening or closing the shutter and
the beam will be scanned horizontally at 10 Hz to cover a sample
entirely.
High Voltage System and Reaction Cell
The reaction cell has a same axis cylinder type main elec-
trode and a negative high voltage is applied to the inside
cylinder part. The outside cylinder has a separate guard elec-
trode which is grounded and the electric current is measured between
the main electrode and the ground. The cell has a CaF2 window and
a stainless steel window for the infrared ray and the electron
beam, respectively.
Experiment Results
Figure 3 presents the relationship between beam application
time and NO concentration when approximately 800 ppm of NO and
2000 ppm of NH3 are supplied (about 1000 ppm of H20 are possibly
included because of system configuration) at the electron beam
acceleration energy of 1.2 MeV and at the beam electric current
of 97 yA. This figure shows that the existence of electrical
field will promote denitrate reaction. Here the relationship
between the reaction velocity (gradient of Figure 3 curve) and
the applied voltage. On the other hand, the voltage-current
characteristics in beam application space is linear as shown in
Figure 5 and this satisfies the Ohm's law. Thus, the relation-
ship between the denitrate reaction velocity and the electric
current between electrodes will be similar to Figure 4. Here, in
order to analyze how much the existence of electrical field con-
tributes to the denitrate reaction, the relationship between the
electric current and the I/nth power of (Vr-225)/225 for the
reaction velocity Vr are obtained (note; i vs n/Tvr-225)/225)
and Figure 6 shows this relationship for n - 2, 3, and 4.
Conclusion
1. In a mixed gas of N2, NH3 and H20, NO concentration
decreases proportionally to the beam application time and the
reaction amount is independent from NO concentration unlike
ordinary chemical reactions. This means that the denitrate
94
-------
-•-¥• aor«v
-.-yioofnv
-.-y.il»
-o-VH.0
TIMC
95
-------
no
I"
MO
Fig. 4
we
I KM
is
IKV]
96
-------
97
-------
98
-------
reaction is done in proportion to the number of emitted electrons
(supplied from a source), thus it is expected that the denitrate
reaction velocity is proportion to the number of emitted electrons
(supplied from a point source), thus it is expected that the
denitrate reaction velocity is proportional to the beam electric
current. This was actually verified by the preparatory experiment
done before this experiment.
2. The electrical field added to the application space ex-
tensively promote the denitrate reaction. It was confirmed that
the electrical field and the electrical current contribute to the
decrease reaction velocity of NO with the order of about third
power and have a good effect in a speedy elimination of the
aerosole product as well as a denitrate reaction, whatever the
cause of this effect will be.
3. In the beam application space, the voltage is proportional
to the current thus the applied electric power contributes to the
denitrate reaction velocity with a power of 1.5th.
4. For the most effective denitrate case which is the 14 Kv
and 2.3 mA, its average electrical field strength (3.68 Kv/cm) is
about the same as that of ordinary electrostatic precipitator
(3 Kv/cm). But the beam application space is abundant with
various ions and electrons which give a high conductivity and the
average current density will be 6.28 x 10~f A/cm2 which is larger
than 100 times that of ordinary electrostatic precipitator
(2 x 10~8 A/cm2) and also the average electric power density will
be 2.34 x 10~2 W/cm3 which again is larger than 100 times that
of ordinary one. Since the conductivity in the application space
has a close relationship with the application amount rate (is
proportional to the square root of the application amount rate),
it is necessary to place the electrode in the application space
which has the best denitrate rate (= denitrate amount/applied
energy) in the actual plant. It will be a future project to
obtain these conditions.
99
-------
Motion of a Microcharge Particle Within
Electrohydrodynamic Field
Electrical Engineering in Japan, Vol. 94, No. 6, 1974
TrmmUted bom Dtnld Galdcat Ronbuuhl, Vol. 9«A, No. 12, D«c«mb«r 1974, pp. 515-522
S. MASUDAandY. MATSUMOTO
Faculty of Engineering, University of Tokyo
100
-------
Electrical Engineering in Japan, Vol. 94, No. 6, 1974
Translated from Denki GaWcai Ronbunshi, Vol. 94A, No. 12. December 1974, pp. 515-522
Motion of a Microcharge Particle Within
Electrohydrodynamic Field
S. MASUDAandY. MATSUMOTO
Faculty of Engineering, University of Tokyo
1. Introduction
Recently the study of the motion of miciro-
charged particles (particle size 0.1 to 100 pm) in
the EHD (electrohydrodynamic) field has become
important in connection with the design of electro-
static precipitators and electrostatic painting.
This study is reduced to the solution of equations
of the motion of charged particles in the EHD
field under given boundary and initial conditions*
However, the analysis is very difficult if the
boundary conditions are complicated and many
problems still remain unsolved.
Steinbigler [1] proposed a charge substitution
method which enables one to obtain approximate
solutions for two-dimensional and symmetrical
three-dimensional electric fields under compli-
cated boundary conditions [2-6]. Using this
method, we analyzed two-dimensional steady
potential field of perfect flow under complicated
• boundary conditions [7]. In this paper, we
analyze the motion of microcharged particles in
EHD field by means of charge substitution and
present a method of calculating collection effi-
ciency of two-stage electrostatic precipitator
(EP-ES type electrostatic precipitator [8]).
2. Equation for Motion of Microcharged
Particles in EHD Field
The motion of microcharged particles in EHD
field is described by
where r is the positional vector of the particle,
m is the particle mass, a is the particle radius;
q is the particle charge, TJ is the viscosity coeffi-
cient of the medium, V(r) is the velocity vector
of the medium at r and £(r) is the electric field
at r. It is assumed that Stokes1 equation holds for
the viscosity of the medium for Reynolds number
Re < 0.5.
In the viscous flow, it is difficult to determine
V(r) under complicated boundary conditions in
both laminated flow and turbulence. Therefore,
V(r) is approximated by the velocity distribution
in the steady-state potential flow. Since both V(r)
and E(r) are vectors in the potential fields, Eq.
(1) is rewritten as
~
=f and
-------
surface distance) b
3. Modes of Particle Motion
,,.(,.»>=-
(6)
(7)
(8)
Vn and EQ are the average flow rate and average
electric field intensity in the region under con-
sideration; e are the velocity potential
and electric potential normalized to the average
flow rate VQ and average electric field EO in
dimensionless space. These are called normal-
ized velocity potential and normalized electric
potential, respectively. Substituting Eqs. (5)
and (6) into Eq. (4), we obtain
) = C {$, (r/4)+K fa (rib)}
where
(9)
(10)
(11)
where U = bEo is the applied voltage ^
electrodes; #EHD EHD
(r/b) and dimensionless parameter
3. 1 Effect of 0EHD (r/b)
In the EHD field with similar boundary condi-
tions (duct shape, electrode shape, etc.), the
distributions of normalized velocity potential
f (r/b) and normalized electric potential e (r/b)
and their magnitudes are independent of the struc-
tural size, average flow rate VQ and average
electric field EQ. Specifically, e
(r/b) are pattern functions representing geometri-
cal patterns of flow field and electric field. How-
ever, normalized EHD potential 0EHD (r/b) which
is a linear combination of EHD » 0 f and the
particle motion depends on the flow field only.
(ii) When K » 1, we have 4>EHD K K«se aw* me
particle motion depends on electric field only.
(iii) When K ~ 1 , the particle motion depends
on both flow field and electric field, and is of the
EHD motion.
3.2 Effect of £
As seeb in Eq. (14), £ represents the effect of
viscosity on inertia in the motion and is called the
viscosity factor
(i) When £
-------
In case of (1) and (ill), particle loci can be ob-
tained by solving Eq. (14). In practice, however,
the condition of t »1 In (ii) holds. In such
cases, the particle loci can be determined from
the lines of force in the EHD field.
4. Examples of Particle Motion
As described in the previous section, the par-
ticle motion is analyzed by obtaining the normal-
ized EHD potential 4>EIID in the dimensionless
space under the given boundary conditions. For
this purpose, we draw the flow path and electrodes
in the coordinate system with b as unit length,
and obtain $1 and <£e, i.e., the velocity potential
and electric potential In the flow field with unit
flow rate and with unit electric potential applied
to the electrodes, respectively. It is not neces-
sary to take the coordinate origin as a reference
point; a point where the flow rate distribution is
uniform is taken as a reference point for EHD for vari-
ous values of K. For deriving
-------
EHD line of forct (K - 2.0)
Fig. 3. EHD line of force and loci of particle
motion starting from the point O for the case of
Fig. 1.
Avenge gas flow ratei; (m,'s)
Fig. 4. Critical particle radius ac for locus
estimation by EHD lines of force vs. average
flow velocity VQ for the case of air flow.
(collecting electrode) to those which start from
points 0, 1 ... 5.
When particles are charged by corona dis-
charge, their theoretical saturated charge is
given [9J by
(17)
when e8 is specific dielectric constant of particle,
eo is dielectric constant of vacuum, and EC is the
field intensity in corona space. Substituting Eq.
(17) into Eq. (10), we obtain electric field factor
:,+2) (18)
Letting ts = 2.5, EC = EO = 5 kV/cm and e = 2.39
x 10~3 Ns/m2 (150°C in air) and expressing a and
VQ in terms of [fjm] and [m/s], respectively, we
obtain K « 0.1 (a/Vo). Therefore, as the particle
radius a increases and as the flow rate Vo de-
creases, the electric field factor K increases.
For the condition K a 1 under which the particle
motion is effectively controlled by electric field
in the EHD field, we have a [/jm] 2= 10 VQ [m/s ].
Therefore, it is difficult to collect extremely
small particles by electrostatic precipitator unless
the value of VQ is reduced below 1 m/s.
4.2 Parallel arrays of channel electrodes
arranged in a zigzag fashion with their
openings facing each other in a uniform
flow
We consider a case (Fig. 5) where channel
electrodes are arranged in a zigzag fashion with
their openings facing each other; the charge has
the same polarity as the upstream electrode and
cylinders are attached at the edges of channel
electrodes. These electrodes are used as a par-
ticle collector of 2-stage electrostatic precipita-
tor; an electric charger by means of corona dis-
charge is installed at the upstream side. In this
electrode arrangement, openings between the up-
stream channel electrodes work as a nozzle to
absorb particles and electric field in a space be-
tween the upstream and downstream channel elec-
trodes force particles to enter openings of the
downstream electrodes.
Assuming that the particle radius is sufficiently
small with the condition of £ a 5, we estimate the
particle collection coefficient from the distribution
of the EHD lines of force.
Figure 6 shows the distribution of the EHD lines
of force for K = 0 ~. 3.0 which was obtained by
means of charge substitution. It is assumed that
the lines of force start from points which divide
the interval between two upstream electrodes
equally by 20. In the field calculation, the charge
to be substituted was not placed behind the down-
stream channel electrodes. Therefore, electric
lines of force and EHD lines of force did not ter-
minate at the downstream electrode and the inter-
ior of the channel electrodes. However, electric
field inside the channel electrodes is so small that
the errors can be neglected'in the estimation of
the lines of force and collecting efficiency. The
collecting efficiency can be expressed in terms of
a ratio of the number of the lines of force terminat-
ing at the downstream electrodes to that of the
lines of force entering the openings. When K = 0
(Fig. 6(a)), the EHD lines of force coincide with
the stream lines and collecting efficiency is zero.
As the value of K increases the distribution of the
EHD lines of force approaches that of electric
lines of force and particles are forced to enter the
interior of electrodes, resulting in an increase
of collecting efficiency.
Figure 7 shows the relation between collection
efficiency and electric field factor; particles
whose loci correspond to lines of force terminat-
ing at the downstream channel electrodes inner
than point P were counted. In this electrode
arrangement, the 100% collection coefficient can
be obtained for K ^ 1.95.
104
-------
Fig. 5. Parallel channel-electrodes arranged in
a zigzag fashion with their openings facing against
each other.
Electric field factor K
Fig. 7. Theoretical collection efficiency rjth vs.
electrical field factor K for the case of Fig. 5.
<»)K-0 (b)K-CW
(line of force)
(e) K-0.+
(d)K-0.6 (e)K'08 (f)K-I.O
(g)K-Z.O
(hi K-3.0
Pig. 6. Distribution of EHD lines of force (case
of Fig. 5).
5. Observation of Loci of Particle Motion
We have observed the loci of particle motion
using the electrodes shown In Fig. 5 with d = 2.17
cin and measured electric charge.
Figure 8(a) and (b) show the experimental setup
and electric charger, respectively. Lycopodium
Charger
-0 DC high volt.
Particle t air
- Duct
—o(--)DC high volt.
—-Collecting electrode
""^Flow- Blower
rate /
meter /
(a) Experimental setup
(b) Electric charger
Fig. 8. Experimental apparatus.
particles (a =* 15 pm, m = 1.5 x 10~H kg) which
are almost spherical were used as test particles.
Their electric charge varies around the average q
= 1.1 x lO'1^ c. The average air flowrate was
Vo = 20 cm/s which allows particles to move in
a viscous mode with the condition of £ > 5.
Figure 9 shows the loci of particle motion for
KI = 0-4.0.
Figure 9(a') shows the flow of cigarette smoke.
The collection efficiency in this case is higher than
that estimated from the EHD lines of force. This
may be (1) because eddys occur below the upstream
electrodes and absorb particles into the interior of
the downstream electrodes, and (2) because the
105
-------
•
I
I
(without charf*)
X.I 0
• Flow of cl|»rett* rmolui
'
(• > x -: i
(D x.tc
(,) X.4 0
Tig. 9. Pictures of particle loci (see Fig. 6).
particle motion becomes ballistic in a region be-
low the upstream channel electrodes where the
average flowrate is 3 VQ.
To examine the above results In more detail,
w« observed the amplitude of particle motion in
the ac field and measured electric charge q(c)
of lycopodium particles which escape from the
dowuBtream electrodes. Figure 10 shows the re-
lation between qmax and EO = U/b where U(V) is
the applied electric field. In this electrode
arrangement an Interval between two adjacent
upstream channel electrodes was identical to an
interval of the downstream channel electrodes 4 d
(d = 2.17 cm) and VQ = 20 cm/s. As seen In Fig.
11, when Kc > 1.4, particles do not escape from
the downstream channel electrodes. From Eq.
(10), the theoretical maximum electric charge of
particles which escape from the downstream chan-
nel electrode is given by
(19)
106
-------
Fig. 10. Comparison of the maximum charge
Qmax °f uncollected particles with its theoretical
value (qmax)th (see Fig. 11).
Fig. 11. EHD line of force starting from the
point P (see Fig. 10).
In Fig. 10, the solid curves represent the mea-
sured relation between (qmax)th and EQ for a = 15
xlO'6 m, TJ =1.83x 10-5 Ns/m2 and Vo = 0.2
m/s. Hie maximum of q is dependent on the elec-
tric charger and is about 3 x 10~14 C which cor-
responds to Eo = 0. Accordingly, the measured
qmax is much lower than the calculated value, and
;the measured collection efficiency is higher than
the collection efficiency estimated from the EHD
lines of force.
(1) The analysis of particle motion can be
simplified by use of EHD potential, and the loci
of particle motion can be studied in terms of nor-
malized time and space.
(2) The EHD potential can readily be calculated
by means of charge substitution.
(3) In an electrostatic precipitator, the vis-
cosity factor £ is always larger than 1, and the
mode of particle motion is viscous. The loci of
particle motion coincide with the EHD lines of
force from which collection efficiency can be esti-
mated.
(4) When K ^ 1, the particle motion can be con-
trolled by electric field.
(5) In practice, the collection efficiency is
lower than that estimated from the EHD lines of
force because of the generation of eddys and ballis-
tic region.
The analysis method described above is applic-
able to not only electrostatic precipitator but also
other fields relating to the motion of microcharged
particles in the EHD field.
REFERENCES
1. H. Steinbigler. Dissertation, T.H. MUnchen,
1969.
2. H. Singer. Ibid., 1969.
3. P. Weiss. 'Ibid., 1972.
4. Masuda, Mitsumoto. Trans. I.E.E., Japan,
Vol. 93-A, 305, July 1973.
5. Masuda, Matsumoto and Uemura. 1971
Tokyo Branch Meeting, I.E.E., Japan, No.
372.
6. Takechi, Masuda, Matsumoto, Nioka. 1973
Nat'lConv. I.E.E., Japan, No. 842.
7. Masuda, Matsumoto. To be published in
Trans. I.E.E., Japan.
8. Shibuya, Masuda. To be reported at the 1975
Nat'l Conv., I. E. E., Japan.
9. H.J. White. AEEE Trans., Vol. 70, 1186,
1951.
6. Conclusions
The motion of microcharged particles in the
EHD field has been analyzed and the perform-
ance of two-stage electrostatic precipitator has
been discussed. The result obtained in this re-
search is summarized as follows.
Submitted July 1, 1974
107
-------
Journal of Electrostatics, 3 (1977) 311—325
© Elsevier Scientific Publishing Company, Amsterdam — Printed in The Netherlands
(Reprinted with permission)
A PRELIMINARY STUDY OF RE-ENTRAINMENT IN AN ELECTRO-
STATIC PRECIPITATOR
J.D. BASSETT, K. AKUTSU and S. MASUDA
Department of Electrical Engineering, University of Tokyo (Japan)
(Received November 11,1976; in revised form March 25,1977)
Summary
Observations were made of re-entrained particles which were deposited electrostati-
cally in a laboratory model precipitator. Photographs of particle trajectories were ob-
tained, the mean gas flow at which re-entrainment occurred, and the structure of the
flow near the collecting electrode, were measured.
A distinct difference between trajectories of particles was noted depending upon
whether or not 'back discharge' was occurring. Consideration of adhesion and removal
forces was consistent with mean flow measurements at which re-entrainment occurred,
and the shape of observed particle trajectories could be explained by a combination of
electrical, gravitational and flow forces. Flow measurements indicated a velocity gradient
in the laminar boundary layer sufficient to explain particle removal, although significant
differences in the velocity gradient for different mean flow velocities could not be ob-
served.
1. Introduction
This paper describes an investigation to observe the processes causing dust
re-entrainment in an elecrostatic precipitator. It was the initial stage of a
study into the overall electrofluiddynamic (EFD) nature of the precipitation
process, including particle motion both before and after being first collected.
Fourteen years ago, White [1] devoted two chapters in his book on industrial
precipitation to gas flow and re-entrainment effects. He quotes an example
where poor gas flow reduced a possible efficiency of 95% to 60-70% because
of re-entrainment and poor particle collection. Recently considerable inter-
est has arisen hi the study of the interaction of fluid and electric fields in
precipitators. Studies of the effect of turbulence on the particle concentra-
tion profile [2] or the effect of ionic wind [3] on the precipitation process show
that EFD phenomena are attracting more attention as precipitators have to
become more efficient.
The experiments to be described can be divided into three parts. Particles
were deposited electrostatically on the collecting electrode of a laboratory
model precipitator. Firstly, the re-entrained particle motion was observed.
Secondly, the mean gas flow velocity threshold was measured at which re-
108
-------
entrainment occurred. Thirdly, an attempt to measure the structure of the
gas flow near the collecting electrode was made. The re-entrained particle
trajectories, with and without b?ok discharge occurring, will be discussed as
well as the relative importance of the forces causing (or allowing) re-en-
trainment to occur. Finally, the flow measurements in the duct will be dis-
cussed, and the main conclusions of the investigation given.
2. Experimental work
Observations of re-entrainment, and determination of the re-entrainment
threshold gas flow, both involved photographic recording of the particles.
Flow measurements were undertaken using a laser Doppler anemometer. All
experiments were conducted using the same laboratory model precipitator.
dust
collecting
electrode
plane of ^
focus
camera
U \discharge
electrode
Pig. 1. Experimental apparatus.
109
-------
2.1 Apparatus
A model of a precipitator was constructed as illustrated in Fig. 1. It con-
sisted of a duct, 18-cm square in section and 1.4-m high. The duct was con-
structed of acrylite plate (1-cm thick) with collecting electrodes on two sides.
Five discharge electrodes, made of 1-mm diameter piano wire, were positi-
oned at the centre of the duct, 12 cm apart. These were connected to a
Ransberg 150 kV negative DC source, with a digital kV-meter for voltage
monitoring. One collecting electrode was connected through a microam-
meter to earth, the other directly to earth. The base of the duct was con-
nected to a blower through a flow-meter and baffle valve, as shown in the
figure. The volume flow rate available was 4.5 m3 min"1, giving a maximum
value of mean velocity in the duct of 2.3 m s"1, and a Reynolds number of
~ 34,000. Dust particles could be introduced into the duct by a simple pow-
der-feed system, consisting of a vibrating fluidised bed, the powder cloud
being transported by air flow. A deposit was formed by precipitating the
particles on the collecting electrode under a low air flow velocity.
2.2 Observations of the occurrence of re-entrainment
These initial experiments involved photographic observation of re-entrain-
ment. An area of the collecting electrode was illuminated by a 2-mm wide
slit of intense white light from a xenon lamp. A mechanical chopping disc
could be used if intermittent lighting were required. The period of chopping
used was either 3* 7 or 1- 8 ms. This slit of light was introduced into the duct
with a lens and mirror system, as shown in Fig. 1. Photographs were taken
using a Nikon camera with bellows and a 105 mm lens giving a linear magnifi-
cation of 1^ 2—1° 4 times. The exposure times used ranged from 1/15—1/125
s. Kodak Tri-X film was used and, as no contrast was required, was force
developed to give maximum speed.
2.3 Determination of re-entrainment threshold
In order to determine the flow rate or mean gas velocity at which re-en-
trainment started to occur, photographic observation was also used. The
lighting system was as described above (without chopping) and a fixed ex-
posure time of 1/1000 s was used. Using a mechanical film transport, 8—10
photographs were taken at the rate of about two per second, after a layer
of particles had been deposited electrostatically. The precipitator voltage
was kept constant and several minutes were allowed to elapse between de-
position and flow application. This was to ensure that the layer had reached
a 'steady state' condition. The number of particles in the gas was counted
from the film record for each value of flow rate. As these results could not
be normalised by the total number deposited, although this was kept as con-
stant as possible, the re-entrained number was expressed as a number per
110
-------
photograph, i.e., it was assumed that the series of short exposure photo-
graphs provided a system of random sampling.
2.4 Flow measurements
The method used for flow measurement was a laser anemometer system
(marketed by Nippon Kagaku Co.). An anemometer provides a voltage which
is directly proportional to the velocity of particles passing through the
crossing point of two laser beams. The particles used for these flow measure-
ments were dioctyl phthalate (DOP), 0*3 /um in diameter. The voltage could
be recorded as a function of time and thus give information on velocity fluc-
tuations and turbulence. Velocity measurements were made for three values
of mean flow rate and at 8 positions in the duct. Electrode geometry did not
allow measurement closer than 2 mm from the collecting electrode. Two
methods of analysing the velocity—time signal were attempted. The first in-
volved using a 'real time* correlation and probability analyser together with
a Fourier transform analyser. The second used an F.M. tape recorder to store
the velocity—time signal, complete with DC component, on magnetic tape.
The recording time used for each set was 20 s. This was then analysed using
an A/D converter and conventional digital computer. The second method
yielded more useful results. The mean velocity, the standard deviation, the
probability histogram (and calculated distribution curve) and the power
spectrum could be obtained for one data set simultaneously.
3. Experimental results
3.1 Observations of re-entrainment
For these initial experiments, observations of re-entrainment caused by
air flow alone were sought. However, some interesting results were obtained
when electrical effects also played a part. The first dust to be used in these
y (mm)i
10
Flow
10
20
x (mm)
111
-------
y (mm)
10
Plow
O 1O 20
x (mm)
Fig. 2. Trajectories of re-entrained calcium carbonate particles at a mean flow velocity of
2.3ms"1. (A) 21 kV on discharge electrode — initial current 13 */A. (B) 25 kV on dis-
charge electrode — initial current 40 n A.
I300
200
1OO
with CoCO3
layer
1O
applied voltage (kV)
Pig. 3. Voltage—current characteristics when using calcium carbonate powder.
experiments was calcium carbonate powder. The diameter of the particles
was between 50 Aim and 130 /jm. The trajectories of re-entrained particles
are shown in Fig. 2. The voltage—current curves shown in Fig. 3 for this case,
and the very uneven appearance of the layer, indicated that back discharge
was occurring. Similar results to CaCO3 were obtained when lycopodium
and nylon powders were used. Finally, a sample of glass powder was tried.
The diameter was 60—70 Aim and the particles were spherical. Although ad-
hesion was low, back discharge did not occur. The V—I characteristic with
and without a layer corresponded to the 'no layer' curve of Fig. 3. Trajec-
tories of re-entrained glass particles are shown in Fig. 4.
112
-------
y (mm)
Flow
10
0
0 10 20
x (mm)
y (mm)
"- Flow
B 1O
O 1O 2O
x (mm)
Fig. 4. Trajectories of re-entrained glass particles at a mean flow velocity of 2-3 m s"1 .
(A) 21 kV on discharge electrode — current 14 pA. (B) 25 kV on discharge electrode —
— current 40 n A.
3.2 Measurement of re-entrainment threshold
The glass powder described above was used for these measurements. As
mentioned in Section 2.3, the layer was allowed to stabilise for a few min-
utes before measurements were made. A total of over 1800 re-entrained par-
ticles were counted to give the graph shown in Fig. 5. The mean velocity
across the duct at which these particles were not dislodged by the gas flow
can be seen to be approximately 1*5 m s'1, regardless of applied voltage.
3.3 Flow measurements
Having established a mean flow velocity at which these particular glass
particles were re-entrained, an attempt was made to measure the properties
of the gas flow which caused re-entrainment to occur. The time-mean velo-
city profile in the duct, starting 2 mm away from the collecting electrode to-
wards the duct centre, is shown in Fig. 6. The turbulent intensity, equivalent
-------
30
t 20
0
1 25 kV
2: 23 kV
3 21 kV
10 20 3O
mean flow velocity in duct (m/s)
Fig. 5. Re-entrainment threshold maesurements at: (1) 25 kV, 40 j*A. (2) 23 kV, 20 »A.
(3) 21 kV, 13
^^
f
20-
10-
velocity values
at 17 mm
1 mean velocity 2.3 m/s
2 - . „ 1.8 m/s
3 1.5 nys
XXO
distance from collecting electrode (mm)
Fig. 6. Velocity profiles in model precipitator at three values of mean flow velocity.
-------
to the standard deviation of the velocity signal divided by the overall mean
velocity, is plotted in Fig. 7.
£
1
01
XXV
• mean velocity 23 m/s
o „ „ 1.8 m/s
• „ „ 1.5 m/s
0 5O 10.0
distance from collecting.electrode (mm)
Fig. 7. Turbulent intensity against distance from collecting electrode at three values of
mean flow velocity.
4. Discussion
The behaviour of particles in a precipitator has to be described by a com-
bination of electric and fluid fields. Recently more and more attention has
been paid to this part of the precipitation process. Examples of this are the
work of Masuda and Matsumoto [4], on improving particle collection by the
use of an electrofluid-dynamic approach, the paper by Adachi [3] dis-
cussing the role of ionic wind, considerable velocities being generated from
this source, and the paper by Cooperman [2] in which consideration of the gas
flow is shown to lead to a more generally applicable theoretical equation to
predict precipitator efficiency. The specific problem of re-entrainment in the
gas requires a consideration of the balance between particle adhesion forces
and the removal force caused by the flow.
4.1 Particle removal by air flow
The removal of dust particles by an air flow has been discussed by. Zimon
[5]. The conditions for detachment of a particle from a horizontal surface are
given as:
F^ntf^+P-Fi) (4.1)
where Ff is the frontal force acting on the particle, jf is the coefficient of
friction, fgd the adhesion force, P the weight and F\ the lift force. For a
115
-------
vertical surface this equation will become:.
P (4.2)
The discussion up to now has only accounted for mechanical forces, but in
the precipitator problem, electrical forces also play an important part. For
conducting particles, the charge induced on a particle on a plane surface was
given by Felici [6] as:
9 = 1'5X ID'10 Ea2 (4.3)
Where E is the electric field applied (undisturbed by the particle) and a the
particle radius. Thus, the electrical force of removal, which will tend to re-
duce the adhesion force, Fa(j, is given by :
Fr = l-5X 10 -10 E* a2 (4.4)
A rough estimate of values of the forces described above, as applied to the
spherical glass particles used in the experiments, will now be given. Zimon [5]
quotes a value for the adhesion force F^ of glass spheres on a steel plate as
2-3 X 10 "8 N for 40—60 /urn diameter particles for 97 % relative humidity,
allowing two hours for capillary forces to stabilise. This is therefore to be
considered a maximum value as particles were normally deposited for a
period of a few minutes. Particle weight, assuming a density of 2- 5 X 103 kg
m~3 is 2-8 X 10 "9 N, and the coefficient of friction for glass spheres on
steel, again given by Zimon, is 0-6. The resistivity of the dust used was
measured and found to be 1*7 X 107 flm, i.e. relatively conducting. The
maximum applied voltage used in the experiments was 25 kV giving a maxi-
mum electric field of 2- 8 X 10s Vm "l . The removal force, because of this
field, calculated from eqn. (4.4), was 10~8 N. Thus, Fa(j would range from
2-3 X 10 -• N with no applied field to 1-3 X 10 '8 N with maximum field.
However, the maximum value of adhesion force was given above. Its mini-
mum value can be estimated from
Fad> J (4.5)
giving F^ = 4- 6 X 10"9 N. It is worth noting that there is only a factor of
5 between the maximum and minimum values of this adhesion force. Thus,
the maximum value of frontal force, Ff , can be calculated, if the lift force,
Fj, is neglected:
Ff>0»6(2-3X 10 '• -1-6X 10'17 V2)-2-75X 10'9 (4.6)
Values of Ff calculated from eqn. (4.6) are tabulated in the second column
of Table 1 below, using the maximum adhesion force. If the minimum adhe-
sion force is used, of course Ff is zero.
At each value of mean flow rate a certain frontal force is applied to the
particles on the electrode. It is reasonable to assume that the re-entrained
number (N) of particles is an increasing function of the difference between
the applied force (Fa) and that required for removal (Ff) as calculated above,
i.e.,
116
-------
(4.7)
where N increases as (Fa — Ff) increases, and Fa is assumed proportional to
flow velocity. The results shown in Table 1, taken from the measured values
of re-entrained particles (Fig. 5), are in reasonable agreement with this.
TABLE 1
Re-entrainment number (from Fig. 5)
V(kV) Ff(X ICT'N) 2-3msM 2-0 m s'1 l-75ms"'
0
21
23
25
11.0
6.8
6.0
5.1
0
11
13
29
0
8
10
12
0
4
10
16
4.2 Particle trajectories after removal
For large particles, as used in these experiments, once removal has oc-
curred, the motion can be described by a combination of electrical and flow
forces, as discussed below.
The Jt-direction is assumed parallel to the flow and the y-direction perpen-
dicular to the flow, as shown in Fig. 8. It is assumed that once the flow
force, Ff, has overcome the adhesion force, Fa(j, electrical forces dominate
particle motion in the y-direction, and flow forces dominate motion in the
jc-direction. Velocity fluctuation due to turbulence in these directions is
neglected because of the large particle size. For smaller particles this should
be considered.
electrode
surface \
Flow
Fig. 8. Coordinates used for calculation of particle trajectories.
117
-------
4.2.1 Electrical force
The maximum initial electrical force on the particle will be given by the
charge calculated from eqn. (4.3) multiplied by the electric field. However,
the particle is always being subjected to the corona-ion current flowing to
the collecting electrode. This will result in a lower value of initial charge and
also, once re-entrained, a decrease of particle charge with time, through zero,
to a value of opposite sign so that it will be deposited again. Thus, the parti-
cle charge at time t, calculated using the equation of Pauthenier and Moreau-
Hanot [7], is given by:
t
9(0 = Qi - (Qi + 9o) (4.8)
t + r
where qj is thlTinitial charge, q0 the saturation charge (eqn. 4.9) and T the
charging-time constant, (eqn. 4.10). For relatively conducting particles, such
as glass
q0 = 12neoa2 E (4.9)
where c0 is the permittivity of free space, a the particle radius and E the un-
disturbed electric field where the particle is being charged.
T= —— (4.10)
where J is the current density.
The equation of motion in the y-direction is given by:
ro —^- = q(t)E - GffTjaVv (4.11)
dt
where m is the particle mass, the field E is assumed constant in the region of
particle motion, 17 is the air viscosity and Vy the velocity. This velocity was
calculated numerically for E time increment of 1 ms. The position value, y,
was calculated from the velocity difference:
y- V'+V' + 1 xiO-3 (4.12)
4.2.2 Gravitational and flow forces
The equation of motion in this case is given by:
• m g + 6iri?a( Vf(y) - Vx) (4.13)
where g is the gravitational constant, and Vf(y) is the flow velocity value at
y at the same time increment for which Vx is being calculated.
As no flow data were available when these calculations were made, eqn.
(4.13) was solved for Vf = 0.
118
-------
4.2.3 Particle trajectories and the influence of back discharge
Particle trajectories calculated by the above method for initial charge
values of 50, 60 and 70 % of that calculated by eqn. (4.3) are shown in
Fig. 9. These compare well with the trajectories of glass particles during de-
position, where only gravitational forces were acting. They are similar to the
trajectories of re-entrained glass particles as shown in Fig. 4.
The trajectories of re-entrained CaCO3 particles shown in Fig. 2 clearly
show the difference when back discharge is occurring. Instead of particles
remaining within a millimetre or two of the collecting electrode surface and
re-deposition occurring, they are ejected to several millimetres (up to 16
mm), and in the time interval of the photographs often showed no sign of re-
depositing. Work on particle charging during back discharge by Mizuno [8]
has shown that a dust layer, when back discharge is occurring, can act as a
surface source of ions of the opposite sign. Thus, particles would be ex-
pected to move further from the electrode, as re-charging would not imme-
diately commence. It would appear that the effect of back discharge on pre-
tipitator performance is not only reduced collection efficiency, but also de-
trimental if re-entrainment occurs, as particles are removed into the full flow
600 •
1000 -.
1500 •
70*'.
Fig. 9. Calculated particle trajectories.
119
-------
of the gas stream, instead of staying close to the collecting electrodes. The
observed trajectories also show an extremely high velocity perpendicular to
the flow. This might be explained by ionic wind occurring from the back-
discharge points. This kind of ionic wind has been observed by Adachi [9] .
The particles used in these experiments were considerably bigger than
those normally precipitated. Domination of particle motion away from the
collecting electrode by electrical forces would probably not occur with
smaller particles and flow forces would also have to be considered in this di-
rection. However, when small particles are electrostatically precipitated,
strong coagulation occurs so that the agglomerates actually re-entrained
probably have a size of several tens of microns. Thus these experiments using
large CaCO3 and glass particles should serve well for predicting how re-en-
trainment occurs in practical precipitators.
4.3 Structure of the flow
During the earlier stages of these experiments, it was hoped that a back-
scatter detection laser anemometer could be used. This would enable the
flow very close to the collecting electrode to be measured either parallel or
(almost) perpendicular to the electrode. As mentioned in Section 2.4, the
particles used to seed the flow for laser Doppler measurement were DOP, 0.3
jum mean diameter, and it was found that the back-scattered intensity was
not sufficient for measurement. The sampling rate of the burst signal was less
than 100 s~! , giving a frequency response of the order of tens of Hz. Thus,
a forward scatter system, as mentioned earlier, was used to give the results
shown in Figs. 6 and 7, and this is the reason for measurements only begin-
ning 2 mm away from the electrode.
From Fig. 6 it can be seen that the thickness of the boundary layer is
several millimetres, and Fig. 7 shows that at all positions the flow was fluctu-
ating. It can be seen from Fig. 6 that the velocity gradient in the boundary
layer close to the collecting electrode was almost the same regardless of the
main flow velocity. A calculation of the frontal force, Ff, as discussed in
Section 4.1 assuming a laminar boundary layer, is given by Schlichting [10] :
. 6^77
Ff = - — (4.14)
where Vj-j is laminar boundary layer velocity and 5 is the thickness of the la-
minar boundary layer. It is unlikely that this model could explain the signi-
ficant differences of re-entrainment observed as shown in Fig. 5, as the flow
data would indicate values of Ff not significantly different for the various
values of mean velocity. However, the value of Ff calculated from eqn.
(4.14) gives 1« 2 X 10 ~9 N, for 60 um diameter particles and a velocity gra-
dient of 500 s"1 , which is of the right order given that the assumptions in
section 4.1 are correct. Flow measurements closer to the electrode are re-
120
-------
quired before more accurate calculations from eqn. (4.14) can be usefully
employed.
It can be seen that the values of turbulent intensity at different flow rates
were all approximately the same. In fact the value 2 mm from the collecting
electrode for the lowest flow rate was highest. The mean value of turbulent
intensity for 1.5 m s"1 applied velocity was 0.16, for 1.8 m s'1 it was 0.11
and for 2-3 m s"1, 0-10. The order of error of the velocity measurement
system was assumed to be 10 %. When the turbulent intensity results were
re-plotted, after subtracting 10 % of the mean value from the standard de-
viation, the same order of results occurred — highest for the lowest flow rate.
Unless a constant value of error was occurring, allowing a fixed amount to
be subtracted from the velocity standard deviation, the turbulent intensity
measured appears to be slightly higher for the lowest value of applied flow
rate.
The power spectra indicated that energy was contained up to higher fre-
quencies at higher flow rates. The maximum frequency for a fixed power-
spectrum value was taken from all the spectra (for the eight values of dis-
tance into the duct). The mean value of this maximum frequency for a !• 5
m s'1 applied velocity was 17- 3 Hz, for !• 8 m s~1 it was 20- 2 Hz and 2- 3
m s"J it was 28-4 Hz. These values are more reasonable than the turbulent
intensity measurements, possibly because the method of obtaining the power
spectrum analyses the shape of the whole velocity—time signal, whereas cal-
culations of standard deviation can be influenced by spurious voltage fluctu-
ations which might have occurred as no filtering was used before the signal
was input to the A/D converter.
5. Conclusions
Observations of re-entrainment occurring showed that the motion of the
glass particles away from the collecting electrode was not influenced by the
flow once they had become dislodged. Electrical forces perpendicular to the
electrode could explain the observed trajectories. The smaller particles nor-
mally precipitated would be expected to be affected by flow forces perpen-
dicular to the collecting electrode. However, electrostatically precipitated
particles often coagulate on collection, the agglomerates having a size of
several tens of microns, similar to the particles used in these experiments.
Considerations of the adhesion and removal forces acting on a particle
were in agreement with experimental measurements of the number of re-
entrained particles plotted against the applied flow velocity.
The influence of back discharge on re-entrainment was observed. This
showed that loss of efficiency when re-entrainment occurs is likely to be far
worse if back discharge is occurring as well, because particles would be
ejected into the full flow of the gas stream instead of staying close to the
collecting electrode.
121
-------
Acknowledgements
Dr. Bassett would like to acknowledge gratefully the support of the Royal
Society for a 10-month fellowship in Japan, and to thank his co-authors for
their kindness and hospitality during his stay.
References
1 H.J. White, Industrial Electrostatic Precipitation, Addison-Wesley, 1963, Chaps. 8 and
10.
2 P. Cooperman, Nondeuschian phenomena in electrostatic precipitation 69th Annual
Meeting of the Air Pollution Control Association, Paper No. 76-42.2, Portland,
Oregon, June 27—July 1 1976.
3 T. Adachi, Ionic wind in the electrostatic precipitator, J. Res. Assoc. Powder Tech.,
Jpn, 12 (3) (1975) 146.
4 8. Masuda and Y. Matsumoto, Motion of a microcharge particle within electrohydro-
dynamic field, Electr. Engi. in Jpn, 94 (6) (1974) 20.
5 A.D. Zimon, in M. Corn (Ed.), Adhesion of Dust and Powder, Plenum Press, New York,
1969.
6 N.J. Felici, The forces and charges on small objects in contact with an electrode in an
electric field, Rev. Gen. Electri. 75 (1966) 1145.
7 M.M. Pauthenier and M. Moreau-Hanot, The charge acquired by spherical particles in an
ionised atmosphere and an electric field, J. Phys. Radium, 3 (1932) 590.
8 A. Mizuno, Studies of back discharge phenomena, J. Electrostal, (to be published).
9 T. Adachi, Ionic wind in the electrostatic precipitator — experimental treatment by the
Schlieren method, Trans. I.E.E., Jpn, 93-B. (7) (1973) 273.
10 H. Schlichting, Boundary Layer Theory, J. Kestin (Trans.), McGraw Hill, New York,
1968.
122
-------
Inst. Phys. Conf. Ser. No. 27©I975: Chapter3
(Reprinted with permission)
Recent progress in electrostatic precipitation
Senichi Masuda
Department of Electrical Engineering, University of Tokyo, Tokyo, Japan
Abstract. Recent progress in the field of electrostatic precipitation, which is one of the
most important applications of electrostatic forces, is reviewed. A description of the con-
struction and general principles of precipitators is given. This is followed by an account
of the progress achieved in both the technicaf developments and the scientific understand-
ing of precipitator performance. Finally, the inherent difficulty in the design of precipi-
tators is explained.
1. Introduction
Electrostatic precipitators play a major role in the emission control of particulate
pollutants, especially when the emphasis is on the removal of very fine particles of less
than 3|xm diameter. These fine particles are important factors in the visibility of stack
emissions, water drop nucleations, the carriage of gaseous pollutants into lungs and they
are a general health hazard. Although both the principles and the construction of electro-
static precipitators are extremely simple, the actual operation is complicated by many
factors which impair the efficiency. As a result, in spite of many research achievements,
precipitator design has long been considered an art rather than an engineering science.
This situation, however, is being improved by increased efforts in research and
development.
2. Principle and construction
The principle of electrostatic precipitation is explained by the system of concentric
cylinder electrodes shown in figure 1. Particles charged by collision with unipolar
ions emitted from the discharge electrode are driven by the coulombic force on to the
collecting electrode, where they are precipitated. The layer of particles is dislodged by
mechanical rapping of the collecting electrode and they fall into a hopper. Meanwhile
the cleaned gas is taken from the outlet to a stack. In practice, a duct-type precipitator
with parallel-plate collecting electrodes and a horizontal gas flow is usually used for
treating large volumes of gas. This is because of its simple and economical construction
and uniform gas distribution. A negative corona is usually used for emission control
because of its higher spark voltage, whereas the positive corona is chosen for the purpose
of air cleaning when the supression of harmful ozone becomes a major factor. In the
latter case a two-stage structure is common in which the charging and collection is per-
formed in different stages. For the voltage sources, only one reference is cited (Hall
1975) and a discussion of this subject is omitted.
123
-------
Recent progress in electrostatic precipitation
To Hv'Source
Insulator
Gas outlet
Ion striom
Gas inlet
Weight
Hopper
Oust outlet
Figure 1. Principle of electrostatic precipitation.
The corona field inside a precipitator having both ion and dust space charges is not
easy to calculate except for the simple case of figure 1. The field at a distance r from
the axis is approximately given (Pauthenier and Moreau-Hanot 1932) by
F /
r=
F*r~\ 21 y*I V \
C0'0\ I II
~r~) I \ml
(1)
for the case when the total surface area of particles, 5"(m2/m3), per unit volume is not
very large. Here,/ = current per unit length of wire (A m~1);e0 = permittivity of free
space; es = relative permittivity of particulate material;n = ion mobility (m2 V~* s"1);
E0 = breakdown field strength of the gas at the wire surface (V m"1).
If the applied voltage is kept constant, the dust space charge causes an increase in E
adjacent to the collecting electrode and a decrease in the vicinity of the discharge elec-
trode, thus lowering the charging current /. This last effect is called 'corona quenching'.
The effect of this quenching on the precipitator is two-fold. Firstly, the charge on a
particle decreases as a result of the drop in charging rate, and thus there is a decline in
efficiency. Secondly, the increase in the collection field strength causes an increase in
efficiency. It was discovered recently by Awad and Castle (1975), that if the initial
corona current was low, the latter effect was more than counteracted by the former
effect and therefore there was a resultant decrease in collection efficiency.
The particles entering the corona field are charged by ion collision by two mecha-
nisms. One is the effect (called field bombardment) of the external field driving ions
towards the particle surface and the other is the thermal diffusion mechanism, in which
collisions result from the thermal motion of the ions, without the aid of an external
field. The theoretical charge acquired by a spherical particle by field bombardment is
expressed (Pauthenier and Moreau-Hanot 1932) by
'fr
1+f/T
(C)
(2)
124
-------
Senichi Masuda
where
3es ,
. = 4tre0 fl £"c = saturation charge (C) (3)
and
4e0 4eo£"c
T = — = - = charging time constant (s) (4)
'
where t = time (s);a = particle radius (m); Ec = charging field strength (V m '); PJ = den-
sity of ionic space charge (C m"3) and i = ion current density (A m~2). Equation (3)
indicates tne importance of the field strength Ec in determining the saturation charge on
the particle, whereas equation (4) shows that the current i governs the charging rate. We
may assume t0= 3r to K)T as the necessary charging time, because 75% and 91% of the
saturation charge are obtained after 3r and 10r. respectively. If we take the typical
values of Ec = 5 x 10s Yin"1 and / = 2 x 10"4 Am'2, we obtain T = 0-088, in other words
the necessary charging time t0 in this case is between 0-26 and 0-88 s. The theoretical
charge imparted to a spherical particle by thermal diffusion is (White 1951):
(5)
where
4-nefflkT
q*= --- = charge constant (C) (6)
4ne0kT
T* = — = charging time constant (s) (7)
k = Boltzmann's constant = 1-38 x 10"23(J K~l); T= absolute temperature (K); e =
electronic charge = 1-602 x 10~19(C); C = RMS value of the ionic thermal velocity =
(3kT/m)^2 (m s~'), m = ionic mass (kg); NQ = number of ions per unit volume (m ~3)
and the assumption is made that a > \ where X= ionic mean free path. According to
equation (5) the charge q initially rises very quickly to become q = 6q* at t = 402r*,
thereafter rising very slowly so that it can be assumed to remain approximately con-
stant. Hence, we may take qM = 6q* as the quasi-saturation charge with charging by
thermal diffusion and t0 = 402r* as the necessary charging time. For 7=150 °C;
a = 0-lnm: m = 5-313 x 10"26kg (for Oj ion) and AT0 = 5 x 1013m"3 we get f0 = 1 -13 s.
These are typical values for industrial precipitations.
If a was as small as 0-01 (im the necessary charging time becomes the large value of
113s. Numerical calculations show that the field bombardment charging is predomi-
nant for particles larger than 2 (xm, whereas thermal diffusion charging dominates for
particles smaller than 0-2fim. In the intermediate range, the sum of the charges calcu-
lated independently by equations (2) and (5) gives a good approximation (Hewitt
125
-------
Recent progress in electrostatic precipitation
1957). There are many detailed studies on particle charging (see Penny and Lynch 1957,
Murphy etal 1959, Smith and Penny 1961, Liu etal 1967, Liu and Yeh 1968, Smith
and McDonald 1975).
The charged particles migrate under the action of the coulombic forces towards
the collecting electrode. Assuming that the viscous resistance acting against the particle
motion is of the Stokes form, then the theoretical migration velocity within a gas at rest
is
(8)
where £"p = collecting field strength (V m !) and rj = gas viscosity (N s m 2). For very
fine particles below 1 (im in diameter the theoretical migration velocity must be modi-
fied by multiplying by the Cunningham correction factor (1 +,4X/a), in other words w
increases with decreasing particle size. For air at NTP, A = 0-86 and X = 0-1 (im (White
1962). The curve A in figure 2 represents the theoretical migration velocity, w, as a
function of particle radius, a, under typical precipitator conditions. It is assumed that
fe-SkVcnr1
Figure 2. Theoretical particle migration
velocity w (curve A) and apparent migra-
tion velocity W (curve B) against particle
radius a. Curve C is thermal diffusion.
01 I 10 100
Particle radius. <7(/jm)
the charge imparted by ionic thermal diffusion is q» = 6q*. The curve clearly indicates
that if sufficient charging time is available, w becomes a minimum when a is between
0-1 and t'Opm; this was verified by Hewitt (1957) experimentally and in field tests by
McCain et al (1975). The remarkable increase in the necessary charging time, fo, required
for particles with a less than 0-1 {xm should be noted, since the available charging time
in practice is normally limited to 5 to 10 s. The theoretical migration velocity given by
equation (8) cannot usually be used to estimate the collection efficiency r?c because of
too many disturbing factors. These include turbulence, which is enhanced by the elec-
tric wind; the partial re-entrainment of precipitated dust, etc. A first order approxima-
tion of TJC can be obtained by using the very crude assumption that, because of the mix-
ing effect of the turbulence, the particle concentration is uniform over an arbitrary
cross section perpendicular to the gas flow, and that the collection rate is governed by a
single parameter called the 'apparent migration velocity', W, for all particles, regardless
of size, throughout the whole collecting region. We then obtain the well known Deutsch
equation:
Tjc=l-e-WF (9)
126
-------
Senichi Masuda
where F = SC/Q = specific collection surface (s m"1) where Q = total gas flow rate (m3
s"1) and Sc = total surface area of the collecting electrodes (m2). Equation (9). because
of its simplicity, is widely used for design purposes in either its original or a modified
form. W is to be considered a design parameter representing all the process factors
except the precipitator dimensions, and should be determined using equation (9) by
experimentally measuring the collection efficiency. The curve B in figure 2 represents
an average value of V calculated from the fractional collection efficiencies measured in
different industrial precipitators. The difference between curves A and B probably result
from the fact that the larger particles tend to re-entrain more easily because of their less
effective adhesion compared to smaller particles (Heinrich 1961). The factors affecting
W are man> in number and usually difficult to estimate in advance. As a result precipi-
tator design requires a lot of experience which is obtained from analysing data on simi-
lar precipitators already in operation. The data often differs, however, from plant to
plant. This situation means that the prediction of precipitator performance is probabi-
listic in nature, especially when sufficient safety allowances cannot be included (Masuda
1966). Another difficulty in the concept of'apparent migration velocity' has been
raised recently because of results from 'large-spacing type' precipitators. .These have
much larger electrode spacings than conventional precipitators and yet the two types
have comparable collection efficiencies even when they are of equivalent sizes and oper-
ate under nearly identical plant conditions. This comparibility was also observed even in
a pilot plant of a wet-type precipitator where no back discharge or dust re-entrainment
occurred (Ago etal 1975). It seems under suitable conditions that the 'apparent migra-
tion velocity' increases in proportion to the electrode spacing. This effect cannot be
fully explained, even considering the increased stability of operation which is a feature
of the large-spacing type, and thus there is a need for more detailed studies on the pre-
cipitation process itself.
In the following section, some of the recent progress achieved in understanding the
precipitation process is described.
3. Ion curtain patterns and their effect on charging efficiencies
It is well known that the negative corona appears on a wire electrode at several
points, from which ion currents in the form of tufts start towards the collection elec-
trode. Hence in the region near the wire, ion dead spaces occur between the tufts and
in these spaces the ion concentration is so small that particles passing through them may
not be charged; in other words the charging time constant of equation (4) or (7) becomes
exceedingly large. The decrease in charging efficiency in the dead spaces was confirmed
experimentally (Masuda et al 1973b) and this led to more detailed studies on the
ion curtain patterns. It was discovered that there was a similarity between the ion
curtain pattern and the electrode configuration. Figure 3 shows that the distribution
of ion current upon the collecting electrode follows a similar pattern to that of the
electrode system (Niioka 1974, Masuda and Niioka 1974). It was also observed that
in cases where there were dead spaces on the collecting electrode, back discharging and
re-entrainment took place and that, for some particles a number of fibre-like pearl chains
protruded from the surface of a dust layer in a dead space. These chains could jump
into space and split into sections (Masuda etal 1973 a).
127
-------
Recent progress in electrostatic precipitation
(/-2cm
10
100 MA
5QM A
d * 10cm
. Z-200/«A
* IOOMA
- 50pA
10/iA
rO 20
Normalized distance, rid
Figure 3. Normalized current density on
plate electrode (i/ip) against normalized
distance (r/J).
The problem of dead spaces may be solved by the use of a special type of positive
corona called 'Hermstein's glow'. This glow occurs uniformly along the wire but still
has a greatly increased spark voltage; sometimes higher than that of a negative corona
(Hermstein 1960). Hermstein's glow occurs when the field strength and its gradient in
the vicinity of the discharge electrode are so high that the shedding of electrons from
negative ions can occur in this restricted region. These electrons diffuse over the elec-
trode surface to form an electron sheath capable of suppressing streamer formation.
Hence, the formation of Herstein's glow is encouraged by supplying the discharge elec-
trode with a small number of negative ions (Hermstein 1960). A remarkable increase
in the average charge on a particle could be obtained by using Hermstein's glow, pro-
duced by knife-edge electrodes, compared to the charge obtained from using a negative
corona from identical electrodes (Niioka 1974, Masuda and Niioka 1973).
4. Electro-fluid dynamic phenomena and particle motion inside precipitators
There are two kinds of phenomena to be studied in more detail from the EFD point
of view; these are electric winds and particle motion and in both the effects of electric
and fluid fields have to be considered. Figure 4(a) is a Schlieren photograph at the
core of an electric wind, taken with a horizontal gas flow with a velocity of 1-25 m s"1.
An approximate estimate of the electric wind velocity obtained from the curvature of
the curve is about 20 m s"1 in the vicinity of the needle point, and the average value is
in the range of several m s"1. This is much higher than expected (Adachi etal 1974). It
follows, therefore, that particles of less than 10 ^m diameter may be transported towards
the vicinity of the collecting electrode by the vortex motion of the gas flow which
is enhanced by the electric wind (see figure 2). At the collecting electrode, the flow
must reverse and only the particles impinging on the boundary layer, either because of
the electric force or the random motion of the particles, may be collected. Therefore,
the motion of charged particles, especially small ones, can only be correctly understood
using the EFD approach, in which the mode of the motion can be approximately esti-
mated by the dimensionless factor
(10)
128
-------
Scnichi Masuda
(a)
Fgure 4. Schlicren photographs ot electric winds (j) Core of elecim wind id - 3 cm.
K = I 8 kV, i = 1 -25 m s~ '; (ft) electric wind by back-discharge (d = 6 un. I Ui kV).
where w0 represents the theoretical migration velocity under the average electric held
between the electrodes, £"<>, and ^represents the RMS value of vortex velocity The
effect of the electric field will become predominant for particle transport for A' > 10.
while the effect of turbulence will play a major role when K < 0-1. More detailed intor-
mation is being collected both concerning the coupling of the vortex motion and the
electric wind, and the structure of the boundary layer under actual conditions I-or
example, figure 4 shows a picture of the electric wind produced by a back discharge
(Masuda and Adachi 1975).
When no vortex motion exists, as in the case of the collecting part ol a sju-cial two-
stage precipitator shown in figure 5 (Masuda et al 1974b, Shibuya and Masuda D75).
the problem can be simplified by approximating the fluid flow to the theoretical How
of an ideal fluid. In this case, both electric and fluid fields can easily be calculated by
the use of the charge-substituting method (Steinbigler 1%9) and its modification
(Matsumoto'1974. Masuda and Matsumoto I974a), even with complicated boundary
Figure 5. A new two-stage typo electrostatic precipitator with the bias-controlled pulse
charging system for the charging parts and the channel electrodes tor (lie collcctmt! parts
1. Charging part. 2. collecting part: 3. discharge electrode; 4, counter electrode. 5 third
electrode. 6. driving electrode; 7. collecting electrode. 8. pis inlet. 1. jzas nutlet
129
-------
Recent progress in electrostatic precipitation
conditions. The particle motion can be analysed using this method by using the con-
cept of the EFD potential, in which the equation of motion is expressed in the form
df2
6flTjadR
"d7
-grad
(11)
where
- [ (6m7flV(R)
. dR
= EFD potential
(12)
where R * position vector, m - particle mass, V(R) = velocity vector of fluid flow.
Whether the particle motion is by the ballistic or viscous mode is determined by a para-
meter £ - (6irnab/mV)in, where b - electrode distance and V~ average gas velocity.
When f > 1, as in the case in most practical systems, the inertia term in equation (11).
can be neglected compared to the viscous term, the motion becoming 'viscous' in nature
and follows the EFD lines of force drawn inside the EFD potential field. Figure 6
shows the EFD lines of force inside the collecting region in figure 5 (Matsumoto 1974,
Masuda and Matsumodo 1974b). With this motion the effectiveness of the electric
force compared tojhat of the fluid force is governed by the dimensionless factor K =
q£/6irnaV where E is the average electric field between the electrodes. One would
achieve a 100% collection efficiency with/: > 1-95. Figure 6(a) indicates the stream
lines of the fluid field when K = 0.
Figute 6. EFD Unfit of forc« inside the collecting parts of the
precipitatcr shown in figure 5,
130
-------
Senichi Masuda
5. Adhesion of particles and dust re-entrainment
The pre-requisite conditions for effective dust collection are both the existence of
enough adhesion between particles and sufficient adhesion between the particles and
the collecting electrode. This allows the build-up of a firm layer, which on rapping can
be dislodged to fall into the hopper without disintegrating and being carried out by the
gas stream. In this sense, an electrostatic precipitator should act as an effective dust
coagulating device in which electrical adhesion plays a major role for particles with a
resistivity pd > 1010n cm (Dalmon and Tidy 1972a). This force is caused by the poten-
tial difference between particles in contact and is proportional to both the apparent
resistivity of the dust layer, pd, and the apparent current density within the dust layer,
/a (Simm 1962). Another possible cause of adhesion of an electrical origin is that pro-
duced through contact electrification (Penny 1975). Low-resistivity particles (pd < 104
SI cm) arriving at the collection electrode are 'inversely' charged by induction even
though they are also bombarded by ions, and pulled back into the gas stream, in other
words there is an abnormal dust re-entrainment of these particles unless the non-
induction adhesion forces are powerful enough to overcome the induction effect. Van
der Waals (London) forces are also involved in adhesion; their effect on coagulation
increases with decreasing particle size (Lowe and Lucas 1953). At comparatively low
temperatures, when the relative humidity is high enough, the capillary condensation of
water molecules on to nucleation centres, may also be an important parameter in
adhesion. It was recently reported that a remarkable increase in collection efficiency for
fly ash could be obtained by enhancing dust adhesion through injecting small amounts
of suitable additive compounds such as triethylamine (Tassiker 1975) or ammonia
(Dismukes 1975). (This ammonia injection is also used to solve the S03 corrosion
problem in boilers burning heavy oil.) It is found that very fine fumes of ammonium
sulphate or ammonium bisulphate are produced in considerable quantities and that this
often results in corona quenching (Dismukes 1975).
6. Back discharge (back corona)
The back discharge is one of the most difficult problems impairing precipitator per-
formance in many large scale industrial plants. The plants affected include ore sintering
furnaces in the steel industry; rotary kilns and clinker coolers in the cement industry;
smelter furnaces in the metal industry and especially boilers in thermal power plants
burning low sulphur coal (these produce high resistivity, mainly metal oxide, dusts).
The effects of the back discharging depend on the value of pj. There are two major
ones: one is the excessive sparking that occurs when the resistivity is between 5 x 1010
and 10I2f2cm. This causes a decrease in the collection efficiency because of the impair-
ment of the collection process. The other is the increase in current which occurs when
pd > 1012n cm. A copious number of ions are emitted from a number of corona points
occurring over the whole surface of the dust layer on the collecting electrodes, and
these ions neutralize the useful charge on the particles so that often no particles are
collected. These phenomena occur as a result of the breakdown of the dust layer because
131
-------
Recent progress in electrostatic precipitation
of its high apparent resistivity, Pd> when the following field condition is locally satisfied
at the layer's weak points:
(13)
where E& = apparent field strength in the dust layer and E^s is the layer's breakdown
strength. Normally the breakdown of the dust triggers streamers in the gas space which,
depending on the field distribution around the breakdown point, proceed either towards
the discharge electrode, or to the space charge accumulated around the breakdown
point, or in both directions. These streamers cause large amounts of carrier multiplica-
tion and photon emission; these processes supplying a copious number of positive
ions which cannot only neutralize the particle charge but also the negative ion space
charge. After the extinction of a streamer, the surface charge is restored by negative
ions supplied either from the discharge electrode or from the residual gas plasma, and
thus the cycle can once again be triggered, each cycle causing a pulse discharge. Figure 7
indicates the effect of the current, /, and the average field strength in the gas space,
normal to the sample layer, /Tgn (Efn - Vgfd) upon the mode of back discharge. Both
/and £gn could be changed independently and a mica plate, with a 0-5 mm diameter
pin hole, was used as the sample layer (Mizuno 1975, Masuda and Mizuno 1975a).
It is shown that a breakdown at the pin hole triggers a surface discharge when
£"gn > 3-3 kV cm"1 and a streamer in the gas space when Zfgn> 5-1 kV cm"1. These
discharges become sparks if a certain limit of / is exceeded, even with a constant Egn.
IUU
5 10
3
K,
1
0
H>
m
Su
dis
rei
rfoce
charg
jion
r
t
\]
y
>
V
«
1
(3
f
;spc
; str
• reg
s.
\
\
«N
*/
ce
earner
ion
^-.-.v./
'
/
$
'\
lv>
«
!*"
$
$
1
i
2345678
Field strength. fr(kV cm)
Figure 7. Effects of the normal field strength,
£gn, and the total current, /, upon the mode of
back discharge (£Vis = 15-6 kV cm"1).
When Egn > 8-4 kV cm 'a streamer becomes a spark. Figure 8 shows pictures of back
discharges which indicate the effects of tangential fields. These fields are caused by the
surface charge of density a
-------
u'hi '•'
.
A..
I:igurc 8. Hlcct ot the taiificntul lidd stfcn.cth upon the modi- ol b.uk discharge
itflAfjs = 22 2 kV cm '. 18kV.-U.-iA i/>)/•..,• J6-5 kVcnrT1. ISkV.
excessive spar king and abnormal rises incur i en t occur witli dillerent values of the apparent
dust resistivity pt). Consider the differences in voltage drops across ihe gas region and
dust layer that occur with differ ; u low p(1. when the applied voltage is raised.
/:'j in the dust layer will remain small, while /:'t5M in the gas ieuioii will sharply rise
causing a rapid gtowth of streamers and thus e\ces-;i\v sparking. When, on the other
hand.pj is higji, the situation will be reversed, in othei words no spaiking will occur in
the gas region, but iheie will be breakdown at a number ot weak points in the dast layer
because ol the rapid rise in /-';). This will resuli i: p rise in the total cuirent.
Figure l> shows an oscillogram of a typical cuiienl waveform of a back discharge pulse
along with the accompanying light emission. l;igure 10. shows the variation of light
emission at various points in space as well as over the surface layer. A glass plate with
a pin hole was used as the sample layer and the spot resolution was 0*3 mm (Mizuno
1975, Masuda and Mi/uno ll>75b). The light signals at the graph origins show that the
back discharge pulse consists of two parts, the primary wave rises very rapidly and lasts
about 20 ns; the secondary wave rises more slowly and lasts about 200 ns. The former
effect corresponds to the first rise in the current waveform which has a small pulse
height and consists of a charge of 1 2 x 10 " C/pulse (shown in figuie 1J). The second-
Figure 9. Wavcforim ot current and lighi signal of hack-ilischari;e pulse
133
-------
Recent progress in electrostatic precipitation
0 200ns
z-axis
0 ZOOns
/•-axis
Figure 10. Waveforms of light signals of back dis-
charge measured at various positions taken along the
z- and r-axes.
ary wave corresponds to the second rise in current which has a much larger pulse height
and a charge content of 2-4 x 10~8 C per pulse. It was confirmed that the primary cor-
responds to a streamer advancing into the gas region with a speed of 4 x 107 cm s"1 and
the secondary to the surface discharges which cause much greater charge multiplication.
It was observed that when the applied voltage was increased the charge per pulse re-
mained the same in both cases whilst the average repetition frequency increased. The
repetition frequency was found to be lower for the surface discharge mode than the space
streamer mode, so that, as far as one breakdown point is concerned, the current rise
becomes steeper for the latter mode in spite of its smaller charge emission per pulse.
Figure 10 also shows that the light signal from the needle point starts almost simultane-
ously with that from the pin hole and that there is a second rise in emission at the needle
with the arrival of the steamer. Figure 11 shows a schematic representation of the
streamer propagation calculated from the waveform in figure 10. Figure 12 illustrates
a picture of the back discharge which occurs when a positive potential is applied to the
discharge electrode. This shows a completely different mode in which discharges occur
from points uniformly distributed over the dust layer. As the applied voltage is increased
no streamers are produced, either from the needle point or from the back discharge
'fc
'30
1
»70
1
JI50
1
10
JL
Uo
1
i*>
1
1200
'20
J^ .
'so
1
|I20
I
^^m
250
Figure 11. Propagation of back discharge streamers (sche-
matic representation).
134
-------
Senichi Masuiia
Figure I 2. link
under positive discharge electrode.
points, but a spark discharge suddenly occurs. The steamer corona at the positive
needle point is fully suppressed, presumably because of negative ions supplied from
the back discharge points (as in the case of Hermstein's glow). As a result, the spark-
over voltage is increased compared to the case when the dust layer is removed (Mizuno
1975, Masuda and Mi/uno I975a).
7. Particle charging
In addition to the theoretical and experimental works already cited, a detailed
examination of Pauthcnier's equations (2)-(4) was conducted in which the field
strength £c, corona current density /, and the charging time / could be changed inde-
pendently (Masuda and Akutsu 1975). With spherical conducting particles a very good
agreement of the measured value of charge with that calculated from equation (2) was
obtained while, for particles such as teflon, with high surface and volume resistivities,
the saturation charge always remained about half as much as that given by equation
(3), except when the particles were subjected to rotational motion. This discrepancy
is evidently the result of the fact that the charges imparted to the insulating particles
by the ion bombardment cannot be instantly uniformly distributed over the particle
surface; this was an important assumption in the derivation of equations (2)-(4). In
the case ol moderately resistive particles (i.e. glass) where the dielectric and surface con-
ductivity relaxation times, rp, ;ue negligibly small compared to the charging time con-
stant, r. or equation (4), then there is good agreement between the measured values of
acquired charge with those calculated from equations (2) (4) for conducting particlci
of the same radius (et - °°). It is evident that in this case the particles behave like con-
ducting paiticles because the distribution time for the charge (from ion bombardment
and polari/ation) is the short time taken for the internal field to disappear. Hence the
term 3e,/(e,+ 2) in equation (3), expressing the effect of the dielectric constant, losei
its meaning. This is due to the contradicting assumptions that the distribution of
polarization charge has the time constant for a dielectric particle and that the distribu-
tion of imparted charge has the time constant for a conducting particle. Under normal
135
-------
Recent progress in electrostatic precipitation
precipitator conditions we can assume that particles having a volume resistivity of below
10n ft cm (this implies pj < 1013 ft cm (Masuda 1965)) to be quasi-conducting particles
(Masudaand Akutsu 1975).
The particle charge imparted by bipolar ions was also measured under back-discharge
conditions. The distribution of positive and negative ion densities, p+ and p_, were cal-
culated by using the data obtained in the relation derived by Pauthenier (1961). It
could be shown that even a weak back discharge could cause a remarkable decrease of
the particle charge to 10% of its normal value. This was predicted by Pauthenier
(1961). In some cases a polarity change even occurred. It was also observed that, in
the space streamer mode, p+ and p_ became nearly equal and almost constant through
the luminous region, which suggests that the carrier production ta'ces place in a fairly
wide region of the gas space. In contrast with the surface discharge mode,p+ and p_
showed an exponential decay in space from both sides. However, their rates of decay
with distance were very low so that a long extension of the positive ion cloud to the
discharge electrode region was observed.
As a novel method of particle charging, the use of gamma rays has been proposed.
The bipolar ions produced by the high-energy electrons are separated by a transverse
electric field, and used to charge the particles which are then collected on to the elec-
trodes (Heinsohn etal 1975).
8. Dust resistivity
Figure 13 exemplifies the effect of pd upon the precipitator performance, expressed
in terms of the apparent migration velocity, W. It is clearly indicated that the prefer-
able range for electrostatic precipitation is for particles with pj between 104 and 1010
ft cm and that the efficiency is limited by dust re-entrainment and back discharge. The
s;
§
s
o>
'§.
Abnormal dust re-entrainment
region
Normal region
Back-discharge
region
I03 I04 I010 10" IOB I013
Apparent dust resistivity p^ (Qcm)
Figure 13. Effect of dust resistivity, pj, upon the apparent migration velocity, W.
value of the apparent resistivity, p
-------
Senichi Masuda
Figure 14. Effects of temperature and humidity of
ambient gas upon the apparent resistivity of high-
resistivity dust.
100 200 300 400
Temperature C'C)
resistivity will be determined by bulk conduction. Current constrictions at the contact
points between the particles will also effect the resistivity (Masuda 1965). The resisti-
vity has a maximum at a temperature between 100 and 200°C which unfortunately is
the temperature range of exit gases in most industrial emission sources. Therefore,
back-discharge troubles often occur. It was recently discovered that alkali metal ions
served as charge carriers in both surface and volume conduction in layers of fly ash
(Bickelhaupt 1975).
9. Technical progress and new development
The large-spacing precipitator with an electrode spacing of 20-50 cm, described
previously, has proved to be successful in many of its applications. This includes cases
with high-resistivity dust where in some cases a cost reduction of approximately 20%
has been obtained. The roof-top type of precipitator is in increasing use. It is installed
on the roof of a plant building from which one has severe dust emission, for example
from an electric furnace. A large hot mass of gas rises to the ceiling where it enters the
precipitator and then passes through it by free convection and is emitted directly into
the open air. Often conductive plastics are used for the collection electrodes because
of their low weight.
The wet-type of precipitator is attracting increased attention because of its very
good performance. It is entirely free from dust r«-entrainment and back discharges
and also performs the additional function of removing gaseous pollutants such as SO2,
HF, etc. The problems with this type have been the necessity of irrigation water, the
treatment of the emitted slurry and the reheating of the cool gas at the outlet in order
to recover gaseous lift. These problems can be solved effectively in the hybrid-type of
precipitator in which the dry and wet types of precipitators were integrated inside a
common casing to make an optimized system (Ago etal 1973, 1975). The major por-
tion of the incoming dust was collected in the dry stage and the remaining very fine
dust was effectively removed in a fairly small wet stage. This meant that a large reduc-
tion in the amount of irrigation water, the slurry emission and the temperature drop,
could be achieved. The slurry, after concentration, could be dried by the use of the
137
-------
Recent progress in electrostatic precipitation
heat contained either in the inlet gas or in the collected dust in the dry stage. The merit
of this system has been found to be emphasised when very high degrees of emission
control for particulate and gaseous pollutants have to be achieved.
As regards the solution of the back-discharge problem, the conditioning of the inlet
gas by use of a water spray has long been used to reduce the value of Pd below about
Sx 10I0flcm. In this case the rapid and perfect evaporation of atomized water has
to be obtained (Masuda and Saito 1966). The so-called 'chemical conditioning* using
suitable additives (Dalmon and Tidy 1972b) has also proved to have been effective in
some applications. For instance, the injection of S03 into the inlet gas is widely used
for the fly ash of low-sulphur coal in order to prevent back discharging (Busby and
Darby 1963, Darby and Heinrich 1966, Cook 1975). The possibility of conditioning
fly ash by the addition of a sodium compound such as Na2C03 to low-sulphur coal, as
it is being burned, was also recently proposed (White 1975). Another solution to back
discharging is the use of the so-called 'hot-side* precipitator in which collection is made
at higher temperatures (300-400 QC) (see figure 14). The key factor in this system is
the consideration of the structural thermal-expansion properties (Walker 1975).
Purely electrical solutions have also been studied. Figure 15*shows one such
approach, where pulse charging is used in conjunction with a third electrode (Liithi
1967). The remarkable features of this method are that the current density can be
Collecting electrode
Third electrode
Figure 15. Pulse-charging lystem with the
third electrode (Lathi 1967).
adjusted independently of the main field strength by changing the magnitude or repeti-
tion frequency of the pulse voltage, and that a very uniform current density is obtained
over the complete surface of the collection electrode because of the expansion of the
dense ion cloud that is produced by the pulses. Thus the condition necessary for the
avoidance of back discharging, /d * Pd < £d»» can be met over the dust surface whilst
the main field strength is always kept at its maximum, A further study revealed that
it is dciirablc to put an additional DC bias voltage in series with the puke voltage (see
figure 5), This ensures the suppression of the DC corona during the pulse-free period,
regardless of fluctuations in the plant conditions (i.e. gas temperature and dust con-
centration), especially when the distance between the discharge electrode and the third
electrode his to be increased to meet design requirements (Masuda ft d 1974b), It
could be shown in the pilot plant tests that the precipitator shown in figure 5, equipped
138
-------
Senichi Masuda
with the bias-controlled pulse-charging system, exhibited an increase in collection effici-
ency from 63% to 93% for dust with the very high apparent resistivity of 10I3n cm
(Masuda etal 1974b). It was also observed that this method might provide an effective
solution for corona quenching because, with the aid of the third electrode, a sufficient
number of ions could be pulled from the discharge electrode, regardless of the dust
space charge existing in the main field. Instead of a pulsed field, an AC voltage could
be used in series with the bias voltage. Another very effective electrical approach to the
solution of the back-discharging problem is to use an AC voltage in conjunction with
an insulating film over the collecting electrode (Krug 1971). The practicability of this
method will depend on progress in the field of insulator materials.
The investigation of EFD motion of charged particles led to a new two-stage preci-
pitator of the type shown in figure 5, in which the charged particles coagulate in the
charging section and are led into the inside of downstream channels by the action of
the gas flow and the use of electric fields, where they are then electrically precipitated
(Shibuya and Masuda 1975). A remarkable reduction in precipitator size could be
achieved by this method under suitable conditions.
10. Conclusions
The recent progress in electrostatic precipitation has been reviewed. The major dif-
ficulty lies in the inherent problem of a multi-variable system, where the many process
variables which affect the overall performance have to be considered together. The gap
between scientific understanding of the elementary processes and the design procedure
has thus been inevitably large. In order to lessen this gap, more detailed studies are
needed of not only the physical phenomena but also the ability of the theories to
correlate the major process variables with the overall precipitator performance.
References
Adachi T, Suyama T, Shimoda M and Masuda S 1974 Proc. Gen. Conf. Inst. Elect. Engrs. Japan
No718
Ago S, Itoh T, Saito H, Furuya N and Masuda S 1973 Proc. Gen. Conf. Inst. Elect. Engrs. Japan
No838
Ago S, Itoh T, Saito H, Furuya N and Masuda S 1975 Proc. Gen. Conf. Inst. Elect. Engrs. Japan
No921
Awad M B and Castle G S P 1975 /. Air Pollution Control Assoc. 25 172
Bickelhaupt R E 1975 /. Air Pollution Control Assoc. 25 148
Busby H G T and Darby K 1963 /. Inst. Fuel 36 184
Cook R E 1975 /. Air Pollution Control Assoc. 25 156
Dalmon J and Tidy D 1972a Atmos. Env. 681
Dalmon J and Tidy D 1972b Atmos. Env. 6 721
Darby K and Heinrich D O 1966 Staub-Reinhalt. Luft 26 464
Dismukes E B 1975 /. Air Pollution Control Assoc. 25 152
Hall H J 1975 /. Air Pollution Control Assoc. 25 132
Heinrich D 0 1961 Trans. Inst. Chem. Engrs. 39 145
Heinsohn R J, Levine S H, Fjeld R J and Malamud G W 1975 /. Air Pollution Control Assoc. 25 179
Hermstein W 1960 Archivfur Elektrotechnik 45 209, 279
Hewitt G W 1957 AIEE Trans, part I 76 300
139
-------
Recent progress in electrostatic precipitation
Kuroda M 1975 Graduation Thesis Department of Electrical Engineering, University of Tokyo
Krug H 1971 Dissertation TU Karlsruhe
Liu B Y H, Whitby K Tand Yu H H S 19677. Appl. Phys. 38(4) 1952
Liu B Y Hand Yeh H C 1968/. Appl. Phys. 39(3) 1396
Lowe H J and Lucas D H 1953 Brit. J. Appl Phys. (suppl) 2 40
Lttthi J E1967 Dissertation ETH-Zurich No3924
Masuda SI965SteuA 25 175
1966 Staub-Reinhalt. Luft 26 459
Masuda Sand Adachi Y 1975 Trans. Jnst. Elect. Engrs. Japan to be published
Masuda Sand Akutsu K l9TSProc. Gen. Conf. Inst. Elect. Engrs. Japan No909
Masuda S, Akutsu K and Mizuno A 1974a Proc. Gen. Conf. Inst. Elect. Engrs. Japan No723
Masuda S, Aoyama M, Shimozono S, Hattori J and Shibuya A 1974b Proc. 19th Ann. Meeting
Static Electrification Group Japan (Inst. Polym. Sci. Japan) p35
Masuda S and Matsumoto Y 1974a Proc. Gen. Conf. Inst. Elect. Engrs. Japan No722
1974b Trans. Inst. Elect. Engrs. Japan 94-A 515
Masuda S, Matsumoto Y and Ohba Y 1973a Proc. Gen. Conf. Inst. Elect. Engrs. Japan No839
Masuda S and Mizuno A 1975a Proc. Gen. Conf. Inst. Elect. Engrs. Japan No924
1975bPlrac. Gen. Conf. Inst. Elect. Engrs. Japan No923
Masuda S and Niioka M 1973 Proc. Gen. Conf. Inst. Elect. Engrs. Japan No847
1974 Proc. Gen. Conf. Inst. Elect. Engrs. Japan No719
Masuda S and Saito H 1966 land E C Process Design and Development S 135
Masuda S, Shibuya A and Ikeno E 1973b Proc. Gen. Conf. Inst. Elect. Engrs. Japan No844
Matsumoto Y 1974 Dissertation Department of Electrical Engineering, University of Tokyo
McCain J D, Gooch J Pand Smith W B 1975 /. Air Pollution Control Assoc. 25'117
Mizuno A1975 Dissertation Department of Electrical Engineering, University of Tokyo
Murphy AT, Adler F Tand Penny G W 1959 AlEE Trans, part 1 78 318
Niioka M 1974 Dissertation Department of Electrical Engineering, University of Tokyo
Pauthenier M 1961 Id Physique des Forces dlectrostatiques et leurs Application (Centre National
de la Recherche Scientifique) p279
Pauthenier M and Moreau-Hanot M 1932 /. Phys. Radium 3 590
Penny G W 1975 /. Air Pollution Control Assoc. 25 113
Penny G W and Lynch R D 1957 AlEE Trans, part 1 76 294
Shibuya A and Masuda S 1975 Proc. Gen. Conf. Inst. Elect. Engrs. Japan No920
Simm W 1962 Staub 22 463
Smith P L and Penny G W 1961 AlEE Trans, part 1 80 340
Smith W B and McDonald J R 1975 /. Air Pollution Control Assoc. 25 168
Steinbigler H 1969 Dissertation TH Munchen
Tassicker O J 1975/. Air Pollution Control Assoc. 25 122
White H J 1951 AlEE Trans, part 2 70 1186
1962 Industrial Electrostatic Precipitation (New York: Addison-Wesley) p!57
1975 J. Air Pollution Control Assoc. 25 102
Walker A B 1975 /. Air Pollution Control Assoc. 25 143
Discussion
MrWEG Plumtree (Rank Xerox)
Particularly with respect to the results in figure 7 (back-discharge plot), how much
dust was present in the system?
Professor Masuda
Instead of a dust layer we used a mica plate with a 0-5 mm diameter pin hole. The
thickness of the plate was 1 mm.
140
-------
Senichi Masuda
MrRHours(CEA)
About two years ago American authors proposed the use of 7-rays to charge (and
thereby remove) very fine particles from the flue gases of coal-fired power stations.
Scientifically, this idea is interesting, but due to the very low yield of 7-ray ionization,
powerful - and therefore dangerous - radioactive sources are necessary. Consequently
the system appears uneconomical compared with those based on corona charging.
What is your opinion about that?
Professor Masuda
I don't think that the use of 7-rays in particle charging will find a wide practical
application in the field of electrostatic precipitation, because of the problems you
pointed out. I comment, however, that a high energy electron beam (0-75-1-5 MeV,
100mA) might be effectively used for the removal of NO* and SO* from the exit gases
out of large-scale industrial furnaces (ore-sintering furnaces, thermal power plants,
etc). It was discovered that these gaseous pollutants are effectively converted by the
electron beam irradiation into aerosol particles within about 1 second. These particles
may be collected by an electrostatic precipitator. Large-scale development work is now
going on in Japan by a research group from the steel industries. It is expected that a
very high initial investment in this method may be balanced by the benefits of a very
small pressure drop and the ease of operation where no catalyst is used.
DrJC Gibbings (Liverpool University)
The design of electrostatic precipitation is a good example, of electrostatics being
very much an inter-disciplinary study in that the study of the electric field must go
hand-in-hand with the fluid mechanics study of the flow pattern for real progress to be
made. Professor Masuda pointed out that his field calculations for K = Q gave flux lines
identical to the streamlines; but these streamlines are for the potential flow and no real
flow in ducts would correspond to these streamlines. In the past, the design of precipi-
tators has been spoilt by lack of concern with analysis and knowledge of the flow.
For example, the electric wind from an electrode will convey a particle towards a
surface, but unless centrifugal and electrostatic forces remove it from the flow on to the
surface, then the return flow, which continuity insists must exist, will equally convey
it away again: in principle this is why the presence of turbulence in the flow is of such
significance in effectively greatly increasing the diffusion coefficient.
141
-------
FLASHOVER MEASUREMENTS OF BACK DISCHARGE
Senichi MASUDA, PhD Department of Electrical Engeneering,
University of Tokyo
Aklra MIZUNO, ME
7-3-1, Kongo, Bunkyo-Ku, Tokyo,
Japan
142
-------
Summary
The mode change and flashover voltage of back discharge under different
gaseous conditions were studied with a special attention to the effects of
dust layer thickness and alkaline components contained In dust. It was
found that back discharge took either streamer or steady-glow mode
depending upon gas mean free path. These modes have their own characteri-
stic flashover voltage as a function of gas mean free path, where Its value
for the former mode 1s much lower than that for the latter. Thickness of
the dust layer and existence of the alkaline components also govern
the Initiation of streamer so that the flashover voltage 1s largely
affected by these factors.
1. Introduction
Back discharge has long been one of the unsolved problems 1n electro-
static precipitation. This Is an abnormal discharge caused by breakdown
of high resistivity dust layer deposited on collecting electrode.
The mode and effect of back discharge differ largely, depending upon the
polarity of corona discharge used. In this paper, however, we restrict
ourselves to the case of negative polarity which has been 1n common use
because of Its higher flashover voltage under normal operating conditions.
If back discharge takes place, flashover voltage falls to about half
the value of that under normal operation, and particle charge will be
neutralized. When the field strength Inside the dust layer exceeds Its
breakdown value, the Initiation of back discharge takes place. A random
or repetitive breakdown appears at the breakdown point, owing to the
continuous 1on supply from corona discharge. With a slight Increase 1n
143
-------
voltage, it turns to a feeble but continuous spot-like onset-glow, refered
to as "onset-glow mode". The current wave form contains that of Trichel
pulse superimposed on a D.C.component. With the further increase in
voltage, this onset-glow either triggers the streamer discharge in gas
space or along the layer surface, or it turns to a pulseless point-like
glow with increased intensity. The former mode should be refered to as
"streamer mode", while the latter as "steady-glow mode". Thus, the mode
of back discharge after onset stage can be classified into streamer and
steady-glow modes, as described separately. ' When gas mean free path
is larcie, the transition occurs from the streamer mode tb the steady-glow
mode with an increase in voltage beyond a certain critical voltage.
This mode transition is reversible and affected not only by the gas mean
free path but also by the thickness of dust layer and its chemical
composition. When the thickness is small, streamer mode does not occur
and the flashover voltage is high. Among the effects of chemical
composition, the most remarkable is that of alkaline metal compounds which
lowers the flashover voltage of back discharge to a great extent.
In this case streamer propagation becomes very pronounced so that it
easily turns to a flashover.
In this paper, the effects of mean free path, dust layer thickness
and alkaline content in dusts on back discharge mode and flashover voltage
are reported.
144
-------
2. Experimental apparatus
The effects of mean free path on the flashover voltage and mode of
back discharge were studied by changing the pressure P under room
temperature or by changing the temperature T under atmospheric pressure.
A needle to plane electrode system with a gap of 50 mm was used inside
a thermostat or vacuum chamber as shown in Fig.l. A mica plate having
a pinhole of 0.5 mm in diameter, tissue paper and dust samples of various
chemical compositions were used as test layer samples to be located on
the plane electrode. As a most important parameter, the resistivity of
the test layer was measured before each experiment. The change of
the resistivity under vacuum condition was enabled by drying the layer by
heat during evacuation to about 1 torr, and thereafter adding dry air, so
that a desired air pressure could be obtained. By this method, the value
of resistivity could be maintained constant at least during each experiment.
However, Its value was delicately dependent upon the drying condition.
Hence, its measurement at a position separate from the centre area using
a fixed counter electrode on the layer surface should be excluded. Thus,
before each measurement, a counter electrode was set on a centre position
facing to a measuring electrode and removed therefrom after the experiment,
with the aid of a remote-controlled crane model. In the case of resistivity
measurement under elevated temperature inside the thermostat, a separate
measuring cell was used, because the resistivity value in this case was
a unique function of the thermostat temperature, so far as the equili-
brium condition was reached. An image intensifier tube (EMI, type
9912) was used to observe a faint glow of back discharge at its initial
stage and to investigate the difference in discharge modes in detail.
145
-------
3. Effect of mean free path
A mica plate with a plnhole of 0.5 mm 1n diameter and 0.45 mm
thickness was used as the test layer. The flashover voltage Vs was
plotted against the normalized mean free path A/fto, as shown 1n Fig.2,
where/^o Is the mean free path at NTP. The solid curves represent the
results obtained by changing P under room temperature (20 °C) while the
dotted curves Indicate those obtained by changing T under atmospheric
pressure (760 torr). The scales for these P and T are also given. It
can be seen that the curves measured by changing T or P agree well with
each other. It should be noted that there exist two different curves
for flashover (curve I and II), each covering the different range of
X/Xo. In the area under the curve I, the onset-glow mode was followed
by the streamer mode with the Increase in voltage, while the steady-glow
mode occurred in the area under the curve II. As a result there are
three different regions In X//lo, A, B and C as Indicated in the figure,
each corresponding to different mode changes. In region A, back discharge
in the streamer mode followed the onset-glow mode and turned to flashover
on curve I. In region C, back discharge in the steady-glow mode followed
the onset-glow mode and no streamers appeared until flashover took place
on curve II. Region B 1s a transition region between A and C. Fig.3
shows the photographs Indicating the mode transition in this region.
The onset-glow mode appeared at first at the plnhole (F1g.3-a). With the
Increase In applied voltage, it turned into streamer mode (Fig.3-b),
bridged across the gap and finally turned to random sparking on the
curve I (F1g.3-c). In region B, the random sparking tended more easily
to occur at smaller value of A/Ao. Slightly above the curve I, streamers
suddenly disappeared to turn to a stable steady-glow (Fig.3-d). This lasted
146
-------
until flashover occurred on curve II. The voltage-current characteristics
in this region were further studied with the use of X-Y recorder, where
the electrode voltage V and the current from the measuring electrode^I,
were recorded. The result obtained with no additional output impedance
is shown in Fig.4-a where the output impedance was only that of the
source (Ro = 15 M ohm) so that it was comparatively small. From the
onset-glow at the pinhole, the space streamer appeared at point A. With
the increase in voltage, it bridged across the gap, resulting in a
transition from point B to C. When the source voltage was further
increased, the space streamer became more luminous and current I increased,
the electrode voltage, however, remaining almost constant (curve ii).
Fluctuations of the current and voltage were large. Around the point D,
random sparking took place. With the source voltage slightly increased
from point D, the transition of the streamer to the steady-glow mode
occurred, accompanied by the transition of the curve from point D* to B*
where D' and B* were very close to D and B respectively. Thereafter, the
voltage and current followed the curve (i') until flashover occurred at
a point beyond E. Then, when the voltaae was lowered, the voltage and
current followed the identical curve E-B' until point F was reached.
The inverse transition from the steady-glow to streamer mode occurred at
point F, resulting in the transition of the curve from F to C. With
the further decrease in voltage, the streamer mode lasted following
curve (1T) until point 6 was reached where 1t turned to the onset-glow
mode again. Thereafter the voltage and current followed the initial
curve (i). Thus, it can be seen that voltage-current characteristics
consists of two curves (1)-(r) and (il)-(ii'), the former corresponding
147
-------
to the glow mode, and the latter to the streamer mode. The curve (1)-(1*)
1s subdivided Into part (1) and (1*), the former Includes the onset-glow
mode while the latter corresponds to the steady-glow mode. It was
observed that these transitions of mode and V-I characteristics were
governed by the output Impedance of the high voltage source* as Indicated
1n F1g.4-b and c. In the case of F1g.4-b, the additional resistance
of R = 27 M ohm was Inserted In series to the output circuit, while
1n the case of F1g.4-c, that of R » 55 M ohm was used. In Fig.4-b, B*
was fairly apart from B, and C* was somewhat apart from C. The transition
from F to C* was unstable, and the Inverse transition easily took place,
even when the source voltage was kept constant. But1! when the voltage
was further decreased, the streamer mode became stable and the voltage
and current followed the curve (11*). When the output Impedance was
excessively high, as In the case of Fig.4-c (R = 55 M ohm), neither
random sparking nor transition took place at point n" where the source
voltage was its upper limit, 50 kV. With the decrease in voltage, the
curve followed (ii) and (ii'). The current wave form in region B in
F1g.2 is shown 1n Fig.5. As the onset-glow started, Trichel pulses
superimposed to a small D.C. current were observed (Fig.5-a), and with
the increase in1 voltage,random pulses having much larger pulse height
appeared, which corresponds to space streamers (Fig.5-b). Finally pulseless
D.C. current appeared with the transition from streamer to steady-glow
mode (F1g.5-c). No Trichel pulse could be observed in this mode. This
suggests the mechanism of electron emission from discharge electrode to
have changed from that for the onset-glow mode.
148
-------
4. Effect of dust layer thickness
The effect of thickness of the layer was studied using three mica
plates with different thickness, each having a pinhole with 0.5 mm 1n
diameter, and three tissue papers with different lamination number. The
resistivity of the mica plates were higher than 10 ohm-cm, while
12
that of the tissue papers were about 10 ohm-cm. Fig.6 and 7 show the
values of flahsover voltage obtained for the mica plates and the tissue
papers respectively, plotted against pressure P with the A/Ap scale
Identical to that in Fig.2. It can be seen in Fig.6 that, with the
decrease in the mica plate thickness t, the transition pressure P. from
the region B to C shifted towards1 the side of higher pressure range (lower
mean free path). Pbc finally exceeded 760 torr before t = 47/
-------
the curve 3 in Fig.6. The curves 2 and 3 in Fig.7 also roughly agreed with
the curve II in Fig.2. It was concluded that the decrease in the layer
thickness shifted the transition pressure between res ions B and C, Pfac,
towards the side of higher pressure range, resulting in the dominance of
the steady-glow mode and the increase in flashover voltage.
5. Effect of chemical composition of dust
In an electrostatic precipitator, flashover must usually be
dominated by streamer mechanism since the values of both pressure and
gap distance are large. It is considered, therefore, that the existence
of alkaline components may help streamer development because of their
low ionisation energy so that the flashover voltage under back discharge
condition may become lower. For example, the exhaust gas from an iron
ore sintering furnace contains fairly high content of alkaline metal
compounds, especially that of potassium, and it has been found that the
collection performance drops as its content increases.
The flashover voltage Vs of a needle to plane electrode system
(gap 50 mm) was measured, with various kinds of dust layers (thickness
2.0 mm) on the plane electrode. The samples used were the first class
agents and shown in Table 1. The experiments were conducted in a
thermostat or vacuum chamber, and the change of dust resistivity was
enabled using the methods described in section 2.
The flashover voltage Vs and the apparent resistivityAl for each dust
layer plotted against temperature T arelshown in Fig.8 and 9 respectively.
In this case measurement was made under atmospheric pressure. The scale
of the temperature in Fig.8 is the same as that in Fig.2. The abrupt
150
-------
rise In Vs in curves 2, 3 and 5 In Fig.8 were due to the disappearance of
back discharge resulted by the decrease in dust resistivity with the
Increased temperature (Fig.9). Curves 1 and 2 in Fig.8 roughly agree
with the curve II in Fig.2, while curve 3, 4 and 5 with the curve I.
For CaC03 and Fe203 dusts (curve 1 and 2), back discharge after the
onset-glow stage was at first of streamer mode, but it changed into the s
steady-glow mode when craters had been formed at the breakdown point and
Vs became high. In these dusts particles were easily ejected from the
area around the back discharge point so that conical craters tended to
be formed. Hence, it was probable that the effective thickness of
the layer at the crater bottom became so small that no streamer could occur.
On the other hand, for KgSO^, KC1 and Nad dusts (curve 3, 4 and 5), all
containing alkaline metals, the streamer mode always dominated so that
Vs was lower.
The flashover voltage under room temperature plotted against pressure
P is shown in Fig.10 where the scale of P is also the same as that in
Fig.2. The apparent resistivity/d could be kept constant inspite of
a large change in pressure by using dry air. It was confirmed that the
Vs curves of various dusts could be classified into curves I, II and III.
Curves I and II agreed with the curves I and II in Fig.2. Curve III was
a transition curve between the curves I and II, like the curve 1 in the
region B in Fig.7. In the dusts corresponding to curve III, pinholes
were easily formed at the breakdown points.
The classification of dusts by their Vs curves is shown in Table 1.
These results show that alkaline dusts are included in Group I (curve I),
non-alkaline dust tending easily to form conical craters in Group II (curve II)
151
-------
and dusts In which pinholes easily appear are classified in Group III (curve
III). With the existence of alkaline compound, only curve I appeared and
no transition to the steady-glow mode occurred, the streamer mode lasting
until flashover took place, at least within the range of T, P and the layer
thickness investigated. The lack of curve II for alkaline! compounds
might be a result of the fact that the transition pressure Pab between
regions A and B was shifted towards the low pressure side beyond the range
investigated. On the contrary, the lack of curve I for non-alkaline
dust investigated might be caused by the shift of the transition pressure
P. between regions B and C towards the high pressure side beyond the
range studied. The characteristics of curve III might be explained by the
propagation model of mode transition assumed for curve 1 in Fig.7.
The effect of alkaline compounds on voltage-current characteristics
under back discharge with a single breakdown point was studied. In this
experiment the thickness of the layer must be kept constant so that its
effect could be excluded. For that purpose a mica plate having a pinhole
(0.5 mm in diameter) was again used, and its surface was painted by water
solution of I^SO^ (10 % in weight) and thereafter dried. Fig.11-a and
b show the difference in voltage-current curves with and without KgSO^
film under the pressure P = 360 torr, where no additional output resist-
ance was used. An X-Y recorder was used also for these measurements.
Fig.ll-a shows the identical characteristics as those in Fig.4-a. In
Fig.ll-b, when K2S04 film existed on the surface, a streamer appeared at
a lower voltage. With the voltage increased, it bridged across the
electrodes and flashover occurred. In this case the transition to the
steady-glow mode did not take place, except for a transient one appearing
rarely at the instant of flashover.
152
-------
6. Discussions
6.1 Summary of back discharge phenomena
The phenomenological behaviors of back discharge so far described
can be summarized as follows. Back discharge starts with the breakdown
of a weak point or pinhole in the layer, when the conditioned x J SEds
ts fulfilled.*'' This breakdown, occurring randomly at first, becomes
nore or less periodical when the voltage is raised. With the voltage
further increased, it turns to a feeble spot-like onset-glow. When
12
the layer resistivity is comparatively low, less than about 10 ohm-cm,
where a sufficient negative ion current is being supplied from the
discharge electrode at the instant of breakdown, the layer breakdown directly
triggers the onset-glow. With the further increase in voltage, the
onset-glow turns either to the streamers (surface and/or space streamers)
or to the more luminous steady-glow. In some cases, a small sized
onset-streamer is observed to appear around the upper edge of the
pinhole prior to the occurence of the well developed streamers. The
difference between the onset-glow and steady-glow lies in the magnitude
and wave form of current, the former current being lower (less than
about 50 jiA/point) and containing both D.C. and Trichel pulse component,
while the latter being higher (more than about 50 ^A/point) and
completely non-pulsive. This difference suggests the change in corona
mechanism occurring at the discharge electrode. The D.C. current
component in the onset-glow increases with the increase in voltage.
In the region A where gas mean free path Xis small, flashover occurs
directly from the space streamer, while in region C where Xis large,
steady-glow turns to flashover without occurence of streamer.
153
-------
In the Intermediate region B, the onset-glow Is followed by the streamers,
which bridge the electrode gap and cause random sparking. The random
sparking tends more easily tor occur atia smaller VJto side 1n region B.
The streamers, however, disappear at a certain critical voltage to turn
to the steady-glow. Flashover occurs from the steady-glow In this region.
The decrease 1n the layer thickness shifts the boundary value of Pb between
the regions B and C towards the smaller side ofJl. The existence of
compounds having low 1on1sat1on potential shifts the boundary value of Pab
between the regions A and B to the larger side of \. In case of the layer
having many,weak points or plnholes, the mode transition between the
streamer and steady-glow modes takes place from one point to another
so that flashover curves in the region B coalesce into a single curve.
6.2 Mechanism of back discharge
There are three discharge districts in back discharge to be
considered separately (Fig.12): (1) the breakdown point in the layer,
(2) the layer surface and gas space, and (3) the corona point at discharge
electrode. The discharge mechanism in these three districts may be
different, but closely connected to each other to characterize the over-all
behavior of back discharge.
After a pre-onset stage, a spot-like glow discharge at the breakdown
point remains to exist either in the form of the onset-glow or in the form
ef the steady-glow, so far as the streamers do not occur. When the streamers
occur, the spots glow repetitively only at the instant of streamer occurence.
Judging from the magnitude of current density, the onset-glow and the
steady-glow should be taken as a kind: of glow discharge having a structure
154
-------
as shown in Fig.12. A most remarkable feature of these glow discharges
at the breakdown point is that they lack a cathode electrode as a source
of electrons which maintain the discharges themselves. There must exist
some origin of electrons at the location S near the top of the glow spot.
The only possible source of electrons may be the negative ions, supplied
from the discharge electrode, from which electrons will be sheded. The
electron shedding, however, requires a value of field strength to pressure
/2\
ratio higher than about 20 V/cm-torr. This field may only be formed by
space charge of highly concentrated positive ions accumulated at the location
P under the shedding zone S. The electrons, shedded from negative ions,
will be strongly accerelated by the positive ion space charge field to
ionize gas molecules at the area G under the region P. This ionizing
region G may correspond to the negative glow in a usual glow discharge
and provide a sufficient quantity of positive ions to the region P. This
I (3)
positive ion space charge region P may correspond to the cathode darkspace.
In region A where gas mean free path A. is sufficiently small, the
nunber of collision for unit length becomes large, while the diffusion
of produced plasma will be largely suppressed. Hence, the positive ion
m
density could become so high that the condition for streamer initiation* '
I (oC-1) cU =* k , k=/0*-20 CIO
•'o *
may be fulfilled. Here c£is the first Townsend coefficient and fy the
attachment coefficient of electrons to neutral molecules. The integration
should be performed from the origin 0, through the breakdown channel!
and along the optimum field line in gas space to the position L where oC=^.
The streamer propagate either into gas space towards the discharge
electrode or along the layer surface, or the both directions. With the
Increase 1n voltage the space streamer finally reaches at the discharge
155
-------
electrode to cause a mighty flashover capable of turning into arc in
this case.
In region B where the gas mean free path \,becomes larger, the
plasma density in streamer channel cannot become'sufficiently high so
that it cannot trigger the mighty flashover when it bridges over the
electrode gap, or even when it triggers a random sparking, so far as the
output impedance of the source is not sufficiently low. In this case
the transition from the streamer mode to the steady-glow mode occurs
at the bridge-over stage of the streamer, but not at the instant of
sparking. At this stage copious positive and negative ions are produced
in the gas space. As a result a strong positive ion space charge
accumulates in front of the discharge electrode resulting in an enhanced
electron emission from its surface because of gamma action (electron
emission by positive ion bombardment). The electron space charge can
effectively compensated by the strong positive ion space charge so that
no periodical choking of electron avalanche occurs. Hence, the Trichel
pulse disappears and a sufficient quantity of negative ion current can
now be supplied from the pulseless negative glow corona. In the meantime,
a strong negative ion sheath is':forme'd near the region S to enhance
the positive ion supply from the layer breakdown point. The highly inc-
reased densities of positive and negative ions in gas space may reduce
the field in the gap so that streamer cannot be maintained. This might
be the situation causing the transition from the streamer to steady-glow
mode. The mighty flashover takes place from the steady-glow only when
the voltage is sufficiently raised.
156
-------
In region C where the gas mean free path J\.is very large, the streamer
initiation condition (1) cannot be fulfilled owing to the decrease
in collision number and the increase in diffusion of plasma. With the
increase in voltage, the positive ion accumulation in front of the
discharge electrode also increases so that D.C. current component due
to gamma action of the positive ion collision also increases. Finally,
when the choking effect of negative ions from the discharge electrode is
offset by the positive ion space charge, the Tirchel pulse may completely
disappear so that the onset-glow mode turns to the steady-glow mode.
The sufficient increase in voltage results in the mighty flashover to occur
directly from the steady-glow. Hence, it is expected that the boundaries
between the regions A, B and C may also be governed by the output
impedance of the source.
The existence of two flashover voltage represented by curves I and
II clearly suggests that, once back discharge occurred, the flashover
becomes solely governed by the mode of the preceding discharge and is
not affected by dust resistivity,/'d, although/to has an essential
effect on the initiation of back discharge itself.
6.3 Effect of layer thickness and ionisation potential of dust
The essential part in the Integration in equation (1) seems to
exist inside the pinhole itself so that the product (pressure P) x (layer
thickness t) should play a major role in the streamer initiation. Thus
it can be understood that, with the decrease in layer thickness t,
the boundary pressure P 'between the regions B and C shifted towards the
DC i
side of higher pressure. In Fig.5, the values of P x t at the boundary
157
-------
pressure between B and C are 35.7 torr-mm for t = HB^/jri and 22.8 torr-mm
for t = 60y(un respectively. Whereas, in case of t = 47/tm, no streamers
appeared at P = 760 torr. This may suggest that streamer will not occur
if the value of P x t is smaller than about 20 torr-mm/
The effective value of oC in equation (1) may become larger if the
layer contains components with low ionisation potential such as alkaline
metal compounds. Such compounds may emit their molecules inside the
breakdown point or even into gas space. The lack of region B and C for
these compounds suggests that equation (1) for streamer initiation could
be fulfilled even at a lower value of P so that the boundary pressure P .
between A and B may be shifted to the side of lower P beyond its range
investigated. Therefore, the streamer initiation condition becomes
easily to be fulfilled inside the breakdown point in this case. There
may exist these components also in gas space, ejected from the breakdown
point. Then, the streamer propagation may be enhanced also in gas space.
7. Conclusions
From the results of flashover measurements under back discharge
condition, following conclusions were obtained.
(1) There are three different regions in the mode change of back.discharge.
These regions are mainly determined by the mean free path of gas. Properties
of back, discharge were almost the same under various values of pressure or
temperature investigated so far as the value of mean free path was the same.
As described in section 3, back discharge takes streamer mode in region A,
whereas in region C, steady-glow mode occurs. In region B, transition
from streamer mode to steady-glow mode occurs.
158
-------
(2) Flashover voltage under back discharge condition 1s also determined
by the gas mean free path. There are two curves of flashover voltage.
The lower one corresponds to the streamer mode (curve I), while the higher
one to the steady-glow mode (curve II).
(3) If the layer thickness 1s very small, streamer mode cannot occur
so that flashover voltage becomes high.
(4) If alkaline dusts are present, streamer causes flashover before
the transition to the steady-glow mode occurs so that flashover voltage 1s
largely reduced.
Acknowledgements
This research was sponsored by the Ministry of Education, Japan, as
its Special Research Project (I) (Project No.011914). Prof. Y. Miyoshi
is gratefully acknowledged for his helpful discussions about the mechanisms
of back discharge mode. Thanks are also due to Mr. R. Shimoda for his
help given to a part of the experiments.
159
-------
Nomenclature
Ji gas man free path
Ao gas mean free path at NTP
T gas temperature
P gas pressure
Vs flashover voltaqe
of. first Townsend coefficient
t electron attachment coeffedent
apparent resistivity of dust layer
layer thickness
160
-------
References
1. S. Masuda and A. Mizuno, Initiation Condition and Mode of Back
Discharge, J. of Electrostatics (to be published)
2. Q. Vuhuu and R.P Comsa, Influence of Gap Length on Wire-Plane Corona,
IEEE Trans. PAS 88 p. 1462 (1969)
3. J.D. Cobine, Gaseous Conductors, p 213, Dover Edition, 1958
4. T. Takuma, Discharge Characteristics of Gases, Part 1,
Central Resereh Institute of Electric Power Industry Technical Report
No. 69015, (1969)
-------
I
KMn04
V°4
Sr(N03)2
(NH4)2S04
NaN03
Nad
II
CaC03
Fe2°3
A1203
Cr203
MgO
III
mica plate
sulphur
sio2
Table 1
162
-------
-D.C.H.V.
I.I.
T: TEST LAYER
P: PLANE ELECTRODE
M: MEASURING ELECTRODE
C: REMOTE CONTROLLED CRANE MODEL
S: COUNTER ELECTRODE
I.I. : IMAGE INTENSIFIER
Figure 1
163
-------
30
20
10
to
0
350
2 3.1
150 550 T ( K )l
•H
760 p560 460 360 260
ab P(torr)
150
Figure 2
164
-------
^ •'
(a) ONSET-GLOW
V = 12 kV
(b) SPACE STREAMER
V = 15 kV
(below curve I)
I^^BHVvQfi^^Bfii^
P^^ffiffifl I..s.-J^V^&'j&f&
(c) RANDOM SPARKING
V = 18 kV
(on curve I)
(ci) STEADY-GLOW
V = 22 kV
(between curve I and II)
Figure 3
-------
300
- 200
- TOO
\
A F
0 5 10
V ( kV )
(a) R = 0
15 20 25
Figure 4
-------
J-*i-i i ttantaniaatninjn
••• 11IIH1Hifll TfHinrlfllI i I I1J
(a) Onset-Glow Mode
( V = 12|kV, I = 2.5,>X
20 pA/div., 20 /^s/div.)
(b) Streamer Mode
( V • 14 kV,
2 mA/d1v., 20 Ms/d1v.)
(c) Steady-Glow Mode
( V • 18 kV, I • 85
20 MA/d1v., 20
Figure 5
167
-------
30
20
> 10
0
0-1: t = 115 (urn)
A--2: t = 60 ( » )
+~3: t = 47 ( " )
be
760 560 160 360
P (torr)
Figure 6
260
160
168
-------
30
20
> 10
A«l: 44 LAYER (2.6 mm )
0-2: 10 ,. ( 0.6 mm )
x-3: 1 .. ( 0.06 mm )
0
760
560 460 360
P(torr)
Figure 7
260
160
169
-------
30
20
ID
*--!: CaC03
0-2: Fe203
A--3: K2S04
+--4: KC1
D--5: NaCI
100 200 300
Figure B
170
-------
14
10
10
£
o
i
H
olO
10
0
100
200
300
T rc
Figure 9
171
-------
30
20
t 10
(A
D--I : K2S(
0--II : Fe2C
+ --III : S
760 560 460 360
P (torr)
Figure 10
260
160
172
-------
600
OJ
400
200
SPACE STREAMER
10
V (kV)
FLASHOVER
20
(a) Hlca Plate with No Coating
Figure 11
600-
400-
200-
FLASHOVER
V(kV)
(b) «fca Plate with K2S04 Coating
-------
* + -
_ 1 + FIELD
* - » LOW
L -
(a) ONSET-GLOW
(Trichel Pulse)
(b) STEADY-GLOW
S: electron shedding zone
P: accumulate zone of positive ion
G: negative glow
D: dark space
C; positive column
N: needle electrode
E: plane electrode
Figure 12
174
-------
.Figure Caption
Table 1 CLASSIFICATION OF DUST
Fig.l ELECTRODE SYSTEM FOR MEASURING FLASHOVER VOLTAGE OF
BACK DISCHARGE
Fig. 2 FLASHOVER VOLTAGE v.s. TEMPERATURE AND PRESSURE
Fig.3 TRANSITION OF BACK DISCHARGE MODE IN REGION B
( T - 450 K, mica plate with a pinhole)
Fig.4 VOLTAGE-CURRENT CHARACTERISTICS WITH DIFFERENT OUTPUT IMPEDANCE
Fig.5 CURRENT WAVE FORM OF BACK DISCHARGE IN REGION B
( P • 360 torr, mica plate with a pinhole)
Fig.6 EFFECT OF LAYER THICKNESS ON P- Vs CHARACTERISTICS
(mica plate with a pinhole)
Fig.7 EFFECT OF LAYER THICKNESS ON P-Vs CHARACTERISTICS
(tissue paper)
Fig.8 EFFECT OF TEMPERATURE ON FLASHOVER VOLTAGE
(atmospheric pressure)
F1g.9 APPARENT DUST RESISTIVITY v.s. TEMPERATURE
Flg.10 EFFECT OF ALKALINE COMPOUND ON FLASHOVER VOLTAGE
(T - 20 °C)
Pig.11 VOLTAGE-CURRENT CURVES WITH AND WITHOUT K2S04 FILM ON THE SURFACE
OF MECA PLATE WITH A PINHOLE
(thickness: 0.2 mm, P • 360 torr, T - 20 °C)
Fig.12 SCHEMATIC REPRESENTATION OF BACK DISCHARGE IN THE GLOW MODES
175
-------
Proc. 4th Int. Clean Air Congress
Paper No. V-47 (May 1977, Tokyo)
BASIC STUDIES ON BACK DISCHARGE MODE
AND STREAMER PROPAGATION
Etudes fondamentales sur le mode de deversement inverse
et la propagation en courant
MASUDA. S. and MIZUNO. A.
Faculty c/Enginttrtng. Tokyo Uniotnity
Tokyo. Japan
INTRODUCTION
Back discharge has long been one of the unsolved prob-
lems in electrostatic precipitation This is an abnormal
discharge cause J b> breakdown of high resistivity dust la>cr
deposited! '- I he mode and effr
back discharge differ largely, depending upon the polanty
of corona discharged used. In this paper, however, we
rcsttK' '!c case of negative polarity which has
been in common me because of the higher fiashover voltage
under normal Derating cc>:- ' ! lakes
place, fiashover s to about half the value of that
under normal .ipentitm. and particle charge will be
neutralized When the field strength inside the dust
exceeds its breakdown value, the initiation of back dis-
charge '. i •
turns
fered :
contains that < ' ;!se superimposed 0:1 a I) (
ponent With the further in^rea'.e in voltage, thu onset
cither triggers the streanu- .? m gas sp
along the layer surface, or it turns to a pulseless ;
like glow with increased intcnsiv. ' :ier mode -'
be refered to a$ "streamer h,c).wlr -
latter mode as "steady-glow modi- ;). In the
streamer mode, lepetitive h^ht and current pulses a;,
while in the steady-glow mode, puhive a
plctoly disappears nie s;i nirs under small
pas mean free j-
•
•;ean free \
'
•i the streamer mode
a certain .
this region, ami the mode transition is rcvcr-iMe Trie
• .
I HACK DISC II \K(,1
MODI
HII SIKI \\1IK
f.mn of sltrjuv
into space streamer ni»Je, surface streamer :
and mived streamer mode as a combination of the former
two as shown in I ig 1 -a. • The most
essential f? !mg the respectivf rn
is the strengths of tangen: • .inJ verii al h-
around the breakdown point as well as corona current
densi: J J, however, are closely coupled to each
other Thus, the effect of Ea and J must be studied
separately by using a grid electrode inserted between a
needle to plane electrode system. By the change of needle
electrode vol:age and grid electrode voltage, J and Ea can
be varied independently. The electrode gap is SOmm and
the distance between grid and plane electrode 20mm. A
pair of glass phtes each having a pinhole are used on top
of one another as the layer located on the plane electrode.
•' the glass plate, pd. is 6 x 10"ficm and
the diameter of the pinht le is 0 5mm, the thickness uf one
.' Omm The breakdown field strength Eds of th_
plate pair can be changed by changing the distance of the
t w o h.
From :! • -charge mode with the
•
(
-------
THK FOURTH INTERNATIONAL CLEAN AIR CONGRtSS
-5
:
-fi
10
E
D
;•
-8
\x
NO flASXOvfP
. iv
SUPFACf i&
cTonufB vf/
ST9EAME?
«
O
A X
III (X$ET-^TOE»."fP Of
1 1 ONSf '
8
0246
Ea ( kV/cm )
Fif. 2 Mode du£jjm of back discharge in field-cuiiem
lid of on imige mtensifier tube and cuucnt wave levin,
the mode diagram of back discharge is depicted
domain as ihown in Fig. 2. where Eds « 20.7 kV/cm. With
the increase in current, the laytr breakdown condition,
Eds nd V. The
region IV, for lower value of Ea, is the surface streiinti
region where the surface streamer n..-*de is predominant
and space streamers a.c few
exceeds about 5kV/cm, both thc
Itrtamer occur to fonr ;!:e mixed streamer rr,
the critical current density for the transition from the re-
gion III to IV and V is nearly constant, except for a
area C. The critical value of the field strength betv,ren
region IV and V (curve H) is about 5 kV/cm under atmos-
pheric condition This value has been taken as s thrcahold
value for the occurence of streamers
The tangential field around the breakdown point wiU
become a function of the surface charge density o0 on the
layer at the instant of breakdown. The value of a0. in turn.
h given by *Eds where t is the dielectric constant and
Eds the breakdown field strength of the layer. Fig. 1-a
shows the pictures of back discharge when Eds - 13.8
kV/cm, while Fig. !< is for Eds - 33 S kV/cm. The surface
streamer becomes especially dominant when the value cf
Oo exceeds about S x 10"' C/cmJ .
Charging efficiency in different regums IV and V are
measured ind it is confirmed that the positive ion source
is considered to be surface-like in the surface streamer
region, but ion generation i.n. the ipji:c or
mixed streamer re.
In an actual precipitator, hovteter, these factors, hi. ha
and J. are closely coupled lo each other, depending upon
the dust resistivity pd The effect of pd is studied by using
tissue paper as a layer, in which p.i can easily be varied by
changing the imbinet hurmditv When pd is between
about 5 x 10'° and 0.9 x 10" Hem, number of breakdown
points is less and streamers proceed into gas space. In this
range of pd, excessive sparking tends to occur bacause the
streamer starting voltage Vst lies close lo the fhshover
voltage Vs. For pd higher than 10|: Sicm, the back dis-
charge streampers start to occur at a much ln*er voltage
and current density In this case, there is a large interval
between Vst and V$ so that the excessive sparking disap-
pears but an abnormal increase in current becomes
dominant. There are more breakdown point with a general
glow surrounding each point.
II. PROPAGATION OF STREAMER21
Propagation of streamers in the mixed streamer
is measured by using photomultipher tube Rg. 3 shows the
propagation of light emission in the normal direction z (gas
space), and in the tangential direction r (liver surface),
respectively The light sipial at the bieakdown point shows
that the back discharge streamer pulv
There is a primary ua\e which ;apidl> and lasts
about 20 ns, followed by a secondary uju- which rise more
0 100 200 ns
(a) z-AXIS
0 100 .
(b) r AXIS
flf. 3
W». red I.
vulout ,VU:I.'RI liken ilonj llic z tnd t j\u
Ipp. }0 mm. umplt, jljn pLile. V • 26 kV. I • |0 »*A)
177
-------
THE FOURTH INTERNATIONAL CLEAN AIR CONGRESS
dowry and Usti about 200 nt. The former proceeds in the
z direction and the Utter in the r direction. The primary
wave corresponds to the first rise in die current wave and
the secondary wave to the second rise in current which has
a much larger pulse height and a charge of 2 -4 x 10"* C/
pulse. The value of charge per pulse remains nearly constant
until just before ftashover takes place, while the period of
each succesive pulse decreased with the Increase in current.
III. MODE TRANSITION OF BACK
DISCHARGE3)
The effects of mean free path on the flashover voltage
and mode of back discharge are studied. A mica plate
having a pinhole of 0.5mm in diameter and 0.45mm thick-
ness b used as a layer. The flashoever voltage Vs is plotted
against the normalized mean free path as shown in Fig. 4,
where* Xo to the mean free path at NTP. The solid curves
represent the results obtained by changing P under room
temperature while the dotted curves indicate those ob-
tained by changing T under atmospheric pressure. The
scales for these P and T are also given. It can be seen that
350 150 550 T(K)
the curves measured by changing T or P agree well with
each other. It should be noted that there exist two dif-
ferent curves for flashover (curve I and II), each covering
the different range of X/XoTln the area under the curve I,
the onset-glow mode is followed by the streamer mode
with the increase in voltage, while the steady-glow mode
occurs In the area under the curve II. As a result there are
three different regions of X/Xo, A, B and C at indicated
in the figure, each corresponding to different mode
changes. In region A, back discharge in the streamer mode
follows the onset-glow mode and turns to flashover on
curve I. In region C, back discharge in the steady-glow
mode follows the onset-glow mode and no streamer appears
until flashover takes place on curve II. Region B is a transi-
tion region between A and C. The onset-glow mode ap-
pears at first at the pinhole. With the increase In applied
voltage, it turns into streamer mode, bridges across the gap
and finally turns to random sparking on the curve I. Slightly
above the curve I, streamers studdenly disappers to turn to
a stable steadyglow mode (Fig. 1-d). This lasts until flash-
over occurs on curve II. The current wave form in region
B at the onset-glow consists of Trichel pulses superimposed
to a small D.C. current component, and with the increase
in voltage, random pulses having much larger pulse height
appears corresponding to space streamers. Finally pulse-
less D.C. current appears with the transition from streamer
to steady-glow mode, no Trichel pulse being observed in
this mbde.
It is observed that no streamer occurs when the thick-
ness of the layer, t, is 47/an even under P - 760 toor. The
boundary pressure Pi between regions B and C shifts
towards the side of smaller XAo when t decreases, and the
value of Pi x t lines in the range between 18.6 and 35.7
torr-mm. It also is observed that, when the dust layer con-
tains alkaline compounds having low ionisation potential,
the region A covers the whole X/Xo range investigated, so
that Vs becomes lower and is given only by the curve I in
Fig. 4. Namely, the existence of alkaline compounds shifts
the boundary pressure between regions A and B towards
the lower pressure side.
760 560 460 360 260 160
P (torr)
Flf. 4 Effect of (ii mean free path on fltihom voltage of back
discharge
REFERENCES
1) Masuda, S. & Mizuno A.. "Initiation condition and
mode of back discharge," /. vf Electrostatics (to be
published).
2) Masuda, S., Recent Progress in Electrostatic Precipita-
tion, Static Electrification 1975, Institute of Physics
Conference Series, No. 27 p. 154.
3) Masuda, S. & Mizuno A., "Flashover measurements of
back discharge," /. of Electrostatics (to be published).
178
-------
Senichi Masuda (Tokyo University) . The Present Status of Elec-
trostatic Precipitator Technology. Presented at the Electrotech-
nical Colloquium of the Technical University of Munich, May 4,
1977.
1. Introduction
The control of atmospheric pollution has become one of the
uost important social goals of today, with the effective removal
of very fine particles (< 3 pm) from waste gases being especially
emphasized. These fine particles are present in smoke emissions
and are the nuclei for the formation of fog and clouds. Because
of their high specific surface, they act as carriers of harmful
gases into the depths of the lungs, and so they are harmful to
health. For this reason, the electrostatic precipitator, which
was born at the beginning of this century, has been given an im-
portant task today. Although the principle and the construction
of an electrostatic precipitator are very simple, its practical
use is often very complicated, because of many interfering ef-
fects that can be avoided only with difficulty. As a result,
electrostatic precipitator technology has for a long time been
considered an art. However, this situation is today slowly ex-
periencing a change as the result of intensive research and de-
velopment .
2. Principle and Construction in General
The principle and construction of an electrostatic precipi-
tator can be explained by reference to the tubular design in Fig.
1. It consists of a grounded tubular electrode (collecting elec-
trode) and insulated wire electrode (discharge electrode), between
which a high potential is applied. On the discharge electrode
there appears a corona discharge, which supplies a unipolar ion
current to the inner wall of the tube. The suspended particles,
along with the exhaust gas, enter the tube from below and pass
up through the tube, in which they are strongly charged with a
unipolar charge by ion impact. They are then driven by the cou-
lombic force to the inner surface of the tube and deposited there,
where they form a dust layer. This layer is dislodged by mechan-
ical rapping of the tube and falls into a hopper below. The
cleaned gas is led out the upper end to the stack.
In practical application, instead of the tubular type, the
so-called plate precipitator is mostly used. It is equipped with
parallel plate collection electrodes. The plate design has a
simpler structure, in which uniform gas distribution is more
easily achieved, especially with larger gas volumes. The corona
in industrial precipitators is usually negative, since its spark-
ing voltage is higher than for a positive corona. In air cleaning,
179
-------
TO HIGH-VOLTAGE SUPPLY
INSULATOR
GAS OUTLET
CORONA POINT
ION STREAM
COLLECTION ELECTRODE
—•i DISCHARGE ELECTRODE
*3 DUST LAYER
DUST OUTLET
Figure 1. Principle of electrostatic precipitator (tube type).
180
-------
a positive corona with appreciably less ozone production is used,
because of the harmful effects on health of ozone. In this ap-
plication, a two-stage construction with separate charging and
collection zones is mostly used (Fig. 2) .
In order to achieve the highest collection capacity, the
precipitator must be operated with the highest voltage possible;
the operation is also, however, determined by the continuous
fluctuation of the sparking voltage with changes in the operat-
ing conditions. This requires two functions of the high voltage
supply. One is the suppression of transition from sparking to
arcing and the rapid rebuilding of the normal operating voltage,
which is achieved with the aid of a thyrister circuit. The other
is the automatic monitoring of the optimum operating voltage.
Hence the spark rate is measured, for example, and the voltage
is so regulated that this rate is kept at a preselected value.
Also the power supply must have a static voltage-current charac-
teristic, which guarantees stable operation of the precipitator.
3. Collection Process
3.1 Equations of motion of charged particles
The collection process in an electrostatic precipitator is
based on the following expression for the motion of the charged
particles in the electrical and fluid-dynamic fields (electro-
hydrodynamic or EHD fields) :
m(d2R/dt2) + 6Tina(dR/dt) = qE + 6TrnaV (1)
where m = particle mass, kg
a = particle radius, m
q = particle charge, coul
p = gas viscosity, Nsec/m2
E = field strength vector, V/m
V = gas velocity vector, m/s
R = position vector of particle, m
For extremely small particles with a < 1 vim, the viscosity n must
be divided by the Cunningham correction factor (1 + AX/a) in con-
sideration of the ion slip effect, where X = free path of the
ion and A = constant. For atmospheric air, A = 0.866 and X =
0.1 jam [1]. Eq. 1 shows that the particle motion in the electro-
static precipitator is strongly influenced by the coulombic force
and by the viscous drag force of the gas flow eirnaV, in which,
181
-------
CHARGING ZONE
COLLECTION ZONE
Figure 2. Two-stage construction.
182
-------
as will be explained later, the corona wind plays a large role
[2]. The saturation charge, which the particle reaches in the
corona field, is proportional to a2. The relation of the coulom-
bic force to the fluid dynamic force is:
K » q E/ 6TTi-|a V = I12ne e E a2/(ee + 2)]/6Tin V a
O SOS
= [2 eoesE/(eg+2)nV] a (2)
Therefore, this relation is proportional to a. This means that
the coulombic force will predominate for the larger particles
with a > several tenths of pm. Then the particle velocity may
be calculated with the resulting "theoretical migration velocity"
=
W.. = qE/6irna m/sec (3)
For the smaller particles with a < several pm, on the other hand,
the fluid-dynamic force 6nna plays the decisive role in the col-
lection process. Thereby the result is that the particle path
in the electrostatic precipitator must in general always take
place with consideration of these two forces, i.e., from the EHD
point of view.
In the following, the magnitude of each factor in Eq. 1 is
considered more closely.
3.2 Corona field strength
The electric field in an electrostatic precipitator is en-
cumbered by the strong space charge of ions and charged aerosol
particles, so its analytical calculation generally is achieved
only with difficulty. An exception is the concentric cylinder
such as in Fig. 1. For it the field strength was represented
by Pauthenier and Moreau-Hanot [3] by the following approximation:
E(r)
h
V/m (4)
where I = ion current per unit length of wire, A/m
e0 = permittivity of a vacuum = 8.842 x 10~12F/m
y = ion mobility, m2/Vsec
e = specific dielectric constant of the particle material
o
S - total surface of the aerosol particles in unit gas
volume, m~l
183
-------
r0 = wire radius, m
E0 = breakdown field strength of the gas at the wire sur-
face, V/m.
plate precipitators one can obtain a rough estimate of the
voltage-current characteristic from Eq. 1.
In this equation the effect of particle space charge is repre-
sented by the value of S, which is determined by both the particle
concentration and the particle size. The higher both these values
are, the more significant the particle space charge effect. As
a result, the field strength increases near the inner wall of
the tube, while it decreases at the wire surface. The former
effect gives rise to sparking, so the sparking voltage is sharply
reduced several fold. The latter effect leads to a reduction
of the corona current, which again with an increase in the parti-
cle charging time decreases the effective value of the achievable
particle charge with limited residence time. This effect is
called the "corona quenching effect". Both the effects mentioned
above can, therefore, generally lower the collection capability
of the precipitator. It has, however, recently been shown that
the increase in the field near the plate can accelerate the par-
ticle collection [4].
3.3 Particle charging
The magnitude of the particle charge plays the decisive role
in the electrostatic precipitator, as it does in all other appli-
cations of electrostatic force. The particles are hit and charged
in the corona field by the neighboring ions through their thermal
motion. Thus a deficiency of ions occurs in this immediate neigh-
borhood, so a continuous supply is required for further charging
of the particles. Two separate mechanisms for ion transport ope-
rate. One is the field effect, in which the coulombic force
drives the ions from the outer regions to the inner regions (Fig.
3 (a)). The other is the diffusion effect, in which the ions
are transported, as the result of concentration differences, to
the inner regions by their thermal movement (Fig. 3 (b)). The
former mechanism is called "field charging" and the latter "dif-
fusion charging". Under practical operating conditions for elec-
trostatic precipitators, field charging is determinative for the
larger particles with a > 1 ym, while for very small particles
with a < 0.1 ym diffusion charging is decisive. The charge for
particles with intermediate sizes is well approximated by the
sum of both charges [5].
184
-------
(a) FIELD CHARGING
(b) DIFFUSION CHARGING
Figure 3. Charging mechanisms in corona field.
185
-------
3.3.1 Field charging
The quantity of charge Qf that is imparted to one particle
in field charging is, according to Pauthenier [3]:
Qf • QM [t/(t + T)] coul (5)
in which
3e
Q * 4ire0 —r, a2E = saturation charge, coul (6)
^ C _ T A C
5
4e0 4e0E
T « -—r » —T—- » charging time constant, sec (7)
where t • charging time, sec
E_ « charging field strength, V/m
C
p. ** ion space charge density, coul/m3
i = ion current density, A/m2
The particle charge increased with time t, finally reaching the
saturation charge Q«, (Fig. 4(a)). This final condition is limited
in that the increased potential of the charged particle repels
all the field lines. After the charging time t = 10 T sec the
particle charge Qf reaches 91% of the saturation value, so one
can assume as the charging time T = 10 T sec. According to Eq.
6, the saturation charge is proportional to the field strength
E , while the charging time according to Eq. 7 is inversely pro-
portional to the ion current i and independent of E . Hence under
practical conditions, with a restricted residence time t, the
magnitude of the particle charge Qf depends not only on Eq, but
also on the ion current density i. At EC = 3 x 105V/m and i =
2 x lO'^A/m2 the value of the charging time constant T = 0.053
sec.
3.3.2 Diffusion charging
The quantity of charge Q, by diffusion charging according
to White [6] is: a
Qd • Q* ln( + t/T*) coul (8)
uho t*o 4 ?r e* _ A IcI*
wnere Q* = *^a*A = specific charge, coul (9)
T* * aCN°e2 = cnar9^n9 time constant, sec (10)
186
-------
1.0
§
§ 0.5
o
4 6
t/T
(a) FIELD CHARGING
8 10
(b) DIFFUSION CHARGING
Figure 4. Increase of particle charge with time (normalized).
187
-------
2 3
J/K
where k = Boltzmann constant = 1.38 x 10
T = absolute temperature, °K
e = elementary change * 1.602 x 10"l9 coul
u
C » r.m.s. value of thermal ion velocity = (3kT/m) m/sec
m = ion mass, kg
N - no. of ions/m3
a » X assumed
According to Eq. 8 the charge increases very rapidly at first,
and reaches 6 Q* at time t = 402 T*. Then it increases appre-
ciably slower, so in a practical sense it remains almost constant
(Fig. 4(b)). Hence one can assume for diffusion charging 6 Q*
as a quasi-saturation charge and t = 402 T* as the charging time
constant. Under typical application conditions of T = 150°C,
x 1013/nr,
a = 0.1 urn, m = 5.3 x 10~2*kg (for Oz ion) and N0 = 5
one obtains/ e.g., a charging time t = 1.13 sec. For yet smaller
particles with a = 0.01 vim the charging time t is as large as
11.3 sec.
3.4 Migration velocity of particles and collection efficiency
of precipitator .
From Eq. 3 and the above-mentioned saturation charge Qm or
6 Q* one can evaluate the theoretical migration velocity. Curve
A in Fig. 5 represents the change of this value Wt^ as a func-
tion of particle radius a, with the assumption of corona field
strength E = Ec = (1-5) x 105V/m and the above-mentioned operating
conditions. It is noticeable that Wtu has a minimum value of
ca 0.1 m/sec in the range of 0.1 - 1.0 ym radius. These state-
ments have been verified both in laboratory trials as well as
in practical installations [5,7]. The increase in the curve for
the even smaller radius region is attributed to the effect of
ion slippage. In practical installations the high migration
velocities for very small particles are almost unobtainable be-
cause of the accompanying increase in charging time, since the
residence time in the precipitator is usually restricted to 5 -
10 sec.
As was already mentioned, these W^.^ values represent only
the electrical effect in the collection process, so they indi-
cate in no way the basis of collection efficiency, on account
of the numerous interferences that operate. Such interferences
include primarily gas flow turbulence, which is greatly increased
by the corona wind. They include the reentrainment of the de-
posited dust layer on the collecting electrode, which arises
188
-------
especially on rapping of the electrodes, and reverse ionization,
which represents the abnormal corona on the collecting electrodes
and which occurs with very high resistivity dust. This strong
gas turbulence produces a uniform distribution of the dust con-
centration on a plane perpendicular to the direction of the gas
flow. From this consideration Deutsch derived the following
formula for collection efficiency:
Collection efficiency * 1 - exp (-wf) (11)
where F = S /G = specific collection area, sec/m
3
S. = total surface of the collection electrode, m2
a
G = gas flow/ m3/sec
w has the dimensions of velocity and is termed the "ef-
fective migration velocity" or "w value"
This formula has a very simple form and has proved useful in
practical installations as an approximation applicable at least
for the same type of dust and operating conditions. Therefore,
it can be used for precipitator design when the w value is con-
sidered only as a measure of collection efficiency in operation.
This w value represents a design parameter that includes
the effects of all the process factors, including dust properties
and operating conditions, as well as" type of gas. Curve B in
Fig. 5 represents the mean value of w for many installations for
removing fly ash as a function of the particle radius a [8] .
The flattening of the curve in the large particle size range is
surprising, and is explained by the tendency for reentrainment
of the larger particles. There are so many influencing factors
that determine the w value that they in no way allow the design
variables to be based on theory. As the result, in the design
of precipitators many examples are required, which are to be had
only by the evaluation of collection efficiency in many practical
installations. It must be emphasized, however, that such practi-
cal data from installations with similar w values sometimes dif-
fer, giving the precipitator design more or less of a statistical
character [9]. Another puzzling w value is found with the so-
called large-spacing type, which has an appreciably larger elec-
trode separation distance [10]. It has been found that the col-
lection efficiency at a constant gas volume remains almost con-
stant, up to a certain limit, with increasing electrode separa-
tion distance. This means that with an increase in electrode
separation the w value must increase proportionately. This ef-
fect was confirmed not only in dry precipitators, but also in
wet precipitators, in which neither dust reentrainment nor reverse
discharge occurs. The basis for this effect cannot be properly
explained even considering the higher stability of the large-
spacing precipitator, and further investigation is required.
189
-------
1000
> 100
sT
10
0-01
W
0-1 1 10
PARTICLE RADIUS a, /im
100
Figure 5. Theoretical and effective migration velocities as functions
of particle radius a.
190
-------
4. Distribution of Ion Current and Charging Efficiency
For the most rapid and uniform charging of the particle,
not only the magnitude of the ion current density, but also its
distribution, plays a decisive role. As is known, the negative
corona on the wire electrode appears in the form of scattered
corona points, from which the tufts of ion flow spread out in
the direction of the collection electrodes. Between these streamers
there are ion dead spaces, in which the charging time constant
is large, because of limited current density, and the charging
efficiency of the particles is sharply reduced. This effect was
confirmed experimentally [11] and gave rise to a detailed study
of current distribution. It was found [12] that the law of
similarity applies to current distribution at the collection
electrode. Figure 6 shows an example. It was further observed
that when such a dead space is formed, dust reentrainment and
reverse discharge can also be present; with the former a zero
current region occurs; with the latter an increased current
region. For resolving these dead space questions, "Hermstein
corona" with positive corona, which is distributed uniformly over
the wire electrode and which has an appreciably higher sparking
voltage offers a possibility [12]. The conditions for this corona
are that the field strength and its gradient near the corona
electrode are sufficiently high. The former results in electron
stripping from negative ions; the separated electrons spread out
with a higher mobility over the electrode surface and there form
an electron sheath suppressing the "streamers". The latter also
retards the development of streamers. Thus, the formation of
the Hermstein corona is accelerated by supplying a small number
of negative ions to the positive electrode [13]. The Hermstein
corona has been used to increase the particle charging efficiency,
although interfering effects of moisture and dust load have been
recognized [14], Hence, its applicability under practical con-
ditions is still an open question.
5. EHD Processes
There are two processes in an electrostatic precipitator
that must be considered from the EHD point of view. One is the
collection process itself, in which, as has already been explained,
the interfering field of the corona wind plays a decisive role.
The other is the corona wind, which is the result of momentum
exchange between the electrically accelerated ions and the neutral
molecules on impact. In principle this flow field can be deter-
mined by the Navier-Stokes equation, but as another force the
coulombic force density, which acts as a motive force per unit
gas volume is drawn in play
F = piE N/m (12)
for which as an ancillary condition the continuity equation of
the gas flow must be considered. The distribution of the ion
191
-------
d = 10 cm
• I =200 //A
A 100/uA
1-0 2-0
NORMALIZED DISTANCE r/d
Figure 6. Normalized distribution of ion current density on
plate electrode.
192
-------
space charge density p^ can be calculated from the Poisson equa-
tion and the continuity equation of the ion flow. For the point-
plane electrode this corona wind field can be determined analyti-
cally [16] but for complicated conditions its mathematical solu-
tion is still open.
On the other hand, many experimental studies of the corona
wind have been made with the aid of laser-doppler velocity mea-
surements and schlieren photography. Fig. 7(a) shows, e.g., a
schlieren photograph of corona wind on which a horizontal gas
flow has been superimposed [17]. The figure shows clearly the
presence of a kind of jet stream with a higher velocity. Figs.
8 and 9 show the velocity components oriented perpendicular and
parallel to the plate along the axis of the point-plane electrode
without superimposition of another gas flow, which was measured
with the use of a laser-doppler measurement instrument up to a
very close distance to the surface (0.05 mm) [18]. In the main
part of the corona field along the axis, there is a perpendicular
velocity of as much as 6-7 m/sec. It is astonishing that this
perpendicular component extremely near to the plate has as high
a value of 4-5 m/sec. and that then its direction of flow changes
suddenly by 90° and it flows along the plate with an equally high
velocity. This remarkable effect can be attributed to the special
character of the ion driving force which acts up to the surface
of the plate. By diluting the boundary layer, this effect can
also produce a significant increase in the gas/plate heat exchange
(corona cooling). Fig. 10 shows the fluctuation of the perpendi-
cular velocity component inside the boundary layer with superim-
position of the corona wind and the gas flow parallel to the plate
[18]. Outside the very near vicinity of the plate surface, the
fluctuation of the velocity in the positive and negative direc-
tions is very large, although the average velocity is positive
(in the direction of the surface). This means that in the main
part of the boundary layer of ca 5 mm thickness, gas turbulence
is present with a velocity of about 0.5 m/sec. As was already
explained in Fig. 7(a), there predominates in the corona field
a kind of jet stream with a significantly higher gas velocity
than the theoretical particle migration velocity according to
Eq. 3, also shown clearly in Fig. 8. Thus, most of the particles
< 10 urn are first transported to the vicinity of the plate by
this jet stream, where the gas stream must change its direction.
There only the larger particles can be collected, those that with
the aid of electrical and inertial forces can overcome the tur-
bulence (Fig. 10) predominating in the boundary layer and finally
reach the plate. The smaller particles, especially in the size
range of 0.1 - 1.0 urn, are captured only with difficulty, except
for the very small particles that reach the plate as the result
of gas turbulence. The main part will flow along the plate with
the parallel flow and again return to the main field. The velo-
city of this reverse flow is of course essentially less than that
of the corona wind.
193
-------
Figure 4
A Schlieren Photograph of Corona Wind
(did not reproduce)
194
-------
10
O
O
"J r-
> 5
O
z
3
00
I
V = -30 (kV)
I = 5
10 20 30 40
DISTANCE FROM CORONA ELECTRODE, mm
50
Figure 8. Distribution of perpendicular component of corona wind
velocity along axis (point-plate; OOP particle 0.03 urn).
195
-------
4
3
u
$
E
*
1—
u ,
O 2
III
| . , . | . | .
V = -25 (kV)
_ 1 = 3 (/iA»
_
^
™
^^
.
Q
Z
1 —
±
x I
10 20 30 40
DISTANCE FROM CORONA ELECTRODE, mm
50
F/jgwre 9. Distribution of parallel component of corona wind velocity
along axis (point-plate; OOP particle 0.03
196
-------
1.0
o 0.5
I
*
o
O
UJ
Q
2
3
-0.5
46
2 [mm]
47
48
V - -25 (kV)
I - 3.5
I
I
o
0
49
46 47 48 49
DISTANCE FROM CORONA ELECTRODE, mm
50
50
Figure 10. Variation in perpendicular current velocity in boundary
layer (corona wind + parallel gas flow).
197
-------
The collectability of the particles in the turbulent boundary
layer can be divided into two parts, which represent the relation-
ship of the electrical and inertial forces to the turbulent force:
KI = Wth/Vo = q V6lTnaVo
K2 = Wm/V0 = [(1/2) (MVk2)/6]/6irnaVo (14)
where V = r.m.s. of turbulent velocity, m/sec
W = mean inertial velocity in boundary layer, m/sec
V. = vertical component of initial viscosity of particle
on entry to the boundary layer, m/sec
M = particle mass, kg
6 = thickness of turbulent boundary layer, m
The value of Vo can be obtained by computer evaluation of
the velocity fluctuations measured with the laser-doppler instru-
ment. For particles with values of Kj or K2 larger than about
10 one can calculate collection in the usual way.
The very high parallel velocity of the corona wind after
impact on the plate surface indicates that it can produce a con-
siderable amount of dust reentrainment . It was established with
the laser-doppler measuring instrument that dust reentrainment
mostly can occur at mean gas velocities above 1-2 m/sec [19] .
Apparently one must find the correct compromise between the above-
mentioned positive and negative effects of the corona wind to
fit the adhesion characteristics of the individual dust. In this
connection the voltage characteristic of the corona wind velocity
in the main field and in the region near the surface is useful
(Fig. 11) . The difference between the curves indicates a possi-
bility that one could well use the transport effect of the corona
wind and a smaller voltage reduction without any considerable
interference by dust reentrainment. Since according to Eq. 12
the motive force for the corona wind is directly proportional
to the ion current, one can regulate the corona wind effect by
correct choice of the voltage-current characteristic. This re-
quires also a correct choice of the corona electrode type or the
application of a special current-regulating charging system, as
in Fig. 13.
Fig. 7(b) represents a schlieren photograph of the corona
wind, which arises in reverse discharge. The wind direction is
perpendicular to each reverse discharge point from above. Fig.
12 shows the distribution of the perpendicular component of the
corona wind velocity along the axis of the point-plate electrode
system under conditions of reverse discharge, where its absolute
value is represented. The wind directed toward the plate by
198
-------
10.0
49.2 [mm]
49.0 [mm]
33.0 [mm]
15.0 [mm]
6.0 [mm]
1.0 [mm]
5.0 [mm]
5 10
VOLTAGE, kV
50
100
Figure 11. Perpendicular component of the corona wind velocity as
a function of voltage.
199
-------
10
8
"E
>
H
1
IU
0
» 5
AIR FLOW FOR D.O. P. SU
I I
10 20 30
DISTANCE FROM CORONA ELECTRODE
40
50
Figure 12. Distribution of perpendiuclar component of corona wind velocity
along axis under conditions of reverse discharge.
200
-------
DC BIAS VOLTAGE
SOURCE
Figure 13. New type of two-stage precipitator with bias-controlled
pulse charging system and electrical screen.
201
-------
negative ions is indicated by "N", while the reversed wind direc-
tion by the reverse discharge ions (positive ions) is marked "BD".
Although the region of the reverse corona wind is relatively
restricted (ca 5 mm), its velocity is astonishingly large, 5-10
m/sec. Experimentally it was established that this strong reverse
discharge corona wind not only can repel the oncoming particles
back to the main field, but also can tear particles from the dust
layer and can enter the main field with a high velocity even up
to 5-10 cm distance [19].
With reference to the EHD treatment of particle motion we
must consider the application of the EHD potential [21] which
is defined as:
md2R , 6TiTiadR , . to\ MR\
Spr + -JJ— = -grad *EHD (15)
from which
*»«i%(R) = ~ I (6irnaV(R) + qE(R) dR = EHD potential, J (16)
EHD J
0
Eq. 15 represents the equation of motion of the charged particle
in the EHD field, in which the gas flow field is assumed to be
that for an ideal gas. The mode of the particle motion changes
according to the parameter:
C = (eirnab/mV)* (17)
where b = electrode separation, m
V = mean gas velocity, m/sec
If c » 1, as in the case of the smaller particles that are being
considered in an electrostatic precipitator, the particle motion
is "viscous", in that the first term in Eq. 15, compared to the
second term (viscosity term), is negligible. In this case the
particle moves along the lines of force of the EHD potential field,
so the particle collection efficiency is obtained as the result
of these lines of force. This calculation can be arrived at
easily with the aid of the charge-substitution method combined
with a computer [22,23]. Fig. 14 represents the distribution
of the EHD lines of force in the collection zone of the 2-stage
electrostatic precipitator shown in Pig. 13 [21]. This distri-
bution of lines of force varies according to the relationship
of the electrical force to the hydrodynamic force, as can be seen
from Fig. 14.
202
-------
K = 0 K = 0-2 K = 0-4
K = 0-6 K = 0=8 K = 1-0
K - 2-0 K = 3-0
Figure 14. EHD lines of force.
203
-------
K3 = qE/6TrnaV (18)
where E = mean field strength between the two electrodes, m/sec.
If K3 = 0, the distribution coincides with the streamline distri-
bution, while with an increase in the K3 value, the lines of force
which terminate on the collection electrode increase, and finally
at K3 = 1.95 give a 100% collection efficiency. Actually one
cannot directly estimate the collection efficiency from the cal-
culation of EHD lines of force in many instances, because the
effects of the moving mass and the viscosity of the jet stream
that is formed and the turbulence operate on the particle motion,
usually in a positive direction.
6. Adhesion and Reentrainment of Dust
One of the most important factors in the effective collection
of dust is the strong adhesion between dust particles and between
the dust layer and the collection electrode. This allows the
formation of a sufficiently strong dust layer that on electrode
rapping will fall into the hopper below without disintegration
and reentrainment. In this sense the electrostatic precipitator
must function as an effective dust coagulator. In this respect,
with high-resistivity dust, which has a specific layer resistivity
of p0 > 1010 ohm-cm, the so-called electrical adhesion plays an
important role [24]. This force arises from the potential dif-
ference at individual particle contacts. It is proportional to
the pd value and the current density i^ in the dust layer [25].
Apparently the roles of the contact charge and electret formation
in the dust layer under the action of the electric field also
come into question [26]. Experimentally it has been established
that an electrically deposited dust layer has a 30-60 times s
stronger adhesion force that one deposited mechanically.
As was already explained, the gas flow causes dust reentrain-
ment if the average velocity exceeds 1-2 m/sec [19]. This limits
the usable gas velocity in a precipitator and results in a large
size for the installation. For air filters, which are used under
atmospheric conditions, the dust layer has a sufficiently high
adhesive force, as the result of absorbed water molecules, so
a relatively high air flow, about 8 m/sec, is usable.
For low-resistivity dust, with p^ < 101* ohm-cm, another kind
of dust reentrainment can occur completely on an electrical basis.
In this instance, the particle on reaching the plate immediately
gives up its charge and finally, despite the arriving ion jet,
becomes strongly positively charged by induction. Then it is
immediately returned to the main field. As a result, an abnormal
dust reentrainment is produced, insofar as it is not overcome
by other kinds of adhesive action. In addition to electrical
adhesion forces, there is also the van der Waals force, which
204
-------
acts more effectively with decreasing particle size [27], Under
atmospheric conditions in which relatively high moisture is pre-
dominant, water adsorption in capillaries of the particles plays
an important role (capillary condensation). Recently it has been
found that one can increase the collection efficiency by increas-
ing the adhesion action by injecting small amounts of triethyla-
mine [28] or ammonia [29] into gas. Ammonia injection is already
used to protect against corrosion by S03 in oil-fired steam power
plants. The corona-quenching effect is due to the formation of
extremely fine particles of ammonium sulfate and bisulfate [29].
7. Reverse Discharge Processes
For very highly resistive dusts with pd = 5 x 10l° ohm-cm,
mostly metal oxides, electrical breakdown occurs due to a high
voltage drop which then leads to a reverse discharge abnormal
corona in the dust layer. This process has long presented one
of the most difficult problems in electrostatic precipitator
technology; in many industrial installations it results in an
appreciable hindrance to precipitator collection efficiency, e.g.,
in ore sintering plants of the steel industry, rotating kilns
and clinker coolers in the cement industry, melting furnaces in
metal foundries, and coal-fired steam power plants. The fly ash
from coal with a low sulfur content, which is required for lower-
ing the SOX content in the stack gas, has an especially high pd
value, so its removal in an electrostatic precipitator presents
great difficulties with reverse discharge. Depending on the
magnitude of the pd value, there are 'two forms of reverse dis-
charge. In the region between 5 x 1010 and ca 1011 ohm-cm, there
is an extremely strong tendency to spark (excessive sparking),
which upsets the stable precipitator operation, and as a result
more or less degrades the collection efficiency. In an even
higher pd value region, above 1012-1013 ohm-cm, the sparking
tendency disappears, and there arises over the entire dust layer
surface a sheet-like glow corona, often with streamer coronas
developing toward the main field, and with an associated strong
increase in the corona current (abnormal current increase). This
current increase is associated with a supply of positive ions
from the reverse discharge points. These positive ions not only
neutralize the useful negative charge of the particles but also
charge them positively so the particles are repelled and are
completely removed from the collection process.
Since, as was explained above, the reverse discharge is
caused by the breakdown of the dust layer, its initial conditions
can be formulated as:
idPd > Eds V/m (19)
205
-------
where id = current density inside the dust layer, A/m2
E, = breakdown field strength of the dust layer, V/m
Impulse-like breakdown occurs repeatedly in the dust layer, by
which the particles are ejected from breakdown point into the
gas space one after the other, and finally a pinhole is formed
in the dust layer. This pinhole allows the formation of a stable
initial glow corona in it. With an increase in the electrode
voltage the discharge goes over an initial streamer corona in
impulse-like streamer corona, which, depending on the field dis-
tribution, develops either at the corona electrode or along the
dust layer surface, or in both directions. The first is a "space
streamer", the second a "surface streamer", and the third a "mixed
streamer" [30]. Fig. 15 is a diagram of the reverse discharge
in atmospheric air, which takes the form of reverse corona as
a function of the field component Ea perpendicular to the surface
and the ion current density i. Thus, one sees that the space
streamer arises first at a higher field strength E > 5 kv/cm,
which agrees with the development conditions of the streamers.
In the mixed streamer region, and in the outer surface streamer
the space streamer is also present; the occurrence of sparking
is determined only by the ion current density i. Since the amount
of charge associated with a streamer impulse remains almost con-
stant, the ion current is proportional to the streamer frequency.
With increase in the ion current, the period between two
streamers decreases and finally reaches the order of magnitude
of a plasma life time, when sparking can appear. In the region
of Ea < 5kV/cm, only the surface streamer occurs, and no sparking
takes place. In this region the ion current depends on its space
charge and is related to the main field strength. As a result
there is in this region an upper limit to the current, which
represents the saturation current. It is notable also that the
breakdown in the dust layer directly goes over into sparking,
if the main field strength exceeds 8.4 kV/cm. The effect of the
tangential field component on the form of the reverse discharge
is shown in Fig. 16. This component Efc is mainly determined by
the density of the surface discharge on the dust surface o^g at
the moment of breakdown, and further the magnitude of a^s is con-
nected to the breakdown field strength of the dust layer in the
following way:
o = eri E, coul/m2 (20)
CIO Vl U9
where e^ represents the dielectric constant of the dust layer.
In Fig. 16 an experimental layer was applied on two glass plates,
one on top of the other, with pinholes, whereby its breakdown
field strength Eds could be regulated by changing the distance
separating the two pinholes [30]. By increasing the Eds value
206
-------
10-5
V"
<
ui
a
10-8
CURRENT SATURATION
MIXED STREAMER
REGION
III INITIAL STREAMER
II INITIAL GLOW CORONA
I NO BREAKDOWN
246
FIELD STRENGTH E,. kV/cm
Figure 15. Reverse discharge diagram.
207
-------
Figure 16
Photograph of Corona From High Resistivity Layer
(did not reproduce)
208
-------
the development of the surface streamers could be greatly accele-
rated. It was established that the conditions for development
of the surface streamers are the presence of a_sufficiently large
surface charge density, greater than ca 5 x 10~9 coul/cm2, as
well as a sufficiently high surface resistivity of the dust layer.
Fig. 15 explains partly the basis for the difference in appearance
of the reverse discharge which occurs with change in the dust
layer resistivity. This change leads to a corresponding varia-
tion in the division of potential between gas field and dust
layer. If the pd value is low, the field strength in the gas
space E increases with increase in the voltage, while the field
strengtn in the dust layer Ed remains relatively low. The de-
velopment of the space streamer is greatly accelerated, which
leads to excessive sparking. If on the other hand, the pd value
is high, the situation is reversed, so no sparks occur in the
gas space, while in the dust layer on account of a rapid increase
in the field strength breakdown occurs at many points one after
the other, before sparking occurs. Thus, the abnormal current
increase described above.
Fig. 17 shows a typical oscillogram of the impulse current
and the accompanying light emission occurring on reverse discharge,
The spatial change in the light emission oscillogram is repre-
sented in Fig. 18, in which the measurements were made along the
electrode axis and the surface of the layer. The spatial resolu-
tion of the measurements was 0.3 mm, and a glass plate with a
pinhole was used as a test layer [31]. .From the light emitted
signal at the original point it may be seen that the reverse dis-
charge impulse consisted of a primary and a secondary wave. The
former represents the first increase in the current impulse and
rises very rapidly. It lasts very briefly (about 20 nsec) and
has a small impulse height. The current impulse connected with
this primary wave has a charge of 1-2 x 10~9 coul/impulse. The
secondary wave corresponds to the second current impulse, which
has an appreciably greater impulse height with a charge of 2-4
x 10~8 coul/impulse and a long duration (ca 200 nsec). With the
aid of an "image converter camera" combined with an "image in-
tensifier", it was established that the primary wave represents
a space streamer, which develops with a velocity of ca 4 x 107
cm/sec at the corona electrode. The secondary wave represents
the surface streamer, which has a velocity of ca 2.5 x 107 cm/sec
and an appreciably higher charge. If the surface resistivity
of the layer is smaller, the surface streamer disappears, and
the space streamer is present. With both the primary wave and
the secondary wave the related charge per impulse remains constant
.independent of the voltage, while the impulse frequency changes
when the voltage changes. In general, the impulse frequency of
the surface streamer is appreciably smaller than that of the space
streamer, so the space streamer, despite its smaller charge per
impulse, has a greater influence on the increase in current.
Fig. 19 shows the reverse discharge for a positive corona elec-
trode, which has a completely different appearance. In this case
209
-------
Figure 17
(Photograph - did not reproduce
210
-------
48
40
E
30
20
10
8
6
4
3
NEEDLE
0 200 ns
Z-axil
E
E
N
15
10
-26 kV
DSC
0 200 ns
r-axii
Figure 18. Wave form of fight signal measured along z- and r- axes.
211
-------
Figure 19
(Photograph - did not reproduce)
212
-------
numerous glow points are uniformly distributed over the dust
layer surface. From neither the discharge points nor the positive
corona electrode is there a streamer, but sparking occurs sud-
denly. The usual streamer from a positive corona is completely
suppressed, presumably on the same basis as with the Hermstein
glow corona, namely because of negative ions supplied from the
reverse discharge points. Experimentally it was established that
the sparking voltage in this case is higher than that'without
a dust layer [30].
Fig. 20 shows the sparking voltage at reverse discharge as
a function of the mean free path of the ions X [31]. As a test
layer was used a mica plate with a small hole, and the magnitude
of X was varied by changing either the temperature or pressure.
Outside the higher temperature region, where no reverse discharge
takes place, the sparking voltage is determined only by the magni-
tude of the X value, quite independent of variation in tempera-
ture or pressure. The sparking voltage is represented by two
curves I and II, the validity regions of which agree in Region
B but not in Regions A and C, above and below. Curve I repre-
sents the voltage at which the sparking develops over the space
streamer, while curve II represents the voltage at which the
sparking takes place directly from the glow reverse corona. In
Region C with a higher value of X no streamer occurs, and the
glow reverse corona is present up to Curve II, at which sparking
suddenly occurs. In Region A with a smaller value of X, which
also corresponds to atmospheric conditions according to Ref. 15,
there appears after the already explained initial glow corona
the space streamer, which goes over to sparking at Curve I. On
the other hand, in Region B the sparking that sometimes takes
place on Curve I disappears, if the voltage exceeds only slightly
Curve I, and suddenly goes over to impulse-free stable glow re-
verse corona. In this case a stronger sparking finally appears
at an appreciably higher voltage on Curve II. The boundaries
between A and B or B and C can be shifted according to the mag-
nitudes of dust layer thickness, alkali content of the dust, etc.,
which determine streamer development. E.g., boundary B/C is
shifted to the left for a decrease in the dust layer thickness,
to finally appear outside the region being considered, so no
streamer will occur in atmospheric air. On the other hand bound-
ary A/B is shifted to the right for an alkali-containing dust,
so sparking in the total region will be controlled by Curve I
at an appreciably lower sparking voltage. This fact may be the
basis for the observation that in an electrostatic precipitator
'for dust removal from an iron ore sintering furnace gas, in which
high-resistivity dust with a high potassium content is encountered,
the collection capability against the strongly abnormal increase
in current and the related lower operating voltage is usually
unsatisfactory. Fig. 21 shows schematically the structure of
the initial glow corona and stable glow jet corona in the developed
213
-------
30
20
10
B
350 450
I I
550
t
T,k
760
460 360 260
P, torr
150
Figure 20. Sparking voltage Vs with reverse discharge as a function of
mean free path of ions.
214
-------
N
N
— -*•
-f -
FIELD LOW
L —
—A- — -f
(a) INITIAL GLOW CORONA (TRICHEL PULSE)
(b) GLOW CORONA AT DEVELOPED STAGE
(WITHOUT IMPULSE CURRENT)
N: NEGATIVE ELECTRODE
S: REGION OF STRIPPING ELECTRONS FROM IONS
P: REGION OF ACCUMULATION OF POSITIVE IONS
G: NEGATIVE GLOW
D: DARK ZONE
C: POSITIVE POST
E: PLATE ELECTRODE
Figure 21. Mechanism of glow corona in reverse discharge.
215
-------
stage (without impulse current) which occurs in Region C and in
the zone between I and II of Region B. In these special glow
corona, which have no negative electrode as electron sources,
the electrons in Region S, which with the aid of strong space
charge field were stripped from the negative ions and produced
the positive ions accumulated in P, must have been supplied to
the glow corona.
Fig. 22 shows the streak photos of the space streamer, which
were taken under a low pressure of 170 Torr on applying the im-
pulse voltage [32]. After applying the impulse voltage there
first appears a short-lived light emission at the corona elec-
trode. About 400 nsec later begin the light emission of the space
streamers from the reverse discharge points back to the corona
electrode. It was established that the dead time between the
two light appearances corresponds exactly to time for passage
of the electrons between two electrodes (d = 50 mm) 132]. With
an increase in the pressure of the air this dead time increases
slowly at first, and at a certain pressure of about 470 Torr there
is a discontinuous increase in the dead time. Finally there
occurs once more a slow increase in the dead time with pressure.
This indicates that at this critical pressure there has been a
change in the charge carriers causing the reverse discharge from
electrons despite their higher concentration to ions, and because
of their high electron attachment probability, to electronegative
molecules.
8. Particle Charging
Experimentally it has been established that the Pauthenier
formula for field charging, Eq. 5-7, applies very well, at least
for spheres of conducting materials. I£ must be emphasized,
however, that the "conducting materials" in this instance include
not only "conductors" in the normal sense, but also such materials,
the relaxation time ep (e = dielectric constant, p = specific
resistivity) for which is appreciably less than the charging time
constant in Eq. 7. In the course of the charging process, the
charges on the sphere are distributed in the same way as they
exist on a true conductor. Since high-resistivity dusts have
a specific layer resistivity generally 100 - 1000 times higher
than their specific volume resistivity [33] , a dust with a speci-
fic layer resistivity < 1013 ohm-cm can be considered as a quasi-
conducting dust because of its relaxation time constant of 1-10
msec. For insulator spheres, the measured value of their satura-
tion charges is always only about half of the theoretical value
in Eq. 6 [34]. If an insulator sphere is reversed by passage
through the corona field, its saturation charge becomes almost
equal to the theoretical value. This is attributed to two con-
tradictory assumptions derived from Eq. 5-7. Namely, for cal-
culation of the external field of the sphere it was assumed that
the sphere is made of insulating material, and that nevertheless
the charge given to the sphere is distributed over its surface,
as though the sphere were a conductor.
216
-------
Figure 22
(Photograph - did not reproduce)
217
-------
Pauthenier has also derived the formula for charging a
spheres by bipolar ions [35,36], which allows an evaluation of
the particle charging under reverse discharge conditions, so far
as the concentrations of positive and negative ions are known.
According to this formula, the particle charge on larger masses
is reduced due to the combined action of small amounts of false
ions (positive ions), which has been experimentally confirmed
[34]. Further, it was established that the surface streamer as
a source of false ions acts as a source of surface ions, while
the space streamer acts as a source of volume ions, and so ions
of both polarities are produced in the field volume. Thus, the
dust particle can under conditions of space streamers often be
strongly positively charged. It is remarkable that the recom-
bination probabilities of both ions under the operating conditions
of the electrostatic precipitator are so low that a considerable
quantity of positive ions can reach the corona electrode. Ex-
tremely high dust resistivities of pd > 1013 ohm-cm can cause
a reverse discharge from the corona electrode to the dust layer
deposited on it. In this way the so-called propagation of the
reverse discharge takes place along both electrodes.
Gas ionization by irradiation with gamma rays or high-energy
electron beams can be considered as a new charging method [37,38].
Aft electric field must be applied to the ionization field, by
introducing parallel electrodes. The positive and negative ions
are electrically separated in order to form at each electrode
a zone of a multitude of ions with a given polarity. There the
dust particles can be effectively charged by the ion multitude
and collected on the appropriate electrode with a high collec-
tion efficiency. These methods have been found to be a very
useful aid in electronic experiments for removal of NOX and SO
from exhaust gases [38]. The gaseous impurities (NOX and SO )
are converted by high-energy electrons and with the aid of NH3
additions to solid aerosols, the particles of which are then
captured with the aid of an applied electric field. It has been
established also that these radiochemical reactions are accel-
erated by the action of an electric field [39].
9. Dust Resistivity
Fig. 23 represents the effective migration velocity w (as
a measure of collection capability) of an electrostatic precipi-
tator as a function of the pd value of the dust layer. The pre-
cipitator usually shows its Highest capability in the pd-region
of 10"-1010 ohm-cm. In the low-p^ region the collection capa-
bility falls off due to the occurrence of abnormal amounts of
dust reentrainment, while in the high-pd region it falls off due
to reverse discharge. The specific resistivity of high-resis-
tivity dust layers is very sensitive to effects of temperature,
gas moisture, and also the presence of small amounts of special
substances, such as S03. Fig. 24 represents schematically the
218
-------
3
*
H
o
3
oc
(9
ABNORMAL DUST REENTRAINMENT
I
NORMAL REGION
REVERSE DISCHARGE
103 104 1010
SPECIFIC DUST LAYER RESISTIVITY pd, ohm-cm
Figure 23. Change in effective migration velocity W as function
of specific dust layer reistivity.
219
-------
I
V)
oc
o
il
o
UJ
o.
V)
1013
1012
io
10
109
108
100 200 300
TEMPERATURE, °C
400
Figure 24. Effect of temperature and moisture on the specific dust
layer resistivity pd of high-resistivity dust.
220
-------
effects of temperature and gas moisture content on the p^ value
of a high-resistivity dust layer. The electrical conductivity
of the dust layer results in general from surface conduction and
volume conduction of the particles, and the conducting path is
also affected by the way in which the layer is formed [33]. In
the low-temperature region, where the relative gas humidity is
sufficiently high and the amount of water absorbed on the dust
surface is sufficiently large, surface conduction predominates.
In the higher temperature region, volume conduction predominates,
and it increases with increase in temperature. As a result of
the equilibrium between the two conduction mechanisms, there is
a maximum in the p^ value at 100-200°C, which corresponds to the
operating temperature of many industrial waste gases. Thus, it
is not seldom in many large industrial installations that inter-
ference by reverse discharge results.
Recently it has been found that alkali metal ions in fly
ash from coal play an important role as current carriers [40].
10. Technical Advances and Developmental Results
The so-called "large-spacing electrostatic precipitators"
with an appreciably larger electrode distance, up to 20-50 cm,
has found more and more applications in Japan in many branches
of industry, including use with high-resistivity dust, usually
with good results. In many instances a reduction in capital cost
of about 20% has been achieved. Fig. 26 shows a "roof-type elec-
trostatic precipitator" which has been built on the roof of a
factory building, from which there is a heavy dust emission.
A large volume of hot, dust-laden gas rises, e.g., from an elec-
tric furnace and enters directly into the electrostatic precipi-
tator. It flows without the use of blowers, only by free con-
vection, through the precipitator, and enters the atmosphere
directly after dust removal. To decrease the weight of the col-
lection electrodes, they are made from a conducting plastic [40],
and they are cooled with trickling water. The wet electrostatic
precipitator has found much interest because of its advantages-
-high collection capability without dust reentrainment or reverse
discharge, plus the added action of effective gas absorption (S02,
HCl, HF, etc.). The high gas absorption capacity in this instance
is related to the already explained dilution of the boundary layer
by the corona wind. In the practical application of wet electro-
static precipitators questions arise as to the large amount of
irrigation water required, the generation of waste water that
must be treated, and the lowering of the gas temperature that
hinder the buoyancy of the stack plume. One of the solutions
for these difficulties is the so-called "hybrid electrostatic
precipitator", in which the dry stage and the wet stage are in-
stalled in series in a common housing. The main part of the dust
(about 90%) is first removed in the dry stage with relatively
-------
COLLECTION ELECTRODE
DISCHARGE
ELECTRODE
THIRD
ELECTRODE
HIGH
VOLTAGE
SUPPLY
Figure 25. Impulse charging system according to Luthi.
222
-------
WATER INLET
INSULATED CHAMBER
HIGH VOLTAGE SUPPLY
Figure 26. Roof-mounted electrostatic precipitator with plastic collection
electrodes (Sumitomo Heavy Ind.).
223
-------
small dimensions, and then the remainder with a very small par-
ticle size can be effectively captured in the wet stage with a
very high collection performance. Most of the waste water is
recirculated to the irrigation after removal of sludge. This
system allows large reductions in the amount of irrigation water
used and sludge generated, with less gas cooling, while simul-
taneously a higher collection performance and proportionally
smaller dimensions of the precipitator are achieved. The advan-
tages of this process are especially worthy if a very high col-
lection efficiency for dust and simultaneous gas absorption are
involved.
Concerning the technical solution of interference by reverse
discharge, humidifying the gas with a water spray at the entry
to the precipitator has been used for a long time; the p^ value
of the dust layer is reduced to about 1010 ohm-cm by humidifying
the gas and by lowering its temperature. The most important
question here is to insure the rapid and complete evaporation
of the water spray [43]. In some applications the injection of
a suitable chemical means for lowering the p^ value has been found
to be a very effective method [44], So, for example, the injec-
tion of small amounts of SO3 (several tens of ppm) into the enter-
ing gas with a high-resistivity fly ash, which arises in the
combustion of coal with an extremely low sulfur content, has been
applied with great success [45,46,47]. Recently it has been
proposed to mix sodium salts, e.g., Na2CO3 with the coal to be
burned [48]. Another solution to the reverse discharge problem
is to operate the precipitator at an appreciably higher tempera-
ture, about 300-400°C, "hot-side operation". This is done by
installing the precipitator ahead of the air preheater in the
power plant. It is clear that the p^ value is greatly reduced
thereby (cf. Fig. 24). In designing the unit the thermal expan-
sion properties of the construction material must be considered.
In addition to the above-mentioned operating precautions
for preventing reverse discharge, purely electro-technical methods
can be used. Fig. 25 represents one according to Liithi [49],
in which a third electrode is installed near the corona electrode
and a pulse voltage is applied across them. Thus, there is be-
tween the third electrode and the collection electrode a high
voltage by which the field is formed. By varying the magnitude
or repetition frequency of the voltage pulse, one can regulate
the magnitude of the ion current quite independently of the main
field strength, which allows the removal of the reverse discharge
according to Eq. 19 at the highest attainable main voltage. The
highly concentrated negative ion cloud is produced. Because of
the strong propagating force, in the course of migrating to the
collection electrode, the cloud expands to provide at the col-
lection electrode a homogenerous distribution of ion current
density. This is one of the most important conditions for avoid-
ing reverse discharge. In the practical application of this
224
-------
method, one must increase the distance between the third electrode
and the corona electrode to at least about 10 cm, in order to
avoid undesirable variations in construction. However, this makes
the screening of the corona electrode by the third electrode
during the pulse-free time period very difficult, which represents
the condition required for avoiding current injection from the
corona electrode during the pulse-free period and the formation
of pulsed corona charging. Also fluctuations in gas and dust
properties cause trouble with screening. One solution is offered
by the application of a uniform voltage, which is superimposed
on the pulse voltage and brings the potential of the corona elec-
trode in the pulse-free time down under the corona potential.
Such a system is used in the charging zone of the two-stage pre-
cipitator in Fig. 13 and is termed the "bias-controlled pulse
charging system" [20]. It was established in an experimental
precipitator like that in Fig. 13 that with this charging system
one can increase the collection efficiency from 63% to 93% for
an extremely high-resistivity dust with pd ~ 1013 ohm-cm. Further,
it has been established in many other installations that this
system, at least up to a p^ value < 10llf ohm-cm, is one of the
effective precautions. Beyond this p<-| limit the above-mentioned
propagation of the reverse discharge effect also takes place on
the third electrode, and also when no pulse current is furnished.
It has been established that in application of this system in
the region of p^ = 1013-10llf ohm-cm, consideration must also be
given to the main field strength and the pulse breadth [50] .
The application of an a.c. voltage combined with an insulat-
ing film on the collection electrode has also proved to be ef-
fective in avoiding the reverse discharge effect [51]. One prob-
lem with that approach is the material for the insulating film,
which must be trouble-free for long time periods at high tempera-
ture.
The investigation of EHD particle migration in the electro-
static precipitators has led to new kinds of two-stage precipi-
tator shown in Fig. 13. This system has enabled a substantial
reduction of the precipitator volume [52].
11. Conclusions
An overview of the present condition of science and practice
of electrostatic precipitators was prepared. In it was recognized
the special character of the collection process, on which numerous
factors operate simultaneously. Each presents its own difficulties
to precipitator practice and the investigation of separate ele-
mentary processes can not completely provide an understanding
of collection as a whole. The gap between science and practice
in electrostatic precipitator technology remains very large.
Considering the present and future needs for even cleaner air
in the environment, increased activity on research and develop-
ment in this boundary region of practice is very desirable.
225
-------
Literature
1. H.J. White: Industrial Electrostatic Precipitation, p. 157,
Addison-Wesley, New York (1962).
2. S. Masuda: Recent Progress in Electrostatic Precipitation,
Inst. Phys. Conf. Ser. No. 27, p. 154, Inst. Phys. London
(1975).
3. M. Pauthenier and M. Moreau-Hanot: J. Phys. Radium, Vol.
3, p. 590 (1932).
4. M.B. Awad and G.S.P. Castle: J. Air Pollution Control Assoc.,
Vol. 25, p. 172 (1975).
5. G.W. Hewitt: AIEE Trans. Pt. 1, Vol. 76, p. 300 (1957).
6. H.J. White: AIEE Trans. Pt. 2, Vol. 70, p. 1186 (1951).
7. J.D. McCain, J.P. Gooch, and W.B. Smith: J. Air Pollution
Control Assoc., Vol. 25, p. 117 (1975).
8. D.D. Heinrich: Trans. Inst. Chem. Engrs., Vol. 39, p. 145
(1961).
9. S. Masuda: Staub-Reinhalt. Luft, Vol. 26, p. 459 (1966).
10. S. Ago, T. Itoh, H. Saito, N. Furuya, and S. Masuda: Proc.
1975 - Gen. Conf. Inst. Elect. Engrs. Japan, Paper No. 921
(1975).
11. S. Masuda, A. Shibuya, and E. Ikeno: Proc. 1973 - Gen. Conf.
Inst. Elect. Engrs. Japan, Paper No. 844 (1973).
12. S. Masuda and M. Niioka: Trans. Inst. Elect. Engrs. Japan,
Vol. A-96, No. 9, p. 411 (1976).
13. W. Hermstein: Archiv f. Elektrotech., Vol. 45, pp 209 and
279 (1960).
14. S. Masuda and M. Niioka: Trans. Inst. Elect. Engrs. Japan,
Vol. 8-95, No. 9, p. 423 (1976).
15. M. Robinson: Trans. AIEE, Vol. 80, p. 143 (1961).
16. T. Adachi, S. Masuda, and K. Akutsu: Velocity Distribution
of Negative Ionic Wind in A Point-to-Plane Corona Discharge,
Journal of Electrostatics, in press, Elsevier, Amsterdam.
17. T. Adachi, T. Shimoda, M. Suyama, and S. Masuda: Proc.
1974 - Gen. Conf. Inst. Elect. Engrs. Japan, Paper No. 718
(1974).
18. S. Masuda, K. Akutsu, and K. Aihara: Proc. 1977 - Gen. Conf.
Inst. Elect. Engrs. Japan (1977).
19. J.D. Bassett, K. Akutsu, and S. Masuda: Preliminary Study
on Dust-Reentrainment in Electrostatic Precipitators, Journal
of Electrostatics, in press, Elsevier, Amsterdam.
20. S. Masuda, Y. Doi, M. Aoyama, and A. Shibuya: Staub-Rein-
halt, Luft, Vol. 36, No. 1, p. 19 (1976).
21. S. Masuda and Y. Matsumoto: Trans. Inst. Elect. Engrs.
Japan, Vol. A-94, p. 515 (1974).
22. H, Steinbigler: Dissertation, T.U. MUnchen (1969).
23. S. Masuda and Y. Matsumoto: Trans. Inst. Elect. Engrs.
Japan, Vol. A-96, p. 1 (1976).
24. J. Dalmon and D. Tidy: Atmos. Env., Vol. 6, p. 81 (1972).
25. W. Sinun: Staub, Vol. 22, p. 463 (1962).
226
-------
26. G.W. Penny: J. Air Pollution Control Assoc., Vol. 25, p.
113 (1975).
27. H.J. Lowe and D.H. Lucas: Brit. J. Appl. Phys. (suppl.)
No. 2, p. 40 (1953).
28. D.J. Tassicker: Air Pollution Control Assoc., Vol. 25, p.
122 (1975).
29. E.B. Dismukes: ibid., p. 152 (1975).
30. S. Masuda and A. Mizuno: Initiation Condition and Mode of
Back Discharge, Journal of Electrostatics, in press, Else-
vier, Amsterdam.
31. S. Masuda and A. Mizuno: Flashover Measurement of Back Dis-
charge, ibid., in press, Elsevier, Amsterdam.
32. S. Masuda and A. Mizuno: Light Measurement of Back Discharge,
ibid., in press, Elsevier, Amsterdam.
33. S. Masuda: Staub, Vol. 25, No. 5, p. 175 (1965).
34. S. Masuda and K. Akutsu: Proc. 1975 - Gen. Conf. Inst.
Elect. Engrs. Japan, Paper No. 90' (1975).
35. M. Pauthenier: La physique des Forces electrostatigue et
leurs application (Centre National de la Recherche Scienti-
fique) p. 279 (1961).
36. N. Ishiga: Master-Dissertation, Dept. E.E., Univ. Tokyo
(1976) .
37. R.J. Heinsohn, S.H. Levine, R.J. Fjeld, and G.W. Malamud:
J. Air Pollution Control Assoc., Vol. 25, p. 179 (1975).
38. S. Masuda, N. Ishiga, and K. Akutsu: Proc. 69 Annual Meet-
ing of APCA, Paper No. 76-71.11 (1976).
39. S. Masuda, K. Akutsu, and Hirano: Proc. 1977 - Gen. Conf.
Inst. Elect. Engrs. Japan, - (1977).
40. R.E. Bickelhaupt: J. Air Pollution Control Assoc., Vol.
25, p. 148 (1975) .
41. Inst. Elect. Engrs. Japan: Technical Report "Present and
Future of Electrostatic Precipitator Technology" (1976).
42. S. Masuda, S. Ago, T. Itoh, H. Saitou, and N. Furuya: Proc.
69 Annual Meeting of APCA, Paper No. 76-42.1 (1976).
43. S. Masuda: I & EC - Process Design and Development, Vol.
5f No. 2, p. 135 (1966).
44. J. Dalmon and D. Tidy: Atmos. Env., Vol 6, p. 721 (1972).
45. H.G.T. Busby, and K. Darby: J. Inst. Fuel, Vol. 36, p. 184
(1963).
46. K. Darby and D.D. Heinrich: Staub-Reinhalt. Luft, Vol. 26,
p. 464 (1966).
47. R.E. Cook: J. Air Pollution Control Assoc., Vol. 25, p.
156 (1975) .
48. H.J. White: ibid., p. 102 (1975).
49. J.E. Liithi: Dissertation ETH-Zurich, No. 3924 (1976).
50. S. Masuda, Y. Doi, A. Shibuya, and I. Hattori: Performance
of Bias-Controlled Pulse Charging System, Proc. 4th Clean
Air Congress (May 1977, Tokyo) (1977).
51. H. Krug: Dissertation, TU Karlsruhe (1971).
52. A. Shibuya and S. Masuda: Proc. 1975 - Gen. Conf. Inst.
Elect. Engrs. Japan, Paper No. 920 (1975).
227
-------
APPENDIX B
ISHIKAWAJIMA-HARIMA HEAVY INDUSTRIES
228
-------
U., M.
Oct. 1977
Teclinical Information
IHI's
NEW PRECIPITATION TECHNIQUES
PAC AND ES
Air Pollution Control Engineering Department
Environment Control Equipment Division
Ishikawajima-Harima Heavy Industries Co., Ltd.
Tokyo Japan
229
-------
IIII PAC and ES TYPE ELECTROSTATIC PRECIPITATOR
Ushers in a new era of dry-type electrostatic precipitation.
Wide Range of Application
Through successful prevention from back discharge and reentrainment which are
caused b'y high resistivity dust and low resistivity dust respectively, the
PAC makes it nossible to collect the dust in a range which can not be covered
by conventional techniques of electrostatic precipitators (EP).
High Efficiency
The PAC and ES an achievement of IHI's new engineering techniques, has been
developed to satisfy the stringent demand for air pollution control.
o
It can technically hold the outlet dust concentration to 10 mg/Nra or below.
This value is virtually difficult of attaining by the conventional dry-type
EP to precipitate the high and low resistivity dust.
Compact
In case of treating too high resistivity dust, the PAC can be installed at a
small or elevated place.
Energy Saving
Electric power consumption is in no vain to effectively collecting the dust
and the draft loss is almost the same as the EP.
Easy Maintenance
Unlike the EP, electrodes of the PAC are readily accessible.
Stable Performance
In the EP, changes in gas conditions can exert a significant influence on its
performance. But in the PAC, the main electric field and corona current are
singly controllable at its charging stage to accord with gas changes.
This ensures stable performance of the PAC at all times.
230
-------
HMvy Mnlrfoi ۥ., IHL
Patents are applied for
Japanese and foreign patents are being applied for, and some have already been
granted.
1) Particle charging device for use in an electric dust collecting
apparatus U.S. Patent No. '*Ol8577.
CONSTRUCTION AND PRINCIPLE
A unit of the PAC consists of a discharge electrodes, control electrodes and
collector electrodes.
Making full use of the electrodes' functions, the PAC has over-come such
drawbacks to conventional electric precipitators as the back charye, the
reent rainment and the corona suppression.
A strong electric field is generated between the collector and control
electrodes by an electric source (l).
* Negative corona ions which are intermittently generated from the discharge
electrode by an electric source (2) are led to the space between the
collector electrode and the control electrode to provide the maximum
charge to the dust in exhaust gas.
r-O
-of
Electrode
Control Electrode
Collector Electrode
231
-------
€•«*
FUNCTION
Charging Dust Particles
Intermittently generated ionic clouds are led through a strong electric
field, which has been preset by the control electrode, to the collector
electrode as they are diffusing.
This method successfully prevents corona quenching and back discharge and
permits dust particles to be charged at a rate far beyond that achieved by
conventional EP's.
In the EP, an electric field is generated partially close to the discharge
electrode. As a result, the electric field is not adequate and the rate of
charging is limited.
In the PAC, independent control of the control electrode and the discharge
electrode is available by individual electric sources which allow each
electrode to operate efficiently with fluctuations of gas temperature, dust
concentration, humidity, electrical resistivity of dust, etc.
Because of its construction, EP's voltage and current are in a fixed,
functional relation and cannot be controlled independently. In contrast,
the PAC can control voltage and current independently according to gas
conditions, and consequently can precipitate all kinds of dust.
Gas Temperature Characteristics
As gas temperature rides, the electric field strength against breaking at
the needle of the discharge electrode will be reduced and spark discharge
will readily occur. To cope with this condition, the electrode voltage of
the PAC is adjustable for controlling the electric field at the needle of
the discharge electrode. Thus stable operation is Maintained*
232
-------
hrf»«tri«» to., LM.
Unit Characteristics
The electrostatic screen created in the collecting stage prevents effects of
ionic wind and permits dust particles to grow larger and coarser by electro-
static condensation.
As a result, in contrast to the EP, the PAC maintains constant high
efficiency at each unit.
ES in progress of IHI's development techniques
In accordance with a conventional precipitation theory, IHI's engineering
groups are developing the dust collection stage of E5 type.
This stage consists of many pairs of guiding electrode and a collector
electrode facing each other at right angles to the gas flow. While negative
high voltage is applied to the guiding electrodes, the collector electrodes
are grounded.
Between a pair of the electrodes a direct current electric field and a flow
field of fluid dynamics are generated. The former moves the charged dust
toward the pockets of collector electrodes and the latter keeps the dust
inside the electric field and forces it into the pockets.
The electric field and the flow field combine to create an electrostatic
screen (electro-fluid dynamics) for effective precipitation.
The dust thus adhered and accumulated in the pockets is dropped into a
hopper by rapping.
•-
_Guidinfl.Electrod« J
.
I—.*
o
-•
•-
»•
-•
Collector Electrode
233
-------
Portable test model of the PAC-ES
ZHI has produced a large, portable test model of the PAC-ES and carried out
testing vith actual exhaust gases from iron and steel productions (gases
from coolers of sintering plants and their environmental gases) and heavy
oil boilers. In these tests the model was proved satisfactory.
"Table 1" shows the performance record of the portable test model.
Table 1 One example of data of performance test
Gas tested
Gas temp. C
Moist in gas %
Gas Quantity _
m /min
Dust cont. .
(inlet) g/NmJ
Oust cont. .
(outlet) g/HmJ
Exhaust i
cooler o:
ing planl
(cooler 3
115 ^
<
48
2.074
0.0037
jas from
' sinter-
t
flue aas)
17O
D
78
1.803
0.0083
Exhaust
boiler
(C heavy
116'-
9.9 ~
40
0.034
O.OO23
gas from
oil)
160
12.3
60
0.034
0.0034
234
-------
APPENDIX C
HITACHI LTD.
235
-------
HIGH TEMPERATURE ELECTROSTATIC PRECIPITATOR
FOR COAL FIRED BOILER
H. Iraanishi, Y. Oataki, K. Ootsuka,
and K. Watanabe
Introduction
The coal fired steam plants have been replaced with the oil
fired ones since late 1950's and currently most of the steam
plants utilize crude oil or gas. This is considered to be an
appropriate consequence in the standpoint of fuel security, com-
bustion restrictions and prevention of air pollution due to fly-
ash. On the other hand, due to the protective policy for coal
industry, some coal fired steam plants are still operating and
some more plants are being planned. Also, the oil crisis which
hit the world in late 1973 had a great impact to our country
(Japan) and the reconsideration of fuel was strongly required.
Therefore, the utilization of coal as fuel to the steam plants
are reviewed seriously.
Because the coal fired boilers produce a great deal of
(10 *> 30 g/Nm3) flyash, it is necessary to install a high effi-
ciency electrostatic precipitator for the prevention of air
pollution. However, the apparent resistivity of this flyash
varies significantly due to the quality of coal this will effect
the performance of electrostatic precipitators. Especially in
the temperature range of 120 'x, 150°C such as at the exit of air
heater (A/H), the apparent characteristic resistance of dust is
so high that is is sometimes hard to maintain a stable performance
This tendency is more obvious for a lower sulphur content in the
coal.
236
-------
Recently, in order to prevent an air pollution due to SOX,
there are very many steam plants, especially in the United States,
which utilize coals with low sulphur content and the high temp-
erature ESP is often considered for this purpose. Namely, the
apparent characteristic resistance of fly ash is almost independent
of the quality of coal and will be the value for a normal operating
range of the ESP in a high temperature range between 350°C and
400°C. Thus, it is possible that the high temperature ESP will
demonstrate a better performance than the low temperature ESP.
Here, we have confirmed a feasibility of the high temperature
electrostatic precipitator by grasping the characteristics of dust
in the coal fired boiler exhaust gases and the characteristics of
precipitating rate. Based on this study, we have performed a
model test at the pilot plant which use the coal fired exhaust
gases. Following is the description of the high temperature ESP
for coal fired boilers.
1. Principle of Electrostatic Precipitator
In order to separate and collect effectively those extremely
tiny flyashes in gas, it is most sufficient to apply the electro-
static precipitator which utilizes the corona discharge. Figure 1
shows the principle of electrostatic precipitator. Normally a
negative high voltage DC is applied to a discharging electrode and
a precipitating electrode is connected to a ground.
237
-------
Figure 1
PRINCIPLE OF ELECTROSTATIC PRECIPITATOR
HIGH VOLTAGE D.C.
DISCHARGING ELECTRODE
IONIZATION ZONE
*<3ASION OR ELECTRON
CHARGED PARTICLES
FALLING PARTICLES DUE TO HAMMERING
238
-------
A discharging electrode uses a wire with a small curvature
radius and negative ions are generated in this vicinity by pro-
ducing a partial insulation break off, then the corona electrical
current will move to the precipitating electrode. Drifting dusts
will be charged by colliding with these negative ions and will
be carried to the precipitating electrode by coulomb force and
be accumulated on its surface.
The migration velocity of dust to the precipitating elec-
trode is, as shown in Equation 1, proportional to the particle
diameter (d) and square of the electrical field strength. Therefore
unlike the case for mechanical precipitator, the precipitating
rate does not decrease drastically for tiny dusts, and even extreme
tiny dusts can be collected with a high precipitating rate in
conjunction with precipitating process by Coulomb force. The
precipitating rate can be expressed by Deutsen's equation
Equation 2).
w - kdE0 (1)
n - 1- exp I-Kwr] (2)
where w » particle migration velocity
d = dust diameter
E0 = electric field strength
k « constant
n = precipitating rate
T • charging time
k » electrode constant
239
-------
Based on these equations, a long charging time and a high
electrical field strength are necessary to improve the precipi-
tating rate. Also, it is necessary to apply appropriate size
and type of discharging electrode and precipitating electrode,
and any past experience was fully incorporated in this aspect.
The characteristics of dusts and gases need to be reviewed in
detail in the following apsects.
2. Dust characteristics and Precipitating Efficiency
There are some factors, such as the concentration, particle
size, precipitation, adhesiveness (adsorption) and apparent
characteristic resistivity of the dusts in boiler exhaust gases,
which will affect the efficiency of EP. These factors vary
significantly depending on the fuel used in the boiler, boiler
firing method and the operating temperature of EP. Table 1
shows the dust characteristics of boiler exhaust gas for the coal-
fired boiler and the crude oil-fired boiler.
2.1 Effect of The Contained Ash Concentrations
The higher the contained ash concentrations are, the lower
the corona electrical current is for a constant charging voltage.
Thus it is necessary to make a charging voltage high in order
to maintain a constant corona electrical current. When the same
corona electrical current is maintained the precipitating rate will
increase accordingly with dust concentrations, because the specific
240
-------
Table 1. Boiler Exhaust Gas Dust Characteristics
Coal fired
Item
Amount of ashes
Average particle size
Dust composition
Si02
A1203
SO,
C
Apparent character-
istic resistivity
Item
g/Nm9
P
fl-cm
17-25
20-30
50-55
27-30
0.3-0.7
0.3-1.0
Ixl012-1:
Crude oil
fired
0.05-0.2
10-12
Ashes 15-20
25-35
50-60
1x10
3-5
241
-------
area of the dust contained per unit gas volume will become larger
and the electrical field strength in the precipitating region
will increase. However, when the electrical field strength
exceeds a certain limit, there will be strong sparks and a stable
charging cannot be maintained.
2.2 Effects of Particle Size and Physical Characteristics
The smaller the particle size is, the smaller the electrical
charges and the migration velocity of the precipitating electrode
are. On the other hand, since smaller particles have more active
movements and different size particles collide with the relative
velocities due to different amount of charges, it is easy to
precipitate them. Actually it is possible to precipitate extremely
small particles such as fume with high efficiency. This precipi-
tating function can be promoted by the particle's moisture
absorbent characteristics. The precipitation characteristics
extensively depend on dust composition and particle size, but
this has not been totally analyzed yet.
A Region Low voltage, large current, stable charge,
low function region
B Region Normal Region
C Region High voltage, low current, decreasing n region
D Region Unstable charge, decreasing n region
B Region Low voltage, large current, stable charge,
low function region
242
-------
Anyway, small particles promote to extend precipitations
and help the electrostatic precipitator improve its efficiency.
On the other hand since tiny particles have strong adsorption
which will cause a problem to adhere to precipitating electrodes
and will be a cause of decreasing efficiency of electrostatic
precipitating, it is necessary to be careful in handling. The
adsorption of dusts decrease with high temperature and the mobility
of dusts increase.
2.3 Dust Apparent Characteristic Resistivity and Behavior of EP
The apparent characteristic resistivity is the most important
factor among those which affect EP performance. Normally the
values of apparent characteristic resistivity with which EP
can maintain a stable performance are, as shown in Figure 2, in
the region of lO^-lO11 fl-cm and it is difficult to precipitate
particles outside of region. In A Region (less than 10* ft-cm) of
Figure 2, the dust whose main composition is carbon such as the
crude oil-fired boiler exhaust gas dust may escape without being
effectively precipitated because of the low apparent characteristic
resistivity causes the reentraining phenomena. In regions C, D,
and E, tiny dusts whose main composition is silica such as fly-ashes,
cement dusts and fine particle emissions from metal refinery will
cause an inverse ionization phenomena (the voltage between the
Surface and the inside of dust layer over the electrode will
Increase, and an electrical breakdown will be created to
neutralize electrons from discharging electrode and the discharging
243
-------
Figure 2. Precipitator Operating Characteristics
As a Function of Particulate Resistivity
102 103 10* 105 106 10? 108 109
244
-------
condition will be tinstable), and the precipitating performance
will be considerably decreased.
Also this apparent characteristic resistivity depends on
a sulphur content of the fuel used in the boiler and a temperature
and it will decrease with higher sulphur content and higher
exhaust gas temperature as shown in Figure 3. Table 2 shows one
example of the apparent characteristic resistivity of coal-fired
boiler exhaust gas dusts.
Table 2. Apparent Characteristic Resistivity
of Coal-Fired Boiler Exhaust Gas Dusts
Temperature
Low temperature region
(120u,130°C)
Air preheater exit
High temperature region
(300~400°C)
Coal saver exit
Apparent EP
Sulphur content characteristic precipitation
in fuel resistivity degree
1.0-2.0%
less than 1% 1012-1013fi-cm
1.0~2.0%
less than 1%
109-.1010J}-cm
Easy
Difficult
Easy
When the EP is used in the low-temperature region, it is
possible to have a normal precipitation for high-sulphur-content
coal dusts. But it is difficult to have a normal precipitation
for low-sulphur-content coal dusts because the apparent
characteristic resistivity is as high as 1012~1013JJ-cm and an
inverse ioni2ation phenomena are created. On the other hand, in
the high-temperature region, since even the low-sulphur-content
coal dusts are affected by the temperature and the apparent
245
-------
Figure 3. Dust Apparent Characteristic
Resistivity and Temperature
•§
*
Ul
cc.
o
CO
£
u
I-
ui
cc
10"
109
S= SULPHUR CONTENT
IN FUEL
I
I
I
100 200 300
TEMPERATURE, °C
400
246
-------
characteristic resistivity is decreased as low as 109~1010ft-cm,
it is possible to have a normal precipitation. Therefore in
high temperature region, it is possible to have a constant pre-
cipitation with high efficiency for any quality of coal.
Thus, as shown in Figure 4, the size of EP due to the sulphur
content of coal varies significantly at 150°C depending on the
sulphur content while it is constant at 320°C. It is therefore
advantageous to apply high temperature EP in order to obtain a
high efficiency constantly regardless with the quality of coals.
3. Characteristics of High Temperature EP
As a purification purpose of coal-fired boiler exhaust
gas, high temperature EP has the following advantages as compared
with ordinary low temperature EP.
1) High performance can be obtained regardless with
quality and sulphur content of coal.
2) It is easy to fall off dusts by hammering and it is
seldom to degrade its performance by dust adsorption
to precipitating electrode and discharging electrode.
3) It has better dust mobility in hopper and has less
trouble with ash stuck.
4) It is possible to keep the A/H in a clean condition
longer than usual and to have less decrease of A/H
performance and require less frequent use of the
sort blower.
247
-------
Figure 4. Sulphur Content in Coal and
and Size of Precipitator
1.0
oc
p
tc
a.
ik
O
III
N
M
0.5
320°C
150°C
I
I
I
0.5 1.0 1.5 2.0 2,5
SULPHUR CONTENT IN COAL, %
248
-------
Including the above characteristics, Table 3 shows a com-
parison of high-temperature EP and low-temperature EP for high
resistivity dusts. As an environmental integrity standpoint and
in order to decrease extensively the fly ashes from the low-sulphur-
content coal-fired generating plant, it is most certain (promising)
to use the high-temperature EP which can maintain high efficiency
without affected by coal quality.
4. Experience of High-Temperature EP
Generally in our country, since the standard limit of fly
ash exhaust with EP was high and domestic coals with relatively
high sulphur content were used, only low-temperature EP was used
in stand alone or in combination with multi-cyclon (MC) and we
have hardly had an experience with high-temperature EP for coal-
fired boilers. But recently in the United States, since the
low-sulphur-content coals have been used more often for larger
volume of boilers and ordinary low temperature can no longer
maintain high performance for these boilers, the application of
high-temperature EP has been widely accepted in order to prevent
the degradation of EP performance for low-sulphur-content coal
and its technology has been well established. According to our
survey, more than 70 units of high-temperature EPs (including
thos under constructions) have been installed. Especially in
the West of the United States where great amounts of low-sulphur-
content coals are obtained, about 60% of constructed generating
plants use high-temperature EPs. Figure 5 presents installations
of high-temperature EPs in the United States (including presently
249
-------
Table 3. Comparison of EP Concept for High Resistive Dusts
Low High
temperature EP temperature EP
Item (140°C) (350°C)
Performance
Real gas amount Base About 1.5 times
Gas viscosity Base About 1.4 times
Dust apparent characteris-
tic resistivity 1011~1013 <10M
Dust moving velocity Small Large
Precipitation performance Pair Excellent
Performance for coal
quality variation Fair Excellent
Re-scattering Large Small
Maintenance
Air heater ash adsorption Large Small
Hopper ash stuck Medium Small
Ash mobility Small Large
Anti-corrision Not necessary Not necessary
Thermal expansion Small Large
Economy
EP volume Small Large
Composing material quality Ordinary steel Ordinary steel
Width for heat insulation 30-50 mm 100-200 mm
250
-------
PENNSYLVANIA NEW JERSEY
t •
(INCLUDING ONES UNDER CONSTRUCTION)
, WYOMING
.
(OKLAHOMA',
CALIFORNIA
NORTH
PACIFIC
OCEAN
NORTH
ATLANTIC
OCEAN
Figure 5. Installations of High Temperature EP for Coal
Fired Generating Plants in the U.S.
(Including ones under construction)
-------
under construction or under planning) . Table 4 is an example
of the operating experiences and as you can see good results have
been achieved.
4.2 Experience in Ordinary Industrial Application
For ordinary industrial plants which create ashes, the ash
compositions and characteristics vary depending on the used main
raw material, the used sub-raw material, kind of fuel and amount
of fuel. The temperature range in which EP is used is fairly
wide and some system (plant) requires to be processed as high as
at 400°C due to its characteristic. There have been some ordinary
industrial high-temperature EPs installed in our country and their
main applications are for exhaust gas purification of such as
cement kiln, metal refinery, and city garbage-burning boiler.
Among these EPs, those EPs such as the former ones require to be
processed at a high temperature in order to return the collected
dusts to production process for a re-utilization of raw materials,
but the main reason for high temperature application is that the
apparent characteristic resistivity is too high at low temperature
and that the applicable region of electrostatic precipitator is
exceeded. Table 5 shows some of these operating experiences.
5. Coal-Fired Boiler High-Temperature EP Model Test at
Ebetsu Generating Plant
The purpose of this model test was to verify precipitation
characteristics and fundamental data for actual system design at
high-temperature regions, and as a result the original objectives
252
-------
Table 4. High Otenferature H> Operating Experiences in the ttiited Stat
A B C
ro
01
u>
ftel (Ooal)
S Content
Ash Content
Beat
(nit No.
Capacity
Gas taxnt
Gas Jeirperature
Entering Ash Anount
Precipitating Bate
Initial Operaticn
0.2 1.2%
6 15%
5,277 kcalAg
1
105 tW
999,260 m'/h
(430,960 tta'/h)
1.5 g/*to'
97.9%
'76.4
2
120 tW
1,138,440 m'/h
(478,145 NnVW
377°C
1.1 g/fcns
97.9%
'75.10
1.3%
18.3%
0.24 0.65%
3.4 22%
6,138 kcalAg 4,555 kcalAg
1
350 Mf
2,948,000 mVh
(1,249,695 Mn'/h)
371 "C
99.5%
Older construction
212
* 447 Mt *
* 5,012,640 mVh *
* (2,079,712 Mn'/h) *
4- 385'C *
1.3 g/tta* »
*• Design Value 99.9% *•
'73 ttider ccnstructicn
0.5% (Average)
7. 9% {Average}
6,000 kcalAg (Average)
1
750 Mf
6,698,000 m'/h
(2,839,000 Nn'/h
371"C
9.2 g/»BJ
99.5%
•74.5
2
4*
4-
4-
4~
4-
4-
•75.3
3
•4-
4-
•4-
4-
-------
Sketches Belated to Table 4
STACK AT CENTER
A/H
EP
I EXHAUST GAS GOES
TO A/H THROUGH EP
EP IS 2 STORIES
A/H IS PLACED BEHIND
EP, AND PDF AND SILO
ARE PLACED FURTHER
BEHIND
STACK
SILO
D
—Q
PDF
PERPENDICULAR PLACEMENT
TO GAS FLOW
254
-------
raoie 5. orainar
Cement Clinker Cooler
Cement Clinker Cooler
Cement Clinker Cooler
Metal Refinery Sulphur
Metal Burning Boner
Metal Refinery Copper
Self Burning Boiler
Sulphate Production Zinc
Metal Burning Boiler
City Garbage Burning
Boiler
City Garbage Burning
Boiler
City Garbage Burning
Boner
City Garbage Burning
Boner
y industrial
Experiences
Gas Amount
775,000
536,300
462,300
89,200
131,800
45,600
81,300
79,000
no, 900
88,000
sign Taupe
in Japan
Gas Tamp.
248
295
237
334
285
319
360
328
302
285
rature EF opei
Precipitat-
ing Rate
99.97
90.4
99.51
99.5
99.6
99.92
98.1
99.7
99.88
99.1
rating
Initial
'73.9
'74.5
•70.9
'68.2
'73.8
'67.3
'72.3
'73.n
'73.3
•72.7
255
-------
have been achieved and the precipitation characteristics have
been grasped as well as various fundamental data have been
verified. The following is the general description of this test.
5.1 Design Specification and Test Condition
Table 6 shows the design specifications for this test.
This test utilized the No. 3 boiler (specifications shown in
Table 7) and coals as fuel (S content 0.2~0.5%) with a normal
burning condition (O2% in exhaust gas and etc.) of boiler, and
we have verified dust characteristics (fly ash amounts, apparent
characteristic resistivity, etc.) as well as precipitation
characteristics due to coal quality. Figure 6 shows the flow-sheet
of the pilot plant and Figure 7 shows its exterior view.
5,2 Test Results
1) Entering Fly Ash Amounts and Precipitating Rate
Even when the entering fly ash amounts varied from 15 g/Nm3
to 35 g/Nm3, the precipitating rate did hardly change. Figure 8
shows the result of measurements.
2) Gas Temperature and Precipitating Rate
When the gas temperature is at 300~365°C, the apparent
characteristic resistivity of dusts is about 108~109fl-cm which
has no effect to precipitating rate. Actually there was hardly
any change in precipitating rate. Figure 9 shows this result of
me as urement.
256
-------
Table 6. Design Specifications
Item
Specification
Process gas amount
Process gas temperature
Entering fly ash amount
Exiting fly ash amount
Precipitating rate
High temperature EP
- Type
- Precipitating electrode
Type
Distance
- Discharging electrode
Type
Size
- Charging facility
Power level
Unit
2,000 Nm3/h
350°C
20 g/Nm3
0.04 g/Nm3
>99.8%
SO-HP12 (steel plate and frame
outside type, horizontal gas
flow 1 chamber 2 sections)
Angular wave type precipitating
electrode plate
300 mm
Frame composing type
4 mm angular
DC 60 Kv DC 60 mA
2 units
257
-------
Table 7. Specifications of Tested Boiler
Item
Power
Boiler type
Ventilation system
Burner system
Burning system
Air preheater
Specification
125 MW
B&W, reheat single body emitter
type
Balance ventilation
Circular burner
Tiny powder coal burning, crude
oil mixed burning
Ljunstrom type
258
-------
Figure 6
FLOW SHEET FOR THE PILOT PLANT
HIGH TEMPERATURE EP
FAN
259
-------
Figure 7. Photograph of Exterior View
(not reproducible)
260
-------
Fi gure 8
2
z
o
Ul
£
>
£
^5
i
rf IUU-
«
99.0-
0.06-
= 0.04-
5 0.02-
| 0-
— a o—
04,500 kcal/kg • COAL (ONLY COAL)
A 6,000 kcal/kg • COAL (ONLY COAL)
^J>~
^ *••»*"**
1 1 1
0 10 20 30 40
AMOUNT OF FLY ASH AT ENTRANCE, g/Nm3
AMOUNT OF FLY ASH AT ENTRANCE
VS. PRECIPITATING RATE
261
-------
Figure 9
Gas Temperature Vs. Precipitating Rate
Ul
u
I
—
Ul
EC
O
CO
E
Ul
CJ
U
I-
z
109
108
ID?
O 4,500 kcal/kg
A 6,000 kcal/kg
• 4,500 kcal/kg
BURNING
A 6,000 kcal/kg
BURNING
x.
0
- COAL (ONLY COAL)
• COAL (ONLY COAL)
• COAL (3% MIXED
RATE WITH OIL)
- COAL (10% MIXED
RATE WITH OIL)
1
100
99.0
UJ
<
cc
u
z
O
100 200 300 400
GAS TEMPERATURE, °C
262
-------
3) Process Gas Amount and Precipitating Rate
In order to verify the relationship between the performance
and the process gas amount which is the important factor to
determine the size of EP, we have done the gas amount variance
test. Its result is shown in Figure 10.
Also, in order to investigate the effect of gas velocity
in the EP to the precipitating performance, we have done tests
by changing duct numbers (passing cross-section) of the pilot
EP. But there was no re-scattering phenomena of dusts due to gas
velocity increases and the expected precipitating performance
was obtained as planned.
4) Charged Voltage and Precipitating Rate
As shown in Equation (1), the relationship between charged
voltage and precipitating rate is that the square of charged
voltage contributes to the performance. The result of this
measurement is shown in Figure 11. It was determined from this
figure that the charged voltage should be more than 35 Kv but
that the charged voltage more than 40 Kv did have little effect
in performance.
Based on these results, it was concluded that the high
temperature EP is the most effective one for coal-fired boiler
exhaust gas fly ashes.
Also, this pilot EP had operated continuously for about
7,000 hours and had always maintained stable performances during
this test period.
263
-------
Figure 10
Process Gas Vs. Precipitating Rate
GAS TEMPERATURE VS. PRECIPITATING RATE
99.99
oc
C9
CL
3
oc
o-
99.9
99.0
O 4,500 kcal/kg - COAL (ONLY COAL)
A 6,000 kcal/kg - COAL (ONLY COAL)
1,000 2,000
PROCESS GAS. Nm3/h
3,000
264
-------
Figure 10
Process Gas Vs. Precipitating Rate
GAS TEMPERATURE VS. PRECIPITATING RATE
99.99
UJ
ec
(D
o
UJ
99,9
99.0
04,500 kcal/kg - COAL (ONLY COAL)
A 6,000 kcal/kg • COAL (ONLY COAL)
1 1
1,000 2,000
PROCESS GAS, Nm3/h
3,000
265
-------
Figure 11
PROCESS GAS VS. PRECIPITATING RATE
100
(9
09
8
£
OS
I
25 30 35 40
CHARGING VOLTAGE, kV
CHARGING VOLTAGE VS. PRECIPITATING RATE
266
-------
6. Problem Area and Counter Plan for High Temperature ESP
Although a high temperature electrostatic precipitator is
very effective for the boiler which utilizes coals of a low sulfur
content, there are following problems when compared with a low
temperature ESP:
(1) Since it is operated at high temperature, real quantity
of gas will be huge.
(2) Placement space including ESP is larger and the duct
work is somewhatv complicated.
(3) Amount of heat diffusion from external surface of ESP
is larger.
(4) Thermal expansion and decrease in material strength
is larger
It is possible to cope with items (1) and (2) by reviewing the
overall placement including boiler in the planning stage. Enough
heat insulation will be required for item (3) . It is also possible
to cope with item (4) based on the design of experienced industrial
high temperature ESP (for cement, metal refinery and normal city
dusts etc.). Table 8 shows a comparison between high temperature
electrostatic precipitators and low temperature electrostatic
precipitators in configuration.
CONCLUSION
This paper has presented a principle of high temperature ESP
and a part of experimental results at the Ebetsu Generating Plant.
This experiment has confirmed the feasibility of designing 350 MW
size high temperature electrostatic precipitators.
We would express our deep appreciation to the following
parties for various assistances and helps in testing the pilot plant:
Mr. Okizaki, Manager, Department of Steam Plant, Hokkaido Electric
Power Co.; Mr. Kobayashi, Plant Manager, Ebetsu Generating Plant;
Mr. Ikemi, Manager, Department of Environmental Technology, Hitachi,
Ltd.; Mr. Arikawa, Hitachi Laboratory; Mr. Kawaike, Manager,
Precipitator Planning, Hitachi Plant Engineering and Construction.
Also we would appreciate all the assistance given by the
Babcock Hitachi Co.
267
-------
Table 8
to
Comparison between High
Item
(1) Steel frame chamber
(2) Pitch between electrodes
(3) Electrode plate
(4) Discharging electrode
(5) Glass tube chamber
(6) Precipitating Electrode
(Plate)
Hammer
(7) Discharging Electrode
(Plate)
Hammer
(8) Position determining beam.
Temperature ESP and Low Temperature ESP in Configuration
High Temperature ESP Low Temperature ESP (Usual)
Separate steel frame (holder) from
precipitating chamber and apply sliding
mechanism.
Apply 300 mm after considering decreases
in spark voltage and strain due to
temperature.
Apply special angular wave type electrode
plate, thermal strain less than 5 mm at
350°C (experimental value).
Use discharging wire of 4mm with frame
type
Glass is made of Alroina and seal air is
put in for anti-stain.
Hammer is set considering expansion of
electrode plate.
Hammer is a vertical shaft type attached
to the discharging frame, and the overall
system is hung from the top.
Steel frame (holder) and
precipitating chamber are
combined.
250 mm
Strain about 10 mm
Same as left
Glass is white ceramic.
Seal air is put in.
No special consideration.
Same as left
Not supported by casing, but placed on the Supported by casing.
electrode (plate) and hung from above.
-------
References
(1) Hashimoto, Taniguchi: "Principle and Application of Electrostatic
Precipitator", Denki Shoin (1965-10)
(2) Society of Electrical Engineering (Electricity) , "Electrical
Technology Special Committee Report of Anti-Pollution" 2nd
Edition No. 45 (1976-9)
269
-------
HIGH TEMPERATURE ELECTROSTATIC PRECIPITATOR
FOR COAL FIRED BOILER
Y. Oataki, K. Ootsuka, and K. Watanabe
(1) Introduction
One of the factors which affect the precipitating efficiency of
the electrostatic precipitators (EP) is a apparent characteristic
resistivity ( p) of dusts. This p varies depending on a dust com-
position and a gas temperature. Especially the fly ash dusts pro-
duced in the coal fired boilers are affected strongly by the sulphur
content in coal and sometimes cannot be precipitated in a stable
manner due to high p at the exit of air heater whose gas temperature
is 130 - 150°C.
Recently, re-evaluation of fuel have been strongly required and
the coal fired boiler is being reconsidered. In this case, the low
sulphur content coals will probably be the main source due to the
standpoint of anti-pollution from SO . Especially in the United States,
Ji
there are very many steam plants which utilize coals with low sulphur
content and the high temperature Electrostatic Precipitators the im-
portant part for processing dusts. This is based on the fact that
p decreases and is independent of the coal quality at high gas tempera-
tures as shown in Figure I, and the precipitation is done within a
high gas temperature region of 350 - 400°C by placing the EP at the
economizer exit of the boiler.
However, because we have not had any experience with high
temperature EP for boilers in our country, we have developed a pilot
plant with a real gas (2000 Nm3/h) and evaluated the operating ef-
ficiency of the high temperature EP.
270
-------
Figure 1. Sulphur Content, Temperature
and Electrical Resistance
10
13
10"
1010
I I
S- SULPHUR CONTENT IN FUEL, wt %
1.0%<8<2.0%
8>2.0%
100 200
QA8 TEMPERATURE, OG
300
400
271
-------
(2) General Description of Pilot Plant
A test was performed by installing experimental apparatuses
shown in Table 1 at the Ebetse Generating Plant Unit No. 3 (125 MW) ,
and the expected performance was experienced. Figure 2 is its flow
sheet. Table 2 shows a composition of coals used for this test.
(3) Description of Operating Results
3-1 Voltage - Current Characteristics
When gas temperature increases, there will be more electrical
current because a relative density 3 of gas decreases and a molecular
movement becomes active.
a = 273 + 20 P (1)
273 + t 760
As a result of this, a spark voltage decreases and it will be dif-
ficult to maintain a high electrical field strength. Also when a
dust concentration becomes high, a total surface area of fly ashes
per a unit gas volume becomes larger and an electrical current will
be restricted. Figure 3 presents a voltage-current characteristic
of gases for air load and at high temperature. Since section 2 has
a lower dust concentration than Section 1, it will be easier to have
an electrical current.
3-2 Gas Temperature and Precipitating Rate
We have changed a gas temperature and investigated a relation-
ship between the apparent characteristic resistivity ( p) of dust and
the precipitating rate. Figure 4 shows its result. p changed from
1.8 x 109 fl-cm (at Tg = 300°C) to 8 x 108 ft-cm (Tg = 365°C) but the
precipitating rate did not almost change. We also have compared the
quality of coal between 4500 kcal/kg and 6000 kcal/kg, there was no
difference due coal quality at a high temperature region. We further
have investigated the case of mixed burning with crude oil but there
was no effect in performance.
272
-------
Table 1. Design Specifications
Items
Gas Flow Rate
Process Gas Temperature
Type
Precipitating Electrode
Type
Gap
Discharge Electrode
Type
Width
Charging System
Power
Unit No.
Spe ci f i cat ions
2,000 Nm3/h
350°C
SO-HP j.2
Angular Wave Type Precipitating
Electrode Plate
300 mm
Frame Composition
Angular Wire
DC 60 kv, 60 ma
(2 units)
273
-------
Figure 2. A Flow Sheet of Pilot Plant
BOILER
(125MW)
A/H
E.P.
HIGH TEMPERATURE E.P.
FAN
274
-------
Table 2. Coal Compositions
Heat 4500 kcalAg 6000 kcal/kg
Sulphur Content 0.43% 0.26%
Ash Content 36.6% 17.6%
275
-------
Figure 3. Voltage-Current Characteristics
t-
Ul
ec
oc
u
40
30
20
10
GAS LOAD
SECTION 1
SECTION
I
20 40
VOLTAGE, kV
276
-------
Figure 4. Gas Temperature Vs. Precipitating Rate
I
tn
tn
ui
oc
u
Ul
O
u
Ul
oc
I
109
108
I _- --J-
V.
\
O 4,500 kcal/kg COAL ONLY
• 4,500 kcal/kg MIXED BURN
A 6,000 kcal/kg COAL ONLY
I
I
100
oc
O
99 5
98
£
100 200 300
GAS TEMPERATURE, °C
400
277
-------
3-3 Performance of High Temperature Electrostatic Precipitator
In order to confirm the relationship between a performance of
EP and a scale factor which is the major factor to determine the
size of EP, we have performed tests by changing the quantities of
process gas. As shown in Figure 5, the result better than the planned
precipitating rate was obtained for the planned gas amount
(2000 Nm3/h). Also as stated before/ it was confirmed that there was
no effect by the quality of coal.
(4) Conclusion
We have grasped the performance of high temperature EP by
using the pilot plant and confirmed that the expected performance
can be obtained regardless of coal quality.
(5) References
1) Hashimoto, Tanignchi: Principle and Practice of Electro-
static Precipitator, Denkishoin
2) Society of Electronics: Anti-Pollution Electrical Technology
Special Committee Report: 2nd Edition Volume 45 (1976-9)
278
-------
Figure 5. Scale Factor Vs. Precipitating Rate
cc
z
s
cc
a.
99.99
99.9
99
90
PLANNED CURVE
I
O 4,500 kcal/kg COAL
A 6,000 kcal/kg COAL
I
2 3
SCALE FACTOR, S
279
-------
ELECTRIC FIELD DISTRIBUTION IN WIDE PLATE
SPACING ELECTROSATIC PRECIPITATOR
T. Misaka, S. Matsubara,
and K. Fujibayaski
This paper discusses the results of an experimental investi-
gation to map the electric field as a function of position for a
variety of corona to collection electrode spacings. The electric
field values were determined by the use of conducting spheres
dropped through a corona discharge into a Faraday cage. The
field distribution measured follows, closely to that expected
from theoretical conditions used in the E.P.A.-SRI computer
systems model.
1) Since it is known that the precipitating rate does not
decrease with expanding the precipitating electrode space in the
electrostatic precipitator (EP), the EPs with wider precipitating
electrode space than ordinary ones are used. This is contrary
to the result of Deutch's equations. We think that a reason
to this is associated with the distribution of electrical field
strength in the EP and have analyzed the relationship between
precipitating electrode space and electrical field strength
distribution.
2) Experimental Apparatus and Experimental Method
In order to measure the electrical field strength in the
EP, we have used the steel ball drop method since it is considered
to be the best method. Figure 1 shows the experimental apparatus.
A flat plate was used as the precipitating electrode and the pre-
cipitating electrode space was made to be changeable from 250 mm
to 750 mm. A 4 mm angular wire was placed in the discharging
electrode with 200 mm space. A measurement was done in the area
indicated in Figure 1.
3) Experimental Results and Review
Since the precipitating electrode spacings are different, a
comparison of the electrical field strength distribution was done
with a same average electrical field strength (Changing voltage
and distance between discharging and precipitating electrodes).
Figure 2 shows the experimental results. In the region of 100 mm
from the discharge electrodes, the electrical field strength has
a same tendency and is independent of precipitating electrode
space. When a distance from the discharging electrodes becomes
larger than that, the electrical field strength will gradually
increase as it approaches the precipitating electrodes. The
electrical field strength in the vicinity of the precipitating
electrodes is about 1.47 times (for 500 mm) and 1.02 times (for
750 mm) that of the precipitating electrode spaces with 250 mm.
280
-------
Figure 1. Experimental Apparatus
TOP VIEW
t
200 mm
SPACE
STEEL BALL
DROP SYSTEM
SIDE VIEW
PRECIPITATING
ELECTRODE (EACH SIDE)
DISCHARGE
ELECTRODE
FARADY
CAGE
ELECTROMETER
281
-------
Figure 2. Electrical Field Strength Distributions
in Precipitating Space 250, 500 and 750 mm
POSITION OP
DISCHARGING ELECTRODE
DISTANCE FROM DISCHARGING
ELECTRODE, mm
320
DISTANCE FROM DISCHARGING
ELECTRODE, mm
iu
PRECIPITATING
ELECTRODE
SPACE
AVERAGE ELECTRICAL
FIELD STRENGTH ,4 kV/cm
DISCHARGING ELECTRODE
SPACE ,200 mm
282
-------
The reason why the electrical field strength increases as it
approaches the precipitating electrode is considered to be due
to a space charge by corona discharge. Also, if the measured
value of the electrical field strength is applied to the Deutch
equation, the precipitating rate is about the same for 500 mm
and is lower for 750 mm as compared with the case of 250 mm
precipitating electrode space.
283
-------
ELIMINATION OF S02 AND NO IN A CORONA DISCHARGE FIELD
Keizoo Ootsuka, Tsugita Yukitake, Makoto Shimoda
Hitachi, Ltd.
Introduction
It is known that in the electrostatic precipitators, nega-
tive ions are produced due to corona discharge, and ion wind and
ozone (Os) will be created. We are analyzing the removal of
sulphurous acid gas (SO2) and nitrogen oxides (mainly nitrogen
mono-oxide) as well as the elimination of ashes using wet
collection type ESP.
Principle
1) Elimination of SO2 - Application of Gas Agitation by Ionic
Wind1'2 promotes contacts between S02 in gas and absorbent
liquid. This turbulence enhances the elimination of S02
according to reactions 1) or 2)
S02 + H20 # H2S03 (1)
S02 + 2NaOH » Na2SO3 + H2O (2)
2) Elimination of NO - Application of Oxidization Process by
Ozone (03)
Since NO is insoluble, NO will be transferred to a soluble
NO2 or N20 5 by applying oxidization by ozone (03) and
eliminated by absorbent liquid. These reactions are shown
as follows:
NO + O3 > NO2 + 02 (3)
2NO + 03 » N20 5 (4)
2NO + H20 » HNO3 + HN02 (5)
N205 + H20 > 2HN03 (6)
284
-------
Experimental Apparatus and Experimental Method
Figure 1 is a flow-sheet of experimental apparatus. A
mixed gas which was arranged to have a similar composition as
oil fired boiler exhaust gas is injected into the wet electrode
type ESP and the concentrations of S02 and NO are measured at
the exit of the ESP. A negative high voltage DC is charged to
the moisture type ESP. The temperature of exhaust gas is 55°C.
Experimental Results
Figure 2 shows one example of sulphur elimination charac-
teristics. Once a discharge is begun, the sulphur elimination
rate is improved by increasing the consumption of electric power.
Figure 3 shows the oxidization characteristics of NO. The
oxidation rate of NO is improved also by increasing the consumption
of electric power. Also it was verified that the oxidized NO
can be absorbed by H2O or NaOH.
Conclusion
It was confirmed that the elimination of S02 and NO is
feasible by using the wet electrode type ESP.
References
1) Uchigasaki and others: The Chemical Engineering 31^ 878 ('67).
2) Adachi: Engineering Laboratory Journal, Yamaguchi University
19, 81 ('67).
285
-------
Figure 1. Experimental Apparatus Flow Sheet
-HV
WET MOISTURE TYPE EP
EXHAUST GAS
SO2 NO
PUMP
ABSORBENT LIQUID
286
-------
Figure 2. S02 Removal Rate aa a Function
of Power Consumption
287
-------
Figure 3. NO Removal Rate as a Function
of Power Consumption
80
70
50
30
300 ppm
5%
10
20
30
40
288
-------
HITACHI EP-SB TYPE ELECTROSTATIC PRECIPITATOR (EP)
1. Principles and Characteristics
Dust sucked from the inlet
EP SB
. receives electric charge at EP.
* • • I N»
^ -I Dust of large particle size
• • T>
will be collected there, while
fine articles with electric
charge are concentrated together
and fed to SB(Shoot Buffle).
SB can collect charged fine particles with high efficiency
making use of a resultant force of electrostatic collecting
force under high electric field formed in two stages and
mechanical collecting force caused by collision of dust
against the shoot buffle.
(HITACHI PATENT 404109)
[Characteristics]
1 The space can be saved by 20 - 25 % to obtain the same
collecting efficiency of EP only.
2 By using a discharge wire for SB to control corona discharge,
power supply for unit processing gas quantity can be
reduced by 40 % as compared to EP only.
3 The shoot buffle with a special form is located vertically
to gas flow, which will hardly cause performance drop due
to re-entrainment.
289
-------
2. Example in Cement Plant
Specifications
Quantity of
processing gas
Inlet gas
contents
Outlet gas
contents
Collecting
efficiency
350,000 m3N/H
18 g/m3N
0.03 g/m3N
99.8 %
Collecting
capacity rate
EP-SB system
i
75 %
EP system
100 %
3. Application
(1) Cement plant
(2) Trash burner
(3) Improvement of performance of the existing
Electrostatic Precipitator (SB only)
290
-------
111-16
MEASUREMENT OF SUSPENDED PARTICULATES
Mesure des particules en suspension
OOTSUKA. K., TSUJI. S. and ARIKAWA. Y.
HitachiResearch Laboratory of Hitachi. Ltd.
Hitachi, Ibaragi. Japan
INTRODUCTION
Emission standards for particulates in a stack gas have
been established to control air pollution. In Japan, the
concentration of particulates must be less than 50 mg/Nm3
for a large scale oil-fired boilers, i.e., the volume of stack
gas of 40,000 Mm1 /h or more.
A standard method for the paniculate measurement is
he dust tube and filter paper method. They are based on
manual gravimetric procedures, making a continuous mea-
surement impossible.
A continuous monitor, therefore, is strongly desired for
controlling the operating conditions of boilers and dust
collectors.
Particulates can be detected continuously and auto-
matically using light beam. The number and size of
particulates can be measured by the scattered tight from the
dust particles, which has been applied to air pollution
monitoring.
A monitor of particulates in a stack gas by the light
scattering method has not been used until now, because it is
difficult to sample a hot and moist gas of high particulate
concentrations.
We have developed a new stack dust monitor, which
continuously measures mass concentration and particle
size distribution of particulates in a stack gas.
I. DESIGN AND DEVELOPMENT OF THE
STACK DUST MONITOR
A schematic diagram of the new stack dust monitor is
shown in Fig. 1. It consists of an iso-kinetic sampler, direct.
1 sample gat
t(rmn)
and diluent sampler, optical system, particle size analyzer
and number-to-weight calculator.
1.1 Optical System
Fig. 2 shows the principle of the forward light scattering
optical system.
With the aid of an incandescent lamp, several pairs of
lenses and a slit, a bright focus (illuminated volume) is
formed. The main beam is intercepted by a light stop and a
light trap. When dust particles in sample gas pass through
sample gas
dean
air
lamp
photo -
multiplier
Exhamr
Fif. I Schematic diatom "* 'tack dust monitor
sensitive volume
— illuminated volume
Fig. 2 Optical system of forward light scattering
this focus, the scattered light in the forward direction is
reached photo-multiplier. The sample gas is surrounded by
a clean air curtain to prevent deposition of the dust
particles on the surface of the lenses.
Yhe output from the photo-multiplier is an electrical
pulse. The number and height of the pulse correspond to
the number and size of particles, respectively. Therefore,
the concentration and size of particulates ar« measured
simultaneously by this instrument.
And then, depending on the results of the analysis of the
particle size with the pulse height, the volume concentra-
tion of mean particle size is obtained for the different size
classes of particles, and thus the total volume concentration
is obtained.
Multiplying this total volume concentration with a
coefficient (mean density of dust particles), the relative
mass concentration of the dust will be obtained.
The relationship between number of particles and
counting errors is calculated theoretically. The counting
error increases with the particle concentration. In order to
obtain S percent or less of the counting error, the particle
concentration must be less than 3.2 x 10* partictes/m*. For
a sample of higher concentrations of particulates, the
291
-------
THE FOURTH INTERNATIONAL CLEAN AIR CONGRESS
sample gas should be diluted with clean air.
In ordej to lead a hot and moist sample gas directly in
the optical system, the protection of the apparatus,
especially the tenses, against temperature rise and mist
formation is necessary.
1.2 Sampling Apparatus
We have developed a new sampling apparatus shown in
Fig. 3. A stack gas is sampled by the ejection effect.
Clean air is supplied to the ejector of the sampling probe
from the outside of the stack. This sampler is set in the
stack.
Only a small fraction of the gas which enters to the
probe is introduced to the optical system, most of the
fraction being returned to the main stream of the stack gas
with the clean air.
The sampling flow rate is proportional to the flow rate
of the clean air supplied to the sampler.
When the mass concentration is less than 10 mg/Nm3,
Ae sample gas is led directly to the optical system (Fig. 3a).
When the mass concentration is higher than 10 mg/Nm1.
the sample gas is diluted with the clean air supplied fo the
sampler, and then led to the optical system (Fig. 3b).
13 Calculation
The calibration of particle size is carried out using
monodisperse polystyrene latex (PSL) aerosols, obtained by
nebulizing aqueous suspensions of uniform latex spheres.
(Dow Chemical Co., Midland, Mich.)
Fig. 4 shows the calibration curve, giving the height of
electrical pulses (e) as a function of the particle diameter
(d). The coefficient is found to be between 0.9 and 1.1, and
the pulse height proportional to the particle size.
jticol
j tytttmj
(stack)
•ample gas
(a) direct
10
- S
>
• I
^
» as
5 001
a.
e«cd
3-9-7
O5
0.5 I 2 5 10 20
Particle size d (pm) (PSL)
Fig. 4 Calibration curve for stack dust monitor
II. APPLICATION
Particles in a stack gas is usually distributed from 1 tun
to 100 f/m and properties of dust are variable in each plant.
This new stack dust monitor can be used for various
kinds of emission sources, such as oil-fired boilers, coal-
fired boilers, kraft recovery and hogged fuel-fired boilers.
incinerators, cement kilns, sintering furnaces, cokes ovens,
ps turbines, glass furnaces, blast furnaces and so on.
Several application data are shown in Fig. 5. A is
coal-fired fly-ash dust, B cement dust. C oil-fired carbon
dust, D coke and E an A12O3 particle. Samples of various
sizes are prepared by a. sieving and sedimentation method.
The nature of these test particles is as follows: A is a
spherical, white particle. B an irregular shaped, gray
particle, C a spherical, black particle, D an irregular shaped,
black particle, and E an irregular shaped, white particle. F is
a spherical, white, standard particle.
From the data, it is clear that there exists an obvious
quantitative relation between the pulse height and the
{optical I
j*ysfem
I I
(stack)
-*omp«t got
(b) diluent tampllng
F%. 3 Sampling app*ntus of ifedt «»
292
-------
THE FOURTH INTERNATIONAL CLEAN AIR CONGRESS
C : • (Carbon)
D * (Coke )
E
160
o 120
I
! 80
7n
•D
* 40
O
(K--8)
carbon dust
5 I
5 10
Particle siie d ( pm )
Fig. 5 Ten d»U vbiih industrial dust pirticlei
.
• ! B
0 40 80 120 160
Standard method
Fij. 6 Relinomhip between «tick dun monitor »nd landud
method
... O
- Opir«iin^T>
- ViM
-:
- -
• •'—
: •
.'.\'.
•-.!-.
"
;
1
.
1:
: I 7 I ; Soot
. i •• I
Burn«r"
._.,,._ k ,.-
^r*y4;t^^r-| -I'-!
Fi». 7 Operituif diu ofiuck dull monitor (Emmion from oil-Cued power boiler)
particle jize, and that the pulse height is very much affected
by the optical properties of the particles. The optical
properties will vary considerably in accordance with the
type of fuel and combustion conditions.
In order to determine the absolute mass concentration
of particulates. calibration curve must be prepared for each
sample by simultaneous measurement with the manual
standard method.
The linear relationship between mass concentration
measured with the stack dust monitor and the filter paper
method is obtained as shown in Fig. 6. The constant. K, is a
conversion coefficient for the absolute mass concentration.
Fig. 7 shows the continuous operating data with a
oil-fired boiler at the outlet of the Electrostatic Ptecipi-
tator. It shows a wide variation of dust particles in the stack
gas with the change of operational conditions of the boiler
It is also shown that when the load of the boiler rises, the
joot slicked at air heater is blown out and the collecting
plates of electrostatic precipitator ate hammered, the dust
concentration increases instantaneously.
293
-------
Hitachi Dust Collection Equipment and Systems
September, 1977
Hitachi, Ltd.
Hitachi Plant Engineering A Construction Co., Ltd.
294
-------
Contents
Page
1 . Preface .......................... . ................. . ......... • .....
2 . Industries and applied dust collection systems
3. Features of Hitachi Electrostatic Precipitatora
4. Wide-Pitch Electrostatic Precipitator ....
5 . Roof Mounted Electrostatic Precipitator
295
-------
1. Preface
Hitachi has been manufacturing dust collection equipment since 1924 and
has a very old history as a manufacturer of electrostatic precipitators in Japan.
Hitachi's supply record of electrostatic precipitators reached to more than
one thousand in the field of thermal power plants, iron industries, non-ferrous
metal refining industries, and so forth, and the total treated gas volume amounted
to more than 150,000,000 Nm /h, including the Japanese record in throughput capaci-
ty of 4,260,000 Nm/h /unit for alQOOMW oil fired thermal power plant.
It is said that design of dust collection equipment, especially electrostat-
ic precipitators, require abundant experiences on the characteristic analysis
of dust, such as particle size, shape, electricalresistivity, etc., and on the nature
of gases. Hitachi, today, can furnish the optimum design and engineering of dust
collection equipment and systems through his experiences.
In addition to such experiences, our researchers have been proceeding with
many improvements on performance from the view point of engineering and economy.
And main theme of Hitachi now are as follows,
l) improvement of dust removal efficiency
2) energy saving
3) cost reduction
Tfcis brochure explains of our recent results in the development of electrostatic
precipitators.
296
-------
2. Industries and applied dust collection systems
Based on its abundant experiences and sophisticated engineering capability,
Hitachi has been manufacturing and furnishing superior dust collection equipment
and systems for various industries.
Table 1. shows the relation between typical industries and type of dust
collection equipment applied.
The following explanations are the outline of dust collection systems for
some typical industries.
l) EP for thermal power plants
Dist collection systems for pulverized coal fired boilers and oil fired
boilers slightly differ from each other.
In the former case, the dry type EP or the combination of the multi-
cyclone and the dry type EP is adopted, because the exhaust gas from the
. ., . . . , . _
boiler contains much dust. The main components of the dust
The dust has sometimes the character of high electric*\resistivity
depending on the kinds of coal. As such high electrioWresistivity nay cause
the phenoma of back ionization. The following technological considerations
to reduce the electricajresistivity are paid;
a. misfiring with high sulfur coal or heavy oil
b. injection of SO into the exhaust gas
c. control of gas temperature
In the latter case, the exhaust gas contains rather less dust which
mainly consists of low electricajresistive free carbon. And the gas also
contains comparatively much sulfur trioxide which sometimes produces corrosive
and adhesive snow fume.
In consideration of such character of the gas and the dust, the dry type
EP adopting ammonia injection is applied.
IMHV
The ammonia injection produces ammoni* sulf&te.which prevents snow
fume and raises the resistivity of dust to optimum value.
2) EP for an iron industry (Fig 2)
Iron industry has various sources of exhaust gases such as, the coke oven,
297
-------
the sintering machine, the blast furnace, the converter, the open hearth
furnace, the electric furnace, the scarfer, and so forth.
The dust contained in these gases mainly consists of iron sulfide.
Character of gas and dust differs from each other, so that most suitable
system should be selected to meet each source. Therefor in planning of
dust collection system for iron industy, much attention should be paid to
gather gas effectively, to prevent corrosion and errosion, to give
effective rapping, to make adequate washing and so forth.
For instance, in case of the blast furnace gas, the venturi scrubber
and the vet type EP are used considering prevention of explosion and high
efficiency f^or the gas from the converter or the open hearth furnace, the
combination of the stabilizer and the wet type EP is used considering the
high resistivity of dust contained.
As an inside-shop dust collection system for the blast furnace, the
converter, the open hearth furnace, and so forth, a bag filter system has
been used hitherto, but recently a roof mounted EPis attracting attention
because of its many advantages like power saving.
3) EP for a cement industry (Fug 3)
In cement industry, there is a lot of gas sources such as raw material
mill dryer, kilns (Lepor method kiln, dry method kiln which includes
suspension pre-heater and so on), clinker cooler and product mill etc.
The exhausted gases usually contains much volume of dust, and its
electrioilresistivity is high.. To prevent back ionization phenomena,
which disturbs dust collecting operation, it is required to reduce the
electrical resistivity of the dust by adding moisture in a stabilizer
installed before EP ox bj treating the gas ia relatively high temperature
condition... Another method to get high dust removal efficiency is to apply
the .constant current, control with thyriator which, contributes .to make
charging stable. .The structure af-.tha.elec.trosta.tic precipitator ia
designed, not to cause, at rain, of electrodes and casing, in..high temperature
condition... And the stabilizer with special spray nozzles is adopted to
reduce drainage.
298
-------
4) EP for aulfuric acid plants (Fig.4)
The operation of iron eulfide calcination and other operations such as
drying, sintering and melting of copper, zinc, lead and so on, produce large
volume of gas which is rich in sulfur oxides. And this gas is used as raw
material to produce sulfusic acid.
Because of high content of dust, this gas is pretreated with a cyclone
and finally fine dust is removed with an electrostatic precipitator.
Mist electrostatic precipitator removes sulfuric acid mist in high efficiency,
which is produced in gas cooling process.
In designing the dust collection systems of sulfuric acid plants,
it is very important to consider the prevention of corrosion and air
leakage, the thermal expansion and the insulation effect etc.
299
-------
Table 1. Industies and Supplied Dust Collection Systems
(tern
Indus trST1
Gas source
Applied system
Remark
Thermal
Power
pulverized coal fired
boiler
heavy oil fired boiler
MC-Dry EP. Dry EP
Dry EP-MC, Dry EP
treatment of high resistive
dust and high content dust.
treatment of fine and low
resistive dust.
Iron
Industry
Blast Furnace (main)
Blast Furnace (inside
-shop gas)
Coke Oven
Sintering machine
Converter(direct gas)
Converter(inside-shop
Electric Furnace
(direct gas )
Electric Furnace
(inside-^shop gas)
Open Hearth Furnace
(direct gas)
Open Hearth Furnace
(inside shop gas)
Hot Scarfer
iold Scarfer
VS-Wet EP
Vet EP, Open BF
Roof EP
Wet EP, SP-Wet EP
Dry EP
ST-Dry EP
ST-Dry EP
closed BF
ST-Dry EP
open BF,Roof EP
ST-Dry EP
Dry EP, Open BF
Roof EP
Wet EP
SP-Wet EP
Closed BF
• Prevention of gas explosion.
• High efficiency gas cleaning
• Advanced gas gathering
technology.
• prevention of CO gas
explosion and adhesion of tar
• treatment of high electric»\
resistive dust
• treatment of fine and high
electrical resistive dust
• advanced gas gathering
technology
• control for variation of gas
volume and temperature.
* prevention of CO gas
explosion.
. advanced gas gathering
technology.
* treatment of fine and high
electlica\resistive dust.
• advanced gas gathering
technology.
* prevention of adhesion of
dust.
gas gathering system for
moving gas source.
300
-------
Industr^N.
Oil &
Chemical
Industry
Cement
&
Ceramic
Industry
Pulp
Industry
Non-ferrous
Metal Ref-
inery &
Sulfuric
Acid
Industry
Gas Source
Fluid Catalytic
Conveter
Oil Gas
Generator
Suspension
Preheater Kiln
Lepor Kiln
Limestone Dryer
Clinker Cooler
Melting Furnace
of Sodium Glass
Carbon Electrod
Calcinating Furnace
Lime Kiln
Soda Recovery
Boiler
Copper Self -melting
Furnace
Copper Converter
Iron Sulfide Ore
Calcinating Furnace
Zinc Ore
Calcinating Furnace
Sulfic Asid Plant
u.
AJjfflina Calcinating
Kiln
Al-Electrolysis
Applied system
Dry EP
Wet EP
ST-Dry EP
Dry EP
Dry EP
MC-Dry EP.Dry EP
ST-Dry EP
Wet EP
Dry EP
Dry EP
Dry EP
Dry EP
Dry EP
Dry EP
Wet EP
MC-Dry EP
Dry EP
Remark
• treatment of hard and high
electricilresistive dust.
• prevention of adhesion of
tar.
• treatment of high electrical
resistive dust.
• prevention of corrosion.
• prevention of corrosion.
• Treatment of high electrical
resistive dust and high
content dust.
• prevention of corrosion and
adhesion of acid dust.
• treatment of adhesive tar
and deposited material.
• treatment of high electric*)
resistive dust.
. prevention of corrosion and
adhesion of tar.
. prevention of adhesion of
dust.
n
II
n
. prevention of corrosion.
. Treatment of high content
dust, prevention of
errosion.
. prevention of burning of
carbon dust and CO gas.
301
-------
^\|tem
Indust^x.
Non-ferrous
Metal Ref-
inery &
Sulfuric
Acid
Industry
Municipal
Gas Source
Al-Smelter
Nickel Kiln
Chromite Kiln
Solid Waste
Incinerator
Applied system
ST-Dry EP, Wet EP
Closed BF
HC-Dry EP, Wet EP
Dry EP
Dry EP, Wet EP
Remark
. treatment of high «±B
electric*|resistive dust .
M
•
n
•
. variety in composition
of gas to be treated.
Note: Dry EP: Dry type Electrostatic Precipitator,
Wet EP: Wey type EP
Roof EP: Roof-mounted EP
Open BP: Open type Bag Filter,
Closed BF: closed type Bag-Filter
MC: Multi Cyclone
VS: Venturi Scrubber
ST: Stabilizer
SP: Spray Tower
302
-------
Hitachi Electrostatic Praclpltators
Fig.1 EP for thermal power plants
Outt Collector (or 000 MW oil flrtd bolltr
Oat volum* 2,510,000m' h
T»mp»r«tur« 14VC
Duit collector (or 220 MW oil flrtd bolltr
Oi* voluma 937,000m' h
T«mp«riturt 141'C
303
-------
Fig. 2 EP for iron industry
Cokes
jCoke oven
'
11 I'M II I larrnist tK
------jD[Ik—-Cp Cgas
Pelletizer
Oust catcher
Stabilizer
dnverter
~
I}"
Open-hearth
furnace Stabilizer
O Cupola
-0- "O-;
EP
B gas
IB—fi-rQ
Fan
EP Fan
: Grounded ore. etc
Pelietizer Klln;
Slag mill
--C 3D—I
IF.
EP
' Raw material
for cement
Fan
i
To Zn
recovery
t
ST-EP dust collector for converter gas
Wet type EP for blast furnace gas
304
-------
Fig.3 EPfor cement industry
Lime stone silo Others
Clay
Raw material
Raw materials
Dry method
EP ™n
Lime stone dryer
Clay dryer
Lepor method
i EP
Product
;_>*,., '
ST- EP for cement industry
305
-------
Fig.4 EP for sulfuric acid plants
turnict Cooler Mitt EP
/Spray tower
Calcinated tine or*
or iron tulfldt
EP for sulfuric acid mist
EP for gas from waste acid concentration plant
306
-------
3. Features of Hitachi Electrostatic Precipitators
flecently Hitachi has carried out remarkable improvement on performance,
manufacturing technology as well as economy. As a result, the cost of a new
model EP became about 83 % of the coventional one.
The typical improvements are outlined below and illustrated in Figure 5.
1) High Performance Collecting Electrode (Pressed Type Electrode)
The collecting electrode is one of most principal elements consisting
an electrostatic precipitator. A newly developed pressafcype electrode gives
higher dust removal efficiency than the coventional electrode manufactured
by rolling does. The outline of the pressed type electrode is illustrated
in Fig. 6.
Moreover, this electrode has feature of small strain at high temperature condition.
2) High Performance Discharging Electrode
Improvement has been given for the frame on which descharging electrodes
are fixed. As the new fframe is constructed by pipes and clamps and there is
technical consideration to reduce strain at high temperature condition,
the new frame contributes to keep the position of discharging electrodes
at the initiated position. Furthermore, another technical consideration to
protect discharging electrodes from breaking.
Fig. 7 shows the outline of the new type frame.
The combination of low strain collecting electrodes and discharging electrodes
assures high dust removal efficiency because the distance between two electrodes
is kept constant. Hitachi successed in reducing the equipment size from ten to
twenty percent through the above improvement on electrodes.
3) Improvement of gas flow uniformity
The outlet dust concentration is regulated severer year by year, and the
requied efficiency of EP becomes higher.
To improve the efficiency of EP, we can not neglect the dicrease in effi-
ciency eaused.by ununiformity of gas flow.
Through earnest research and test we established technology of improving
the uniformity of gas flow in EP.
307
-------
We are introducing some examples of test results.
Figure 8B shows the relation between open area ratio of gas flow distri-
bution plate and deviation of gas velocity.
Figure 6A shows the relation between deviation of gas velocity and dust
collecting efficiency.
Applying these test results, we can improve efficiency of EP resulting
about 10 per cent reduction of size.
4) Saving Energy Type Insulator Chamber.
Since the insulator is one of the most important parts in the EP, it
should keep up the sufficient insulating function under any operation condi-
tions.
Fig. 9 shows the outline of the newly developed saving energy type
insulator chamber which secures the above mentioned function.
As the insulator is tightly covered with a small and light shelter which
isolates the insulator from the atmosphere and prevents from contaminations.
The inside of the insulator is kept clean with the seal air which prevents
the treating gas from coming into the inside, because the seal air is introduced
into the inside then is exhausted into the EP through the specially designed
guard stool. Dewing on the insulator is prevented with a small capacity heater
derectly installed on the insulator. As the seal air flow in this new type
makes 60 per cent decrease and the heater capacity 85 — 90 per cent decrease,
electric power consumption of the whole EP results in 10 — JO per cent decrease.
5) Adoption of Parts Manufacturing with Press H**M
Adoption of pressed manufactured parts such as electrodes, casings, rotary
valves and so on improved the quality and reliability of EP, with the effect of
mass production.
308
-------
Pig. 5 Features of Hitachi Electrostatic Precipitators
U)
o
vo
1.
2.
3.
4.
Pressed Type
Collecting
Electrodes
Clamp Frame Type
Discharging
Electrodes
Improvement of
Gas Flow
Uniformity
Stable High Valtage
because of Less Strain
Less Gas Velocity
Deviation
Higher Dust
Collecting Efficiency
(10 ""• 20# Size Reduction)
Higher Dust
Collecting Efficiency
(Further 10£ Size
Reduction)
Saving Energy
lype
Insulator
Chamber
60 % Decrease of Seal
Air Flow
85 ~.90# Decrease of
Heater Capacity.
»
10 — 30# Decrease of
Electric Power
Consumption
5.
Parts
Manufacturing
with Press
Mass Production
Effect
X.
Higher Reliability
-------
Section of Waved Plate
a
CM
vg
Collecting
Electrode
Fig 6. Outline of High Performance Collecting Electrode
310
-------
T- Shaped Clamp
LisoTnarge Electrode
Frame
Pig 7. Outline of High Performance Electrode
7-
311
-------
u>
H
10
5"
8
£
*->
M
s
o
99
90
0.2 0.4 0.6
Gas Velocity Deviation S
3s
0.6
4 o.4
I
» 0.2
§
20
40
60
80
Open Aif Ratio of Distributer
A. Relation between Gas Velocity
Deviation and Dost Removal Efficiency
B. Relation between Open Area Ratio of Distributer
and Gas Velocity Deviation
Pig 8. Gas flow uniformity and Dust removal efficiency
-------
Shelter
Air How
Seal Air
Guard Stool
Hanging Bod
Cone-Shaped Insulator
Beater
Header Plate
Fig 9. Outline of the Saving Energy Type Insulator Chamber
313
-------
4. Wide-pitch Electrostatic Preoipitator
A wide-pitch electrostatic precipitator (hereinafter refer to EP) stands
for an EP having vide distance between a collecting electrode and a discharge
electrode more than 200Bn. The distance of the conventional EP is about 125 .
In spite of some expected advantages, such as reduction of investment cost,
easy maintenance, reduction of total weight and ao forth, the idea of the wide-
pitch EP had not appeared until Hitachi's development because there was lack of
sufficient theory on collecting mechanism in the wide-pitch EP.
Fundamental study of the driving force observed in a space of EP and charging
character of dust has been carried out with a Schlieren device and some kinds of
pilot plants. The study gave us enough information on the mechanism of dust
collection in the wide-pitch EP to start design of a Hitachi wide-pitch EP.
l) Principle
It is said that dust removal efficiency (1?) is given using the following
Deutsen's equation.
1 = 1 - .
where Ye: migration velocity of dust
t : charging time
p : distance between a collecting electrode and
a discharge electrode
This equation shows the narrower P brings the higher T7 . On the other hand,
our testing results show the increase of the distance P in a certain range gives
rather higher efficiency. The above phenomena could be explained by introducing
the idea of ion wind.
Hitachi considers;
(l) The driving force to move charged dust to the collecting electrode might
consist of not only the Coulomb's force, but also the force of ion wind.
(2) In a narrow-pitch EP, the force of ion wind is too strong to scatter dust
accumulated on the collecting electrode resulting in low efficiency.
314
-------
On the other hand, in the case of a wider-pitch BP having the distance of more than
about GOQ10"1, the contribution of the force of ion wind decreaces.
These data means that is the optimum distance of the two electrode.
It is required to raise applied voltage propartionally to the distance of two
electrodes.
2) Structure
Fig. 10 showes a conceptual figure of the wide-pitch EP in contrast with
the conventional one. The former has only half the total number of the latter's
elements (collecting and discharging electrodes, etc. ), though the former's
size is much the same with the latter's one. On the other hand, the applied
voltage to the former would be as high again as one to the latter.
3) Features
On the performance
(l) CftLgh efficiency is obtainable] even for the gas containing small amount
of dust or fine dust because scat&ing of accumulated dust does not occur
in the wide-pitch EP.
(2) CStable charging is maintainable]) because the effect of strain of
electrodes is smaller than that of the conventional EP even for hot gas.
On the economy
(3) Ccost reduction is obtainable]) because number of collecting electrodes
and discharge electrodes are reduced.
(4) CCost of civil work is reducible]) because the weight of the wide-pitch
EP becomes lighter than that of the conventional EP.
On the running and maintenance fee
(5) In case of the wet type wide-pitch EP, [the consumption of washing
water is reducible!) because the total surface area of the collecting
electrode is small.
(6) [Inspection of the interior is easy]) because the distance of the two
electrodes is vide.
315
-------
60 KV
Power Supply
Collecting electrode Discharge Electrode
120 K\T
Poxer Supply
Collecting electrode
CTl
Gas inlet
Gas inlet
y
Discharge electrode
/ / /
/ s s
/
/ s s
/
/
X /A--J
'
s s
s /
f
Conventional EP
Wide pitch EP
Fig 10. Conceptual Figure of Conventional EP and wide Pitch EP
-------
5. Roof Mounted EP
In iron industry, exhausted gases from converters and electric furnaces are
spread in the shop and are discharged from ventilator.
In order to keep the working environment clean, bag filter system has been
used, which needs large capacity fan and long ducts.
Ve have developed roof mounted EP aiming at saving of electric power consump-
tion. We have two types: one is dry type, and the other is wet type.
The roof mounted EP is based on following technology.
(l) abundant experiences of gas gathering technology
(2) wide pitch EP having high efficiency
l) Principle
The dust containing exhausted gas from the furnace is gathered effectively,
with specially designed hood, and then introduced into the roof mounted EP.
By planning of effective gas gathering system, working environment in
the shop are free from contamination, without using a fan.
2) Structure
Fig.11 shows outside views of dry and wet type EP#, which have a gae
gathering hood at the bottom.
3) Features
(l) Features in Performance
a. Both wet and dry type are applicable depending on the operating
conditions. In the case of low electric^resistive dust or in the
case of installation on the remote roof from the source, we recommend
dry type.
On the other hand, in the case of hi*h electriojresistive dust or
in the case of wet gas, we recommend wet type.
b. Effective gas gathering hoods is combined.
Ascending gas from the source are gathered effectively with hoods
to be introduced into the EP.
c. Wdde pitch EP is adopted.
Adoption of the wide pitch EP enables the reduction of load on the
shop structure and the easiness of maintenance.
317
-------
d. Standardized unite are adopted.
Planning is simplified and Construction is made easy.
(2) Features in Operation and Maintenance
a. Cost of Equipment
The cost of equipment is decreased compared with that of the bag
filter which requires a large capacity fan and long ducts.
b. Cost of Operation
The consumption of electric power is over 70 per cent less than
that of the bag filter.
c. Maintenance
The maintenance of equipment is easy. The roof mounted EP has no
filters, which requires troublesome maintenance.
d. Other Features
The roof mounted EP does not require any space, .on the ground,
resulting in compact layout.
(3) Appliance
The roof mounted EP is applied for following facilities in various
fields of industry, such as iron industry, non-ferrous metal refinery
plants, mechanical industry and so on.
Blast furnace
Converter
Open hearth furnace
Electric furnace
Others (Coke Oven, cement plant, etc.)
318
-------
u>
Dry Type
Wet
Fig 11. Hitachi Roof Mounted EP
-------
Large Capacity and Special Application List of Hitachi Electrostatic Prec'ipitator
NOV, 1, 1977
Hitachi Plant Engineering and Construction
1
2
3
Application
Thermal Electric po-
wer Generation
oil fired boilor
Coal fired "boiler
Cement Industry
Suspension Pre-
heater Kiln
(Dopole Kiln)
Air quenching Coo-
ler
Iron and Steel In-
dustry
L.D Converter
sintering machine
Client
Tokyo Electric
Power Co, Inc.
Kyushu Electric
Power Co, Inc
Tokyo Elictric
Power Co, Inc
Kansai Electric
Power Co, Inc
Osaka Cement
Co, Ltd.
Osaka Cement
Co, Ltd.
Nippon Steel
Corp
Nippon steel
Corp
Site locution
& Plant NO.
Kashima Power Sta-
tion No,6unit
(1000MW)
Buzen Power Sta-
tion No,lunit
( 500MW)
Yokosuka Power
Station No,l,2unit
(265MW)
Amahigashi Power
Ststion N6,l,2uni1
(156MW)
Kochi No, 7 plant
Kochi No, 7 plant
Kuroran
Muroran No , 5
plant
Capacity
NrnVh
2,820,000
1,500,000
845,000
481,000
420,000
400,000
250,000
400,000
Applied
Temp
C
140
145
128
130
170
180
720
80
Outlet dust
Content
g/Nm3
23
12
1.2
0.6
0.05
0.05
0.1
0.1
Collecting
efficiency
%
82
88
96
98.
99.8
99.8
99.7
98.6
Year of
Completion
1973
1977
I960
1963
1973
1973
1969
1969
-------
u>
4
5
6
7
8
ocarfing machine
Nonferrous Metal
Industry
Copper flash
Smelter
Copper Converter
Copper Pyrite Bed
Roaster
Zink fluid Bed
Roaster
Municipal Incinera-
tor
Wide space BP
Petroleum Industry
Pluideric Catalytic
Cracking
Glass Industry
Glass smelting
furnace
Gas Industry
Coke oven
environment
(plate type wet EP]
Sanyo Special
Steel Co, Ltd
Nippon Mining
Co, Ltd
Nippon Mining
Co, Ltd
Nippon Mining
Co, Ltd
Nippon Mining
Co, Ltd
Tokyo
Metropolitan
Office
flaebashi City
Koa Oil Co, Ltd
Nihon Taisan Bin
Kogyo Co, Ltd
Tokyo Gas Co, Ltd
Himeji
Hitachi
Hitachi
Hitachi
Tsuruga
Adachi
Maebashi
Osaka
Oogalci No, 3 plant
Tururai
72,500
99,000
45,000
48,250
34,300
102,000
75,600
289,6000
23,000
300,000
44-66
350
400
350
350
300
250
200
250
40
0.05
0.2
0.2
0.2
0.2
0.03
0.1
0.05
0.02
0.02
98.75
99.2
87.5
94
99
99.4
94.2
75
99
99.3
1972
1972
1972
1972
1968
1977
1976
1973
1974
1969
-------
APPENDIX D
SUMITOMO HEAVY INDUSTRIES, LTD,
322
-------
R.Ep
ROOF-MOUNTED ELECTROSTATIC PRECIPITATOR
1977. Oct.
SUMITOMO HEAVY INIXJSTRIEa UTD.
323
-------
INTRODUCTION
Recently, it is believed that the most important matter in planning
a production facilities is how to make a energy saving program.
Meanwhile as for the dust collecting system of buildings the Bag Filter
Type systems, consuming such a large amount of electric power for forced
suction as used in a blooming mill, have been applied so far in many cases.
However, our Electric Building Dust Collector, which we are going ID explain
now, is an completely different type of building dust collector. Namely, it is
Roof-mounted Electrostatic Precipitator (hereinafter called R-EP.) mounted
a compact and light electric dust collecting system directly on the upper
part of building which collects dust contained gas rising by natural
ventilation. The present machines have been adapted, since we put on sale on
1973, to many equipment such as converters, electric fur nance, blast furnace
pouring places, foundry shops, etc., and are enjoying good reputation.
We believe it will greatly contribute to energy saving of your company.
We wish to explain here general characteristics and matters to be kept in
mind in planning, especially, about R-EP for steel making convetor and a com-
parison to other types of building dust collecting system.
324
-------
1. Workshop dust collection system
The systems and their general appraisals are as shown in the following
table.
Table 1 Merits and demerits of various workshop dust
collection systems
(1) Canopy hood
System
(2) Closed workshop
system
(3) Close workshop
system with
optnlno-cloting
type monitor
(4) Canopy hood and
closed workshop
combined system
(SIR- EP
Opening
closii
type
Opening
closing
By pass
Dember
, A,
(1)
Conopy
hood
system
(2)
Closed
workshop
Merits
(1) The collection efficiency
of dark colored gas is
good.
(2) The working environment
is good by the combind use
of the monitor.
(1) There is a gas storage
effect.
(2) Cost of installation is cheap
(Will dp with a small
amount of trating air
capacity.)
(3) There are hardly any
colored gas leaks.
Demerits
(1) Those which cannot be
collected by the hood
leak into the open air.
(The effect on open air
distrurbance is great.)
(2) Due to installation of a
great remodeling becomes
necessary from the aspect
of structure and strength
of the workshop top.
(3) A great amount of treating
air capacity is necessary.
(1) The gas will originate
an inversion phenomenon
when the balance of the
storage capacity and
the suction capacity is
offset and there lies a
fear of harming the
working environment.
325
-------
(3)
Closed
workshop
system
with
opening-
closing
type
monitor
(4)
Canopy
hood and
closed
work -shop
combained
system
(5)
R-EP
Merits
(i) The gas storage effect
can be utilized.
(2) Since it is of an opening-
closing type of monitor,
the working environment
is good .
(1) The collection efficiency
of dark colored gas is
good and there is also
a strage effect.
(2) Since it is of an opening-
closing type monitor, the
working environment is
good.
(1) Installation areas of
by-pass, etc., become
unnecessary.
(2) Operating and installation
costs are cheaper in com-
parison to the bag-filter,
(3) Pressure loss can be
extremely minimized.
Demerits
(1) Cost of installation is
expensive.
(2) Inspection and maintenance
of the monitor part are
necessary.
(1) Cost of installation is
expensive
(2) Due to installation of
hood, a great remodeling
becomes necessary
from the aspects of struc-
ture and strength of the
workshop top.
(3) Inspection and maintenance
of the monitor part are
necessary.
(1) A problem point exists on
the dust collection ef-
ficiency.
(2) Actual results of this
system for electric arc
funances are presently
very scarce.
In the smoke collectkng system of workshop precipitators, there are
presently 5 systems undertaken.
1) Canopy hood system
This system is a method where a canopy hood is attached to a
position which does not interfere with the crane operation, etc.,
and the generating dusts are instantaneosly suctioned by the
hood. In order to suction and treat the dusts generating from
326
-------
the furnace intactly, a large capacity exhaust blower which cor res-
i
ponds to the momentary maximum value of the generating dusts is
necessary. Since the natural ventilation monitor of the workshop
is left opened, the hot air and steam of the heating device, heated
lumps, etc., within the plant are discharged in their condition from
the monitor. In case the dusts generating from the furnace are of
mass volume and the treating air capacity is small, the dust and
hot air leaking from the hood will be discharged from the exhaust
monitor and it will not be favorable from the point of environ-
mental pollution prevention, however, cases of extremely
deteriorating the environment within the workshop are nil.
2) Closed workshop system
This most widely adopted system stores the dusts generating
from the electric furnace at the top of the closed workshop
and performs gradual exhaustion within a fixed period. Since
the top of the workshop is used as a smoke stop, the capacity of
about 1/2 - 1/3 of the momentary maximum dust volume will be
sufficient as the treating air capacity of the precipitator even
when a great volume of dust is generated.
If the airtightness is perfect, there is no fear of the dust leaking
from the workshop, however, in case the treating air capacity is
too little in quantity, there lies a fear of the dust flowing within
the workshop and deteriorating the working environment.
Particularly, there are heat sources such as ladle, dryer, etc.,
within the closed workshop and when this hot air remains at the
327
-------
top of the workshop, the low temperature dusts will hand over
in the workshop so it becomes necessary to consider extra
treating air capacity. Moreover, in case the storage capacity
becomes insufficient due to the balance offset of the workshop
closed capacity and the treating air capacity of the precipitator,
the dust descending phenomenon will originate and the environ-
ment within the factory will be deteriorated.
Generally, when determing the treating air capacity of this system,
factors of ventilation time, storage capacity, fixed ascending current,
maximum dust generating time and ascending speed are con-
sidered.
3) Closed workshop system with opening-closing type monitor
This is an interim system of the above 1) and 2) systems and in
case there are no generating dusts or in case of dust-free hot air,
natural ventilation is performed by using the workshop monitor and
in case of mass volume dust generation, the workshop monitor is
closed and the stored dust is suctioned and this operation is re-
peated. However, there remain problems on the opening-closing
mechanism, operation and reliability of the monitor. Since this
system is of the monitor opening-closing type, the working en-
vironment is superior in comparison to other systems.
4) Canopy hood and closed, workshop combined system
Similarly as in system 3) above, this is an interim system of the
above 1) and 2) systems and a canopy hood is provided directly
on top of the electric furnace and the opening-closing of the work-
328
-------
shop monitor is made possible. In case the generating dusts are
few, they are auctioned by the canopy hood and natural ventilation
is performed with the workshop monitor left opened and in case of
large volume dust generation, the monitor is closed and suction is
performed from the canopy hood and it is a system where suction
and treatment are made upon temporarily storing the leaked dusts
within the workshop. Similarly as in system 3) above, there re-
main problems such as reliability, intricacy, etc., of the mechanism
in this system also.
5) R-EP
The electrostatic precipitator is an equipment which provides
electric charge to particles within the dusts and separately collects
them to the collecting electrodes and this system is a method in
which this electrostatic precipitator is mounted on top of the work-
shop and dust collection is performed.
In comparison to the bag filter, it does not necessitate installation
spaces of a precipitator, exhaust blower, etc., and the operating
cost is cheap.
329
-------
2. Electrostatic precipitator on the top of workshops
1) Characteristic of general electrostatic precipitators
As well-known, the principle of the electrostatic precipitator is
based upon generating « corona discharge between the discharge
electrode and the collecting electrode and charging the suspended
dusts within the gas with electricity by means of negative corona
discharge, collecting the charged dusts to die collecting electrodes
by die Coulomb's force, releasing die collected dusts by means
of hammering, washing, etc., and collecting at me bottom part and
removing out from die vessel.
a. Electric charge layer
b. Locus of particles
c. lonization sphere
d. Discharge electrode
e. High voltage DC power
source
Fig. 1 Principles of electrostatic precipitator
330
-------
(1) High collecting efficiency
The dusts are almost instantaneously electric charged by
the numerous negative ions, electrons, etc., between the
electrodes and by means of the high voltage power source
which has the electric charging capacity of about 100 each
for those of about 1 micron size and about several ten each
for those of 0.1 micron being applied to the charging and
discharge electrodes, the dusts are powerfully collected to
the collecting electrodes by the Coulomb's force.
In this case, die dust is to move while receiving a resistance
due to viscosity of the gas, however, in comparison to coarse
dust, it is found that the finer the dust, the greater the charge
and since the viscosity resistance is small, they can be ade-
quately collected. Resultantly, since die electrostatic pre-
cipitator is capable collecting coarse dust as well as fine dust,
a high collecting efficiency can be obtained. It is common that
a collecting efficiency of 99.99% is obtained in the wet type and
that of 99.9% is obtained in the dry type.
(2) Operating cost is cheap
The Coulomb's force works on the dust as aforementioned
but since it does not practically work on the gas, only a slight
ionic wind is originated. Since the inner part contains only
die discharge and collecting electrodes and is free from other
obstacles, etc., die gas pressure loss is extremely small and
even when including the perforated gas distribution plate which
331
-------
is provided so that the dust is uniformly distributed flow,
there is only a loss of 10 - 20 mmAq and resultantly, die
capacity of the exhaust blower which suctions the gas can
be made small..
Moreover, electricity consumption amount will vary ac-
cording to applications, however, it is about 0.05 - 0.5
KWH/lOOOm3.
(3) The maintenance fee is cheap
There are hardly any movable parts in the inner part so
when it is used under an ordinary good condition, it is
practically maintenance-free.
1. Inlet conditions
GM
Gat volume
GM temperature
GBS components
Gas pressure
I Dust concentration
I Particle sit* distribution
Outt1 Specific gravity
I Dust components
2. Operating conditions
Electric charged condition
Hammering condition
3. Flue conditions
Gas distribution
Oust distribution
Collection
effedencv
Pig. 2 Performance factorial diagram of electrostatic
precipitate r
332
-------
On die factors which the electrostatic dust collecting action
has, there are many factors which provide influence
to the collection efficiency, however, the following 5 items
are resultantly important.
(1) Control of voltage and current
It lies in how great a voltage is placed, how great a corona
current is flown and how great the Coulomb's foice is made
so it is necessary to provide sufficient care in the controlling
method. There are the following control methods which are
currently employed.
Saturable reactor system
Thyristor system
(2) Invesed ionization phenomenon
This is a phenomenon in which electric discharge is made
from the surface of the dust which has been collected and
accumulated at the collecting electrode in case of a dry type
and as causes, the specific resistance of the dust is dominant.
Generally, it generates at the value of more than 10 - 10^
n - cm and results as shown in the figure below.
ii
ii
Collection eff«iency\ / pig. 3 Relations among me
Current
10* W WW* 10"
Specific electric resistance waive fl cm
333
specific electric
resistance value and
the collection efficiency
and discharge current
-------
The collection efficiency will drop by about 10 - 20% and an
extremely great amount of current will flow and moreover,
the voltage will drop up to 20 - 40% of the normal
voltage.
Since mere are various studies made by many persons on
the elucidation of this phenomenon, details will be omitted
here, however, it is believed that under the present state,
there are no remedy means on the current dry type pre-
cipitator, itself. Therefore, the method of varying the gas
conditions at the precibitator inlet and lowering the resistance
value of the dust, itself, has been employed.
io»
10"
Si
| 5 10"
| 2
II 10"
10"
100
200
300
Toorcj
a. Limestone kiln moisture 10%
b. Generating boiler moisture 3%
c. Sintered dust-proof moisture 3%
d. Sintering machine main exhaust moisture 5%
e. Sintering machine main exhaust moisture 10%
Fig. 4 Relations among the specific electric resistance
value and temperature, moisture, etc.
334
-------
As shown in the above diagram, there are the raising and
lowering of the gas temperature, increase of moisture in die
gas, addition of sulfuric acid, etc. It has recently been des-
cribed that the ultra high-voltage wide pole pitch type, field
screen type, etc., are types in which inversed ionization is
difficult to generate, however, the actual results and details
are unknown.
(3) Re -entrainment of dust
The re-entra inment of dust also originates on the dry type
and in this case, the phenomenon occurs when the electric
resistance of the dust is less than the value of 10* ft - cm.
Under the present condition, there are no progress on the
remedies for this case and it is approximately the same as
mentioned for the preceding item (2).
In case of heavy and special boiler dusts, the addition of
ammonia is performed as the additive into the gas for com-
bining it with 863 in the gas and forming ammonium sulfate
and raising the resistance value up to above 10* ft - cm.
Moreover, on the rapping method of the rapping device, that is,
on the strength, frequency, rapping pieces of collecting
electrodes, etc., the method with the least re-entrainment is
being studied and on the strength too, the actual necessary
gravitational acceleration (g) is being measured and efforts
are exerted for grasping the suitable value.
335
-------
Furthermore, upon considering that re-entrainment is un-
avoidable, there is also the method of providing a damper at
the back of the rapping section and intercepting the gas flow
during the period. The effect of re-entrainment due to rapping
of the front chamber, etc., is descreased by minutely sepa-
rating the gas flowing direction chambers or lowering the
height of the collecting electrode.
(4) Flow regulation of gas and dust
It is necessary to prevent the drift of the gas and dust within
the precipitator. Normally, there is a drift of about 30 - 40%,
however, this has been fairly improved by preparing a model
prior to designing and preliminarily installing it in a suitable
position through testings. Moreover, with the development of
the measuring technology, actual measurement by an actual
machine and labor adjustment have become possible.
In case this drift is great, the effect placed on the efficiency is
great and there are times when the reentrainment increases,
the adhesion amount of dust to the collecting electrode is biased
and the outlet dust concentration increases by about
10 - 20%.
2) Points which have been considered at development
(1) Inversed ionization phenomenon
The electric resistance value of the dust to be handled by the
workshop precipitator is about 109 - 1013 0 - cm and high
so mere lies a danger of originating mis phenomenon.
336
-------
(2) Weight reduction
As measures for obtaining a lightweight body, plasticization
of the collecting electrode, the strength reduction due to re-
moval of the rapping device and removal of the rain entering
prevention roof have been performed and a great weight re-
duction has been made,
(3) Re -entrainment prevention
The prevention of re-entrainment of collected dust since they
are mainly less than 100 y.
(4) Explosion
Since there lies a danger of coal dust and carbon monoxide
explosions, make it an opened type to extent possible and
avoid an airtight structure.
The semi-wet type (intermittent cleaning) electrostatic precipitator
on the top of wrkshops has been developed upon providing considera-
tions on the abovementioned points.
3) Characteristics
(1) Simplification of facilities
By directly mounting the electrostatic precipitator on top of
the workshop, the hood, duct, blower, motor, chimney and
foundation work are not necessary.
(2) Extremely cheap running cost
Treatment by natural outflow of the exhaust gas eliminates
the necessity of a large capacity blower so treatment can be
made by extremely cheap running cost.
337
-------
(3) Purges noises
Noises do not practically originate as s blower is not neces-
sary.
(4) Economizes on installation space
No installation apace la necessary for the precipitator, blower,
etc.
(S) Inspection and maintenance are simple
Since movable parts such as the rapping device, blower, motor,
etc., are few in number, not only are inspection and maintenance
simple but their frequency is less.
(6) Highly stabilized collection efficiency
A high collection efficiency can be maintained as an electrostatic
precipitator Is employed.
(7) Dust removal system is semi-wet type
Removal of dust is done by an intermitted washing system
that to separate the R-EP into many small blocks and wash
dust away block by block.
Washing of 1 block takes only about 10 minutes a day, and
the quantity of water required is only about 600 1/min.
As it is a semi-wet system, there is no tear for back colona
effected even for the dust of high resistivity further, it Is
effective for re-scattering.
In addition, as there is no such equipment as rapping system,
dust valve, screw conveyor, etc., the building is not subject
to a*ny vibration, and maintenance is easy.
330
-------
(8) Natural ventilation type
As R-EP utilizing hot air and building draft, the building shall
not be filled with gas even in case of power failute, thus, there
will be no trouble for ventilation.
4) Structure and function
Construction of R-EP
Water supply unit
Insulator chamber
Spray nozzle
Curtain
Discharge
electrode frame
Silicon rectifier
Detail of Hopper
339
-------
(1) Gas flow
The dusty gas ascends by the heat current and the draft effect
of the workshop, however, when the resistance increases and
the ascending current is obstructed by installing a precipitator
and the dusty gas remains within the workshop, the working
eu/ironment will become unfavorable. There force, it is
necessary to select a brand which has possibly small re-
sistance and is capable of collecting minute particle dust when
installing a precipitator on top of workshops.
This equipment employs the electrostatic precipitator with
a small resistance and which is capable of collecting minute
particles and the hopper and louver possess a flow regulation
effect so that the dusty gas volume uniformly flows within the
electrostatic precipitator.
(2) Shell
By drastically employing special lightweight steels, the
structure of the workshop beams are those which require
minimum reinforcement.
(3) Collecting electrode
Although practically all collecting electrodes of conventional
electrostatic precipitators are of steel plate property, it is
necessary to use material quality with a small specific gravity
for weight reduction.
The temperature of the gas flow within the workshop is
practically all below 80°C and by using conductive synthetic
340
-------
resins with heat resisting and corrosion resisting properties,
the problem of weight reduction is solved.
(4) Discharge electrode
In this equipment, the gas temperature is below 80°C and
the heat distortion is comparatively small and the frame
structure which does not necessitate rapping due to adoption
of the intermittent spraying system proves to be extremely
convenient for installation, centering, adjustment, etc., and
it is also a great merit construction aspect.
(5) Dust shaking systems
In the dust shaking system (rapping system) of dry type
electrostatic precipitators, there are 3 systems; the pneumatic,
electromagnetic and the machine systems.
All systems shake off the dust by means of impacts and
vibrations. In the electrostatic precipitator on the top of
workshops system, however, upon considering reasons such
as maintenance of movable parts being troublesome, it would
be undesirable to apply vibrations to die workshop, the collected
dust is extemely fine, etc., the intermittent spraying system
was employed.
For determining the spraying conditions which indicate die
maximum removal efficiency under minimum water volume
during intermittent spray, experiments were performed on
die* configuration of the spray nozzle, spraying hydraulic pres-
sure, spraying angle, nozzle pitch, etc., and die optimum
spraying condition was discovered.
341
-------
0234566789 10
Spray time (min)
Fig. 6 Variations of suspended solid and pH in waste
water by spray time
(6) Waste water treatment method
The dust collected at the collecting plate is water washed and
removed by intermittent spraying and become slurry. This
slurry passes through the louver hopper and is guided to the
waste water treatment installation by the drainage gutter.
As an example, variations by time of suspended solid and pH
in sprayed waste water in a converter plant is shown in Fig. 6
above.
In can be understood from the chart mat the suspended solid
and pH are extremely great immediately after spray commence-
ment but several minutes later, they have considerably declined.
Thus, the setting of the spray time is facilitated by this curve.
The waste water treatment installation differs according to the
slurry composition, treatment object (Determined by the waste
342
-------
water disposal standard), waste water volume, etc.
An example of the waste water treatment method is shown
in Fig. 7 below. In the diagram, the sprayed waste water is
sent to the settling tank after being lowered up to the pres-
cribed pH at the storage and neutralization dual-purpose
tank and after the clear water in the settling tank is further
adjusted to the stipulated pH, a portion of it is reused as the
intermittent spray liquid and the remainder is discharged as
final effluent. On the other hand, the sludge is sent to the
hydroextractor via the sludge storage tank and here it is
hydroextracted and discharged as cakes.
rH
10
2 Chemical addition
1 Sprayed waste water
3 Storage and neutralization 4 Sedimentiation (Settling)
both-purpose tank
5 Chemical addition
7 To spray equipment
(Recycling)
9 Sludge storage tank
11 Cake
tank
6 Neutralization tank
8 Final effluent
10 Hydroextractor
Fig. 7 Example of waste water treatment
343
-------
5) Consideration points on installation of R -EP
It is necessary to provide care on the following points when planning
the installation of the electrostatic precipitator on the top of work-
shops.
(1) Generating gas volume
Since the ascending current from die high temperature dust
generating source is a heat current which ascends while mixing
the ambient air, it can be considered mat the ascending speed
will differ according to the degree of the gas temperature.
Therefore, it decreases in proportion to the distance from the
high temperature dust generating source and at the tap of the
workshop, it lies in a tendency of becoming relatively uniform.
a) In case to install on an existing building
It is safer to make an plan based on the results of
measurement of discharge gas before hand at the
monitor position of existing building.
b) In case to install on a new building
Find out the theoretical amount of gas to be
generated using the following formulas and
determin the amount based on the resulting value
after including the value of our experience.
344
-------
Amount of gas generated from convenor (Qz)
(Design standard of hood)
Qz » 1.95 Z3/2 x
H1 (Theoretical amount of gas)
„•- if- As (At)4/3
Q = (1.5 - 2.5) qz
(Generated gas volume = EP treated gas volume)
n
As = —T— D_ (Convenor area)
4 0
•^ t = 1400 - 20 (Temperature difference)
Blower
345
-------
Pig. 3 shows the method to explain throretical amount of gas to be generated.
Pig. 3 Diagram of Theoretical volume of gas generated
E 14.000-
£ 13.GOG-
S'
g 12.000
g
•3 ll.OOOh
« 10.000-
| 9,000-
u
« 8,000-
g 7.000
& 6,000-
g 5,000-
4,000
O 3,000
I 2,000
1.000
I
H=30m
H=25
H=20
H=15
H = 10
1
3456
Dia. of heat source Do (m)
8
346
-------
2) Generating dust
(a) Particle size distribution
The suspended particle diameter of the suspended dust within
the workshop will differ according to the ascending speed and
the workshop height.
If the generating dust is a minute particle, it is necessary
to prevent the re-entrainment of collected dust upon taking
into consideration the water spraying frequency of dust
shaking.
(b) Dust generating volume
Since the dust generating volume is related to the dust
generating source and the gas ascending speed, it is neces-
sary to preliminarily perform measurement. In most cases,
however, it seems that it does not matter to presume it
as less than 1 g/Nnv*.
(c) Composition and electric resistance of dust
The gas temperature within the workshop is normally less
than 60°C and the moisture in the gas is approximately the
same as the moisture in the atmosphere under that temperature.
Thefore, the electric resistance of the dust is controlled by
the composition and particle size of the dust.
It would be preferable to select a more economical size
equipment by preliminarily measuring the composition and
electric resistance of the dust.
347
-------
Result of analysis of accumulated dust on a converter building
monitor mesured at the planning of R-EP for converter delivered
by our company are shown in Table No. 1
Quality Test Table of dust 1
1. Analysis Results of constituents
Nimeof
Sample
A Co.
BCo.
CCo.
T.F.-
29.6
45.07
:« ^.9 2
SlOt
H5
H.7l»
6.92
AhOk
3.35
:i.:« o
2Z 1)
T.S
1.1
01 54
1.23
MgO
1.45
132
5.57
Na«0
1.65
0.32
0.5 7
KtO
0.17
0.17
0.16
OaO
22*5
8.91
1 3JJH
ZnO
0.7
1.99
0.5 7
PbO
0.1 5
0.36
0.0 8 6
T.C
4.65
9.93
2.16
NiO
0.02
MnO
4.44
2. Grain size distribution Table 1 - B
••-^^
A tf
B Co.
C Co.
•~l 0/«
1 5 5
4
1 3. 5
1 0 ~- 2 0 ft
1 9. S
1 1
a o 5
2 0~5 0 li
5 4
5 3
4 2. 0
5 Of»~
1 1
3 2
1 4. 0
True specified
gravity
3. 4 9
4. 2 7
3. 9 4
3. Electric resistivity & difficulty in dust collecting Table 1 - C
\
A Co.
B Co.
C Co.
3 0 C
6. 7 8 x l o1 '
5. 3 x 1 0 *
7 3 1 X 1 fllz
6 0 C
6. 8 3 x I 0 ' ~
9.6 X 1 0 5
91 1 > 1 0 ' 2
9 0 TJ
!.» \ * 1 O13
5 >. 1 0 *
1.0 7x 1 O13
Moisture
Atmocphe
//
H
Difficulty in
dust collecting
re Normal
Good
Normal
348
-------
3) Installation place of die electrostatic precipitator
It is not economical to mount an electrostatic precipitator to all
monitors on top of the workshop. It is advisable to install it only
at the part where the dust passes upon considering the diffusion
distance of the smoke from the dust generating source and the
variation width due to the direction of the wind.
Show the general arrangement of R-EP delivered by our company.
349
-------
FJg. 2 General arrangement of buildings dust collecting system for A company
uv
o
7,000
—I—
Converter
36,000
CM
«* y46.000
20X2,500=50,000
1LD
2LD
60,000
3LD
-------
Pig. 3 General arrangement of buildings dust collecting system for B company
63,800
to
Ul
10.000
-------
Fig. 4 General arrangement of buildings dust collecting system for C company
*
in
J:
u>
Ul
ro
f
1
SfcrT
CM|
t
— •
•
12570
6.2356,235
,
1
o
1 20TQHCI
Q
U
f —
l_
12,000
VF
VF
mm
N
J
«=;
/
L+51,850
L +46,050
7
Lance winch room
— ^
Work deck
Converter
pWMIM^B.
23.000
50
21,
5JH
) 1,500
3
REP
21,500 1.
2
REP
5i
y—
— v.
3CV
23,000
Work deck
J—
— v.
2CV
23.
000
90
21
,50
0 1.51
1
REP
s~
— v.
1CV
23,000
» 21,500 750
0
REP
i
L
Metal mixer
23,000
-------
Fig. 5 General arrangement of buildings dust collecting system for D company
2,850
10,000
21,200
7,0007,0007.000
NalR-Ep
Center of
No. 1 converter
21,200
7.0007,0007,000
1
Na2R-EP
Center of
No.2 converter
102,200
-------
(4) Beam strength
It is necessary to investigate beforehand on the beam
strength of the existing workshop when mounting on existing
workshops.
(5) Installation term
In case of newly constructed workshops, it is merely a high
place installation work, however, in case of existing work-
shops, caro should be provided since the installation work
term will differ according to the conditions of location.
(6) Ventilation
As our building dust collector is of an electric dust collector
of natural ventilation type, the pressure loss is very small
as equal to or less than that of the monitor.
For ventilator, we take a) number of ventilation and b) the
value of A/V.
In general, number of ventilation of less than 20 (A/V
1 • 4 x 10"3) is considered as bad
a) Number of ventilation
Amount of gas treated m^/H/building volume
b) A/V
A (Building upper opening area m2)/B (Building volume R)3)
Ventilation data of our delivered R-EP are shown on the
Table - 2.
354
-------
Ventilation condition Table - 2
Delivered
to
A Co.
B Co,
C Co.
DCo
Tide of converter
80T/CHx3L.D.
Converter
250T/CHx3L.D.
Converter
160T/CHx3L.D.
Converter
230T/CHx2
Q-BOP
No. of revolution
inmoiii-
toriaf
1 M
15.6
22.6
—
after
R-EP
intuited
14
15.6
17.1
20
A/V
in
moni-
torini
1.12
X10"'
1.73
X10~'
2.71
X10"1
—
After
R-EP
iniUlfed
2.2
X10~*
2.42
X10"1
5.94
X10"'
3.35
X1Q-'
Ventilation
condition
(visual inspect)
Slightly bad
Normal
Normal
Normal
355
-------
(7) Comparison between designed value and result classified by
delivered place
Design values and actual results of our delivered R-EP
for converter are shown in the Table - 3
Comparison between designed value and actual result
Delivered to
A tt
80Tyt)hX3
B Co.
250X/OhX3
C Co.
160T/DHX3
D Co.
230iyt)hx2
Design
Result
Design
Result
Design
Result
Design
Result
Quantity of
total gas
nt/*
24.000
13,600-28,600
43.800
51,000
30.936
31.902
2 7.0 0 0
27,000
Quantity of
gasper
converter
nt/*>
8.000
4.500~9.500
14.600
17.000
7.734
10.634
1 3,500
13.500
R-EP
flow
rate
m/8
1.77
1.0~2.1
1.78
1.9
as
1.1
1.66
1.66
Inlet dust
contents
f/Hnt
ai
0.2 6 9 (max)
0.4
0.33
0.4
ass
a4
0.3 - 0.8
Mainly 0.4 unda
Outlet
dust
con ten)
9/ttnt
0.02
a047
aos
0.02
0.02
ao2
aos
0.03
under
Remarks
Fitted with
aux-dust
collector
LD
»»
M
Q-BOP
356
-------
6) Comparison in general between
R-EPand Bag Filter
Up to the present, forced suction type dust collectors have mainly
been used, which collects dust by installaling a suction hood at
the upper part of building and a suction on the ground which leads
the gas to the bag filter.
Merits and demerits of this tipe and our R-EP are shown in the
Table - 4 below;
Comparison in general between bag filter and R-EP
No.
1
2
3
Item
Dust collecting
system
Separable
grain
Scope of
Application
(1)
Gas temperature
(2)
Moisture
contents of
gas
(3)
Dust
properties
Bag filter
Dust separation by
collision, contacts
diffusion, and
filtering action
Below 1 u
in case of building
dust collection, less
than 60°C
No problem
No problem for
building dust
collection
Not proper for
hygroscopic gas
R-EP
Separation of dust from
gas by static current in
electrifying grains with
the corona discharge
Below 1 u
Same as left
Same as left
When electric? resistance
is between 106 - 1013 n -cm
no problem
357
-------
No.
4,
Item
Functions of
dust collector
(1)
Dust collecting
capacity
(2)
Pressure
loss
(3)
Removal of
grasped gas
(4)
Water
(5)
Draining
Work
(6)
Hood
(7)
Ventilator
(8)
Pump
Bag filter
Over 99%
Exhaust gas density
is less than 0.03
g/Nm3
Bag filter 150 - 200
mmAq
piping etc. 150 - 250
mmAq
Total 300 - 450
mmAq
Removal of dust
grasped on cloth
filter by filtrage
1 Re-verse washing +
mechanical vibra-
tJon
2 Pulse pressure by
high pressure air
Not required
Not required
Hood at the upper
part of building
ceiling and pipe duct
from the hood to the
bag filter are
necessary
Air pressure
300-450 mmAq
is required
Not necessary
R - EP
Exhaust gas density is
0.03 g/Nm3 - 0.02 g/Nm3
Same level as monitor
Dust adhered to
electrodes are washed
away by water once
a day
Water required for the
above
600 1/min x 10 mln x 4
Unit No x 1 day
Drainings for the above
is required.
Not required but,
piping for water is
required
Not necessary
Ventilation fan for anti-
polution for insulator is
required (2.2KW - 3.7KW)
Pump
600 1/min is required
358
-------
No.
4
5
Item
(9)
Area for
installation
Utility
Bag filter
Instllation area for
bag filter on the
ground Is required
Electricity charge,
exhaust for running
cost are very large
amount
R -EP
R-EP is installed directly
on the building
No exhaust fan is re-
quired, as natural
ventilation .
Power for cottrel,
ventilation fan, and
water pump is about
1/10 - 1/20 of bag filter
7) Running cost
As explained previously, our R-EP is of natural ventilation system.
Thus, no large type suction tan is required as bag-filter type, which
reduces electric power consumption to a large extent or about
1/10 - 1/20 of the latter.
Washing water used for dust removal is also very small amount or
0.4 - 0.6 m3/min in addition, the wasted water can be used again as
circulating water after water treatment, therefor, the water really
required to be supplied is only for replacement for dewatering
cakes carried away and evaporated amount, and it is very small
amount.
Comparison of electric power consumed between our delivered
R-EP for converter and bag-filter in the Table - 5.
359
-------
Comparison of consumed electric power between
R-EP & bag-filter Table - 5
Delivered to
A Co.
B Co.
C Co.
D Co.
Treated
gas volume
m^/min
2 4. 0 0 0
4 3. 8 0 0
3 0. 9 .'{ 6
2 7. 0 0 0
Consumed electric
Bag-filter (assumption)
2. 8 0 0 KW
5. 1 0 0 KW
3, 6 0 0 KW
3. 1 0 0 KW
KW
R-E p
R-EP source 3 2KW
Motor 9 KW
4 1KW
R-EP source l l OKW
Motor 4 OKW
1 5 OKW
R-EP source 8 9 KW
Motor 8 OKW
I 6 9KW
R-EP source 6 1 KW
Motor 2 7KW
8 6KW
Note: Power consumption of bag-filter is assumed as;
A p = 450 mmAq.
8) Prior investigation methods
Since mere are no suitable measures on the measuring method
of exhaust gas from the workshop in JIS (Japanese industrial
Standard), the followings can be listed methods which tentatively
conform to die present situation.
(1) Gas volume
A multi-point anemomaster (anemotherm) is installed at the
monitor discharge port and continuous recording is made
in the relation with operation and the gas volume is decided
360
-------
from the timeelanse variation and the discharge
volume.
Besides this, there are the windmill system and the Pilot tube
system, however, the anemomaster is suited for the discharge
speed (2-3 m/s) from the monitor.
(2) Dust concentration
Since it is not an uniformed dust generation and mere are
many cases where dust generation is made in a short period,
it is considered suitable to measure a high volume air sampler
o
which h;is the suction volume of about 1 - 2m°/mm.
(3) Measuring method of electric resistance value
In this measuring method, there is the method of measuring
the dust collected at the actual site upon taking it back to the
laboratory and the method of performing direct measurement
at the actual site flue. We will indicate below an example of
s
the dust apparent resistivity measuring equipment which is
used at the laboratory.
1 Electric heater
2 Heat retaining water bam
3 Baffle
4 Dust casting port
5 High voltage (-) teminal
20KV
6 Pump
7 Heater
8 Blower
9 (1) Needle-Plate dust collect-
ing compartment housing
10 (2) Thermal refining port
11 Thermometer
12 Thermostat
8
Fig. 9 Dust resistivity measuring equipment
(race track method)
361
-------
^x<2) Imulattd iniulitor
(4) Fr-mt
f
Needle ihapwd
electrode
Plata electrode
Fig. 10 Dust resistivity
measure
Circulate the dusty gas In the dust collecting compartment
housing (1) of Fig. 9 while maintaining the dusty gas at a
desired temperature and humidity and collect and accumulate
the dust on top of the plate electrode by corona discharge of
the needle electrode and calculate the dust resistivity by
measuring the current which flows within.
The gas temperature can be freely maintained from room
temperature up to 300 - 400°C and the humidity can also be
regulated up to the range of 0 - 40% by volume ratio.
Moreover, the dust collecting housing interior can also be
maintained at an atmosphere which is close to the actual con-
dition by imoregnating specific gas from port (2).
1 Guard electrode
2 Dust layer
3 Main electrode
4 Ohm Meter
Pig. 11 Dust layer resistivity measuring method
(Parallel plate electrode method)
362
-------
As shown in the drawing, when the resistance measuring
opposed plate is lowered and closely contacted to the dust
surface which has been collected and accumulated on top
of the plate by the corona discharge between the needle-plate
electrode, and voltage is applied between both electrodes and
the current of the main electrode is measured, the apparent
resistivity of the dust can be computed by the following
formula.
v = i • q • a
q - TTT (° ~cm>
whereas,
V : Voltage placed on dust layer (V)
i : Current density (Reading of ampere meter
— Area of main electrode) (A/cm^)
S, : Thickness of dust layer
363
-------
U.K.P. SIWtYLIST
Umvy ludvutriri, LU.
Mtmb, 1077
Uwpaay
kl.bi- Klr.'l, 1.1*1.
kubi- V»rk»
k'.*.«lllfHI
•III III 111' In. ,1 Id.
I»«.l|l»-Uurlill
Pan fir Hi-l«U
<•»., I. lit.
ILn-li iimbf—lfa*"!* «
1'Uul
Mi'l'l
niiit/prlrr
I.I.IIHX
1 ifi-l ric
f «ll lull 1*
k-lllC SIW., I.U. | -.I...I
kit*** Vuikf ) r.tiivrrli**'
I'.irifir M.- 1^1 1
r..., i.i.:.
| if f gffy-kiti rk «
Ymmtlll K*lt*V«* l«i,t
1.1*1. lilltlluHil Vt»lk
K..IM* Sl.M-t . |.|i|.
iHk.iK.tr" V<»r»»i«
Hnt»aH»lti >!(••• I
I'III-|MM.|| ion
Mi /HKhiMi-Vtii kit
« Kli-rl
N«n«r»iriuriu*i
C«., 1.1*1.
I^rrnf-V.r1l.
R»k« 8t»»l, LU.
blwf*M.H»rk«
i i r<- 1 r i ••
1-ll'lMK-
•
«
«
Mi-rl
iwnvprlri
.
(J-IWj!
Klft-lric
furiwre
m
m
m
m
m
HlMt
furaM*
NiMbrr
(rcvi rum)
•J
•J
I
I
1
1
1"
A
ii
1
2
1
I
a
l
8
Sine
6.a»iiWk'i.iNMu.
. tV.lVl)
1,WiOWalU,i*Ktl
ll,l«MMtllll'*IHM.
(lil«'JMt«L>)
(.,M(HW>TIHII
(l.l,*Lli^Ul)
(l,H>HIW>ll).IM<»|
(lil.t^l'lllil)
i|,(*KPWvVI,(HHI|.
(liU^T'llH))
1C "l»I)
u*. iTWvai.HM)!.
« s
III,IMMIW\^I,IMK)I.
(liU^-.KHl)
7l(MWVlH.IMMH.
((iUIVJKl)
7,IKHIW«M,IKMH.
i a
(r,u-_>7iii)
6|t«NIWi7fMI
IIVi
2MOO>2
2*1(10
>H;O
MVI
K!MI
«VnN)
HNHNItS
DJOUI2
267U
371«i3
5000
24MI
MM
W»0
1)000
10000
T7-
to
M)
(ill
M
M
M
fiO
4iO
Ml
liO
bO
bO
bO
Ml
00
Wl
W
t
90
Ulct Du*il
Uwl
«/**••«
O.SHA
n.i
".I
U.I
«.T
II. T
tt.S
O.t
0.4
«.*
O.fcl
o.>
0.2
O.I
o.l
0.1
0.6
OullH Unxl
UtMl
U/ktaT
o.ow
u.tn
0.01
0.02
0.01
O.lll
ii. oa
O.lll
0.04
0.0)
0.01
0.01
0.02
0.03
O.Ul
0.01
0.01
Ilklr
Inxtullrd
N«r. I97J
Jul. l'l»l
*M«. I'»71
N»v. 1971
H»v l>)7t
Srpl. I071l
Jwi. I'lTJ
Mm-. I'lTI'
Nxr. 1
Brc. 197<>
Scpl. |i)75
IW. 1«I71
M»v 1976
H«r 1'iTb
Apr. I'I7<>
IW. 1-177
K«r. 1978
364
-------
APPENDIX E
NIPPON STEEL-KIMITSU WORKS
(no papers)
365
-------
APPENDIX f
SHINHA TRADING AND ENGINEERING CO,
(no papers)
366
-------
APPENDIX G
ISOGO POWER STATION
(no papers)
367
-------
APPENDIX H
KYOTO UNIVERSITY
368
-------
DYNAMICS OF NATURALLY COOLED HOT
GAS DUCT
KAZUYUKI HOTTA, NOBORU YAMAMURO
AND KOICHI IINOYA
Department of Chemical Engineering, Kyoto University, Kyoto
(Reprinted with permission)
Reprinted from
JOURNAL OF
CHEMICAL ENGINEERING
OF
JAPAN
Vol. 7, No. 6 (1974)
Page* 455—461
369
-------
DYNAMICS OF NATURALLY COOLED HOT
GAS DUCT*
KAZUYUKI HOTTA, NOBORU YAMAMURO**
AND KCHCHl IINOYA
Department of Chemical Engineering. Kyoto University, Kyoto
characteristic* of a pnhewbe teatpsratare peak aloag a aatarally cooled hot gas
ladled exyernaaatally with possible applications to bag BUsr systems fa mind. In bee of
fa gas pfopsrtiss, dM to the temperatore change, a Uaear node! Is fonad nseftri for
< prediction of tbs tnuMleat Charts for asdmattag tbs peak height of the outlet
gas teapsratare an couUliad to bs readily assd by dialp and operaHoa laihirm.
In bag-filter operations for higher-temperature
gases, the gates must be properly cooled prior to
filtration. For this purpose, a spray tower or other
positive means may be inserted in the duct system
which leads the gas from the dust source to the
filter. It is, however, possible to expect a con-
~RMdvod on November 2S, 1973
Prorated at the 38th Annual Meetlnf of The Soc. of Chem.
Entn., Japan, April 4, 1973
•• The JapanoM G«on Co., Ltd.
TOW fl.«m*K* 13 *• 8 Tl
siderable natural cooling along the duct if the source
and the filter are separated by some distance. In
this latter case, prediction of the temperature drop is
not so simple as in the case of incompressible fluid,
since the physical properties of the gas change with
the temperature along the duct. Moreover, the gas
temperature of the source often undergoes pulse-wise
changes like those shown in Fig. 1.
The particular example shown is taken from an
automatic record of the air temperature from a
furnace plant in batch-wise operation.
Formulae for predicting the damping charac-
teristics of the temperature peak are not yet found in
VOL. 7 NO. 4 1*74
370
-------
—.ZOO'c
15C
thr.
Table 1 Data used for estimation of statics
I I ._ JKO
«— time
Fig. 1 Typical example of temperature chaage
at dust sources
HI
heater
S blower
Fig. 2 Laboratory test model
standard handbooks. The system is naturally de-
signed amply on the safe side, resulting in wasteful
redundancy.
To supply the designers of such systems with some
practical information, experimental as well as theo-
retical investigations were made and the results are
presented in this paper. Even though the study was
motivated by the problem in dust-collecting systems,
we treated the problem as that of naturally cooled ducts
in general without remarkable pressure drop along
them. The effect of solid particles is not explicit,
since it has an effect only on the inside film coefficient.
Nor did we use dust-laden gas in the laboratory ex-
periments. We assume here that the values of the
heat transfer coefficients in the equations derived in
this report are known from different sources.
1, Experiments
Experimental data were obtained through two
different sources, one from a laboratory test model
sketched in Fig. 2, the other from a full-scale industrial
duct system for a bag filter system.
In the laboratory test model, the air flow rate was
automatically controlled and the heater Hi was used
to set up a steady state. In dynamic test, this steady
state was upset by adding another preheated heater
HI to the pipe line and also boosting the power
supply to the heater H,. The gas temperature was
measured by thermocouples with quick response at
the inlet TR1 and the outlet TR2 of the test section.
The outlet wall temperature was measured by pasting
two identical sheet-thermocouples TR3 on the outer
surface of the pipe at equal distance of about 5 cm
up- and downstream from the outlet point This
Mass flow
rate
[kg/m'-min]
15.53
15.57
15.49
15.51
15.47
15.52
ambient
air
302.0
302.0
301.5
301.0
302.4
298.8
l emperai
inlet
409.5
395.0
409.0
362.1
407.0
371.6
ure i *.]
outlet
348.6
343.3
347.1
329.5
347.8
331.5
outlet
wall
334.7
332.3
334.3
321.5
337.2
324.0
was done to avoid erroneous measurements due to the
cooling effect of the nozzle tapped for gas temperature
measurement. The mean value of the outputs of the
two was adopted as the wall temperature at the
outlet. Random breeze was generated by three
fans to keep the average ambient condition constant.
The industrial system tested had a diameter of
582mm with 3.2mm wall thickness. The test sec-
tion was 50.4m long and the mass flow rate was
around 8.8 kg/m'-sec. For this system, most of the
data were obtained from the automatic records of the
control systems.
2. Statics
To estimate the static characteristics of the labo-
ratory test model, numerous sets of steady-state
data, with varying inlet temperature and air flow rate,
were obtained. However, we show in the following
only the results obtained from several sets of data
for which the air flow rate is around 15.5 kg/min-m1,
since the dynamic test was performed with this flow
rate. The data used are shown in Table 1. As the
static model of the system, the following simplified
equations were used.
-r.)=o (i)
-r.) (2)
As to the notations, the list at the end must be
referred to. The underlying assumptions are obvious
from the forms of the equations.
[case 1] As the first step, both the inside and outside
film coefficient ht and />. were assumed to be constant
and their values were sought by a computer search
program so as to minimize the differences between
the calculated and observed values of the outlet gas
and wall temperature 7, and Tw. The criterion
function used in the search program was
where J7« and Jrw( are the differences mentioned
above in the /-th run of the experiments.
The converged values for ht and A. are shown in
the first row of Table 2, together with the value of 0.
In the model calculation for this case, the specific
heat of the gas was estimated by Eq. (5) at the mean
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
371
-------
case 1
case 2
case 3
case 4
Hi [kcal/'C-m*-se
0.020253
0.02054 (r/r.i)« »
0.019930
0.019930 (TIT.if
Table 2
c]
a
m
Al, instead of the constant A<,
was sought and the results are shown in Table 2.
[case 3] Thirdly, A, was set back to constant and
the constancy restriction on A. was relaxed. To the
value of A., natural convection, forced convection
and radiation were supposed to contribute. But the
attempt to fit a phenomenological equation for A.
was abandoned after some trials, because inconsistent
results, such as the blackness of radiation beyond
unity, were obtained by the computer search program.
Prediction of the film coefficient being rather trivial
for the purpose of this study, the following empirical
equation was adopted.
The search program converged to the optimal values
shown in the third row of Table 2. The drastic im-
provement of the value of the criterion function f>
over the previous cases indicates the controlling
influence of the nonlinearity of A. on the system
statics.
[case 4] Finally, both h, and A. were allowed to
change in accordance with Eq. (6) and Eq. (7),
respectively. The results obtained are listed in Table
2.
The results of the static experiments are summarized
in Fig. 3. Here, the curve showing the relation
between T and Tf is obtained by solving Eq. (2) for
the case where 7. is equal to 300°K (as was approxi-
VOL. 7 NO. 6 1974
Mil
O020
3M 400
gas temperature
Fig. 3 Statics of laboratory test model
mately the case in experiments) and Eq. (6) and Eq. (7)
are inserted for A, and ht.
3. Dynamics
To investigate the dynamic dependency of the outlet
temperature on the inlet temperature, pulse testing
was used. The input pulses generated as depicted
in section 1 of this paper were by no means similar
in shape to those observed in practical systems, but
were found good enough to stimulate the dynamics
of the laboratory test model. A typical input and
the response at the outlet of the test section are shown
in Fig. 4 and Fig. 5.
In spite of the finding in the static study that the
nonlinearity effect is significant, Fourier transforms of
the input and output were tried to obtain approximate
linear dynamics in the form of Bode diagrams, an
example of which is shown in Fig. 6. The striking
similarity of this diagram to those of heat exchangers
for incompressible fluid, where nonlinearity plays
but a minor role, suggested that the dynamics of the
present system could be approximated by a linear
model, at least for a small input change.
Thus a linear model was developed in a manner
described in section 4 of this paper. First, the
parameters of the model were estimated based on
Case 1 of the static study and obtained the model
frequency response shown in Fig. 6. The agreement
with the observed response is only fairly good. But
by more elaborate choice of parameters, which will
be described later, the model behavior improves.
Examples are shown in time domain in Fig. 1
372
-------
100
no
lime (»»0
Fig. 4 Typical test input
Fig. 5 Response of outlet temperature to the Input of
Fig. 4
0001
angular frequency Crad/sec)
001 01.
-90
Fig. 6 Bode diagram of laboratory test model at
G=15.5 kg/mln-m"
Furthermore, the same type of linear model was
formed for the industrial system and the response is
compared in Fig. 7 with the observed response. Due
to the rather complex geometry of the system and
lack of exact static data, agreement is not as good
as that of the test model, but seems good enough for
most of design and operation planning purposes.
This becomes more convincing when one considers
the enormous time and memory space needed to solve
a nonlinear partial differential equation mode!.
time
Fig. 7 Example of transient response of an industrial
system and its linear model
timer,-)
Fig. 8 A normalized step response of G (s)
4. Derivation of Linear Model
Here we treat the gas as if it were an incompressible
fluid with constant physical properties. Then the
discussion can be started with the following simplified
equations.
- T.)
(8)
(9)
r.=l-r«
A set of dimensionless variables and parameters are
defined as follows:
T=t/(LP/G)
(10)
T.=(GILP)CJ(A(ht+AJi.)
Then Eqs. (8) and (9) reduce to
a-3T&-(airxr-Tu) (ii)
(12)
The transfer function between inlet and outlet tem-
perature is obtained" from these equations as
J7i(j)= exp [—is+a+h.s/(\+Tms)}]
(13)
JOURNAL OF CHEMICAL ENGINEERING OP JAPAN
373
-------
where X=*arjrt.
A normalized step response of G(s) is schematically
drawn in Fig. 8 and the mean delay time Tm for
G(s) is, as is easily found,
The first term of Tm corresponds to a pure delay of
one residence time of the gas. The second term ac-
counts for the lag due to the heat capacity of the
duct wall. When the inside fluid is a gas (liquid), the
second (first) term usually dominates the other. In
the industrial system investigated, the second term
is approximately equal to 180 and in the laboratory
test model it reaches as high as 1200, making it
possible to neglect the first term without any loss of
accuracy.
It then becomes important to know the charac-
teristics of the transfer function
GJ?) A exp[-;r.j/(l+r.5)] (15)
which forms the essential part of the system dynamics.
It is, however, not easy to calculate the response of
this function to an arbitrary input. So an attempt
was made to derive an approximate ordinary dif-
ferential equation, which behaves similarly to Eq.( 15),
and enables us to use a computer routine such as
the RK.G method to calculate the response.
Instead of handling Eq. (15) directly, another
related function H.(s) defined by
is introduced. Meaning of this separation of G,(s)
into two parts is obvious from Fig. 8. Checked in
terms of the frequency response, a first-order lag was
not satisfactory in accuracy to simulate #„($), so the
following function was adopted.
Hm(is)=H'f(s)=K(l +r.s)/(l +vs+us*) (17)
The unknown parameters in H'w(s) were fixed so as
that the following values coincide between Hw(s)
and H'w(s).
i) Initial slope of the step response.
ii) Up to second moments of the impulse response.
Introduction of the first condition is of advantage
over the ordinary moment method in two respects.
First it gives improved accuracy in the higher frequency
range. Secondly, we can obtain explicit solutions
for the unknown parameters as shown in Eq. (18),
since the use of third moment, which inevitably
introduces a cubic equation, is avoided.
The solutions thus obtained are:
(18)
Then
C.(j)
where
09)
Letting ATtl(t) denote the input change
delayed in dimensionless time by unity or 4r
-------
c
I
angjUr trtquwiey
ifc-u in, tit.
Fig. 9 Schematics! gain diagram of linear model
value is fairly reasonable, the value of /i< is by no
means in accordance with the results of the static
experiments. For a mean temperature (=106.8°C)
of the inlet, ht should be around 0.02002, so the above
value is some 6% higher. We suppose this was
caused by the added turbulence .due to valve operation
to produce the temperature pulse in the dynamic test.
Difficulty associated with the wall temperature
measurements may be partly responsible for this dis-
crepancy.
Furthermore, mathematical models generally tend
to show quicker response than the real counterparts,
since many minor factors causing delay, such as
finite rate of heat conduction across the duct wall
and mixing of the gas in axial direction, are neglected
in model formulation.
In the present application of the linear model,
however, a slight overestimation of the peak height
is certainly tolerable. So we propose here using
the peak temperature of the input pulse in the
estimation of the parameters of the linear model.
A computer calculation based on Eq. (21) will
yield the response to a input AT*i of arbitrary pulse
shape. However, in the situation where use of a
computer is to be avoided, the peak height of the
outlet temperature can be estimated based on the
frequency response of the linear model. The gain
diagram shown in Fig. 4 is essentially Z-shaped and
can be schematized as Fig. 9. It is then expected
that, if the angular frequencies of the major com-
ponents of the input are less than l/(erH), the ratio of
the output peak height to that of the input is ap-
proximately e~"(^fifm»»). As the frequency range
of the input components extends beyond
-------
» 20 S>
a/t. (-)
c,
D
C
0»x
r.l.
A
K
Ar«
L
*
/i
Pi,ft
4
r
i
T
T«
7*jr
Tm
I
*
(a) triangular input
Fig. 10 Diagrams fore**
[kcalTC-kfJ
= specific heat at constant pressure
— diameter of circular duct
= mass flow rate of gas
= *- H
= «-<«+!) [_)
= film coefficient of heat transfer (kcal/<>C-m1-secl
«- a constant in Eq. (17) [—]
= a constant in Eq. (7)
- total length of duct [m]
«= a constant in Eq. (7)
= peak height ["Q
— constants defined in Eq. (19)
** heat resistance rc-sec/kcal]
«= resistance ratio defined in Eq. (10) (—J
- Laplace transform parameter H/secJ
«• gas temperature (°K]
= ambient temperature [*K]
» representative temperature of gas [C°]
= wall temperature (°K]
• time [sec]
= constant parameters in Eq. (17) (sec*)
= constant parameters in Eq. (17) [sec]
= flow rate of gas (kg/*ec]
— distance from inlet of duct [m]
or
S
1
t
1
P
Tf
Tw
o
2
st
at
o-z
(b) square input
t for triaagabr an* sqare bavt
= defined in Eq. (10)
•= duration of input pluse (refer to Fig. 1(9
= defined in Eq. (22)
= conductivity of heat for gas [kcal/°C-m-sec]
- arjn t—1
- viscosity of gas [kg/m-sec)
- xlL M
=- density of gas [kg/ufl
- time constant in Eq. (17) l«ec)
- wall time constant defined in Eq. (10) [sec]
— criterion function
— change from a steady state
- inner surface of duct
" outer surface of duct
- inlet
- outlet
— standard condition
Uteratore Ctttd
1) Bankson C. A. and D. M. McEligot: Int. /. Heat A Mau
Trans., 13, 319 (1970)
2) Hotta, K. and M. Imaeda: Kagaku KOgaku, 29,980 (1965)
VOL. 7 NO. 4 1974
376
-------
PARTICLE SIZE CLASSIFICATION BY DEPOSITION ANGLE
IN A GAS CENTRIFUGE AT REDUCED PRESSURE
ZENNOSUKE TANAKA, HIROAKI TAKAP, NORISHlGE OKADA
AND KOICHIIINOYA
Department of Chemical Engineering, Kyoto University, Kyoto, Japan
(Reprinted with permission)
Reprinted from
JOURNAL OF
CHEMICAL ENGINEERING
OF
JAPAN
Vol. 4, No. 2 (1971)
Pages 167—171
THE SOCIETY OF CHEMICAL ENGINEERS, JAPAN
377
-------
PARTICLE SIZE CLASSIFICATION BY DEPOSITION ANGLE
IN A GAS CENTRIFUGE AT REDUCED PRESSURE*
ZENNOSUKI TANAKA, HIROAKI TAKAI**, NORISHIOI OKADA***
AND KOICHIIINOYA
Dtpattmnt of Chimical Engumring, Kyoto Univtrsity, Kyoto, Japan
Centrifugal particle size classification by a gas centrifuge which produces a forced vortex
wot Investigated at reduced preuures. In this method, different particle trajectories cause
the formation of a continuous gradation of particle size on the rotor wall. Here, particles
In the subsleve and submlcron ranges 'were classified with good resolution because the 'Cun-
ningham correction' Increases as pressure decreases. Also, the computed solutions gave ex-
cellent agreement with the experimental results. The numerical solutions with and without
the Integral term for non-uniform motion have been compared, the approximate equation of
best fit for the drag coefficient of spherical particles being used In both cases.
Introduction
The investigation of particle size classification
based on the difference of deposition angle on the rotor
wall of a gas centrifuge under forced vortex condi-
tions at atmospheric pressure has been reported by
Kriebel*>, and Burson et al.1), but no study at reduced
pressures has yet been reported.
In this method, the centrifuge produces centrif-
ugal force fields within a forced vortex in the classi-
fication chamber. The particles travel from near
the center to the rotor wall. In the classification
chamber, the particle motion is delayed by Coriolis'
force opposite to the direction of rotation, causing
a gradation of particle size on the rotor wall. At
reduced pressure, the mean free path of the gas mole-
cules is of the, sanfe order of magnitude as, or greater
than, the particle size. In that case, the effect of
the slip factor in the 'Cunningham correction' is sig-
nificant and the fine particles in the subsieve and
submicron ranges can be classified with good resolu-
tion.
A comparison of experimental results with calcu-
lated results is presented for experiments at various
pressures using glass beads, zinc powder and tung-
sten powder.
Experimental Apparatus and Procedure
A sketch of the centrifuge and associated equip-
ment used to investigate particle size classification
is shown in Fig. 1. As the centrifuge has an oil seal
' Received on November 17, 1970
Presented at the 36th Annual Meeting of the Soc.
Chetn. Engrs., Japan at Tokyo, April 3, 1971
** Yamanouchi Pharmaceutical Co., Ltd., Tokyo
*** Toray Industries Inc., Otsu
on the shaft to make it air-tight, it can be used at
reduced pressures.
As shown in Fig. 2, the rotor is made of high-
strength duralumin and the classification chamber is
19 cm inside diameter and 2 cm high. Two webs
are installed in the rotor to ensure forced vortex
conditions. The particle inlets to the classification
chamber are mounted on the rotor, and rotate at the
same angular velocity.
The powder is charged in an acrylic resin tube
fitted with a screen of 325Jf or lOOOtf mesh on the
bottom for dispersion. As shown in Fig. 3, this tube
is placed inside a glass tube with a rubber plug to
make it air-tight.
After the centrifuge attains the specified pressure
and speed of rotation, a 60Hz vibration is applied
to the feeding device to make the particles disperse,
pass through the capillary and enter the classification
chamber. Hence at the particle inlets, the particles
attain the same angular velocity as the rotor.
In the classification chamber, the particles move
from near the center (r0 = 1.19 cm) to the rotor wall
(r = 9.5 cm). On the way, the particles are class-
ified and deposit on the rotor wall in accordance with
particle size. As shown in Fig. 2, the motion of the
particles relative to the rotor is opposite to the rota-
tion of the rotor. The deposition angle, which is
the difference between the angular displacement of
the rotor and the particle on the rotor wall, that is
( — 0)|r.|.i, increases with increasing particle size.
12 mm transparent double-sided adhesive tapes
are mounted on 15 mm transparent plastic strips
and placed on the rotor wall. In the case of sub-
micron particles, sheet meshes for the electron mi-
croscope are mounted on these strips.
After centrifuging, the particle .sizes and the an-
gular location are measured using an optical or elec-
V
-------
1 Centrifuge 4 Recorder for vacuum
2 Rotor 5 Stroboscope
3 Vacuum gauge 6 Vacuum pumo
Fig. 1 Scheme of the experimental apparatus
1
DtpeiltUn angl*
Fig. 2 Schematic deilgn of the rotor
tron microscope.
Glass beads (p, — 2.5g/cmf), zinc powder (p, =
7.0g/cml), and tungsten powder (pr = 19.2g/cm')
were used as the test materials. Hollow particles
in the glass beads were removed by flotation with
a mixture of acetone and 1, 1,2,2-tetrabromo-
ethane.
Theoretical Considerations
The particle motion in the centrifugal fields is al-
ways changing in both speed and direction, that is,
the motion is non-uniform and curvilinear. The
RubDir plug
w-Cliss
"Acrylic ritln
Screen nesh
(3251 or 10001)
To rotor
Fig. 3 Feeding device for
powder
equations of motion for a spherical particle in polar
coordinates are given by the following expression1-".
" * D'0vr(d0 -d*
; 8 Dtt>'vr\dt dt
in which
Ct =
0. 653 X 10-« X 760[ P
(5)
(6)
[cm] for air (7)
These equations contain the slip factor 'Cunning-
ham correction'4-7) for the increase in mean free path
of gas molecule and the approximate equation**1 for
the drag coefficient of spherical particles.
The simultaneous integro-difTerential equations
(1) and (2) are non-linear, so no analytical solution
can be obtained. Thus numerical techniques using
a digital computer were employed instead to obtain
the solutions of Eqs.(l) and (2).
Let us now compare the following alternative solu-
tions. One is a numerical solution of the simulta-
neous second-order ordinary differential equations
without the last intergral terms of Eqs.(l) and (2) by
the Runge-Kutta-Merson method').
The other is a numerical solution of the intergro-
differential equations by the method given in the
Appendix.
379
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
-------
1.0*
1.05 .
1.00
14 « 8 10
Mdtus r C » ]
Fig. 4 Pressure distribution in a gas centrifuge
1(0
140
100
(0
40
61«ss buds
H'JOOOrpn
PR- Z.Sg/cn'
I
• 8 10 12 14 16
Pirtlclt dl«»»t«r by optlcil alcroscopt
Op [•(crons)
Fig. 5 Variation of deposition angle with pressure for
glau beads (experimental and computed results)
Strictly speaking, the pressure distribution in the
rotor should be considered, but this effect appears
to be negligible. The pressure distribution of gas at
a constant speed of rotation may be expressed by
(8)
For an ideal gas, the general relation between density
and pressure is
Pi = PP, JP. (9)
Substitution of Eq.(9) into Eq.(8) and integration of
Eq.(8) with respect to r gives
uo
140
100
so
40
20
e 8 10 12 14 16
rirttclt dlntttr by optfctl •(crotcopi
0 [itcrons]
Fig 6 Variation of deposition angle with pressure for
glass beads (experimental and computed results)
160
140
"120
N»9000rp»
»p'2.5g/c»'
— Conputli) results Kith
Inttgn! ttra
— Co»putid nsults Kith-
out l«ttjr«l ttri
' » 10 12 14
Pirtlclt dUcittr by optlcil itcroscopl
Fig. 7 Variation of deposition angle with pressure for
glass beads (experimental and computed results)
(10)
using the boundary condition
B.C. at r = 0, p = p, (11)
The calculated pressure distributions are shown
VOL. 4 NO. 2 1971
380
-------
uo
140
120
100
10
40
20
0 2 4 68
Particle diameter by optical Microscope
0 [nlcrons]
Fig. 8 Comparison of experimental and computed results
for Zn powder
160
140
"S 120
•9
M
7 100
4r
7 BO
a*
C
•
c
£ (0
4*
M
e 40
20
Tung(ten ponder
N'iOOOrp*
up'19.2g/cn'
0.4 0.8 1.2 1.6 2.0
Particle diameter by electron •Icrosiopt
Op [ulcront]
Fig. 9 Comparison of experimental and computed results
for tungsten powder
in Fig. 4. The deviations of relative pressure
between outer and inner parts of the rotor are less
than a few percent
*£.-.: ';
• »*«
• • •
depoiition angle 80° deposition angle 130"
Gla« bead. P»20mmHg JV=6000rpm
deposition angle 80° deposition angle 140"
Zinc powder ^=20mmHg Af=6000rpin
fly. 10 Photomicrographs of experimental results
Experimental Results and Discussion
Figs. 5, 6 and 7 show the comparison of the ex-
perimental results and computed solutions for glass
beads at various pressures and rotor speeds of
3000, 6000, and 9000 rpm. Figs. 8 and 9 present the
results for zinc powder and submicron particles of
tungsten powder, respectively.
The solid line represents the numerical solutions
computed without the integral term in Eqs.(l) and
(2), and the broken line represents the computed values
of Eqs.(l) and (2). The solid and broken lines are
computed at the following boundary conditions B.C.
and initial conditions I.C., respectively.
" = * (const-)>
*»
gas rotates at
the same angular velocity as the rotor. (12)
at i=0, r = rt (13)
(14)
(15)
0 = 0
dO
~dT
(16)
The experimental results agree well with the com-
puted values. Fig. 10 shows photomicrographs for
a couple of the experimental runs.
From the numerical calculation, the deposition angle
calculated with the integral term included is slightly
greater than the value obtained without the integral
term. The effect of this term on the calculated value
of the deposition angle is less than 10%. Furthermore,
the effect decreases with decreasing pressure. Below
20mmHg the deviation is less than 1%, so the effect
of the intergral term may be considered negligible,
If possible, the powder should be fed to the classifier
in a well-dispersed aerosol, but the inital radial veloc-
381
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
-------
ity of the aerosol at the inlet ports reduces the depo-
sition angle. This has been confirmed by both the
experimental and theoretical results. Also, it is diffi-
cult to ensure that the initial radial velocity of each
particle is the same. Consequently, this non-uniform
velocity results in poor classification. This problem
may be solved in future, but deserves further attention.
Conclusion
Particle size classification in a gas centrifuge at
reduced pressure has been investigated theoretically
and experimentally, resulting in the following conclu-
sions.
1) Particles down to the submicron range can be
classified with good resolution, a gradation of
particle size occurring on the rotor wall.
2) Agreement between experimental results and
computed solutions is shown to be excellent.
3) The slip factor 'Cunningham correction' ob-
tained by using Millikan's data for oil drops in
air may be applied to various materials suspended
in air and at various pressures.
4) Numerical solutions of the simultaneous integro-
diflerential equations of a non-uniform curvi-
linear motion have been obtained.
5) The effect of the integral term for fine particles
suspended in air is not very large.
As a consequence of these results, the development
of a new particle size analyzer appears possible.
Appendix
The method of successive approximations is employed
to solve the simultaneous intcgro-difTerential equations (1)
and (2).
The calculation of the integral term is carried out using
the relative velocity between the particle and the fluid, u,
to simplify the notation.
The first stage of average acceleration is defined by
in which (rf «/«*/)«,/ denotes ;'-th approximation of i-th
step and the subscript m denotes the mean value. The
first approximation of the integral term after a imall incre-
ment of time At is
Substituting Eq.(2a) in Eqi.(l) and (2), the second approxi-
mation (rf«/rf/)I§i can be calculated. Then from Eq.(la),
(rfN/rfOi*,! may be evaluated. If the »-th approximation
is nearly equal to the (M — l)-th approximation
(rf«/rf/),<1,-(rfi«/rf/)1.._l-tolerance limit (3a)
Then defining the final approximation of the first step
(rfn/rf<),-(rfi./rf/)i.. (4a)
and the mean acceleration (du/dt)lm is calculated.
Thus from Eq.(2a), the value of the integral term at
the end of the first increment of time can be evaluated.
Similarly, (du/dt),m and the integral term can be calcu-
lated. For the purpose of calculation, the time is divided
into k steps (it is not necessary that each step be an equal
interval) and the integral term is computed as follows
du
<5a)
Applying this method to the radial and angular directions,
the numerical integration of Eqs.(l) and (2) by the RKM
method gives the required results.
Odar*) has proposed an equal-time interval method, but
in that case, only the Rungc-Kutta-Gill routine may be
used for integration of the equations. For the method
described here, the RKM routine is applicable.
Acknowledgement
We are very grateful to Professor C. Orr and Dr. J. H.
Burson of the Georgia Institute of Technology for their
assistance in the preparatory work for this study.
[Additional Note]
The numerical solutions presented in this paper were
calculated with the FACOM-230-60 digital computer at
the Data Processing Center, Kyoto University.
Nomenclature
C, = Cunningham correction factor
CD = drag coefficient
Dp = particle diameter
g, = gravitational conversion factor
N = revolutions per minute
P = pressure
f = pressure
f = radius of gyration
Kt = Reynolds number = (DTup,/fi)
/ = time
[ — '
[ —
[cm
[(ff/G) (cm/see1)
[mmHg
[G/cmr
[cm]
[ — ]
[sec]
u — relative velocity between particle and fluid in
general [cm/sec]
v — relative velocity between particle and fluid defined
by Eq.(3) [cm/,*]
* = integral variable [sec]
8 = angular displacement of particle [rad]
la* = mean free path of gas molecule [cm]
P, = density of gas ' [g/cmr
Pj, = density of particle [g/cm*
{i = angular displacement of gas [rad
a = angular velocity of rotor [rad/sec
Literature cited
1) Basset, A.B.: Phil. Trans, of the Royal Soe., 179, 43
(1888)
2) Burson, J. H., E. Y. H. Keng and C. Orr: Pewdir
Technology, 1, 305 (1967/68)
3) Fuchs, N.A.: "The Mechanics of Aerosols", p. 75,
Pergamon Press (1964)
4) Kennard, E.H.: "Kinetic Theory of Gas", p. 310,
McGraw-Hill (1938)
5) Kriebel, A.R.: Trans, of ASMS, J. of Basic Engr.,
83, 333 (1961)
6) Lukehart, P. M.: Communications of tin ACM, 6, 737
(1963)
7) Millikan, R.A.: Phyt. Rn., 22, 1 (1923)
8) Odar, F.: U.S. Army Cold Regions Research and
Engineering Laboratory, Research Report 190 (1966)
9) Schiller, L. and A. Naumann: Z. VDL, 77, 318 (1933)
VOL. 4 NO. 2 1971
382
-------
Journal of Electrostatic!, 2(1976/1977)341-350
O Elsevier Scientific Publishing Company, Amsterdam — Printed in The Netherland*
(Reprinted with permission)
ELECTRIFICATION OF GAS-SOLID SUSPENSIONS FLOWING IN STEEL
AND INSULATING-COATED PIPES
HIROAKI MA8UDA, TAKAHIRO KOMAT8U, NAOHIRO MITSUI, and KOICHIIINOYA
Department of Chemical Engineering, Kyoto University, Kyoto 606 (Japan)
(Received September 24,1976; in revised form December 10,1976)
Summary
Electrification of gas—solid suspensions flowing in steel and insulating-coated pipes are
experimentally studied. It is found that the currents generated on insulating-coated pipes
are higher than the currents generated on a steel pipe and the sign of the currents follows
a kind of tribo-electric series. The currents are proportional to powder flow rate, propor-
tional to mean ah- velocity to the power 1.4—1.9, and inversely proportional to the mean
particle diameter. It seems that the currents are proportional to the pipe diameter. The
effects of a bend and powder feeding inlet are also studied.
1. Introduction
In gas—solid pipe flow, particles are charged through their collisions with
the pipe wall [1]. The wall is also charged, and the charge generated per unit
time, which can be measured as a current to earth, is a function of several
variables, such as the number of collisions, area of contact and duration of
contact [2]. The current may also be affected by the wall material. In practi-
cal application of gas—solid flow, the pipe wall is sometimes made from insu-
lators such as transparent glass, acryl and polyvinyl chloride.
In the present work, a steel pipe and pipes coated with various insulators
are set in a pneumatic conveyor line, and the currents generated on those pipes
are measured for several kinds of powder. The effects of a bend and a powder-
feeding inlet on the generated current are also studied.
2. Experimental apparatus and procedures
Figure 1 shows the suction type pneumatic conveyor used in the experi-
ment. Test pipes are listed in Table 1. Both ends of these test pipes are insu-
lated with polyvinyl-chloride flanges (12 mm thick) and the pipes are set in
the conveyor line at positions A, B and C in Fig. 1. The current generated is
measured by a galvanometer. A 2000 pF condenser is, if necessary, connected
in parallel in order to suppress fluctuations in the current. The conveyor line
before and after the test section is grounded at one point.
383
-------
Bag filter
Blower
Fig. 1. Experimental setup (Lengths are shown in millimeters. A, C: horizontal, B: vertical).
As a preliminary experiment, test pipe A (steel, 15 cm long, 5.3 cm bore) is
set at various positions and currents are recorded. Effects of the length of the
test pipe (Ax) and pipe diameter (D) are also studied by use of steel pipes.
Powder-flow rates (W)aie measured by a direct-measuring method. A cali-
brated diffuser [3] and a Pitot tube are used to measure the air-flow rate.
Powders used in the experiments are listed in Table 2.
TABLE 1
Test pipes
Symbols
A
B
C
D
E
Materials
steel
teflon
polyvinyl chloride
glass
plastic*
Insulator-
thickness
(mm)
_
<0.1
1.6
2.2
0.1
Inside
diameter
(mm)
53
53
50
51
53
Length
(cm)
15
30
15
30
15
•Mitsubishi Paper Mills Ltd., Hishirapit.
384
-------
TABLE 2
Properties of powder materials
Powder
Quartz sand
ultra fine
No. 8
No. 5
Morundum
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Flour
PVC
Glass beads
Mass median diameter
0pso (Mm)
16.8
51
440
58
73
93
126
180
340
515
760
57
115
55
Mean particle
diameter* £»p (Mm)
14
48
329
50
63
93
126
180
340
515
760
37
111
53
Density
pp (g/cm})
2.65
2.65
2.65
3.97
3.97
3.97
3.97
3.97
3.97
3.97
3.97
1.44
1.41
2.42
//
3. Results and discussion
/
3.1 Current generated on steel pipes as a function of axial distance
The results are shown in Fig. 2. Higher currents are generated on the steel
test pipe adjacent to the powder-feeding inlet and the bend. The effect of the
inlet on the current extends downstream by 2QD, while that of the bend
extends upstream by 10D and downstream by 2QD. Excess current generated
between x = 0 and x = 20D equals that generated on the pipe line of length
1QD. The main cause of the higher current seems to lie in the higher number
of collisions arising from the flow disturbance produced by the bend and the
powder inlet.
10
20
30
50
60
70
15
10
§ 5
3
WCg/53
•«• 0-36
•A- 0-27
•O- 01
Horizontal
Bend
Vertical upwards
50
100 150 200 250 300 350 400 450
Distance from pipe inlet XCcm]
Fig. 2. Effect of a bend and a powder-feeding inlet on the generated current (test pipe A).
385
-------
3.2 Effect of the length of test pipe
Figure 3 shows the effect of the length Ax on the current. The measure-
ments are carried out between x = 90 and 195 cm where the currents shown
in Fig. 2 are almost constant. However, the relation between the currents and
the length Ax is not linear, but is expressed by the following empirical equation
k st 0.024 cm
or as a first-order approximation
Im Si KW(&x •»- A/), A/ a 27 cm
-i
(1)
(2)
Eqn. (2) suggests that the polyvinyl-chloride flanges at either end of the test
pipe may have some effect. However, results obtained by use of 3 mm flanges
are almost equal to those shown in Fig. 3. Further examination of the effect
of the length of test pipe may be necessary.
3.8 Current generated on test pipes
When insulating-coated pipes are used, the sign of the current sometimes
reverses in the initial period of powder flow. Figure 4 shows one such result.
The negative current generated initially becomes positive after a short time.
Test pipes and powders which showed such sign change are listed in Table 3.
For test pipe D (glass) and quartz sand the sign changed several times. For
test pipe C (polyvinyl chloride) with quartz sand, the sign change in the initial
period was seen in the first experiment, but did not occur on repeating the
experiment after about one month. However, on repeating the experiment
after seven months, the sign change was again observed.
Test pipe B (teflon) showed no sign change for any powder. Experiments'
in which no sign change was observed are:
20
^
2,5
quirts Mnd No.8
flour
285 • 1t7
(•55 » 0.625
087 T 0365
Eq.( 1)
0 40 60 120
Pip* ItngthAxCcml
Fig. 3. Generated current u a function of length of toit pip*.
386
-------
10 nA
— quartz sand No.8 -
test pipe; C
(-)«• - 0 -
Generated current
Fig. 4. Sign change of current in the initial period of experiment.
test pipe B (teflon) ... all powders,
test pipe C (polyvinyl chloride) . . . glass beads, morundum, flour,
test pipe D (glass) ... PVC, flour.
Experiments were carried out in the following order:
test pipe B: quartz sand— PVC— flour— morundum,
test pipe C: quartz sand— PVC— glass beads— flour— morundum,
test pipe D : quartz sand— PVC— morundum— flour.
The above experimental facts are summarized as follows.
(1) Sign of the current generated may change when the charge of a particle
leaving the preceding pipe line has the same sign as that generated in the test
pipe.
(2) Sign of the generated current does not change when the particle charge
generated in the preceding pipe line has opposite sign to that generated in the
test pipe, nor when these particle charges have the same sign provided the
current generated on the insulating-coated test pipe has the same sign as in
the preceding experiments with a different powder.
TABLE 3
Sign change of current in the initial period of experiment
Pipe
Powder
Sign change
C
C
D
D
E
quartz sand (— ) -*• (+)
PVC ( + )-»(-)
quartz sand (+) -» (— )
morundum (+) -* (— )
quartz sand (— )-* (+)
387
-------
150
nlOO
li
E
0 2 4 6 8 10
Powder flow rate WCgM]
Fig. 5. Relationship between the generated current* and powder-flow rate (test pipe C,
position B, u • 20 m/t).
200
g 50
3 20
I10
2
1
0-5
1-4
pip* powder
B flour 1.35
B morundum 54- 6-6
C flour 1.3*1-9
4.1-4-7
* C glass
beads
x D flour
o B quartz
sandNo.8
•» A glass
beads
1-8- 2-1
3.3
4-0*44
B PVC 1.7
C morundum 8-7- 9-0
C quartz 38 40
sand No.8
-f D PVC
A D quartz
sand No 8
1-9-2-0
3.3
D morundum 5-4-6-0
7 10 20 30 40
Mean air velocity 0Cm/si
Fig. 6. Current generated per unit powder-flow rate as a function of the mean air velocity
(position P).
388
-------
From these results, it seems that the sign change depends on the polariza-
tion of the insulator (electret formation), or on the change of the physical
property of wall material.
In the steady state, however, the current is reproducible and constant. The
currents are proportional to powder-flow rate as shown in Fig. 5, provided the
mass-flow ratio is less then about unity. Figure 6 shows the relation between
the generated current per unit powder-flow rate and the mean air velocity.
The results are represented by straight lines on log—log paper. These lines
may be divided into two groups, one with a slope of about 1.4 and the other
about 1.9. One possible explanation is that the contact area varies with the
mode of wall deformation, elastic or plastic. The slope for elastic deformation
may be about 1.4 and for plastic deformation about 1.9 [2]. In practice,
collision will be partly elastic and partly plastic.
It is also confirmed that the currents are inversely proportional to the
mean particle diameter. The results are shown in Fig. 7; It is found that the
larger the pipe diameter, the higher the current generated. This fact is shown
more clearly in Fig. 8.
It has been shown theoretically [2] that the current is expressed by
An
«m
mpz0T Ax
where the initial particle charge is neglected. Number of collisions per unit
length of a particle An/Ax is represented by [2]
An
— -mp7r£$
Ax *
By substituting eqn. (4) into eqn. (3), the following equation is obtained:
(3)
(4)
1
i
~EO.S
0.2
0.1
Morundum
• D* 2.8cm
SO 100 200 500 1000
Dp [yum 3
Fig. 7. Effects of particle diameter on the generated currents (steel pipe, position b,
u - 20 m/s).
389
-------
MorundumO)
2r
o>
E
= 0-5
0-2
2 5 10
Pipe diam. DfcmJ
Fig. 8. Relationship between the generated currents per unit powder-flow rate and the pipe
diameter (position B, steel pipe).
We,
Z0T
From the empirical eqn. (1) or (2), eqn. (5) may be modified as:
/m = -
where
and
Z0T
(5)
(6)
(7)
(8)
Equation (6) means that the current is proportional to the pipe diameter pro-
vided g is constant. Experimental results are represented by the following
semi-empirical equation:
or
1 +
A/
Ax
U
where
and [4]
e0
(9)
(10)
(11)
390
-------
TABLE 4
Constants in eqn. 9*
Test
pipes
A (steel)
B (teflon)
C (poly vinyl chloride)
D (glass)
E (plastic)
Powder
quartz sand
PVC
glass beads
morundum
flour
quartz sand
PVC
morundum
flour
quartz sand
glass beads
morundum
flour
quartz sand
PVC
morundum
flour
quartz sand
aXlO1*
1.6
-4.9
-19
-11
-14
-88.5
-157
-138
-220
9.1
-8.0
-23.3
-18.6
4.6
28.5
-14.7
-27.1
18.1
0
(-)
1.9
1.9
1.8
1.4
1.9
1.3
1.3
1.6
1.6
1.7
2.1
1.8
1.9
1.3
1.4
1.3
1.5
1.8
* ,-7
u (m/s), Dp (cm), W (g/s), A A (cm1), Im (A).
Constants a and 0 are listed in Table 4. Absolute values of a for insulating-
coated pipes are larger than that for the steel pipe by one or two orders. A
teflon-coated test pipe generates the highest current. These high currents may
be due to higher contact-potential differences and larger contact area. It is
also noted in relation to the danger of dust explosion that flour is highly
charged.
Table 5 shows a tribo-electric series obtained in this study.
TABLE 5
Tribo-electric series
{+) flour - morundum - glass - PVC - steel - quartz sand - teflon (-)
4. Conclusions
Electrification of gas-solid suspensions are experimentally studied and
the following results are obtained.
(1) Electrification of powder depends on the wall material of the pipe.
391
-------
Insulator pipes may cause greater electrification of powders than a steel pipe
by one or two orders of magnitude.
(2) The sign of electrification follows a kind of tribo-electric series.
(3) The effects of powder-flow rate, air velocity and particle size on the
current generated by insulating-coated pipes are similar to those in a steel
pipe. They depend on the mode of collision.
(4) Sometimes the sign of current changes in the initial period of experi-
ments with insulating-coated pipes.
(5) A bend and the powder inlet affect the electrification. Effect of the
inlet extends downstream by 2QD, while that of the bend extends upstream
by 10D and downstream by 20£>.
Nomenclature
AA
D inside diameter of pipe
Dp particle diameter
fo) particle-size distribution (weight basis)
A/ constant length introduced in approximate eqn. (2)
Im current generated on an insulated pipe
K, k constants in eqn. (1)
mp mass of a particle
n number of collisions of a particle
S area of contact
At duration of contact
u air velocity
Vc contact-potential difference
W powder-flow rate
x axial distance from inlet
z0 gap between surface of particle in contact and pipe wall
a constant in eqn. (9)
0 constant in eqn. (9)
€0 dielectric constant of air, 8.85 X 10"n F/m
TJ defined by eqn. (7)
g number of collisions per unit area and unit mass of powder
r relaxation tune
References
1 B.N. Cole, M.R. Baum and F.R. Mobbs, Proc. Instn. Mech. Engrs., 184 (1969-70) 3C77.
2 H. Masuda, T. Komatsu and K. linoya, AIChE J., 22 (1976) 558.
3 H. Masuda, Y. Ito and K. linoya, J. Chem. Eng., Japan, 6 (1973) 278.
4 H. Masuda and K. linoya, Memoirs of the Faculty of Eng., Kyoto Univ., 34 (1972) 344.
392
-------
JAPAN
K. Maki.no
K. linoya
M. Shibamoto
S. Toyama
M.Ikazaki
Experiments on the electrical
dislodging of a dust layer
A new method for dislodging a dust layer from a flat surface by a non-
uniform electrical field was studied experimentally. This was found to
be effective for industrial dusts, such as sintering furnace dust and
AVS resin powder, especially when a corona discharge was applied.
Appropriate design factors were described.
Introduction
electricity causes various troubles In
many processes for treating powders, so that the
technology of coping with them has become signifi-
cant as an engineering subject. On the other hand,
electric charges have been utilized as in electro-
static precipitatora. It has also been reported that
the electric properties of powders are being applied
to measuring the flow rate of powders {L.2], the void
fraction of the powder bed [3], and the particle-size
distribution. Charged particles in a non-uniform
alternating electrical field produced by electrodes
are driven in the opposite direction, against the
electrodes. Masuda et al. [5] analyzed theoretically
the characteristics of electric screens of the stand-
ing and moving wave type, and verified it experi-
mentally. However, this effect of the electrical
screens falls sharply when either the particle size
and the electrical charge are small or the velocities
of the particles are high. This tendency makes direct
application to dust collectors difficult. In this study.
a new trial method for electrically dislodging a dust
layer is examined in order to overcome the defects
of this electrical curtain. This method enabled duat
particles to be captured and deposited on the filter
set-up, with electrodes in a suitable arrangement,
and to be dislodged electrically. The typical char-
acteristics of this method were examined at various
points, and found to be applicable- to high-efficiency
dust collectors.
This article was flnt published in KagaJai KogaJai Ronbtuaku
(Proceedings of the Society of Chemical Engineers, Japan),
Vol. 2, No. 1, pp. 31-37 (1976). Kazutaka Makino. Koichi
linoya, and Masami Shibamoto are associated with the De-
partment of Chemical Engineering. Kyoto University, Kyoto,
and Shigeki Toyama and Fumikazu Ikazaki with the National
Chemical Laboratory for Industry; Tokyo. Translated by one
of the authors. Professor Kazutaka Makino.
1. Experimental apparatus and procedure
Lycopodlum, ABS resin powder, and sintering-
furnace dust, shown in Table 1, were used here as
test dusts for the following reasons: Lycopodium
has electrical charges of 1-4 x 10~u Coulomb/particle-
and it is convenient to examine the basic electrical
phenomena on the powder. As for sintering furnace
dust, it is difficult to capture it in electrical pre-
cipitators, due to the high electrical resistance of
steel used; therefore, it can be captured and dis-
lodged efficiently by our new method. This method
is expected to be widely utilized for the same kind
of dust. ABS resin powder was used in order to ex-
amine the effect of the number of electrical charges
of the particles, by the corona-discharge method, on
the dust-dislodging efficiency.
The experiments are performed in two ways. In
experiment No. 1, the filter paper and the electrodes
are lined up horizontally, as shown in Figure 1. Dust.
captured and accumulated on the filter from dust-
laden air is electrically dislodged under some pre-
determined conditions by the application of a high
AC voltage to these electrodes. In this experiment,
the effects of the superficial air velocity, the
quantity of electrical charges of the particles by
the corona discharge method, etc.. on the dust-dis-
lodging efficiency were studied mainly from the
practical point of view. In experiment No. 2, as
shown in Figure 2, the acrylonitrile board and the
electrodes' are-arranged horizontally.- The sample-
Table 1
Characteristics of test powders
Hrtld* Dntlfl lIC/ *Uun>|
"~ 2S.O1-4x10-1* 1.10
7.2
SUutrtas
runttc*duu
1.0)
I-S
April 1977
(Vol. 17. No. 2) INTERNATIONAL CHEMICAL ENGINEERING
(Reprinted with permission)
393
-------
M
>to> *I>MI|U
>u.
t Film i«r*»
) TIM! «mr
Kg. 1. Schematic diagram of arrangement of electrode and
filter (Experiment No. 1).
4. TtM
*.
Fig. 3. Experimental apparatus No. 1 for tltctric duit dl*
lodging.
BH«-
•-c-t.imm
I.
Fig. 2. Experimental apparatus No. 2 for •lactrle duit di»
duit deposited naturally on to* sheet from the stove
ta dislodged •Itotrieally in th» sara* way aa d«-
•erlbtd abov«. Exporintnta on tha afftota of tha
dlamatar of tha alaotroda win 2r, tha ratio of tha
tlMtroda dlamatar to tha dlataaea bctwaao tha
eaatara of tha alactrodaa R, tha dlatanoa batwaaa
tba doat layar and tha aurlaoa of tha alaotrodaa ybi
tto., on tha duat-dialodfing alOolaney wara par-
foraad la order to-aaalyn tha-.baale oharaotarlaltea
of thla mathod.
b ajqparlraant No. I, aampla duat partlelaa wara
niUrtd and aeoumulatad on tha taat-flltar papar
•adar a constant auction of air ualnf a blowar and
• eoutut faad of parttolM from a faadar, u ihown
in Flgura 3. Than, a thraa-phaaa AC voltaga (oom-
marolal fraquaney. maxlnum applied voltaga 20 kV)
*•• sppllad to tha alaotrodaa wUh tha auotton of pura
•It, whan tha praaaura drop through tha filter had
rtaohad a pradatermlnad laval. Tha dust partlelaa
wara dislodged eleotrlcally In thla way. The pres-
sure drop through the teat filter waa measured after
It had reached some predetermined level, and the
residual duat load on tha filter waa estimated from a
relatlonehlp experimentally obtained In advance,
between the preeeure drop of a dust-laden filter and
the dust load, using the air velocity aa a parameter.
Therefore-, tha electrical dust-dislodging efficiency
can be defined here as (1 - m^m^, where ra( and
mi are the duat loadia on the filter, eetlmated from
the relationship before and after the application of
the AC voltage, respectively. A constant flow rate
of air waa attained by regulation of the valve (Figure
3). Insulated electrodes were used with a diameter
of S.8 nun, and the diameter of the electrode wire
waa 1.8 mm.
In experiment No. 2, copper pipes covered with
vinyl-chloride tubing were used as electrodes, and
the distance between the duat layer and the surface of
the electrode y0 (in Flgura 2) could be changed by
varying the height of the leg attached to the aorylo-
nltrile sheet. The definition of the electrical
dust-dialodging efficiency waa the same as la ex-
periment No. 1, but thla time mt and ms were weighted.
2. Calculation of strength of eleotrlo field
In thla calculation, the electrical charges due to
dielectric polarisation must be considered, since
thousanda of volte are applied to electrodes lined up
at intervals of a few millimeters. In the arrange-
ment of the electrical ebarge and the dielectric ma-
terial, the linear charge q la arranged parallel to a
dielectric cylinder of diameter a with a dielectric
constant ki, la a medium with a dielectric constant
k|{ the'Intensity of the electrical field la equal to a
INTERNATIONAL CHKMICAl ENOINKEIUNO (Vol. 17. No. 2)
April 1977
394
-------
situation in which the linear charges are arranged in
the way shown in Figure 4. q at x = d. q' {= qx(kr -
(kj + k;)) at x = aVd and -q' at x = 0 (see Appendix).
It is reasonable to assume that the linear electrodes.
are infinitely aligned in a plane in our experiment,
so the intensity of the electrical field is regarded
as that formed by the linear electrical charges, by
the dielectrical polarization mentioned above and the
true charges.
In general, the potential function U is given by
Equation (1)
for an Infinite row of linear electrical charges at the
regular interval d [6]. The intensity of the electrical
field in the x and y directions can be given from the
definition £ = -grad U.
(2)
(3)
For this arrangement of electrical lines, Equation
(1) (or (2) or (3)) can be added to give'the appro-
priate electrical charge, q In Equation (1) is calcu-
lated from the boundary condition that the potential U
mast be equal to the applied voltage V at any surface
of the electrode wire.
From the above considerations, under the condition
of a constant applied voltage, the variables are
known to be y/d, x/d, a/d, d and r; these affect the
intensity of the electrical field. Consequently, the
dust-dislodging force Increases with the increasing
intensity of the electrical field; the dust-dislodging
ability is discussed here aa a function of the in-
tensity of the- electrical field or the initial duat-
dialodging voltage.
3. Experimental result* and discussion
Experiments were performed under the conditions
of a temperature between 17 and 25*C and a relative?
humidity between 62 and 75% The experimental
results of Figure 1 are described first. Figure 5
shows the initial dust-dislodging voltage with respect
to a- and b-type filters shown in Figure 1. This
figure shows that a b-type filter, in which the dis-
tance yjj between the dust layer and the surface of
the electrode is less, is more efficient that the a-
type. The calculated dimensionless intensity of the
electrical field W = E/(V/1) is displayed in Figure 6
where W is known to be very heterogeneous near
the electrodes, but depends on y only if it is rather
far from the electrodes. W decreases linearly with-
increaslng y. This tendency can explain the results
in Figure 5.
Figure 7 shows a better dust-dislodging fraction
with increasing dust load, due to the existence of a
fixed undislodged zone in the dust layer, even in the
electrical field. The initial dust-dislodging voltages
in Figure 7 were observed to be almost the same.
In Figure 8, the dust-dislodging fractions are shown
for various distances between the centers of. th»
electrodes d. The dust-dislodging fraction Is higher
1."
•"6
1
3-ph*»»
10
InHWdi*
Fig. 5. Experimental initial dust dislodging voltage.
s
Fig. 4. Electric field by linear charge and dielectric.
April 1977
1 t J 375
Jt(mm)
Fig. 6. Dimeniionless strength of electric field near electrode.
(Vol. 17. No. 2) INTERNATIONAL CHEMICAL ENGINEERING
395
-------
and the initial dust-dislodging voltage is lower with
smaller values of d. This shows that the term p/4dk2
in Equations (2) and (3) is the principal factor in
determining the intensity of the electrical field.
Therefore, the capacity for dislodging the dust layer
falls with increasing d. The effect of superficial
air velocity on the capacity for dislodging the dust
layer is shown in Figure 9. The result, showing a
high efficiency, even under the severe conditions of
Fig. 7. Effect of dust load on dust dislodging fraction.
1.0
0.1
d-lmm
d-Smm
0 • y 10 »
ApplMMtaplkV)
Fig. 8. Experimental dust dislodging fraction In cot d « 5 and
vflVflt*
a 30— 50 cm/sec air velocity, is of technical interest,
taking account of the usual superficial air velocity of
a few cm/sec in bag filters; This suggests the pos-
sibility of dislodging the dust layer continuously, dur-
ing the filtering of dust-laden air through the filter
paper.
The effect of the magnitude of the electrical
charge on a particle due to a corona discharge on
the capacity lor dislodging a dust layer was examined
next. The capacity is almost doubled with a corona
discharge, as can be seen in Figure 10. It has been
confirmed that the particle charges have a great ef-
fect on the dislodging capacity; in other words, our
new method would be also applicable to dust particles
having only a few intrinsic electrical charges.
The results of experiment No. 2 are as follows.
Figure 11 sbowa the limiting distance between the
dust layer and the electrode surface to obtain a
high dislodging efficiency. Then, It is shown that,
with an applied voltage of 10 kV, a layer of sintering
furnace dust can be electrically dislodged with an ef-
ficiency of 80% or more, In the case of yfa (see*
i.o
o.s
Plllwt 0-HOO
b-typ* fill*
O wild
m.0.«k«
-------
Figure 1) less than 1 mm with a diameter of the
electrode wire 2r of 1 mm, yj, less than 2.5 at 2r of
4 mm, and yo less than 7 mm at 2r of 7 mm. This
result may be regarded as fairly good in compari.-
son with those for an ordinary bag filter with a 30~ .
40%dislodging fraction [7]. All the results on
Figure 11 show a rapid drop at y^, greater than some
given value. The relationships between the dislodg-
ing efficiency and the frequency of the applied voltage
are shown on Figure 12. The dust-dislodging fractions
are almost the same at a frequency lower than the
•commercial one, while, at a higher one, they exhibit
a considerable drop. This tendency was similar to
that for other sample dusts. This is because the
electrical force on a particle moving freely in an
electrical field is proportional to {(2irf)1 + (3jrDpu/
M)'}-> [8]. In a dislodging experiment with ABS pow-
der at 500 Hz the powder covered the filter evenly
and the pressure loss did not decrease. This is be-
cause the adhesion of dust particles is stimulated by
the heat evolved In the dlelectrlcal material by high
frequency, as observed in our experiment. Figure
13 shows the duat dlslodglng-voltage for lycopodlum
as a sample dust with respect to yb and R, and the
ratio of the distance between the centers of the elec-
trodes d to the electrode diameter 2a, In the case of
a three-phase applied voltage. The optimum diameter
of the electrode wire is determined for each yD, and
Increases with an increase in y^. The optimum diame-
ters of electrode wire are almost the same for sin-
tering furnace-dust, since they are determined by
the relative value of the Intensity of the electrical
field near the electrode, and are not affected by the
characteristics of the dust Itself, in our experimental
range. On the other hand, the dust-dislodging voltage
Itself differs from that for lycopodlum because of
the different chargeablllty. I.e., furnace dust la apt
to receive more charge by contact and friction than
lycopodlum.
The dust-dislodging efficiency turns out to be lower
at R • 2 than at R • 1, according to Figure 1. Almost
the same dislodging efficiency as for lycopodlum was
obtained for slnteriog-furnaoe dust and this new
method Is also applicable to such Industrial dusts.
Under the definite conditions of the relative humidity
in our experiment, there was no appreciable effect
on the efficiency. It was also confirmed that moving
the filter paper, while the electrode was fixed, was
much, more effedtiver this' new dislodging method us-
ing moving filter paper suggests the direction of
future research and development.
Conclusions
The following results were obtained by experi-
ments on the electrical dislodging of a dust layer.
1) The main factors In the arrangement of the
electrodes and the filter paper which affect the dust-
dislodging efficiency are the distance between the
dust layer and the electrode surface, the diameter of
the electrode wire, and the ratio of the pitch of the —
electrode to the electrode diameter.
2) The b-type filter turns out to be better than the
a-type.
i.o
fO.5
n so 1001 KO soo «co MOO
Friqutn«y *l »ppU«d voll>g»( Hi)
Fig. 12. Effect of frequency of applied voltage on dust dis-
lodging fraction.
20
IS
10
NO
0 1 10 IS
Olwntltr s* ftalroft v*t Zr(mm)
FI0.13. Effect of dlsmtttr of electrode wire and distance
from altetrode surface on dust diilodglng voltage In case of
lycopodlum.
so
I"
I
1
o s to u
Ditmrtcr el ((tetrad* win j'r(mm)
Fig. 14. Dust dislodging voltage in case of sintering furnace
dust.
April 1977
(Vol. 17. No. 2) INTERNATIONAL CHEMICAL ENGINEERING
397
-------
3) The dislodging efficiency increases linearly
with the increasing initial dust load and with a de-
creasing ratio of the pitch of the electrode to the
electrode diameter. It is not much reduced in the
range of scores of cm/sec of superficial air velocity.
•i) A higher dust-dislodging efficiency was obtained
by increasing the electrical charge on a particle us-
ing a corona discharge.
5) A dust-dislodging efficiency of more than 80 pet
for sintering furnace dust was obtained by the appli-
cation of a 10 kV AC high voltage, in the case where
the distance between th.e dust layer and the electrode
surface was less than about 3.5 mm.
6} The applied frequency of the electrical power
should be less than commercial frequency to avoid
decreasing the performance.
7) The optimum diameters of electrode wire are
most efficient in dislodging a dust layer electrically
with a constant ratio R of the pitch of the electrode
to the electrode diameter. They depend on the dis-
tance yb between the dust layer and the electrode
surface and increase linearly with an increase in the
distance. An optimum diameter of electrode wire of
3~ 5 mm was obtained with a value of yD of 1 mm and
a value of R of 1 mm.
8) A higher performance was obtained with a rela-
tive movement between the electrode and the filter
paper.
Finally, it is still necessary to study the adhesion
and cohesion of dust in an AC electrical field and
the time-dependent performance after a long run in
order for the method to be widely applied.
Appendix
Under the condition that the linear electrical
charge q is at the position P«(r0,0() as shown in
Figure 15, the potential Up at the distance of R from
PI is written in the form of Equation (i)
(i)
Equations (ii) and (ill) are derived with the applica-
tion of the sine and cosine theorem to Equation (i).
X(co* /i* cos n*,+sia nS sin /i»0)-ln '
(ii)
X(cos ir» cm ntfa
-In r,J(r,>r) (ill)
.-A[J,v(t)"«—-H
Ha the case of the existence of a linear charge q. — q'
and q' at x - r0. 0 and aVr0, respectively, on the x-
axis, the potential U in the position P(r,0) is ob-
tained as follows, considering that r < r0 and r > aV
Tt in our case.
(iv)
(v)
(vi)
Summing Equations (iv), (v) and (vi) obtained above
for q' - qx(k, - krf/(ki -f
(vll)
On the other hand. Equation (viii) is obtained from
the Laplace equation in cylindrical coordinate, by the
method of separation of variables,
•f E («.
. sinfl»XJV
(viii)
Fig. 15. Electric field by linear charge.
where a. /}. y. «. on, 0n. yn, and fin are integral
constants. This equation exhibits the potential dis-
tribution of the linear electrical charges.
It can be seen that Equation (viii) is obtained by de-
termining the integral constants.
Acknowledgment
The assistance of Izumi Sano is acknowledged.
Nomenclature
a ... radius of covered electrode, m
Dp ... particle diameter, m
d ... pitch of electrode, m
E ... intensity of electrical field, volt/ra
f ... frequency of applied voltage, Hz
k ... dielectrical constant, F/m
k* .. .specificdielectrical constant
I ... distance between adjacent electrode wire
surfaces, m
M ... mass of particle, kg
m ... dust load. kg/ms
INTERNATIONAL CHEMICAL ENGINEERING (Vol. 17. No. 2)
398
April 1977
-------
R
r
U
u
V
w
x, y
yD
. electrical charger per unit length of elec-
trode, m
. d/2a
. radius of electrode wire, m
. potential of electrical field, V
. superficial gas velocity, cm/sec
. applied voltage, V
... coordinate from center of electrode, m
... distance between bottom of dust layer and
top of covered electrode surface, mm
... viscosity of air, kg/(m)(sec).
Literature cited
1 . Nakajima, H., Goto, K., and Tanaka, T., Kagaku KSgaku
(Journal of the Society of Chemical Engineers, Japan) 31,
p. 504 (1967).
2. Masuda, H.. Ito, Y., linoya, K., and Sakai, K...J.
Assoc. Powder Tech. 10, p. 1SI (1973).
3. Nakajima, H., Nakamura, A., and Tanaka, T.,/
Auoc. Powdtr Tech. 10, p. 3 (1973).
4. Nakajima. H., Mitsui, R., Kuramae, M., and Tanaka,
Ko&aku Kogaku (Journal of the Society of Chemical Eng.
neers, Japan) 36, p. 1243 (1972).
S. Masuda, S., Fujibayashi, K., and Ishida, K., Stat,b-Rlu>
haltung dtr Luft (Dust-Air Cleaning) 30, p. 449 (1970)
6. Oberhettinger, U., "Anwendung der elliptischen Funk
tlon in Physik und Technik" (Application of Elliptic Functio
to Physics and Engineering), p. 67 Berlin, Springer Verlai
(1949).
7. Tanaka, N., Makino, K., and linoya, K., Kagaku KSgaJa
(Journal of the Society of Chemical Engineers, Japan>37
p. 718 (1973).
8. Masuda, S., 'Textbook of 17th Powder Technology*
Count," p. 23, Res. Astoc. Powder Tech., Japan (1970)
April 1977
(Vol. 17, No. 3) INTBMNATIONAL CHKMICAU BN.OINBBR1NCT
399
-------
COMPARISON OF DUST CLEANING PERFORMANCE OF COLLAPSE
AND MECHANICAL SHAKING TYPES OF FABRIC FILTERS
K. linoya, K. Makino, Y. Mori,
J. Okutani and H. Kawabe
Kyoto University, Chemical Engineering Dept.
Introduction
The bag filter, one of the high performance dry type
collectors, is currently increasing in importance. The main
investigation of ordinary bag filter is characterized by the
collection efficiency versus pressure drop during operation and
there are not many investigations concerning dust cleaning char-
acteristics. l'2 However, in order to investigate the operation
of a bag filter, the dust cleaning performance associated with
this system is an important characteristic in determining its
operating condition as well as an economically optimal design.
Therefore this report, in recognizing this point, applies reverse
collapse and mechanical shaking type dust cleaning operations
which are widely used for filter cloths and presents the result
of experimental comparison of them in terms of pressure drops.
1. Experimental Apparatus and Method
1.1 Sample Fly Ash and Filter
Two kinds of ashes (fine calcium carbonate I and II) are used
as sample ashes and the measurement result of their particle size
distribution is shown in Table 1. As filters, we used nylon
(long fiber NR-9A) and teflon (long fiber TR-9A, short fiber
TR-2020S) which are widely used in the industry. The bag is a
cylinder type bag with inner diameter of 170 mm and length of
1,800 mm. Also the reverse collapse type bag filter was adjusted
with spring to have a constant initial tensile force of 12.5 kg.
For mechanical shaking type, the bag tension was adjusted so that
the bag can move vertically about 20 mm while the middle of the
bag is seized. Although it is generally believed that dust
cleaning performance depends on type of ashes, the purpose of this
report is to verify the qualitative characteristics of dust
cleaning performance when the above mentioned sample ashes are
used.
1.2 Experimental Apparatus
Figure 1 shows the experimental apparatus. Ashes travel
from the feeder through the rotating impact type dispenser and
diaphragm valve (V2) to bag filtering surface where they are
collected. The cleaning air which passed through the bag travels
400
-------
through the valve (VI) and the flow measuring orifice and is
exhausted. On the other hand, for reverse collapse type dust
cleaning operations, four valves (V1^V4) are operated reversely
and air for reverse dust cleaning goes through valve 3 (V3) and
bag surface in a reverse direction to the collected ash accumu-
lated at the inside wall of the bag/ and passes through valve
(V4) and orifice before exhausted. For mechanical shaking type
dust cleaning operations, only one valve (VI) is closed, as
compared to the collecting operation, to terminate the air flow
after passing through the bag. The shaking equipment (450 rpm)
on top of the system is started to clean the accumulated ashes.
The cleaned ashes fall into the dust chamber.
1.3 Experimental Method
1) Supply sample ashes quantitatively (at a constant rate)
and start collection, 2) stop collection when the bag pressure
drop reaches at pre-determined value, 3) start dust cleaning and
stop after pre-determined time T, 4) weigh ashes which fell into
dust chamber, 5) restart collection with only cleaning air passing
through bag and measure pressure drops, 6) repeat above operations
3), 4), and 5) and continue measurements until there will be no
ashes. Assume the final dust cleaning pressure drop as Ap^.
2. Experimental Results
2.1 Final Dust Cleaning Pressure Drop Ap°°
Table 2 shows a relationship between the final dust cleaning
pressure drop Ap» and the initial dust cleaning pressure drop
Apmax for reverse collapse type and mechanical shaking type
systems. According to this, the final dust cleaning pressure
drop for reverse collapse type system is considerably higher
than that of mechanical shaking type system. On the other hand,
when the dust accumulating condition of the filter cloth surface
is observed at the final dust cleaning operation, the former one
has a considerable amount of secondary accumulation layer and
ashes on the surface are in a mottled pattern while the latter one
is observed with only a primary accumulation layer. Also the
value of Ap,,,, is independent with Apmax and is almost constant.
2.2 Effect of Local Dust Cleaning Duration Time T
Figure 2 shows a representative relationship between the
cumulative cleaning duration time to (= n-r) and residual fraction
of pressure drop after cleaning,Ap,for the reverse collapse type
system and Figure 3 shows that of mechanical shaking type system.
Based on this, the former one will have the minimum value for re-
sidual fraction after cleaning Ap at a certain value of T(10 ^ 30
sec) while the latter one has a smaller Ap with a shorter T.
401
-------
2.3 Effect of Initial Dust Cleaning Pressure Prop
Figure 4 and 5 show a relationship between Ap and to as a
parameter of the dust .cleaning initial pressure drop Apmax for
both systems and it is seen that/ when tc is constant, the value
of Ap is smaller with higher Ap^x* This can also be predicted
from above 2.1.
2.4 Dust Cleaning Curve
The dust cleaning curve is defined as the relationship between
pressure drop and dust load when collection is continued until
the pressure drop becomes the same value as that of initial dust
qleaning after ashes were uniformly accumulated on filter cloth.
The dust cleaning operation was continued until no more ash was
removed. Figure 6 and Figure 7 show typical dust cleaning curves.
For the reverse collapse type system, the larger the dust load
at initial cleaning is, the larger the residual dust load is.
Also when the bag pressure drop at reversing decreases to less
than a certain value, there will be no dust cleaning function.
It is also noticed that the increase ra,te of pressure drop,
when the system was switched to collection after achieving
pressure drop ApM, is much bigger than that of cleaning filter
cloth. For the mechnical shaking type system, the residual dust
load is a constant value which is independent with initial condi-
tion of dust cleaning and most of the secondary dust layer is
cleaned out. This is explained by the fact that the increase rate
of pressure drop after the system is switched to collection from
final achieving pressure drop is consistent with that of collection.
Namely, the latter one is different from the former one and is
perfect in cleaning out secondary layer. These characteristics
also apply to short fiber materials and its example is shown in
Figure 8.
3. Consideration
3.1 Estimation of Dust Cleaning Process
When the sample ashes are uniformly accumulated on filter
cloth, the initial cleaning pressure drop Apmax and tne final
cleaning pressure drop Ap^ are expressed in the following
equations,*
APmax = U(A -f Byraax 6) (1)
Ap» = UA (2)
* Pressure drop characteristic coefficients A,B and 6 in equations
(1) and (2): In this experiment, A is measured at the condition with
very little residual dust load after complete cleanings. Therefore,
A has a close value as a (the pressure drop characteristic coefficient
only for clean filter). B and 6 have close values as b and 6
respectively. Please refer Reference 3) and 5) for definition of
a, b, and 6.
402
-------
If dust cleaning is started from equation (1) 's condition,
the bag pressure drop gradually decreases from Apmax and reaches
at Apo,. When the dust accumulation condition at a certain point
of dust cleaning process was observed, it was determined that the
dust layer consisted of the part where dust had not fallen,
the part only with initial dust layer after others fell off
and the part which is in between of above two parts and has a
part of secondary layer. It was also determined that the mechani
cal shaking type has .mostly the former two cases and has mottled
falling. **
Considering afcove results of observation, introduce the
model like Figure 9 in which 9 is defined as the ratio of inter-
mediate part with residual secondary layer to all filtering area.
Using this model, the pressure drop Ap at a certain point during
cleaning process is obtained by the following equation (refer
Appendix) .
AP - - - - (3)
(1-0
)-
(1-a)
( l\-1
IA -
e
)
+
0/
e
a66
a06
-I- (1-a) x6
dx
A Poo/ Ap and A in Equation (3) can be measured experimentally.
Also if 6 is experimentally determined with same dust and filter
material, 0 at a certain t0 can be calculated by Equation (3).
Figure 10 shows a typical result of reverse collapse type.
According to this, 6 has a tendency to increase from 0 to a certain
value when t0 increases. This means that the cleaning process
initially has only the mottled fallings and has less mottled
fallings later, but even at a final stage the mottled fallings
still exist. Also when t0 is constant 6 has a minimum value and
especially under Figure 11's condition 0 has a minimum value at
Apmax = 120mm H2O. Causes of this phenomenon will need to be
investigated fundamentally in the future. On the other hand the
result of observation shows 6=0 for the mechanical shaking type
and this can be seen that the calculated cleaning curve with
0 - 0 in Equation (3) is consistent with experimental value as
shown in Figure 8. Also similar results were obtained for
calcium carbonate and metallic silicon which have different
particle size distributions. As shown in Figures 2-5, mechanical
shaking type can complete dust cleaning in a much shorter time
(10-30 sec(practical region)) than reverse collapse type.
**"Mottled falling" in this report means that the dust residual
condition at cleaning process consists of one part where no
dust had fallen off like (1-6) section of Figure 9 and the
other part where only initial dust layer is left.
403
-------
Namely mechanical shaking type precedes dust cleaning by mottled
fallings and eventually most of the secondary layer will be
cleaned off. Therefore, the drastic increase in pressure drop
due to mottled falling during re-collection does not exist.
3.2 Estimation of Optimal Single Cleaning Duration Time TOpt
The cleaning mechanism characteristics of accumulated dust are
known to be approximated by one step later (4). Thus, if it is
cleaned for time T and the time T cleaning is repeated for n times
after a stop, the drag coefficient (drag resistance) due to dust
falling at ith process can be obtained by following equation.
Rf, - R_. = (R<=°, - R . ){l-exp(-T/T.)} (4)
£1 a 1 J. al . J.
Here, R ., R,. and R^. are the initial drag coefficient at ith
cleaning process, the final drag coefficient after cleaning and
the hypothetical drag coefficient after infinite cleaning processes
Since R-. is equal to R . ,, in Equation (4) is varied to obtain
Equation (5).
Rfn " R*i = z t(R«>i ~ R J{l-exp(-T/Ti)}] (5)
ill a 1 * — 1 1 cil
Since the result of experiments show that Ti is constant and is
independent with i, Ti = T leads to the following Equation.
Rfn ~ Ral =
-------
Therefore, G can be experimentally obtained by using Equation (8)
and our review for the combination of the dust and the filter
material used in this experiment shows that it can be determined
by the following equation.
G - k.log n + k2 (k., k2; experimental constant>0) (9)
Thus, G is proportional to a logarithm of number of cleanings n.
The comparison of experimental results and the calculational
values of Ap which was obtained by substituting Equation (9) into
Equation (7; is shown in Figure 12. It is seen that both match
fairly closely. Also this result suggests that, if k^ and k2 are
determined by certain experimental values with a constant T, it
is possible to estimate the optimal single cleaning duration time
Topt-
Conclusion
After comparing the cleaning phenomena of reverse collapse
type and mechanical shaking type bag filters, the following
results were obtained.
1) When the cumulative cleaning duration time to is constant,
the reverse collapse type has the most optimal value for a
single cleaning duration time T but the mechanical shaking type
has a better cleaning performance with smaller T.
2) The final cleaning pressure drop Apw is generally higher for
the reverse collapse type than for the^mechanical shaking type.
For both cases, the final cleaning pressure drop Ap has a constant
value independently with initial cleaning pressure drop Apmax.
The higher initial cleaning pressure drop Ap^x is the more
residue fraction of pressure drop after cleaning decreases.
3) The reverse collapse type initially has only mottled falling
and its ratio will decrease with time to have more intermediate
cleaning layer. However, fairly large amounts of mottled falling
(9 = 20-50%) will also be left. The mechanical shaking type
rapidly proceeds cleaning by mottled falling and most of the
secondary dust layer is cleaned off eventually. Namely there is
almost no rapid pressure drop increase at collection during mottled
falling.
4) One method to determine experimentally the most optimal value
of single cleaning duration time T for reverse-collapse type was
obtained.
The cleaning performance should also be reviewed in conjunction
with collection performance and this will be one of the important
topics in a future.
405
-------
Appendix
Suppose the accumulating condition at certain point of
cleaning process is expressed by the model shown in Figure 9,
6 part; Ap = u (x) (A+BM(x) 6) (i)
where M(x) = Mmax«F(x)
(1-6) (1-A) part; Ap = Aui (ii)
(1-6)A part; Ap = u2(A+BMmav6) (iii)
max
Since air flow rate Q is considered to be almost a constant,
Q/S = /9 u(x) dx + [u. (1-A) + u2A](l-8) = um = umav (iv)
0 J- luclX
Following Equation is obtained from Equations (1) , (2) , and (iv) ,
Iv)
Substitute Equation (v) into Equations (i) and (iii) ,
Ap = u(x) [A + — i- (APinav - Aft,) F(x)6] (vi)
"max max
Ap = u2 [A 4- -i— (Ap - Apte) ] . (vii)
max
Also obtain u,, u(x) , and U2 from Equations (ii) and (vi) and
(viii) and substitute them into Equation (iv) ,
dx
"w
xl
u. - (viii)
406
-------
Substitute um&
-------
ON THE
ECONOMICALLY OPTIMAL DESIGN OF BAG FILTER
K. linoya, K. Makino and N. Tanaka
Kyoto University
Dept. of Chemical Engineering
Introduction
He have already done a theoretical review of the operation
of multi-compartment bag filter and verified how the pressure drop
and the cleaning cycle of bag filters are affected by cleaning
residual fraction, number of bag compartments and pressure drop
characteristics. Based on these, this report gives a theoretical
review concerning economically optimal design of bag filter.
There are already some reports which deal with economical
design of bag filter as a large system including cooling system
or which deal with economical design by assuming that the life of
filter cloth is a function of filtering velocity. Our economical
design report of air-filter is essentially the same as these.
Based on a relationship between operating pressure drop and
cleaning cycle obtained in previous reports, this report provides
optimal operating pressure drop, optimal filtering velocity and
optimal cleaning cycle as variables for the optimal problem
by expressing a strict relation of process amount and by assuming
that the life of filter cloth can be determined by number of
cleaning or operational time. The pressure drop of the bag filter
is expressed as a sum of the pressure drop of filter itself and
the pressure drop due to dust load. An analytical solution can
be obtained when the former one can be neglected against the latter
one. The parameters used in this analysis were determined after
contacting several users and manufacturers in order to get more
practical calculations. Also by obtaining relative sensitivity
of each parameter to optimal solution, it was shown that the cal-
culations in this report can be easily applied to actual cases and
is the important factor in economical design.
1. Derivation of Equation for Economical Calculation
Generally the fixed and operating costs of instrument are
necessary to know for resolving the economically optimal problem
of systems. The design of bag filter requires information for the
fixed and. installed costs of bag filters, the fixed installation
costs of the fan and the cost of replacing filter cloth due to
damage. The operating cost of bag filter is mainly the cost of
power for cleaning and this can be neglected as compared with
other costs. Although it is also necessary to consider the man-
power cost for changing filter cloth, it is difficult to formulate
this and it can be included in the cost of filter cloth.
408
-------
Here, the fixed cost of the device is expressed as a pro-
duct of the cost of instrument (including installation cost) and
the certain annual rate, including tax, interest, repair cost
and depreciation cost. However, the fixed cost of bag filter is
considered not to include changing cost of filter material.
1.1 Fixed Cost of Fan
As a result of cost analysis of fan (mainly turbo-fan) in
the ranges of 200~500 mm H20 air pressure and 200~5,000 m3/min
air flow, there is a linear relation between motor power cost
(including costs of motor and installation) as shown in Equation
(1). Also as shown in Equation (2) there is a linear relation
between motor power and the product of air pressure and air flow.
Yi = hiP = 2.7P (1)
P = h2Ap-Q = 0.019Ap-Q (2)
If the annual rate of tax and depreciation is set as kj, the fixed
cost of fan is as follows:
Ci = kihaP = KjAp-Q (3)
Now as a standard value, ka = 0.2 is considered. Substitute
Equations (1) and (2) into Equation (3).
= 0.010
Here, the coefficient of Equation (2) is consistent with the report
by linoya.
1.2 Power Cost of Fan
The power cost of fan can be expressed by the following
equation
Cz — e*e«P (4)
where e is a unit power cost and e is an annual operating time.
Substitute Equation (2) into this and you will get the following
equation.
Cz = e-e-hzAp-Q = KzApQ (5)
When hz = 0.019, e = 8,000 hr/yr, e = 5 yen/Kw«hr as standard
values, K2 = 0.076.
409
-------
1.3 Fixed Cost of Bag Filter
As a result of cost survey (including installation cost)
of bag filter, it was found that the cost is about the same for
mechanical shaking cleaning type and reverse air-flow cleaning
type and that the cost is in exponential relation with filtering
area as shown in Figure 1. The pulse air type is slightly higher
in cost than these. Assuming the annual rate is k3, the fixed cost
can be expressed by the following equation using the cost of bag
filter in Figure 3:
Cj - kj(h3Sqs)SKsSqj (6)
As a standard value, assume kj - 0.2. Also from Figure 1, when
S>300 m2 it is seen that hs • 2.5 and 3s • 0.89 and when S<300 m2,
h, - 12.3 and qs - 0.61. Thus, Ks = 0.50 (S>300) and K3 - 2.46
(S<300).
1.4 Filter Cloth Replacing Cost
Although filter damage depends on complicated factors and
there is not a complete definition of the lifetime (6) of filter
cloth, filter damage can be considered due to following two cases;
damage caused in a certain time 60 due to long-time exposure to
high temperature gases, and damage caused by a certain number of
cleanings R due to numerous dust cleaning operations. These two
cases can be expressed respectively as follows:
8 = 9o (60 = constant) (7)
6 =« RT (R - constant)
where T is the total period of the cleaning cycle required to have
cleaning operation through all compartments. Also, the allowable
repeated number of cleaning operations is about 10-20 thousand
although it may depend on filter material and cleaning type. The
lifetime is 1~2 years for ordinary glass fiber and 2~3 years for
mixed fiber.
Now the filter cloth replacing cost can be expressed by the
following equation.
C^ = fSe/8 (9)
where f is a cost of filter cloth and is about 1,000 yen/m2 except
for certain more expensive ones.
1.5 Cleaning Cycle and Operational Pressure Drop
The pressure drop through the bag filter can be expressed as
a sum of pressure drops of filter cloth itself and of dust load.
Ap = u(a + bmq) (10)
410
-------
where a, b and q are constants determined by experiments. The
values of constants are different if they are determined by above
equation using new filter cloth or old filter cloth which was used
for some cleaning operations and the constants used in this report
are those of latter case. Generally the pressure drop due to
filter cloth itself is considered to be minimal and we assume
a =2=1 o by considering that the big advantage of this is the optimal
solution can be analytically obtained. The case when the value of
a cannot be neglected is reviewed in Section 2.3. Now Equation 10
can be expressed as follows:
Ap ^ bumq (11)
When the filter is constantly operated in a certain cleaning cycle
(it is generally called as a timer type) , a relation between cleaning
cycle T and operational pressure drop Aps can be expressed by the
following equation.
Aps^=i psbu(cnuT)q (12)
On the other hand, when it is constantly operated with a certain
maximum operational pressure drop Aps (it is called as a differ-
ential pressure type) , the cleaning cycle T is expressed as
follows.
where Aps and NTS are functions of the constant q which can be
determined by cleaning residual rate £, number of bag compartment
N and characteristics of filter cloth and dust.
(Aps)'/q
+ N<
q/1+q~"1/q
~l"1
J
Since the relation between ps and T in Equations (12)-(14) is the
same for both timer type and differential type; the following
solution can be applied for both operations. Also, the dust load
in the bag compartment immediately before the cleaning operation
(Xfj3) can be obtained by the following equation.
(l-C)xN3 + cnuT (15)
411
-------
1.6 Annual Total Cost
Based on above/ the annual total cost can be obtained.
Cm = (Ki+K2)Ap -Q+KjSq3+feS/e (16)
* s
It was assumed here that the pressure drop through the duct does
not affect the optimal solution and that the air pressure of fan
is equal to operational pressure drop of bag filter. Although the
above equation has three variables (filtering air velocity,
operational pressure drop and cleaning cycle) , it is actually a
two variables function since there is a relationship as equations
(12) and (13) for these variables. Also, when the lifetime of the
filter cloth can be determined by its operational time as described
later, it will be a one variable function since the cleaning cycle
can be determined easily.
1.7 Calculation of Optimal Solution
Now we have a simple consideration concerning annual cost.
First, when operational pressure drop Apg is assumed as a
constant and the filtering air velocity u is changed, the fixed
cost of bag filter is cheaper with large u but the replacing cost
of filter increases with frequent cleaning cycles. Thus, the optimal
filtering air velocity exists. Similarly when the filtering air
velocity is assumed as a constant, the optimal operational pressure
drop exists.
When a = o, the optimal condition can be analytically obtained
from Equations (12) and (16). However, there are two analytical
solutions (2-1 and 2-2) for two filter cloth lifetime equations
(Equations (7) and (8). Also, even if there are upper and lower
limits in variables Aps, u and x^j, solutions can be easily obtained
(2'l'lr2«l«3). However, when the lifetime is determined by
operational time 80 (one variable optimal problem) , the solutions
are naturally the same as limiting values when variables Aps. u
and XNJ exceed upper and lower limits. Thus, it is not included
here.
2.1 Optimal Solution When Lifetime of Filter Cloth is Determined
by Number of Cleanings (a *=t o)
Place 3CT/3u = 3CT/3Aps = o in Equation (16) and substitute
Equations (8) and (12) into this to obtain solution. However,
there are two solutions because the fixed cost equation of bag
filter has different coefficients at S = 300 m2.
i) S V 300 m2.
412
-------
qqs/a
m i J-e. i
P8,0pt \Kt
U - ffeJ
"opt \ Rq J
(17)
-q/a
U+q+qs)/a
(19)
K2)bApc(cn)q}q8/a(q,K3)1/aQq3(2*q)/a (20)
where 3 = 1 + q3 + qq3 (21)
The parameters in the first parenthesis on the right side of
Equation (20) indicates that the ratio among the fixed cost and
power cost of fan, the fixed cost of bag filter and the filter
cloth replacing cost is l:(l/qa):q.
ii) S = 300 m2 (=S. )
b
(22)
413
-------
uopt = Q/Sb (23)
UK
a lfo+1
K2)bApg(cn)qr1(q+1 (-) (24)
q/(q+D
{(Kj
a l/fa+:M/0\q <*
+ K2)bAps(cn)q}1/(q+1)(#-) (25)
\ b/
The result of numerical calculation for various Q and en is
shown in Figure 2. The values of used constants are those standard
values described earlier and parameters fe/R = 300 and bApg = 10s
are for cost of filter cloth f = 0.2 ten thousand yen/in2, annual
operational time c = 8,000 hr/year, allowable repeated times of
cleaning operation R = 2 x 101* and for these constants in
Equation (12) such as operational pressure drop Aps = 200 mm H20,
collecting dust amount cnuT = 0.1 Kg/m2 and filtering air velocity
u = 0.02 m/sec, and these are standard values. According to the
Figure, the effect of Q to the optimal solution is minimal except
for a small Q but the optimal condition of operational pressure
drop and filtering air velocity needs to be changed for dust con-
centration of carrying gas.
2.1.1 When Filtering Air Velocity is Limited
When the filtering air velocity has upper limit u or lower
limit u from restrictions of filter collection performance and
system design and when the solution Uopt obtained .with assuming
no restrictions (limits) exceeds these upper and lower limits, the
optimal solution is when Uopt is consistent with U or u. Therefore
n or u needs to be sutstituted for Q/Sb in Equations (?2)-(25)
2.1.2 When Operational Pressure Drop if Limited
When the operational pressure drop Aps has an upper limit
Aps from filter cloth strength and when the solution obtained with
assuming no limits exceeds Ap~8, the optimal solution is when ApOpt
is consistent with Aps. Other solutions can be obtained by dif-
ferentiating Equation (16),
414
-------
uopt " \RqqsKJ/
-q/d+qqs)
bAps(cn)
q .-l/d+qqs)
B
Qq(q3-i)/d+qqs)
(26)
opt
Q(q+l)d-qi)/d+qqi)
(27)
T,min
x Q
qsd+q)/(l+qq3)
bAp (CT))
o
q
qa/d+qqt)
(28)
Figure 8 shows the result of this calculation. This figure indi-
cates no significant changes due to limits in operational pressure
drop. Especially T0pt is not affected by this.
2.1.3 When Dust Load is Limited
There may be some limits in dust load on filter cloth in
order to have an effective cleaning operation. Therefore, we
review the case when there is generally an upper limit H or lower
limit M and the dust load obtained with no limits exceeds these
values.
The following optimal solution can be obtained from Equation
(16) under a condition of Equation (15).
415
-------
s,opt
(29)
u
opt
(30)
*-Ns
opt
cri
l/(H-q3)
Q(l-q3)/d+q3)
XNQ = M, or M
(32)
T,min
feQcn
(33)
VNS
The result of this equation is shown in Figure 9. As seen
from the above equations, under a conditon of constant dust load,
Aps,opt and Uopt are constant and are independent with en and Topt
is inversely proportional to en.
416
-------
2.2 Optimal Solution When Lifetime of Filter Cloth is Determined
by Operational Time (a No)
In this case, the minimum cleaning cycle Tmj.n possible to
have for the system will naturally be the optimal solution. There-
fore it is a one variable optimal problem. The following equation
can be obtained by substituting Equation (12) into Equation (16)
and setting 3CT/3u = o.
q 3K3Qq 3"
(34)
By substituting the optimal solutions U0pt obtained from above
equation into Equations (12) and (16), Aps,0pt and CT,min can be
determined respectively. And the following approximate solution
can be obtained by
( o3*3^8"1 U/(l+q+q3)
U '=i\ - 9 K g - [ (35)
opt
Ap >{bAp (cnT .
^ *s ' mm
d+q) (K,+K2)
C
T,min I q3 6
.
mm
(37)
417
-------
Figure 3 shows an example of numerical result for various Q and
en by using Equation (34). Also it shows approximate solutions
by Equations (35)"(37). It is seen from the Figure that the
approximate solution is a good approximation for Equation (34).
The parameter fe/9o = 0.05 in this Figure is for f = 0.1 ten
thousand yen/in2, e = 8,000 hr/yr and 60 = 16,000 hr.
2.3 Optimal Solution When Pressure Drop of Filter Cloth Cannot
Be Neglected.
As discussed in Section 1.5, when the pressure drop Equation
(10) is experimentally determined, the value of a may not be
neglected. In this case those solutions obtained in Sections 2.1
and 2.2 cannot be obtained and you have to obtain solutions by an
iterative method. Now the pressure drop Aps through multi-
compartment bag filter can be expressed by following equation.
Ap = Ap u{a+b(cnuT)q} (38)
8 S
Also Aps cannot be obtained by Equation (14) and has to be deter-
mined numerically by the model in another report.
The calculational result of the case when the lifetime of
filter cloth is determined by cleaning numbers is shown in
Figure 4 with variation of the value of a. a = 1,000 is an
according value for Apsau = 40 mm H2O when u = 0.02 m/sec and
Apg a 2. The figure indicates some effect of neglecting a when
dust concentration is small but the tendency of solution does not
change. On the other hand, Topt seems to be affected considerably.
3. Effect of Parameters to Optimal Solution
Since the solutions in Equations (17)~(20) are most important
among those solutions obtained in Section 2, we will calculate
a sensitivity of each parameter to optimal solution by using these
analytical solutions.
3.1 Effect of Coefficient Parameters KI, K2, K3 and fe/R.
Table 1 shows the relative sensitivity of parameters (Ki+K2)
Ks and fe/R to the optimal solution using Equations (17)-(20).
Figure 5 represents the calculational results for q3 = 0.89 and
0.61 by varying q. Based on the figure, the relative sensitivity
to annual cost is about 0.3 from (Ki+Kz) and fe/R and is about 0.4
from Ks, and it is indicated that the annual cost is changed in
a same degree with changes in each cost such as the fixed cost and
power cost of fan, filter material replacing cost and the fixed
cost of bag filter. Topt is considerably affected by fe/R and K3,
but not affected by (Ki+K2).
418
-------
In order to show how the optimal solution is affected by
improving filter material specific cost and endurance limit, the
calculational results for several fc/R is shown in Figure 6.
Namely, by reducing fe/R, u0pt will be larger and Aps,opt be small
and thus it will be possible to reduce the size of system.
3.2 Effect of Operating Conditions Q and en
The relative sensitivity of Q and en to optimal solution,
from Equations (17)~(20), is in Table 1. Figure 7 shows the
calculational results for q3 = 0.89 and 0.61 by varying q. The
annual cost is, according to the Figure, proportional to the 0.8
power of process gas amount for a small system and to the power
of 0.9 for the big system. Also Topt cannot be affected by Q and
en except for the case of small gas flow amount (qa - 0.61).
Aps,opt is roughly proportional to (en)°*3 and u0pt is inversely
proportional to (en) . This was also shown in Figure 2.
3.3 Effect of a, b and q of Pressure Drop Characteristics Equa-
tion
The relative sensitivity of b to optimal solution obtained
from Equations (17)~(20) is shown in Table 1 and Figure 7. T
is not affected by b while Aps,opt and CTfinin are roughly pro-
portional to 6°'3 and u0pt is inversely proportional to 60*11.
This is shown in Figure 8.
The effect of a to optimal solution, as already discussed
in 2.3, cannot be neglected when en is small.
Figure 9 shows the example the effect of q. It is seen
from the Figure that q has a significant effect. Thus, it is
important to accurately obtain pressure drop characteristics,
including a and b for determining optimal operating condition.
3.4 Effect of Cleaning Residual Rate £ and Bag Compartment
Numbers N
Since the parameter Aps is a function of bag compartment
numbers and cleaning residual rate, the effect of N and £ to
optimal solution was obtained as a parameter of Aps and the
result is shown in Tables 2 and 3.
According to Table 2 which shows changes in optimal solu-
tion when only N changes, it is noted that N does not have any
effect. Especially, the annual cost can only be reduced by
multi-compartment with at most 20% for £ = 0 and by several
percent for £ = 0.8.
According to Table 3 which shows changes in optimal solu-
tion when only £ changes, it is noted that 5.has a considerable
effect. Especially, when £ is large, it will require the annual
419
-------
cost of twice as much as that of 5 - 0. However, Topt is not
affected by £ when Q is large (q3 = 0.89)..
Conclusion
We have derived the equation showing that the annual cost
of the bag filter consists of the fixed cost cost and power cost
for the fan, the fixed cost of bag filter and filter cloth re-
placing cost. We assumed that the lifetime of filter cloth can
be determined by cleaning times (numbers) and operating times.
Since analytical solutions can be obtained when the pressure drop
of filter cloth itself is neglected, the analytical solutions were
obtained for various cases including the case when there are limits
in operational pressure drop, filtering air velocity and dust
load.
Then numerical solutions were obtained for a practical case
using constant values obtained from users and manufacturers. Also
relative sensitivity was calculated in order to estimate the
effects of each parameter on the optimal solution.
The following are major conclusions obtained by this research:
1) The optimal operating condition is not affected by gas flow
amount except for the case of small flow amount a about
1 m3/sec. Also the annual cost is roughly proportional to
gas flow amount with a power of 0.8 for a small system and
with 0.9 for a large system.
2) The optimal operating condition is affected by dust concen-
tration in process gas except for the optimal cleaning cycle.
Namely the optimal operating pressure drop and the annual
cost per unit flow amount increase with an increase in dust
concentration, while the optimal filtering air velocity
decreases in an increase in dust concentration.
3) The ratio among the fixed cost and power cost of fan, the
fixed cost of bag filter and the filter cloth replacing cost
in the annual cost is I:(l/q3):q3.
4) The relative sensitivity of proportional constants Kj, K2 ,
Kj and fe/R of each cost to the optimal solution is about
0.3-0.4. However, the relative sensitivity of (Kj+Kj) to
optimal operating pressure drop, the relative sensitivity
of K3 to the optimal filtering air velocity and the relative
sensitivity of K3 and fe/R to the optimal cleaning cycle are
big and are about 0.8-1.2 in absolute value.
5) Since the pressure drop characteristic equation (Equation (10))
has a significant effect in optimal operating condition, a
determination of pressure drop equation is important.
420
-------
6) An increase in number of bag compartment does not change
optimal operating condition. Especially the annual cost can
only be reduced by multi-compartment with at most 20% for
£ = 0 and with a several percent for £ = 0.8.
7) The effect of residual rate to the optimal operating condi-
tion is considerable and the improvement in residual rate has
a significant effect in reducing annual cost.
Example Problem
Assume that the cost f of the selected filter cloth is 0.4
ten thousand yen/m2 including manpower cost for replacement and
that the allowable repeated times of cleaning operation is 20,000.
Also the pressure drop equation can neglect the pressure drop at
no dust load and assume Ap = 50,000 um (a = 0, q = 1, b = 50,000).
Number of bag compartment N is 4, collection efficiency n is
100% and cleaning residual rate £ is 0.6. When the operating
condition is at the process gas amount Q = 10 m3/sec and the dust
concentration c = 2 q/m3, determine the economically optimal
filtering air velocity and operating pressure drop.
Solution
Assuming the annual operating time e is 8,000 hr (2.88 x
107 sec), fe/R is fe/R = 0.4 x (2.88 x 107)/(2 x 10") '=i 600.
Using the given condition (N=4, a=0, q=l and £ = 0.6),
Aps *=• 2 from Equation (14). Therefore bAps =5 10s. The follow-
ing optimal conditions are obtained using Figure 6 or Equations
(17)~(21). Also, the constants discussed early in this report
were used in this calculation except those constants given in
above problem.
Aps,opt = 153 mm H2O (Maximum Operating Pressure Drop)
uopt =1-0 m/min (Average Filtering Air Velocity)
T0pt = 45 min (one cleaning cycle)
XNs,opt =0.23 Kg/m2 (Maximum Dust Load Before Cleaning)
CT,min = 411 ten thousand yen/yr (Annual Total Cost)
Based on these values, the lifetime of filter cloth 0/e is
6/e = RTopt/e =* 1.9 year and the filtering area S is
S = Q/u0pt = 600 m2.
421
-------
NOMENCLATURE
a,b,q = constants used in Eg. (10)
Ci = fixed cost of fan, including maintenance cost [101* YEN/year]
C2 = electrical cost of fan per a year [10" YEN/year]
C3 = fixed cost of bag filter, including maintenance
cost except C.J [10" YEN/year]
Ci, = renewal cost of filter cloth per a year [10" YEN/year]
CT = total cost per year [10" YEN/year]
c = dust concentration at inlet [kg/m3]
e = electric power rate [10" YEN/kW-hr]
f = price of cloth [10" YEN/m2]
hifk^Ki = constants used in Eqs. (1) and (3)
h2,K2 = constants used in Egs. (2) and (5)
h3,K3,k3, q3 = constants used in Eq. (6)
m = dust load on bag filter [kg/m2]
N = number of compartments I—]
P - power of fan [kW]
Ap = pressure drop [mmAq]
Aps ' = pressure drop across bag filter [mmAq]
Aps = dimensionless pressure drop defined in our previous
paper [—1
Q = gas flow rate [m'/sec]
R • allowable repeated times of cleaning operation [—]
S * filtering area of bag filter [m2]
T = total period of cleaning cycle [sec]
u = average filtering velocity (=Q/s) [m/sec]
XNS = maximum dust load on bag through all compartments [kg/m2]
422
-------
Y - cost of equipment [10" YEN]
a = parameter defined by Eg. (21) [—]
e = operating time for a year [sec/year]
n = collection efficiency [—]
8 = lifetime of filter cloth [sec]
80 = allowable duration time of filter cloth [sec]
T = dimensionless partial period of cleaning operation
defined in our previous paper [—]
£• = residue fraction of dust load after cleaning [—]
min - minimum
opt = optimal condition
~", = upper or lower value
423
-------
PERFORMANCE OF A MICRO-CYCLONE*
Koichi linoya**, Aishi Nakai***
1. Introduction
The cyclone, having a simple configuration and low instal-
lation cost, has been widely used in various fields but has a
disadvantage of low efficiency. It is also not known how much
the collection efficiency can be improved by significantly re-
ducing the size of cyclone. Although it is believed that greatly
reducing its size is unrealistic, the measurement of the micro-
cyclone performance, which we have designed, has given interest-
ing results.
2. Experimental Apparatus and Method
Figure 1 shows the size and the concept of the cyclone and
Figure 2 shows a flow sheet of the experimental apparatus. The
sample aerosol is produced by generating stearic acid particles
through the Sinclair-Lamer type monodisperse particle generator
and by mixing them well with clean air (about 10 1/min) in a
mixing chamber. The aerosol is sent to the cyclone entrance
through the high concentration side of dust concentration meter,
and passes through the low concentration side of dust concentra-
tion meter, filter paper and glass orifice by the action of the
vacuum pump. The dust concentration meter (indicator) used in
this experiment is the relative concentration indicator (Shibata
Chemical Instrument Industry Co.) using light scattering. The
glass orifice was monitored by the wet test gas meter. The
milipore-filter was used to protect the orifice and vacuum pump.
The pressure drop across the cyclone was measured by a water or
mercury manometer.
3. Experimental Result and Consideration
3.1 Pressure Drop
The pressure drop across the cyclone can usually be expressed
as a function of cyclone inlet velocity.
* Report received on 1/21/69.
** Regular member, Department of Engineering, Kyoto University,
Yoshida Honmachi, Sakyoku, Kyotoshi.
*** Research Institute for Production Department, Morimotocho,
Shimogamo, Sakyoku, Kyoto.
424
-------
Fig. 1. A micro-cyclone.
425
-------
Difl-lol du»« iodicotor
Pig. 2.
chomMT
Vbcuumpump
•—Ctaonair
'-r-J From monodiMrM
Mretol ommtar
Manometer for ibeolute prtmirt
Manometer for prenure drop
M»nomrt
-------
Ap = F x
In this case, the cyclone is used on the suction side of vacuum
pump (namely, with negative pressure) and the inside of the
cyclone is under a considerable low pressure as compared to am-
bient pressure. Therefore, the measured pressure drop Apm is
actually the apparent pressure drop and it is desirable to correct
the specific weight of gas at inlet (y) to a standard condition.
This correction can be done by the following equation.
= APm ^ (2)
As shown in Figure 3, the relationship between the measured Apm
and U^ is that Apm is not proportional to the square of Ui as indi-
cated in equation (4). But Aps which is corrected by equation (2)
is proportional to the square of Uj_. If this correction is made
for the case when the cyclone is used in a considerable range of
pressure, it will be seen that the pressure drop is proportional
to the square of inlet velocity. On the other hand, there have
been many simple equations reported to estimate the pressure drop
coefficient F from the configuration of cyclone, and when F is
calculated by using the typical linoya and First's Equation, the
value of pressure drop coefficient Fs = 20.5. Also the Reynolds
number Rei at cyclone inlet is about 103-10I* which is the range
of turbulent flow.
3.2 Collection Efficiency
Figure 4 shows the relationship between inlet velocity Uj,
and collection efficiency n as a parameter of particle size using
the experimental results. Based on this, it is indicated that
the ordinary cyclone inlet velocity (about 20 m/sec) with a
smaller cyclone will lead to about 90% of collection efficiency
for about 1 y dust particles (Pp = 1.0 g/cm3) and that the high
velocity (50 m/sec) will even lead to 90% of collection efficiency
for 0.25 y dust particles without having an efficiency reduction
for the high velocity. Therefore, if many small cyclones can be
designed to operate in parallel, it is possible to have a high
efficiency multi-clone.
On the other hand, only the inertia force has been considered
to be the controlling factor in the collection mechanism of the
cyclone. But when the collection efficiency n of this experiment
was plotted as a function of inertia parameter
,. _ PPUiDPCm
* ~ 18yD '
it will give a different curve for a different particle size.
When n is plotted as a function of V/(Dp/D) or f/Nsc, it gives a
427
-------
c Mcoturvd tpK
• Modified »P.
I03 5 i I0~203050
InM v«locily u.lnviac)
Fig. 3. Rebtion between inlet velocity ut
and pressure drop Jp.
428
-------
Fig. 4.
10 20 30 4O
Inlet velocity Uj (m/fcc)
Relation between Inlet velocity
and collection efficiency.
429
-------
single curve. When it is plotted on a log-log paper, one linear
graph is obtained as shown in Figure 5. In this case the effect
of particle concentration may be considered as another factor
and it may be necessary to consider the particle diffusion phe-
nomena as one of the collection mechanisms. This is considered
to be a question in the collection mechanism of cyclone.
4. Conclusion
As a result of measuring the performance of sample micro-
cyclone, the following conclusions were obtained.
1) It was confirmed that the pressure drop is accurately
proportional to the square of inlet velocity, by an appropriate
correction, even when the cyclone is used over a considerable
large pressure range.
2) Even submicron particles can be collected with a con-
siderably high efficiency for an extremely small cyclone and a
high velocity.
3) Although it was believed that the inertia force was the
only controlling factor in the collection mechanism of cyclone,
it was recognized that the effect of particle size also needs to
be considered.
430
-------
. ^
1
g
V
Vo
V*
1
\
(
i
\
5
10;
* ?oi
" 40
I60
t *>•
1 sot
i
99
10 2030 5070100 200
Modified ir«Hia porarr^if ^/(C^/D)
Fij. 5. 7 and V/(Df/D') (tog-normal distribution).
431
-------
International Seminar on
Dust Collection
I Date and Time
May 24 (Tuesday) 1977, 1.00p.m. - 7.00p.m.
II Place
Kyoto International Conference Hall
Takaragaike, Sakyo-ku, Kyoto, Japan 606
Phone: 075-791-3111
III Lectures 13.00 - 15.30
1. Introduction K. linoya, Professor, Kyoto University
2. Concept of Research in Particle Gas Separation
F. Loffler, Professor Institut fur Mechanische
Verfahrenstechnik der Universitat Karlsruhe,
Richard-Willstatter-Allee, 7500 Kerlsruhe, 1,
Germany 434
3. New Ideas in Electrostatic Precipitation Technology
E.G.Potter, Leader
Process Chemistry Section, Division of Process
Technology Minerals Research Laboratories,
Commonwealth Scientific and Industrial Research
Organization, Australia, P.O.Box 136, Delhi Road
North Ryde, N.S.W. Australia, 2113,
4. Current Research on Particle Removal at the Harvard
Air Cleaning Laboratory
M. First, Professor, Department of Environmental Health
Sciences, School of Public Health, Harvard University,
665 Huntington Avenue, Boston Massachusetts 02115,
u. s. A 452
5. Size-Selective Aerosol Collection with Centrifuges,
W. Stober, Professor, Tnstitut fur Aero Biologie,
5948 Schmallenberg-Grafshaft, Germany
432
-------
IV Coffee Break 15.30 - 16.00
V Discussion 16.00 - 17.00
Chairman, S. Masuda, Professor Tokyo University
Co-chairman, T. Yoshida, Professor Osaka
Prefectural University
VI Party with International Communications
17.30 - 19.00
Organizer, K. linoya, Department of Chemical Engineering
Kyoto University, Sakyo-ku, Kyoto, 606 Japan
Phone: 075-751-2111 ext. 5566 (5586, 5576)
Research Activities on Dust Collection 453
Directory of Foreign Attendants 471
Attendants (Japan) 473
433
-------
Concept of Research in Particle-Gas-Separation
Prof.Dr.-Ing.P.Lfiffler
Institut fUr Mechanische Verfahrenstechnik der UniversitSt Karlsruhe
Particle-Gas-Separation is considered as one of the important unit
operations of "Mechanische Verfahrenstechnik" (Process technology).
Apart from its technical significance it plays an important
role in general and political interest.
We are working since many years on the problems of separation of
particles (solid particles or liquid drops) from gases at the
Institute founded in the year 1957 by Prof. Dr.-Ing. Hans Ruxnpf
who unfortunately passed away a few months ago.
A section for dust collection and cleaning of air has been in
existence since 1965.
Research and development works are not pursued having only environ-
mental pollution control in view, but also in view of material
recovery within the processes of production.
434
-------
We are at present working on the following separating principles:
Centrifugal separator (cyclone), wet scrubber and fibre filter.
Besides this, some general topics from other fields other than
dust collection are also considered; for instance/ dust feeding
and desagglomeration of fine dust particles in gases, or agglo-
meration and deposition on the walls under the consideration of
the effect of electrostatic charges. The works done in the mea-
suring technique group will also be counted along with these
themes.
On the one hand, basic studies are being done in these fields
With an object of explaining the processes by a detailed study
of single phenomena and also to see if they can lead to reliable
methods for design,or to imporved or new techniques. On the other
hand, problems in practice in direct collaboration with the
industry are also treated. A brief illustration regarding a few
problems are given below and as far as possible with representation
of typical results.
Centrifugal separators - also called as cyclones - are simple
in construction, cheap and sturdy. They have become more popular
as they are reliable and safe to operate. But still the quota
of cyclones in the total turnover (1974) for dust collectors
was 15 % that corresponds to a higher part of flow rate because.
of the low investing costs of cyclones.
Problems lie in the proper design and in recognizing the possible
range of application. The particles are separated by the centri-
fugal force acting in a rotating flow. In the apparatus looking
simple outwardly a very complicated, three dimensional, turbulent
rotational flow predominates which evades an exact calculation.
This holds even .more when one considers the effect of material
on the flow.
Pressure drop,the collection efficiency and preferably the grade
efficiency curve are to be determined as characteristic parameters
for each of the separators. The grade efficiency curve gives the
collection efficiency in terms of particle size.
435
-------
Simplified models of flow field have to be used provided one does
not forego a calculation from the beginning and counts only on
the pragmatic value, and also realizes the Importance of cal-
culation possibilities for the design especially, the optimization
of the set up.
This was done by Rumpf and his co-workers |1j for flow without
dust. The object of the investigation that is running at present
is to verify such models experimentally. In particular the appli-
cability on dust loaded flows is verified. This is possible by
measuring the loss in the moment of momentum at different parts
of the surface of the cyclone to obtain the wall coefficient
of friction in terms of operating parameters,the particle size distri-
bution, dust load and rate of air flow.
Measurements of velocity in cyclone and the determination of the
grade efficiency curve will also be added to this. The works on
the very complicated experimental set up are }n operation and re-
sults will be reported in the near future.
Studies on a high efficient cyclone for the separation of fine
graphite particles from a gas flow is mentioned as an example
for a problem in practice. The object was to obtain a separating
curve as fine as possible.
In figure (1) the degree of separation T(x) is plotted against
particle size. V, the volume flow rate is the parameter. The
curves show the typical behaviour for cyclones. As usual the
cut size is defined as those particle sizes from which SO %
are separated. The figure shows that a cut size of about
2 urn was obtained. However, it was hardly possible to remain still
under this value, although the theoretically -predicted values
were below 1 yra. This experience, on one hand, points out the
necessity of demarcation of experiments from the theoretical
models/and on theother hand shows an important effect of turbulence
on the separation. Turbulence acts against separation particularly
436
-------
Nr.
A 84/85
•81/63
» 77 /7a
-------
This condition is not only unsatisfactory scientifically, but ic
also disadvantageous with respect to the design of collectors in
practice; moreover, the calculation of optimization cannot be
made. Our object, therefore, is to explain the physical fundamental
phenomena so that we could arrive at the equations of design.
The investigations take place in three stages: Single drop processes,
behaviour of drop swarms, effect of collectors.
Till now we were dealing with the experimental work particularly
about the production of drops with the specified size and velocity,
and intensively with the measurement of drop size distribution, by
different methods as well. The chosen range of drop size ranges
from some hundreds of ym till about 0.5 yra.
Theoretical collection efficiency calculations for single drops
were verified experimentally by suspending the drop on a 5 pm
diameter wire and photographing the paths of the dust particles
with the help of a high speed camera (till 320OO pictures/sec.)*
It could be observed directly that all the impinged dust particles
on the drop could not be retained there, some of them were
rebounding.
Experimental results of the studies made on a scrubber, designed
together with Prof. Leschonski, Institut fUr Mechanische Ver-
fahrenstechnik, Technische Universitat Glausthai, agreed with the
theoretically predicted high collection efficiencies.
Fibre filters are employed in the technology in two basic forms.
In the case of deep bed filters the separation takes place within
the fibre layer that has passed through the flow. After a particular
dust embedment, that is, when a predetermined pressure drop has
reached the filters are extended and are thrown away frequently.
Some types can be cleaned by washing or by blowing put. This
filter, therefore, can be used mainly in the case of low dust
contents up to a few mg/cm , so in the ventilation system for
laboratories, production plants, (clean rooms) or operation rooms.
438
-------
The second type o£ filters are the surface or uleanable filters.
Here the separation takes place mainly on the surface of the
filter cloth. A dust layer (filter cake) will be build up which
separates effectively. Because of the raise in the flow resistance
this dust layer which occurs periodically during the operation by
different methods has to be removed frequently. This filter can be
employed in the case of high dust content (up to a few hundreds
of g/cm ) and also in the branch of industrial dust collection or
in production recovery. There, it has been proved as the most
efficient fine dust collector.
Deep bed filters are highly porous filters in which the part of
fibre volume amounts to less than 1 % and seldom more than 5 %.
The spacing of fibres are big compared with that of particle size.
Therefore the particles are not separated by sieving, but they
will have to be brought to the fibre surface by different transporta-
tion methods and then must be retained there. The collection
efficiency is* therefore, a product of degree of impingement
and probability of adhesion.
We are trying to explain the transport and adhesion phenomena by
our studies and investigations using single fibre, fibre grids as
model filters and also on technical filters. At present we are
concentrating on the particle size range say above 0.5 - 1 pm
and flow velocities from say 10 cm/sec, onwards, and in this
range the Brownian movement of the particle can be neglected.
Forces of gravity, inertia and electrostatic forces are at the
disposal of transport phenomena.
Muhr |2,3| in his theoretical and experimental studies, was able
to point out the important significance of electrostatic charging,
in particular for the separation of particles, below say 3 yro and
he could describe this quantitatively.
Fig. (2) shows some curves of an exai pie where single fibre
collection efficiencies */* , that were calculated back from the
measured values for the total model filter are plotted against
the so called inertia parameter \l>. In this representation the
439
-------
Reynolds Number calculated using fibre diameter and free flow
velocity is the parameter. For comparison two theoretical curves
a.re drawn where electrostatic forces are not taken into account.
We can see that in the range \J> < 1 the measured collection
efficiencies are much higher than the predicted theoretical
ones for the particles and fibres that are free of charge. These
high values can be accounted for with an explanation that the
particles as well as the fibres are charged and therefore the
Coulomb-forces had an effect.
t.4
1.3
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
5.85
2.75
2.15
1.25
symbol
X
o
particle and fibres
charged
Stenhoust
o=0.08 Re~0
0.0 0.1 0.2 0.5 1
10
. m Pp Dp Up
"""**" 18 v OF
Fig. (2): Collection of NaCl-particles in model filters
This was proved by the measurements of charge as well as by the
experiments with discharged particles and fibres/ and was also
established by the theoretical model calculations.
In addition to this it can be seen in fig. (2) that at the
increasing particle size and/or increasing velocity the collection
efficiencies were lower than the predicted theoretical ones -
440
-------
this in particular can be clearly seen on the curve tor Ke = 2.5.
This, as we already know from our experience in practice and
from other investigations, would lead to the ovservatjon that the
particles bigger than 3 pm do not positively adhere to the fibre
surface so that the probability of adhesion is less than 100 %.
If the particle has to remain adhere, firstly it should not
rebound and secondly it should not be detacted subsequently.
In order to explain the second stipulation we made a detailed
study on adhesive force some time ago and found out that when
once the separated particle adheres it can be hardly detached 14 | .
In one of the works that has been completed recently, resistance
and buoyancy forces exerted by a flow on the sphere-shapec. particle
deposited at a surface have been determined 151. The results of
this investigation are not just restricted to the filter problem,
but, is can be»for instance,applied to the wall depositon in
pipes and machines.
Measurements on filter and adhesive forces showed clearly that
the rebound action is very decisive. Vie, therefore, developed
an equipment with which we could study the impact process on a
single fibre photographically. The films could be evaluated
quantitatively. An example of such a result is shown in fig. (3).
In this figure flow velocity v is plotted against probability of
adhesion h, where h is defined as the ratio of the number of
adhering particles at the first impingement to the total number
of impinging particles. Experimental parameters were particle
type, its size and fibre type.
Particles begin to rebound obviously at velocities between 5 and
15 cm/sec. Probability of adhesion decreases quickly with the
increasing velocity and reaches below 10 % at 1 m/sec for 10 um-
particles. 5 ym-particles rebound more than 10 ym-particles.
441
-------
Round glass spheres rebound more vigorously than the irregularly
shaped quartz particles which is obviously an effect of geometry.
Quartz particles can have multi-point contact or surface contact
Whereas' glass spheres can have only single-point contact.
Studies on solid wax particles and oil drops of sizes from 5 to
10 urn showed surprisingly that partly these particles also show
elastic behaviour and rebound likewise in the velocity range- shown
here. These investigations will be continued.
l\l \\ I0 ^
Potyomidfasar 20|jm
x Quarzpartikel
o Gtaskugeln
Glasfascr 20pm
10 20 30 40 50 60 70 80 cm/s 100
Fig. (3): Probability of adhesion of rigid particles
on a polyamide-fibre
In addition to the studies on single fibres and fibre grids, we
are making some studies on the separating behaviour of commercial
filter media for different types of particles in a filter testing
apparatus. At present the separation of clouds (oil clouds, fat
droplets) carries and special weight.
442
-------
In the branch of cleanable filter, an experimental set up is
being arranged with an object of studying the cleaning behaviour
of filter bags. As already mentionedithe pressure drop raises
with increasing dust deposition and at the same time the collection
efficiency becomes better. Now when it is cleaned the collection
effciency also falls down \6\ . Hence,in this work which is very
close to the conditions in practice optimum cleaning conditions
are investigated for filter bags in technical dimensions.
Dosage and dispersion of solids are not only important for sciencit-
fic studies,but also from the technology point of view they have
a significance. Here the difficulties arise especially when the
particle size is less than 1O urn, as the adhesive forces for
these particles are bigger than their own weight by a few orders
in magnitude. Therefore, these particles have a characteristic
tendency for agglomeration and adherence.
Since the dust feeders available in the market or the ones
displayed in the publications do not function satisfactorily in thf;
range below 1O jm, we developed a device for dust feeder whose
schematic "representation is shown in fig. (4).
Luft
Fig. (4): Schematical diagram of the dust feeder
443
-------
Solid is carried away by a rotating brush and a jet of air by
the surface of the packing. A piston takes care of the dust
supply. Using this principle mass flows in different quantities
between O.01 and 6000 g/h with higher stability can be produced (7|,
Investigations on the conditions for desagglomeration using forces
of flow or wall impact |8| on a baffle plate showed that the
agglomerates could be effectively desagglomerated from the primary
particles of 2 ym on a baffle plate at velocities of 200 m/sec,
whereas this was possible without a baffle plate only up to
the primary particles of 5 ym. The required jet velocity decreases
with the increasing particle size.
In the measuring technique branch of our studies we emphasize on the
optical methods that are suitable for the determination of the
distributions of particle concentration, particle size and velo-
city, and also on the visuadization of the movement of flow and
particles. Partly it is also to investigate the applicability and
the limit of error of the known methods and if the necessity
arises to adapt them to our problems. Partly some methods and
instruments are also developed up to commercial specification.
This often contains the clarification of basic mechanisms.
Studies were made and are still beiiig made on the following
methods individually:
- Photometric determination of concentration in particle
deposits and in flowing gases |9|;
- Particle size determination using scattered light techniques |10J;
- Measurement of flow velocities according to spark-tracing
method |11j;
- Particle velocity measurement according to Laser-Doppler method;
- Determination of the paths and velocities of the particles using
high speed photography ;
- Determination of particle concentration, size and velocity
using impulse holography.
444
-------
Some hints v/ith regard to only two of these methods are given.
The method of photometric determination of concentration is
employed technically, say for dust deposition in a standard
filter test technique and for the emission control of slacks.
The well known Lambert-Beer law is the basis for this method.
Difficulties always arise when the concentration and particle
size distribution change simultaneously, and if these changes
are uncontrollable which, for example is often the case in
brown coal-power plants. With this the basis of calibration
also changes. A highly uncertain inference on the dust content
is bound to be drawn in casa of not carrying out additional ex-
perimental runs. This was verified by our experiments.
Some special advancement has been made with regard to the develop-
ment and application of scattering light method for the determination
of particle size distribution. This method can be used directly
in the flow to make a very quick measurement of the local size
distribution of particles larger than 0.3 pm and concentrations
up to 10 particles/cm , and this is possible without the particles
are being separated. The particle size distribution is obtained
by measuring arid analysing the impulse of the scattering light which
are produced by single particles while passing through a
small and pure optically circumscribed measuring volume in flow
field. The theory of Mie is the basis for this which is about the
scattering of light on spherical homogeneous particles. The
special feature of this method (apart from the fact that the
distribution state of the particle is not disturbed) is that the
particle distribution can be seen at high measuring velocity.
Depending upon the concentration and velocity nearly 100 000 particles
per minute can be counted and measured. Very often it is sufficient
to count small quantities of particles for the analysis of the
processes so that the measuring times could lie in the range of
10 sec. The evaluation and graphical representation of the measurement
can be done very quickly as the processing of the impulse can be
done with the help of a calculator and also electronically.
445
-------
Typical example? of application are the investigation of dis-
persion and desaggloroeration processes, of dust injectors/ the
determination of drop size distributions of atomizer nozzles.
o
e
.c
o
M
Zwaistoffdiis*
Druck : a , b 4,0 atu
Otirchsatz : a 0,2llr/min
b 2,0 Itr/ min
0.1 0.2 0.5 1 2 5 10
TsilchengroRe x/um
23
50 100
Fig. (5): Drop size distribution of an atomizer nozzle
The cumulative number distributions of water drops that are pro-
duced in an atomizer nozzle is shown as an example in fig. (5).
The rate of flow of water was varied in this experiment. In the
above figure the measuring arrangement is sketched on the left.
If the number distributions represented here will be calculated
in terms of volume distribution a more clear difference can be
seen than that of shown in fig. (5). Quick evaporation of the drops
is a problem in the measurement of distribution of drops. Hence
an immediate measurement is absolutely necessary and this is
possible with scattering light equipment.
446
-------
Only a few examples are mentioned here from the numerous appli-
cation possibilities of this method. On the other hand,it should
be mentioned that more investigation is to be done for answering
some basic questions like for example, the effect of particle
shape. Some efforts regarding the further technical developments
are performed.
Nevertheless, with the already available equipment at the moment,
grade efficiency curves of the dust collectors down to 0.3 ym
can be obtained quickly and reliably. This was not the case till
now since in the previous methods and above all in the range
below 5 ym, uncontrollable changes because of agglomeration and
other processes had to be accepted.
Studies in this particle size range are necessary considering to-
day's high demand for fine dust separation.
References
|1| Rumpf, H., K.Borho, H.Reichert:
Optimale Dimensionierung von Zyklonen mit Hilfe vereinfachendsr
Modellrechnungen.
Chemie-Ing.-Technik 40 (1968) 1O72-1082
|2| Muhr, W.:
Theoretische und experimentelle Untersuchung der Partikelab-
scheidung in Faserfiltern durch Feld- und TragheitskrMfte.
Dissertation Universitat Karlsruhe, 1976
|3| Muhr, W., P.Lfiffler:
Abscheideverhalten von Faserfiltern bei elektrostatischer
Aufladung.
Maschinenmarkt 82 (;976) 669-672
|4| LSffler, F.:
Uber die Haftung von Staubteilchen an Faser- und Teilchenober-
flachen.
Staub - Reinhaltung der Luft 28 (1968) 456-462
447
-------
|5| Rubin, G., F.Loffler:
Widerstands- und Auftriebsbeiwerte von kugelformigen Partikeln
in laminaren Grenzschichten
Chemie-Ing.-Technik 48 (1976) 563
|6| Loffler, F.:
Abscheidograd und Druckverlust von Filterstoffen versohiecJencr
Struktur bei unterschiedlichen Bedingungen.
Staub - Reinhaltung der Luft 30 (1970) 518-522
|7| Zahradnicek, A., F.Loffler:
Eine neua Dosiervorrichtung sur Erzeugung von Aerosolen fuvj
vorgegebenen feinkornigen Feststoffen.
Staub - Reinhaltung der Luft 36 (1976) H. 11
|8| Zahradnicek, A.:
lint err;uchuny zur Dispergierumj von Quarz- und Kail k.ste.i.nfr.(!)tt..iojiO)i
inn Korugrcfiuuberoich O.5 - 1C) IJIQ in stromenden Gasen.
Dissertation Universitat Karlsruhe, 1976
| 9| Umhauer, H., F.Loffieri
EinfluO der Partikelgroficnvcrteilung bei der fotometrischon. bo-
stimiriung der Konzentration industrieller Staube in
Gas-Feststof f--Stroroungen
1 . Europaisches Symposium "PartikelmeBtechnik", Niirnberg,
17. - 19. September 1975, hrsg. von Prof .Dr.-Ing.tURxvmpf,
Karlsruhe und Prof.Dr.-Ing.K.Leschonski, Clausthcil, in
Zusammenarbeit mit der DECHEMA
K)| Umhauer, H.:
Ermittlung von PartikelgrdQenverteilung in AerosolstrSmuncjea
hoher Konzentration mit Hilfe einer StreulichtmeBcinrichtuny
CIT 47 (1975) 7, 297
448
-------
I 11 I Bernotat, S., H.Umhauer:
Applications of spark tracing-method tc flow measurements
in an air classivier
Opto-electronics 5 (1973), 107/118
|12| Loffler, F., H.Umhauer:
Eine optische Methode zur Bestimmung der Teilchenab-
scheidung in Filterfasern
Staub - Reinhaltung der Luft 31 (1971) 2, 51/55
449
-------
In the previous report the works done at the "Institut filr
Mechanische Verfahrenstechnik" in the field of dust collection
and its measuring techniques were described.
The great interest of the public in the field of environmental
protection led to a plenty of research projects in industry and
research Institutes. A review about these projects given in
"Umweltforschungskatalog 1976" which contains about 1300 poges may
not be complete, but it is impressiv and informativ. Publisher:
Uraweltbundesamt, Berlin. Many projects have been dealt with
special processes of production.
A list of the names of some research workers and Institutes which
is certainly not complete is given below.
It is emphasized that the nair.es given in the list are purely
incidental and are in no way in order of importance of the works.
Prof .F.Mayinger, Institut ftir Verfahrenstechnik, UniversitSt Hannover,
(Ventury scrubbers-studying, atomization and energy-dissi-
pation in the venturi-throat)
Lit.: F.Mayinger, W.Neumann, DECHEMA-Monographien
Nr. 1639-1669, Band 80, Teil 2, Seite 637
Prof.E.Weber, Institut fUr Mechanische Verfahrenstechnik,
Universitat Essen
(Wet scrubbing with emphasis to absorption of gases,
distribution of dust concentration in elektrostatic
precipitators etc.)
Prof.R.Quack, Institut fUr Verfahrenstechnik, Universitat Stuttgart
(Electrostatic precipitation)
Staubforschungsinstitut Bonn, Direktor Dr.A.SchUtz
(Testing of HEPA-Filters, development of dust measuring
equipment etc.)
450
-------
Dr.Schikarski, Kernforschungszentrum Karlsruhe
(Deep bed filtration in sand filters)
Prof.Stflber, Institut fur Aerobiologie der Fraunhofer-Gesellschaft,
Grafschaft
(Different projects on air pollution control and measurement)
Techn.Hochschule Leuna-Merseburg (DDR), Sektion Verfahrenstechnik,
Prof.Jugel
(Fibre Filtration with emphasis on bag filters)
Dr.Stenhouse, Dep. of Chemical Engineering, University of Loughborough
(Deep bed fibre filters, adhesion probability, i.e. similar
work as we are doing)
Prof.Papai, Institut fur Stromungsmaschinen, Technische Universitat
Budapest, Ungarn
(Big technical bag filters)
Dr.Benarie, IRCHA, Frankreich
(Different works on different subjects)
Prof.Dr.K.Leschonski, Institut fiir Mechanische Verfahrenstechnik,
Techn.Universitat Glausthal-Zellerfeld
(On-line measurement of particle size distribution)
Dr.J.Gebhart, Gesellschaft fiir Strahlen- und Umweltforschung,
Paul-Ehrlich-Str. 2O, 6 Frankfurt/Main
(measurement of particle concentration and size
distribution)
451
-------
CURRENT RESEARCH ON PARTICLE REMOVAL AT THE
HARVARD AIR CLEANING LABORATORY,
Melvin W. First
1. Fabric Filters
Our current fabric filter research program has three major objec-
tives: (1) Basic mechanisms of cake formation for a full range
of filtration velocity, pressure drop across the cake, and dust
characteristics that are typical of industrial fabric filter appli-
cations. Filter cakes have been fixed in place by infusions of a
liquid plastic monomer and, after polymerization, the rigid cakes
may be sectioned, polished, and examined under the scanning elec-
tron microscope to quantify the internal geometry of the cake and
relate this to the principal filtration parameters. (2) Reten-
tion and penetration characteristics of full-scale pulse jet
cleaned felted filter bags. Dust penetration has been found to
occur by straight through passage, by seepage, and by pinhole
plug losses from the formed cake. These several penetration mech-
anisms have been quantified in relation to filtration velocity,
particle size, and the period in the cycle since the last clean-
ing pulse and it has been found that different dust penetration
mechanisms predominate at different periods during the cycle.
(3) Decreasing filter size by increasing filtration velocity through
the fabric. Higher filtration velocities reduce equipment cost
but result in increased filter resistance and dust penetration
in conventionally constructed units. Substantially higher velo-
cities with low dust penetration depend upon the development of
inproved cake removal methods that are gentler and more efficient.
2. Incinerator Off-Gas Cleaning
Interest in hot gas filtration by moving granular beds is being
studied in our laboratory because it has several applications to
nuclear waste treatment. For conventional incineration of com-
bustible materials contaminated with low levels of radioactivity,
particle size will be relatively large and removal will be prin-
cipally by inertia; favoring high filtration velocities. The
temperature resistance and cleanability of mineral or ceramic
granules make them attractive for this service. Efficiency may
be regulated by selection of granule sizes, thickness of the bed,
and filtration velocity. Application of moving granular beds
for vitrification of high level wastes in preparation for storage
Is a more complex application because of the very high tempera-
tures generated (825°C) and the very small size of particles
formed by vaporization and subsequent condensation. Investiga-
tions are underway to employ a moving granular bed as a cooler
as well as a filter and, In the process, utilize the strong thermo-
phoretlc particle separating forces that can be generated inside
a bed of cool granules when a hot gas passes through it.
452
-------
RESEARCH ACTIVITIES ON
DUST COLLECTION
453
-------
Study cxi the moving behaviour of dust particles in a precipitator by laser
Doppler velocimeter
Research Institute of Applied Electxiai ty, Hokkaido University
Sapporo, Hokkaido
Professor Toshiinit su Asakura
Assistant professor Hircmichi MisKina
Research Associate Yasushi Kawase
Research Associate Yoshio Shindo
(1) Construction of laser Doppler velocimeter for measurements of the moving
velocity of dust particles in a precipitator
A laser Doppler velocimeter with high spatial resolution and high accuracy
has been constructed to measure the moving velocity of dust parH.cl.es in
a precipitator. To achieve the high sampling rate, the i.xrricd-nieasuring
system is used for obtaining the velocity data frcm the Jnser Doppler velocimeter
which are directly transferred to a minicomputer and are. analyzed. A n
-------
1) Department of Electrical Engineering University of Tokyo
2) Senichl Masuda* Professor
Kensuke Akutsu. Assistant
4) Electrostatic Precipitation of Aerosol Particles
inside an Electron Beam Irradiated Field
5) SO and NO pollutants are rapidly converted Into
aerosol particles by the Irradiation of high energy
electron beam. These particles can be effectively
removed by an electric field formed inside the irradiation
space. The distributions of positive and negative ion
concentrations, field intensity, particle charge and
particle migration velocity were calculated. As a result,
the theoretical possibility of particle collection was
advanced. Experiments, using CaC03 powder and the aerosol
particles produced from S02 pollutant by electron beam
Irradiation, confirmed the high collection performance of
this method. Also the desulfuration or denitration rate
seemed to be enhanced by simultaneous field application.
455
-------
DEPARTMENT OF CHEMICAL ENGINEERING
THE FACULTY OF ENGINEERING
UNIVERSITY OF TOKYO
BUNKYO-KU. TOKYO, JAPAN
1. Department of Chemical Engineering, University of Tokyo
2. Akira Suganuma (Associate Professor)
3. Research and Development of Dust Cloud Generator
3-1: measurement of aerodynamic particle size distri-
bution of agglomerated airborne dust
3-2: dispersion of agglomerated fine powder by high
speed air stream
3-3: development of dust cloud generator for testing
dust collectors
456
-------
[1] Nagoya University, Department of Chemical Engineering,
(Furocho, Chikusa-ku, Nagoya, 464)
[2] GENJI JIMBO, Professor.
RYOHEI YAMAZAKI Assistant
JUN-ICHIRO TSUBAKI Assistant
[3] * Measurement of adhesion force of pew der particles.
The adhesion force of powder particles is measured by
several methods including centrifugal separation method,
vibration separation method, split cell method and
fluidized bed method. These results are compared, and
the cause of very big difference between these measured
values is investigated.
* Measurement of deagglomeration phenomenon of
agglomerated powders In air stream.
The size reduction phenomenon of agglomerated powder
particles is measured ,by pulverizing these particles
in negative acceleration field of air stream. The
results obtained are investigated with the data of
adhesion force of powders.
* Dislodging mechanism of particle layer collected
on fabric filter.
The structure of particle layer on a fabric filter is
investigated, and the size distribution of agglomerated
powders dislodged from a fabric filter is also measured.
457
-------
1. Daido Institute of Technology
2. Yasunori MIYOSHI (Professor)
3. Researches on corona discharge characteristics in various
types of gap geometries and approach to improvement of
ELectrostatic-Precipitator Design
i Discharge characteristics and discharge regions of negative
point-to-plane gap in air.
ii Onset of coronas.
Hi Transition from negative corona to spark.
iV Sparkover characteristics of negative point-to-plane gap
with a minor auxiliary discharge on the plane.
458
-------
SUZUKA COLLEGE OF TECHNOLOGY
SHIROKO-CHO. SUZUKA-SHI. MlE-KEN. JAPAN
(1) Norio KIYURA, Professor
(2) Suzuka College of Technology,
Department of Industrial Chemistry
(3) Research subject:
# High temperature gas filtration by granular moved bed:
This subject is studying on the thermophoretic deposition
of aerosol particles by a granular moved bed.
# Collection efficiency of fibrous filter with dust loading;
The collection efficiency of an air filter increases with
the filtration time by the interference effect of collected
particles. In this study, the collection efficiency is obtained
by experiment and a simple theory.
# Dust collection performance of Louver type dust collector;
The particle separation mechanism of the louver dust separator
is mainly inertia and particle rebound on the blades. We
produced many blades of two dimensional type and tested its
efficiency, the theoretical collection efficiency is given
by a simulation method.
459
-------
Kanazawa University, Department of Chemical Engineering
Faculty of Engineering, Kodatsuno, Kanazawa, Japan
Hitoshi Emi (Professor, Dr.)
Chikao Kanaoka (Assistant Professor, Dr.)
Subject of Research
1. Air Filtration
@ Filtration of Aerosols by Fibrous Filter
0 Collection Efficiency of Aerosols by Microscreen
@ Inertial Deposition of Aerosols on the Surface of Micro-
perforated Plate or Nuclepore Filter
@ The Effect of Mist or Dust Loading on the Performance of
Fibrous Filter
2. Particle Deposition on the Wall from Moving Aerosols
@ Deposition of Aerosols in Fully Developed Turbulent Pipe Flow
@ Deposition of Aerosols near the Entrance of Pipe
@ Deposition of Aerosols in a Bifurcation Tube
3. Measurement of Adhesion Forces
@ Measurement of Adhesion Forces between Two Particles by
Centrifugal Method
@ Detachment of Particles from a Cylinder by Aerodynamic Drag
460
-------
Staff Members and Research Project in linoya Laboratory
Department of Chemical Engineering, Kyoto University,
Sakyo-ku, Kyoto 606., Phone (075) 751-2111 Ext. 5566-9
as of spring in 1977
Staff Members
Professor
Assistant Professor
Research Assistant
Research Assistant
Ph. D. Candidates
Koichi linoya Dr. of Eng.
Kazutaka Makino Dr. of Eng.
Hiroaki Masuda Dr. of Eng.
Kenichi Ushiki
Hideto Yoshida, Yasushige Mori,
Michitaka Suzuki
Research Projects
(1) Dust Collection Performance of Bag Filter
(2) Dislodging Characteristics of Powder Cake on Filter Fabric
and on Collecting Electrode of Electrostatic Precipitator
(3) In Stream Measurement of Flow Rate in Powder Pneumatic
Conveyor
(4) Dust Sampling Techniques in Stack and Environment
(5) Electrification of Particles
(6) Measurement of Dust Concentration by Electrification
(7) Virtual Impactor for Particle Size Measurement
(8) Mist or Inertia Separator
(9) Mist Size Measurement
(10) Powder Layer Mechanics and Stability
461
-------
AEROSOL RESEARCH BY PROFESSOR YOSHIDA'S GROUP
(Department of the Chemical Engineering, University
of Osaka Prefecture, Sakai 591, Japan)
STAFF Professor Tetsuo Yoshida
Assist. Professor Yasuo Xousaka
Research Assist. Kikuo Okuyama
Our researches are about "Particle growth of aerosol
particles of sub-micron diameter by condensation and coagula-
tion, and its application to industrial dust collection".
In these studies, the particle size distribution and
the particle number concentration were determined by the
measurement method using an ultramicroscope developed by us.
•"• PARTICLE SIZE ANALYSIS OP AEROSOLS
USING AN ULTRAMICROSCOPE
CONDENSATION £-*. PARTICLE -^ *
PRECONDITION!
COAGULATION
GROWTH
SING OF
DUST COLLECTION
J., A New Technique of Particle Size Analysis of Aerosols and Pine Powders
Using an Ultramicroscope.(Ind. Eng. Chem. Fundam., 14, 47(1975))
2, Condensation of Water Vapor on Aerosol Particles
i) Condensation Growth of Aerosols by Nixing Hot Saturated Air with
Cold Air.(Ind. Eng. Chem. Fundam., 15, 37(1976))
ii) Condensation Growth of Aerosols by Injection of Steam into Air.
(unpublished work)
iii) Dependence of the Evaporation Rate of Micron Order Droplet on Particle
Number Concentration.(unpublished work)
iv) Experimental Study of Thennophoresis of Aerosols
(J. Chem. Eng. Japan, 9, 147(1976))
3, Coagulation of Aerosol Particles
i) Change in Particle Size Distribution of Polydisperse Aerosols
Undergoing Brownian Coagulation.(J. Chem. Eng. Japan, 8, 317(1975))
ii) Turbulent Coagulation of Aerosols in a Stirred Tank.
(J. Chem. Eng. Japan. 10, 142(1977))
iii) Turbulent Coagulation of Aerosols in a Pipe.(unpublished work)
iv) Behavior of Aerosols Undergoing Brownian Coagulation, Brownian
Diffusion and Gravitational Settling Between Two Horizontal Halls.
(J. Chem. Eng. Japan, 8, 137(1975))
v) Behavior of Aerosols undergoing Brownain Coagulation, Brownian
Diffusion and Gravitational Settling in a Closed Chamber.
(J. Chem. Eng. Japan, 9, 140(1976))
vi) Effects of Brownian Coagulation and Brownian Diffusion on Fine
Particle Size Analysis by Sedimentation Method.
(J. Chem. Eng. Japan, 10, 46(1977))
Jj, Application of Particle Growth to Industrial Dust Collection
i) Application of Aerosol Growth by Condensation to Industrial Dust
Collection.(to be presented at The Second Pacific Chemical
Engineering Congress(Pachec* 77))
ii) Effectiveness of Particle Growth in Dust Collection by Wet Scrubber.
-Venturi Scrubber, Sieve Plate Type Scrubber and so on-
5. i) Dispersion of Powders into Air.(unpublished work)
ii) Aerodynamic Diameter of Non-spherical(Needle-like and Aggregate)
Particles Using an Ultramicroscope.(unpublished work)
462
-------
1. Okayama University
Department of Industrial Chemistry
Faculty of Engineering
3-1-1, Tsushima-naka, Okayama, Japan, 700
2. Zennosuke TANAKA (Assistant Professor, Dr.)
3. Performance of Centrifugal Dust Separator
Developing the new equipment .of hybrid dust separator
( centrifugal dust separator with rotating fibrous filter
i.e. centrifugal bag filter).
The device is designed to remove particulate matter
by filtration and centrifugation. Centrifugal forces
play the roles in precleaning a heavy loading of large
particles and dislodging the collected material on filter.
463
-------
1. Department of Electronic Engineering , Tokushima University
2. Prof. Y. GOSHO
3. Increase in Breakdown Voltage of Non-Uniform Field Gap by Adding
Electronegative Gases
In a non-uniform field gap such as Electrostatic Precipitator,
when the potential across the gap is raised, corona occurs at a co-
rona onset potential and the corona current increases with increas-
ing the potential. At a certain potential breakdown occurs across
the gap. In these conditions, by adding a small amount of electro-
negative gases, the breakdown voltage was found to be greatly in-
creased. With a point-plane geometry with a 15 mm gap in air, the
increments of the breakdown voltage and the corona current prior to
breakdown were 20 percent and 70 percent respectively with the ad-
dition of 3 percent of SFg. It will be expected that the efficiency
of precipitation of E P is improved by applying this means.
464
-------
1. Faculty of Engineering, Yamaguchi Uni\ezsity.
2. Takayoshi Adachi, Associate Professor.
3. Ionic Wind and Behavior of Particles dragged by Ionic Wind in
an Electrostatic Precipitator.
(]). Velocity distribution of ionic wind in tho corontt discharge
space consisting of point to plate electrodes has txxsi
observed by means of laser-doppler flowmeter and schlieren
photograph method, and also theoretically studied by
treating Poisson's equation, equations of. electric current
continuity, Navier-Stokes equations, and the main cxjaatlan
by use of computer.
(2). The drastic influence of ionic wind ori suLtnicron particles
was experimentally confirmad in its collection process.
(3). The effect of the force acting on particles was discussed
in terms of two components of the ionic wind force and the
Coulomb force which were analyzed from the observation in
the EPsystem with a sham ionic wind.
465
-------
1. Kyushu Institute of Technology
2. Shinichi Yuu (Assistant Professor)
3. Particle turbulent Diffusion in Dust Laden Flows
Attention has bcjon focused on the diffusion of snail particles
in various flow fields. Such a phenomenon is of interest in
numerous cler-ning devices (electrostatic precipitator etc.)
and atomized fuel injection systems. The principal purpose of
this study is to reveal the mechanism of the particle turbulent
diffusion in dust laden flows by predicting particle turbulent
diffusivities theoretically and measuring them experimentally.
Turbulent particle Lagrangian trajectories and velocities in a
round and a plane air jets were calculated by using fluid
integral scales, intensities and average velocities. From
the calculated results the turbulent diffusivity and the local
time-averaged velocity of particles are obtained. Measurements
of local mean aerosol concentration were made with a photo-electronic
dust counter and a dust tube. From the measurements tho experimental
particle diffusivity is obtained and compared with the theoretical
calculation. The results indicate that the particle diffusivity
decreases with the increase of the particle inertia. In general
the turbulent diffusivity of particles is smaller than that of
fluid scalar quantities. Tho particle inertia and the fluid
large eddies, which are expressed by the Stokes number and the
integral scale, respectively, play an important role in the
transport mechanism of particles in the dust laden flow.
466
-------
1. Department of Applied Science, Faculty of Engineering,
Kyushu University
2. Terutoshi Murakami ( Professor of Applied Physics )
3. Scavenging of Aerosol Particles by Liquid Droplets
4. As a basic research of scavenging of aerosol particles
by>liquid droplets, the interaction between the particles
and the droplets are investigated in two cases.
(i) Aerosol particles of 5yu n in diameters of Rhodamine
B aqueous solution are generated by ultrasonic method,
and the amount of the particles collected by falling water
droplet is measured by spectrophotometer. To obtain correct
collection efficiency, the distribution of the small particles
and the aerodynamic flow patterns around the falling droplet
are observed by pulse laser holography.
(ii) A small cylinder of 1 mm in diameter are set up in
the high speed aerosol flow produced in a shock tube, then
the flow patterns around the cylinder are investigated by
means of the laser holography or shadow photographs. We
intend to scavenge the aerosol particles by water droplets
dispersed in the shock tube.
467
-------
1. Prof. M. AKAZAKI
2. Kyushu University, Faculty of Engineering,
Department of Electrical Engineering
3. Discharge Phenomena in the Electrostatic Precipitator
(a): Characteristics of DC Corona Discharge in the
Particle Collection Space
(High Temperature, High Humidity and Flying
Particles)
(a-1) The effect of Experimental Condition* for Corona
Discharge (*Voltage Waveform, Electrode Shape
and Gas Condition)
(2-2) Characteristics of Corona Pulses from the
Polluted Electrode Surface
(b): Charging Mechanism of Dust Particle by Corona
Streamer
(b-1) Mechanism of Streamer Propagation in the Air
Containing Dust Particles
(b-2) Mechanism of Electric Breakdown in the Dust
Layer on the Electrode
(c): Mechanism of Particle Reentrainment
(d): Mechanism of Back Discharge
468
-------
AUSTRALIAN COAL INDUSTRY RESEARCH LABORATORIES LTD.
P. 0. Box 83. NORTH RYDE. N.S.W. 2113. AUSTRALIA
Telephone; 88-0276; 888-5341
Australian Coal Industry Research Laboratories Limited
comprises five laboratories, each registered by the National
Association of Testing Authorities, and located in the
Australian States of Queensland and New South Wales.
The laboratories are staffed by over 100 employees and
provide facilities for analytical investigations and pilot
plant studies which support the specialised consulting
capabilities that have been developed in the fields of coal
preparation.
A.C.I.R.L. through its testing and consulting activities
has been particularly active in the area of pollution control.
These activities have caused research to be undertaken into
various areas, which has resulted in the development of
new techniques.
Included in these new techniques in the unique facility
that was developed, which enables small samples of coal,
usually derived from bore cores, to be processed into a
laboratory fly ash by controlled firing through a micro furnace.
The laboratory fly ash which is similar to a Power Station
fired fly ash, is then electrically, chemically, physically
and microscopically examined in order to assess its potential
capability to be electrostatically precipitated. The electrical
assessment includes determinations of both Resistivity and Voltage
Current Corona characteristics over the range of potential
operating temperatures and under varying mo.isture contents
in a simulated flue gas environment.
469
-------
This whole technique enables an investigation to be
carried out of a coal area which is intended for use in a
proposed power station, without having to first mine the
area. It is the only means known whereby an estimated
assessment of the combustion and electrostatic precipitation
characteristics of such a coal area may be made at the
pre-development stage of a proposed Power Station project.
J.W. Baker, B.Sc., B.E. -Mechanical Engineer, ACIRL Ltd.
K.M. Sullivan, B.E., FIEAust., FInstF., -Principal Fuel
Engineer, ACIRL Ltd.
470
-------
Directory of Foreign Attendants
1 P. Lflffler
2 E.G. Potter
3 M. First
4 W. Stttber
5 R.W. Mcllvaine
May, 1977
Professor, Institut fur Mechanische
Verfahrenstechnik der Universitat
Karlsruhe
Richard-Mil3statter-Allee, 7500
Karlsruhe 1, Germany
Leader, Process Chemistry Section,
Division of Process Technology Minerals
Research Laboratories, Commonwealth
Scientific and Industrial Research
Organization, Australia
P.O.Box 13f. Delhi Road North Ryde,
N.S.W. Australia, 2113
Professor, Department of Environmental
Health Sciences/ School of Public Health,
Harvard University
665 Huntington Avenue, Boston,
Massachusetts 02115, U. S. A.
Professor, Institut fttr Aero-Biologie,
5948 Schmallenberg-Grafshaft/ Germany
President, The Mcllvaine Company
2970 Maria Avenue llorthbrook,
Illinois 60062, U. S. A.
6 K.M. Sullivan Principal Fuel Engineer, ACIRL Ltd.
P.O.Box 83, North Ryde,N-S.W. Australia/
2113
471
-------
7 J.W. Baker Mechanical Engineer, ACIRL Ltd.
P.O.Box 83, North Ryde N.S.W.
Australia, 2113
8 R.H. Horning Vice President, Combustion Power Co.,
Menlo Park, California, U. S. A.
472
-------
International Seminar on Dust Collection
Attendants (University) -
24, May 1977
Univ. or College
Name
Adrress
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
Hokkaido Univ.
Res. Ins t. of Applied
Electricity
Tokyo Univ.
Dept. of Elec.Eng.
ti
Tokyo Univ.
Dept. of Chem.Eng.
Nagoya Univ.
Dept. of Chem.Eng.
M
Daido Inst. of Tech-
Dept. of Elec. Eng.
Suzuka Technical
College
Kanazawa Univ.
Dept. of Chem. Eng.
it
Doshisha Univ.
Dept. of Chem.Eng.
n
Kyoto Univ.
Dept. of Chem. Eng.
n
n
n
n
M
ii
Hirotatsu
Mishina
Senichi
Masuda
Akio
Akutsu
Akira
Suganuma
Genji Jinbo
Junichiro
Tsubaki
Yasunori
Miyoshi
Norio
Kimura
Hitoshi
Emi
Chikao
Kanaoka
Satoshi
Okuda
Hiroshi
Takano
Koichi
linoya
Kazutaka
Makino
Hiroaki
Masuda
Kenichi
Ushiki
Hideto
Yoshida
Yasushige
Mori
Michitaka
Suzuki
Nishi 6-chome, Kita 12 Jyo,
Kita-ku, Sapporo (060)
Kongo, Bunkyo-ku, Tokyo (113)
it
»
Furo-cho, Chikusa-ku, Nagoya (464)
tl
2-21, Daido-cho, Minami-ku,
Nagoya (457)
Shirako-cho, Suzuka, Mie (510-02)
2-40-20, Kotachino, Kanazawa (920)
it
Karasuma Imadegawa, Kamigyo-ku,
Kyoto (602)
ii
Yoshida Honmachi, Sakyo-ku,
Kyoto (606)
ii
it
n
n
n
it
473
-------
20. Osaka Pref.Univ. Tetsuo
Dept.of Chem. Eng. Yoshida
4-804, Mozuume-cho, Sakai, Osaka
(591)
21.
Kikuo
Okuyama
22. Okayama Univ. Zennosuke
Dept.of Ind.Chem. Tanaka
3-1-1, Tsushimanaka, Okayama
(700)
23. Tokushima Univ. Koshichi
Dept.of Elec.Eng. Goshb
2-1, Minami Tsunemishima-cho,
Tokushima (770)
24. Kyushu inst. of-Tech.
Research Inst.of Shinichi
Powder Tech. Yuu
1, Sensui-cho, Tobata-ku,-
Kitakyushu (804)
25. Kyushu Univ.
Dept.of Applied
Physics
Terutoshi
Murakami
Hakozaki, Higssni-ku,
Fukuoka (812)
26. Kyushu Univ.
Dept.of Elec.
Eng.
Masanori
Akazaki
27. Kansai Univ. Takuzo
Dept,of Chem.Eng. Matsuyama
Senriyama, Suita, Osaka (564)
474
-------
International Seminar on Dust Collection
Attendants (Company) -
24, May 1977
Company
Name
Adrress
1. Hosokawa Iron
Works Ltd.
Teruaki 9, 1-chome, Shodai Tajika,
Suzuki Hirakata, Osaka (573)
2.
Takashi
Kitamura
3. Nitta Gelatin Co., Yoshihiro
Ltd. Nonaka
Futamata, Yao, Osaka (581)
4. Matsushita Seiko
Co., Ltd.
Toshio 4811, Marunouchi, Takagicho,
Shibahara Kasugai, Aichi (486)
5. Sankyo Dengyo
Co., Ltd.
Kazuo
Saito
6. Nippon Donaldson, Tsutomu
Co. , Ltd. Shibuya
1-8-11, chuo-cho, Meguro-ku,
Tokyo (35?)
100, Iinadera, Ornc, Tokyo
(198)
7. Mitsui Miike Co., Hideo
Ltd. Noziri
1-1, Kokubu-cho, Tochigi
(328-03)
8. Denka Consultant Ryusuke
& Engineering Co., Araki
Ltd.
1-4-1, Yuraku-cho, Chiyocla-ku,
Tokyo (100)
9. Izumi Kakoki
Co., Ltd.
Chiaki 3-7, Nakanoshima, Kita-ku,
Shiota Osaka (530)
10. Nippon Felt
Co., Ltd.
Ikuo
Vasni
2-2, 2-chome, Marunouchi, Chiyoda-
ku, Tokyo (100)
11. sinto Dust Collector Takeshi 1, Nishinagane, Kodacho,
Ltd. Yoneda 'Sakazaki, Nukata-gun, Aichi (441--03)
12. Sanko Engineering Akio 4-6-29, Namamugi, Tsurumi-ku,
& Construction Co., Furukawa Yokohama (230)
Ltd.
13.
14.
15.
16.
17
18.
n
Hitachi Plant
Engineering &
Construction Co. .Ltd.
Sumitomo Kinzoku
Kozan Co. , Ltd.
Yamamoto Industries
Co. , Ltd.
NGK Insulators
Co. , Ltd.
Kobe Steel Ltd.
Toshio
Seki
Hiroshi
Yamada
Ken
Takimoto
Rinkan
Kawamura
Shigeharu
Kito
Hiroyuki
Kohama
n
1-13-3, Kitaotsuka, Tokyo (170)
5-11-3, Shinbashi , Minato-ku,
Tokyo (105)
1-2-2, Kawashiro, Tobata-ku,
Kitakyushu (804)
1, Maegata-cho, Handa (475)
1-3-18, Wakihama-cho , Fukiai-ku,
Kobe (651)
475
-------
19. Kobe Steel Ltd.
Akira 1-3-18, Wakihama-cho, Fukiai-ku,
Wakabayashi Kobe (651)
20. Sinto Dust Collector
Ltd.
Takeo 1, Nishinagane, Sakazaki,
Hisatsune Kodacho, Nukata-gun, Aichi (441-01)
21.
22.
23.
Kurimoto Tekkosho
Co. , Ltd.
"
Hosokawa Research
Inst. of Powder
Tech.
Ryota
I to
Akira
Hama
Tohei
Yokoyama
1-56, Oike-dori,
Nishi-ku, Osaka
11
Kitahorie,
(550)
9, 1-chome, Shodai Tajika,
Hirakata, Osaka (573)
24. Kawasaki Heavy
Industries, Ltd.
Kimihiro
Funahashi
16-1, 2-chome, Nakamachi-dori,
Ikuta-ku, Kobe (650-91)
476
-------
ERROR IN MEASUREMENT OF GAS FLOW RATE
IN GAS-SOLIDS TWO-PHASE FLOW BY USE OF
A HORIZONTAL DIFFUSER*
HIROAKI MASUDA, YOSHIFUMI ITO
AND KOICHI IINOYA
Department of Chemical Engineering, Kyoto University,
Kyoto, Japan
(Reprinted with permission)
Reprinted from
JOURNAL OF
CHEMICAL ENGINEERING
OF
JAPAN
Vol. 6, No. 3 (1973)
Pages 278—282
477
-------
ERROR IN MEASUREMENT OF GAS FLOW RATE
IN GAS-SOLIDS TWO-PHASE FLOW BY USE OF
A HORIZONTAL DIFFUSER*
HIROAKI MASUDA, YOSHIFUMI ITO
AND KoiCHf IINOYA
Department of Chemical Engineering, Kyoto University.
Kyoto, Japan
Measurement of gas flow rate in gas-solids two-phase flow is studied both theoreti-
cally and experimentally by use of a horizontal diffuser. Since the pressure recovery in
the diffuser varies with both the solid flow rate and diffuser length, the gas flow rate
measured by the diffuser generally has some error. The relative error for the measure-
ments is a monotone-increasing function of the measuring length, but never exceeds the
value of the mass flow ratio. It is also shown that there is a length at which the error
vanishes. This length can be estimated using an analytical equation. The effects of
powder properties on the magnitude of the error are also discussed in detail.
1. Introduction
The measurement of the mass flow rate of solids, or
of the flow rate ratio for solids to gas in a two-phase
system is complicated by the fact that the measured
value, such as pressure drop along the pipeline, is also
a function of gas flow rate. Therefore the gas flow rate
must be measured simultaneously by an independent
method. Farbar", Earth et a/.2>, Goto et a/.», and
Sakata4', examined the possibility of using a diffuser
to make such measurement on a two-phase system.
They concluded that pressure recovery in the diffuser
decreased with increasing solids rate. However, the
manner in which the error of the measured gas flow
rate varies with the distance along the diffuser has not
yet been determined, as was mentioned by Sakata.
This study will examine the nature and extent of
error, and, where the error offers a problem, methods
for estimating the error theoretically and design
methods for minimizing the error. In particular, it is
shown that there is a design method for reducing the
error to zero. Following the suggestion of Goto et a/.,
the investigation is carried out for the horizontal part
of a pneumatic conveyor, and is mainly concerned with
particles smaller than 100 microns.
• Received on May 6, 1972
Presented at the 37th Annual Meeting of the Soc. of Chem.
Engn., Japan, April 5, 1972
T606
2. Theoretical Approach
If it is assumed that the particles are uniformly
suspended and in rather low concentration, the wall-
friction of particles may be regarded as negligible in a
diffuser*'". Therefore, the following momentum
balance equation is obtained51.
—dPm=paudu+mpaudv (1)
The relative error on the basis of the pressure difference
is defined by the following equation**.
uv'dx
(2)
uu'dx
where ua and v0 denote the gas velocity and the particle
velocity at the inlet of the diffuser, respectively. The
prime represents the derivative with respect to the co-
ordinate x. In this equation, the denominator shows
the pressure recovery for gas flow alone, and the
numerator shows the additional recovery by the solids
momentum.
Assuming the incompressibility of gas flowing in the
diffuser section, the velocity u can be expressed by the
equation
u—
(l-ra*)«
tan0
o= - •-•
(3)
** see also Eq.(18)
On the assumption that the wall friction of particles
may be neglected in the diffuser, the particle velocity u
is given by the following equation of motion":
478
JOURNAL OF CHEMICAL ENGINEERING
JAPAN
-------
,.>_
where C, and C, are constants which are given by
and
(4)
(5)
(6)
Fif. 1 show* the velocities u, v and the trend of the
relative error calculated by numerical integration using
Schiller and Naumann's8* drag coefficient. Particle
velocity vt at the inlet is smaller than the corresponding
gas velocity u,, because the particles have been af-
fected by wall friction in the straight pipe section just
before the diffuser. The relative error monotonously
increases with the coordinate x. This trend of the error
is analytically explained by Eqs.(2), (3) and (4).
From Eq.(2), it can be seen that the error is zero at
x=x0 or 0=»8»
(7)
Applying 1'Hospital's theorem to Eq.(2), the limiting
value of the error as x-*oo is given by the equation
(8)
*-.- uu
If the position x0 can be estimated in some way, the gas
flow rate in suspensions may be measured very accu-
rately. Now, assuming that x0 is determined, the re-
sidual error arising from the difference between the
measuring point x and x0 may be estimated from the
gradient of the error curve at the position x0. Taking
the equation Ett=0 into consideration, the following
equation holds in the neighborhood of x0;
uv'dx
uu'dx
(9)
If the relation »c—u~|i£,|(x—x0) is substituted in the
above equation, Eq.(9) reads
Note that EXt=0, and hence the differential form of
this equation is
dE
~2m-
(11)
u» _ ..I i »'
"o «»
Eliminating v'n by use of the equation of motion (4),
a more explicit form of Eq.(l 1) is obtained as
(12)
U.'«0 KtfHC
39 */MC
pciilion
ml«
Flf. 1 Velocltlei «, » and the trend of the relative
error (numerical Integration)
In practical applications, the second term in the
brackets may be regarded as negligible. From this fact
and the relation u0>f0>uJo, Eq.(12) may be simplified
as follows;
dE
•2T
(13)
This equation shows that the error presents a problem
when the inertial force of a particle is smaller than the
viscous force of the fluid.
The next problem is to estimate the position at which
the error vanishes. From the above results concerning
the gradient of the error, the second term of Eq.(4) is
seen to be negligible. Then Eq.(4) may be rewritten
as
t,'=C,(U/»-l) (14)
A first approximation by Picard's method7' gives
]-«] (15)
/ J
Now, the equation for xt is obtained from Eq.(15)
where o=r0:
, where 0,s-3s.
(16)
This equation shows that the first approximation of x0
is determined by the dimensions of the diffuser and the
velocity ratio 0. The characteristics of the powder are
not included explicitly in this equation.
3. Apparatus and Experimental Procedure
In this experiment, a pneumatic conveyor line oper-
ated under negative pressure is used. The experimental
setup is shown in Fig. 2. The diffusers used are shown
in Fig. 3. The selected diffuser is placed in the hori-
VOL.6
1973
479
-------
• ^. A
r^—~r i r
*—I—p^T western
typeFMot
tube
cyc.cn.
How
control
valve
through
bag filter
to
motor blower
Fig. 2 Schematic diagram of the experimental setup
i
Fig. 3 Diffusers used in the experiments
99.9
100 1000
Particle size Dp : moons]
Fig. 4 Size distribution of the solid mate-
rials (Log-normal) (determined by sieving;
for floor, the sedimentation method It
used.)
Table 1 Properties of the powder materials
Materials Mass median dia" N-Ioan Particlc dia" part*!? *
flpso [microns] uf [microns] pp [g/cma]
Quartz sand
No. 8
Glass beads
Vinyl chloride
Flour
Quartz sand
No. 5
il
55
115
57
380
a
M
:
.••
295
2.f.5
2.42
1.41
1.44
2.65
zontal section of the piping, allowing for the appropri-
ate approach length. To charge the solid particles into
the conveyor, a table feeder is used. Materials used are
quartz sand No. 8 (fine), glass beads, vinyl chloride
powder, flour, and quartz sand No. 5 (coarse). The
properties of these powders are shown in Table 1, and
their size distributions are presented in Fig. 4.
Air flow is monitored by a Pilot tube, and strain-
gauges, transducers, and an on-line hybrid computer
(CLAOP 2000) are used to measure pressure differ-
ence. The system used for these on-line studies is shown
schematically in Fig. 5.
4. Results and Discussion
The analytical values calculated by Eq.(13) are
compared wi'.h the results of numerical integration ob-
tained by a digital computer (FACOM 230-60). Fig. 6
shows a set of the results on the gradient of the error.
Fig. 7 shows a comparison between xa calculated by
Eq.(16) and x0 obtained by numerical integration.
When particle size is large, the coincidence of the ana-
lytical and numerical results is not good. The dis-
agreement is not serious, however, because the gradient
of the error itself is very small in this case.
The relative error E has been defined in relation to
the pressure difference measurements. Now, the corres-
ponding relative error for measuring the gas flow rate
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
480
-------
Hybrid
computer | | Digital computer |
10
Fig. 5 Schematic diagram of the on-line system
15
10
8
a
6=25°
30
-Eq.(16)
0.4 06 0.8 1
Velocity ratio tj>,[-]
Fig. 7 Comparison between jro calculated by Eq.(16) and
that obtained by numerical integration
is introduced, as follows;
As the pressure difference is in proportion to the square
of the gas flow rate, Eq.(l 7) may be rewritten as
The results presented in Fig. 8 indicate the magni-
tude of the error. Calculated values are shown by lines.
The mean particle diameters defined by the following
* When the particle size distribution is log-normal, Eq.(19) is
equivalent toS) £>j>=exp(ln/)j>j(i —<7S), where
-------
developed from Eq.(13).
(20)
'
Note: *o indicate! the position at which the error vanishes.
The values of*t*a:e. are determined by UK of Eq,(16).
Fig. 9 Relative error •>• « fsimrti'ini of axial position
Fig. 8 d) shows some of the results obtained using
diffuser III (0=5°). In this case, the error becomes
negative with increasing particle velocity »0 and mass
flow ratio m. This phenomenon may arise from the
separation of the fluid from the wall*. As the flow
disturbance is aggravated by the particles, the sepa-
ration occurs more readily. When the diverging angle is
fixed, this phenomenon occurs more easily at smaller
mass flow ratios as the particle velocity becomes higher.
As shown in Fig. 8 e), however, this phenomenon can-
not be seen in the experiments with smaller mass flow
ratio and lower particle velocity, even when 5=7.5°.
Further, this phenomenon depends on the shape of the
particle, and occurs more readily for quartz sand No. 8
than for glass beads.
Fig. 9 shows the experimental results proving the
existence of a position x, at which the relative error is
zero. Calculated values, *o,«aic.i are *»«> indicated in
this figure. From these results, it is found that the co-
incidence between the experimental and the calcu-
lated ATO is very satisfactory. Fig. 9 c) shows the experi-
mental results obtained when the measuring position
x is shorter than the x, estimated by Eq.(lG). In this
case, the error is negative, as is estimated by the theory.
These data are similar to those of Farbar, Earth et at.,
Goto tt al., and Sakata.
The following discussion is concerned with the effects
of powder properties on the magnitude of the error.
The difference in the magnitude of error obtained by
several authors using coarse particles may arise from
differing values of (x—*0). The equivalent difference
(*—*§), which gives rise to an error of equal magnitude
for the systems I and 2, is given by Eq,(20), which is
Summarizing the above discussion, the published
data for large particles and the results of this study for
small particles have been explained consistently, taking
into account the nature of the variation in magnitude
of the error and the fact that there is a position at which
the error vanishes. Eq.(8), showing that the error does
not exceed the mass flow ratio, will prove useful when
a low mass flow ratio system such as a dust collector is
involved. In a system using large particles, the velocity
ratio 00 is constant1*-'4' without recourse to the
mass flow ratio m and the gas velocity u. The position
x0 can, therefore, be estimated knowing the properties
of the particles and the dimensions of the diffuser.
Nomenclature
Cj, Ci - constants, Eqj.(5) and (6J
Dp — particle diameter
Dp = mean particle diameter, Eq-( 19)
E => relative trror defined by Eq.(2)
i = relative error defined by Eq.(l 7)
/i •• particle size distribution
APm. = pressure recovery of gas-solids suspensions flowing in
the diffuser
JPa = pressure recovery of gas flow alone
= gas flow rate calculated by APm
= actual gas flow rate
=• inside radius of the diffuser inlet pipe
= gas velocity
= particle velocity
= coordinate in axial direction
= position at which the error Eon vanishes
* If we assume that this phenomenon arises from future of the
original assumption, that the wall friction of the partlclei it
negligible in the diffuser, we cannot explain the fact that the
phenomenon is not observed when a diffuser with small diverg-
ing angle is used.
(2.
Q
r
u
c
x
*o
a = constant determined with 8 and r, defined by Eq.(3)
8 =- half of the diverging angle of the diffuser
fia — viscosity of the gas
pt — density of the gas
pf " density of the particle
00 — velocity ratio defined by Eq,( 16)
Literature Cited
1) Farbar, L.: TVani. tfllu ASME, 75, 943 (1953)
2) Earth, W., R. Nagel and K, van Waveren; CHmit-Ing.-Ttthn.,
29, 599 (1957)
3) linoya, K. and K. Goto: Kagaku KSgaiai (Chm. Eng., Japan),
27,80 (1963)
4) Sakata, M.: Traiu. of tht Japan Set. tfMieh. Engrs., 37, 1560
(1971)
5) nawiKKKft H. B. H H. H. CUPOMBTHHKOB: unxtmpHO-
fivsuttCKuytl xvpnajt, 17, 26 (1969)
6) Schiller, L. and A. Naumann:£. V. D. I., 22, 318 (1933)
7) Lapwoodi E. R.: 'The International Encyclopedia of Physical
Chemistry and Chemical Physics", Topic I, Vol. !, p. 179,
Pergamon Press London (1968)
8) Maiuda, H. and K. linoya: Memoirs of the Faculty of Eng,,
Kyoto Univ., 34, 344 (1972)
9) Maiuda, H. and K. linoya: J. Chm. Bag. Japan, 4, 60 (1971)
10) Weidner G.: Forsduaig, 21,145 (195.*>)
11} Boothroyd, R. G,: Trent. o/Uu ASMS, sen B, 91, 303 (1969)
12} linoya, K., T. Kamimura and Y. Tsukada: KtgaJat Klgaku
(Cktm. Eng.t Japan), 23,400 (1959)
13} Barth, W.: OWmw-/^..^^., S», 171 (1958)
14) Konno, H., S. Saito and S. Maeda: Kataht ftgaJtu, 31, 243
(1967)
482
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
-------
DUST CLEANING DYNAMICS IN REVERSE COLLAPSE TYPE BAG FILTER
Kazutaka Makino, Yasushig Mori
Naomi Takado, Koichi linoya
The bag filter which is one of the typical high efficiency
collectors requires a dust cleaning operation. However, research
on bag filters so far are mainly about collection efficiency and
pressure drop characteristics,3'1* While there are not many studies
concerning dust cleaning characteristics which is an important
factor in design of filter operations. Furthermore, there are
very few reports concerning the dust cleaning dynamics which is
especially necessary in the determination of dust collection
charateristics.
Therefore, this report related to the reverse flow type
collector, a typical dust cleaning method, reviews and investi-
gates its dynamics experimentally and presents the results. Here,
the dust cleaning dynamics mean the variance of a combined flow
resistance of dust and filter material in time due to reverse
flow after starting dust cleaning operations.
Estimation of Transfer Function for Each Component in Measurement
System
Figure 1 shows the block diagram of dust cleaning Dynamics
Measurement System in this experiment. In order to estimate the
transfer function of recorder and transmitter, five kinds of
pressures were applied for each case as an input and the step
responses were measured. The result of this measurement shows
that there is a linear relationship between the above mentioned
inputs and the outputs (response value indicated at recorder)
and that it can be approximated by one step later {one order later)
The transfer function for reverse air flow value was obtained by
experimentally approximating the general transfer function which
is the product of transfer function of recorder and transmitter
and above transfer function, after doing step response experiment
for certain air flow with 5 different positions of flow control
value and after assuming the transfer function of reverse air
flow value to be dead time 1 order later. Each transfer function
estimated this way is shown in Figure 1.
Estimation Result of Dust Cleaning Dynamics
For dust cleaning operation due to reverse flow, if the
combined flow resistance of dust and filter material of the bag
due to reverse air flow is assumed to be R(t). R{t) will probably
483
-------
rtcordwiI tronwnlnvr tar0,(l)
KLIJ\m _.'•"* _
0|W)* 1.1.14
central volv* far
rtvtrt* air
(Ml)
»
G](4)
0}(4)
Aft(l)
1
1.2 Z^
Oo(l)
f-.fj*. .. 1 ,
°'C4>' 1.2 24
AWI),
t*r rtcardcr t Ironsmiller
for &P.(I)
Fi|. I Block diatram of fabric filter system Including
measuring instruments
484
-------
decrease as dust cleaning proceeds. However, it is difficult at
the present time to directly measure this process quantitatively
Thus in this report a model is set and dust cleaning dynamics
are estimated. Also, it should be noticed that the transfer
function G2 (s) of bag filter itself is not a Laplace transforma-
tion of R(t).
Here the bag pressure drop Apr(t) during reverse air flow
can naturally be expressed by the following equation.
Apr{t) = L-1 (^ G1(s)G2(s)] = Qr(t)R(t)/S
(1)
Now the reverse air flow rate Qr(t) is obtained as the following
equation from Figure 1.
Qr(t) = L-M^ G,(s)] = Qrf00E(t-l){l-exp(- £|) } (2)
Also the apparent pressure drop Apa(t) across the bag filter shown
in recorder during reverse air flow is expressed as follows
using tranfer function Gi» (s) .
Apa(t) = L"1^ G,(s)G2(s)G^(s)]
S
= L-l[L[Apr(t)] G^s)] (3)
Now for the combined flow resistance R(t) of dust and filter
material during above mentioned cleaning process, consider a
model to approximate by dead time and 1 order later.
(4,
This is the case when the dimensionless resistance coefficient
l=(R(t)-Rmax)/(R00-Rmax) 1 can be approximated as 1 order late.
The reason why a dead time 0 was considered was because there is
a case when a start of reverse air is not consistent with a
start of cleaning. In this case Apa(t) is expressed by the
following equation from Equations (I)- (4).
-=i)> E(t-l)
+ Qr,o0(R30-Rniax)E(t-9) [1 + (fif^ - ^f exp (-
- exp(-(9-17/1.3) , , t-6 13 (_ t-1,
exp ( 2.2 * . 9 exp { 1.3
__ L. t^ij +- .
T-2.2 expv T ; (-2.2(1/1.3 + 1/T)
x exp {-(t-9) (yyj + i)}] (5)
485
-------
Now if you set T+0, T(t) can be approximated only by a dead time.
Thus it is assumed as one order late model with dead time 0
when T 7* 0 and as dead time model (dead time 0 = A) when T-»-0 .
If you plot the results of actual measurement on the Apa(t)
vs t line figures which were drawn by assuming various cleaning
time constant T, cleaning dead time 8 or X, T,6,X during actual
dust cleaning process can be estimated. The fine calcium
carbonate (Dpso = 5.4 ym) was used as sample dust and we have
changed the most typical filter materials (tetron and nylon
fibers) as well as number of so-called rings which are installed
to prevent the bag insides from contacting during reverse collapse
operations. The installation of bag is similar to that of previous
report (bag with 1,800 mm long and 170 mm $ inner diameter).
Figures 2 and 3 show typical examples of this experiment
results. Based on these, it is seen that the approximation by
the dead time 1 order late system in Equation (5) is appropriate.
Table 1 shows the ranges of T, 9 and X obtained in this experi-
ment. According to this, each case is within several seconds
and it is indicated that several ten seconds will be enough as
a single cleaning duration time for reverse bag filters.
For tefron long fabric filter cloth, when one ring is
installed, the time constant is minimum and it is suggested
that there is the most optimal number of rings , namely the most
optimal ring installation distance, in light of cleaning dynamics.
Also when it is switched to collection process after reverse
cleaning and to have only clean air, the pressure drop, as shown
in Table 1 will be smaller as the number of rings increases.
This will probably require future investigations. It was also
confirmed that the dust load changes at the start of dust cleaning
does not considerably affect the above mentioned T, 9 and X.
Conclusion
After investigating the cleaning dynamics of reverse col-
lapse type bag filter, the following conclusions were obtained.
1) After the combined flow resistance variation of
dust and filter material due to reverse air flow was
investigated with 1 order late and dead time models
at reverse air flow, it is relatively well approxi-
mated with a time constant less than several seconds
for this experimental region (1.0^ur<1.42 m/min,
<150 mm H2O) .
2) The effect of ring numbers is not significant to
the time constant but is fairly important to the
cleaning final pressure drop.
Although this report experimentally investigated the qualitative
characteristics of reverse bag filter cleaning dynamics, a quanti
tative consideration as well as a scale factor consideration will
be necessary in a future.
486
-------
T»
resec
TsUsec
5a Nyton(tong)
2"osec wilh «*«
— first order lag model,6«1sec
--dead time model
• experimental data
10 15 20
time. t (sec)
25 30
Fig. J
A Biting cximpto of two modtli (Pint order lag
model * Deed tlm« model) to experimental
data (Nylon fabric, One calcium carbonate
duit, «,-! m/mln, JpmiI-93mmHiO)
487
-------
A=5sec
.T=2sec
/-T=3sec
• experimental data
first order lag model
9=2 sec
dead time model
20
25
0 5 10 15
time . t (sec)
Fig. 2 A fitting example or two models (First order lag
model & Dead time model) to experimental
data (Tetron fabric, fine calcium carbonate
dust, «r »1 m/min, Jpm»i =95 mmH2O)
488
-------
Table 1 Estimated values of time con-
stant Tand dead time 0, i
fiat order Dead time Pressure loss
lag model model given by
Filler Number filtering air
media of flow after
rings time constant dead time infinite
T (sec) j [sec] cleaning
I-lux «->2sec [nnnH;O]
0 4 2 4 70.0
Tetron . • • <* *•* <
(long) 1112 67.5
3 J 2 3 40.5
~Nyton T IT I 2 44.0
(long) 3 1.3 0.7 3 2S.5
489
-------
A METHOD OF MEASURING PRESSURE DROP PARAMETERS FOR
MULTI-COMPARTMENT BAG FILTER
—MECHANICAL SHAKING TYPE AND REVERSE COLLAPSE TYPE—
K. linoya, K. Makino, K. Ueshima,
M. Lin and Y. Mori
Kyoto University, Dept. of Chemical Engr.
Introduction
Generally bag filters will have a long continuous operation
of one or two years after a clean filter cloth is installed.
Therefore, it is necessary to investigate the collection
efficiency and variation of pressure drop characteristics with
time in order to estimate bag filter performance and to determine
a method for design. Especially the variation of pressure drop
characteristics with time is important for economically optimal
design and operation of bag filter. However this pressure drop
characteristic is difficult to measure in the laboratory because
of the following two reasons:
1) The pressure drop characteristics necessary for bag
filter design and operation are usually for one or two years of
operation when pressure drop parameters are stable after instal-
lation. It is very difficult to have such a long investigation
in the laboratory. However, the initial characteristics are
naturally important in the fundamental base and this report will
also investigate this case,
2) Even if the same composition of some dusts and filter
materials as actual field is obtained, it is technically and
economically difficult to reproduce the actual (field) dust dis-
tributions, dust supply condition, gas composition, temperature
and humidity in the laboratory.
j»
Therefore the pressure drop characteristics have to be ob-
tained directly from the actual pressure drop data of multi-
compartment bag filters which is operating in the field. This
method has never been utilized.
We have established the method to qualitatively estimate
pressure drop parameters, which are necessary for economically
optimal design and operations, using the field measurement data
of pressure drop variations of mechanical shaking type and reverse
collapse type bag filters. This report presents this method as
well as the results of actual applications.
1. Fundamental Theory
Many experimental equations are reported for bag filter
pressure drop characteristics and this report applies the following
equations
490
-------
Ap = U(A+BM6) (1)
Here, A, B and 6 are the pressure drop parameters determined by
the characteristics of dust and filter material and by dust
cleaning method. Especially A is the value which, in addition
to filter material and primary accumulation dust layer, includes
residual secondary dust layer after dust cleaning.
Generally the collection characteristic for the multi- com-
partment bag filter is the repeat of (N-l) compartment collection
with one compartment being cleaned while the N-l compartments are
collecting. Now we call the former the cleaning time and the
latter the collection time. There is a following relationship
between the dimensionless dust load xi and the dimensionless
filtering velocity u^ in the ith compartment at the dimensionless
time t in this cycle.
dxi = iiidt (2)
Also the dimensionless pressure drop Ap(t) is given by the follow-
ing equation,
l-a)xi'}ui (3)
where a = A/{A+B(Cnun'T) fi) . (4)
Figure 1 shows one example of the typical pressure drop variation
pattern for collection and cleaning cycles for the mechanical
shaking type and reverse collapse type multi-compartment bag
filters. Generally the process gas quantities during cleaning
collection time are approximately equal. Thus, there is a
following relationship between the average filtering velocity at
cleaning time and collection time.
un_i = {N/(N-l)}un' (5)
Thus the following equations are obtained.
n-1
Z Ti = N (6)
n
Z xi = N (7)
Integrating equation (2) from £=0 to £=TO (cleaning time) and
from £=TO to £=T (collection time) and add the former one from
i=l to i=(n-l) and the latter one from i=l to i-N to obtain the
following equation:
491
-------
cleaning
(fmensionless lime. 1 (—)
Fig. I Conceptional diagram of prcuun drop of
> multi-compartment bag filter
492
-------
n-1
N*. (8)
(9)
There is generally the following relation between the dust loads
at each specific point of collection and cleaning cycle.
(cleaning finish time) (collection starting time)
xi'To = Xi+l,T0 (10)
(cleaning start time) (collection finish time)
*i,o • Si,f
where i = 1,2,..., N-1
Consider equations (10) and (11) and add each side of equations
(8) and (9) .
x_ ?-x' , =N? (12)
n» T i ,TO
Consider x" 1 TO=O and T=NT here to obtain the following equation:
Also the following equation can be obtained from equations (2)
and (3)
{a+(l-a)xi<5}dxi = Ap(€)dt (14)
Integrate equation (14) as described above over cleaning time and
collection time, and add the former one from i=l to i=(n-l) and
the latter one from i=l to i=n.
493
-------
*n ~)-y(xr ¥ ))=N{
" i T n / T o |~
•'T
TO
where y (x)=ax+{ (l-a)/(l+a) }x1+6 (17)
Consider equations (10) , (11) and (13) and add each side of
equations (15) and (16).
y(x'n~) = yd) = CH- ^!
T° ~ ~ f¥
Ap(t)dt+Nl Ap(€)d£
= (n-l)\~Ap(t)dt+\" Ap(£)d£ (18)
* IM
fT ~ - fT
-l)l Ap(t)dt+V
•*o JT
Next, we will describe the determination method of each pressure
drop parameter.
1.1 Determination of Ap (t) and a
N
N
n-1 .
Z {a+(l-a)xi °}~1-1 (20)
Equation (20) can generally be changed to the following in con-
sidering equation (13).
Z
i=l
NApn(T)
N-Apn(T)
Obtain Apn(f) from this,
494
-------
ip ,<0)-ip (T)
=N6 <22>
Now make same consideration for Ap (TO) and Ap _-,(TO).
- <23>
Obtain a from this
*P- Lli ,24,
Here, since the right sides of equations (22) and (24) can be
easily measured, these can be used as the equations to determine
Apn(t) and a*
1.2 Determination of A, B and 6
Based on the definition, the pressure drop characteristics
parameters A and B have the following relationships with a.
, (25)
r> c u T)
n ni n n n
Ap.-u'A Ap- (l-a)Ap-
n • - = - = - L— . (26)
As mentioned above a and AfL(f) can be obtained from equations (22)
and (24) , thus equations (25) and (26) are the fundamental equa-
tions to estimate A and B. Also, using above results, £ can also
be determined by equation (18) . But the equation (18) requires
the time integration of pressure drops and is actually very dif-
ficult. Therefore, the actual region of pressure drop was
considered and numerically reviewed, and it was determined that £
can be considered physically as 1. (See Appendix.) This deter-
mination will make this measurement method very easy and practical.
2. Application to Actual System
495
-------
2.1 Mechanical Shaking Type System
This measurement method was applied to the mechanical
shaking type bag filter for an electric boiler collection and its
pressure drop was measured. Table 1 shows its operating and
installation conditions while Figure 2 presents its pressure drop
measurement results. The time regions 1, 3, 5, 7 and 9 in Figure 2
are so-called cleaning times and all process gas amounts are those
when collection is occurring in the compartments other than cleaning
compartments are collecting. Our method theoretically requires
the variation in pressure drop of one partial cleaning cycle,
which means the data from five measurements. The pressure drop
parameters estimated by applying our approximation method to these
data are shown in Table 2 along with other various values. The
pressure drop characteristics in this case can be expressed by the
following equation.
Ap = u(7.63x!03 + 1.27xl05 M) (27)
Alsq when the same data were analyzed by the numerical integration
method, this will be expressed by the following equation which
gives almost the same result as equation (27).
Ap = u(7.53x!03 + 1.42xl05 M1'08) (28)
2.2 Reverse Collapse Type System
This method was applied, similarly to the previous section,
to reverse collapse type bag filter for electric boiler particulate
collection and its pressure drop was measured. Table 1 shows its
operating and installation conditions and Figure 3 shows the results
of pressure drop measurements. (a) in Figure 3 is the pressure
drop change of other compartments when one compartment is during
cleaning and (b) in Figure 3 is the pressure drop change of the
compartment which is being cleaned at the same time as (a) .
Although for (a) and (b) in Figure 3, we have programed the period 6
in which very small amount of air is introduced at the initial
stage of collection, its existence does not essentially affect our
method. However, the collection amount can be neglected in period 6,
it is necessary to consider the total period of the cleaning cycle
after subtracting period 6. By applying our method to Figure 3,
the pressure drop parameters are obtained as shown in Table 2.
The pressure drop characteristics in this case can be expressed by
the following equation
Ap = u(l.40x10* + 6.40x10* M) (29)
Also the results of equations (27) and (29) can be estimated as
reasonable based on the operating condition of Table 1. Now we
have to review the accuracy of these measurement results statistically
According to this, if the measurements can be done, in considering
actual experience and five measurement results of pressure drop
496
-------
Table 1 Operating conditions of an in-
dustrial multi-compartment bag
filter
Type of dust cleaning Mechanical Shaking Reverse air
Filler medium
Number of compart-
ments A' I— J
Filter area \m-\
Gas flow rate [m'/min]
Total period of cleaning
cycle 7" [min]
Average filtering
velocity fi.v'[m/min]
Dust concentration
r Is/mi]
Dust
Tetron 2020S
10
3150
3840
45
0.74
1.7
withdrawn by suc-
tion directly from
ferrochrome elec-
tric furnace
Tetroa 5203
«
5489
5880
120
1.07
0.30
withdrawn by
suction from the
roof of a factory
housing of a
steelmaking elec-
tric furnace
Total hours operated
after installation [hrs] 408 1600
497
-------
-180
l60
- 1
•V
to
200 400 600 800 1000 1200 1400
tim* , I (we)
Fl|. 2 Examplt of pnuura drop eyeU of in Induttrial
raultl-compartmtnt bt| fllttr (Mechanical
Shaklni Type)
498
-------
Table 2 Examples of Estimation of Pressure Drop Parameters in
Multi-Compartment Bag Fitters
mfy* w« «••*•*• »•*••••
Item* / Run
*H
ImmHtOJ
^Jx-jW)
[mmHtOJ
4fjr.l(rt)
ImmH.OJ
^/» (rt)
ImmH.O]
tfy(l)
|mmH,0]
•••
1
182
143
131
129
133
2
180
146
153
131
135
3
192
149
155
133
138
4
189
150
157
135
139
— «••
5
175
153
158
137
141
Avcrai
1(4
148
155
133
137
P t
286
242
247
207
212
2
267
232
247
202
207
3
286
242
247
207
212
4
286
242
247
207
212
5
286
242
247
207
212
6
225
235
247
196
200
7
230
243
247
193
200
Avcrafe
267
238
247
203
208
0.73 0.75 0.72 0.73 0.80 0.75
• 0.48 0.53 0.47 0.52 0.58 0.52
M
AxlO-* 7.05 7.69 7.27 7.93 8.23 7.63
[mmHiO-tct/m]
BxlO-< 13.4 12.0 14.4 12.8 10.5 12.7
lmmH,0.(m«/kg)l
Bf 228 204 245 218 179 215
|mmH,0/m]
0.74 0.78 0.74 0.74 0.74 0.89 0.87 0.78
0.87 0.89 0.87 0.87 0.87 0.78 0.78 0.84
14.0 13.5 14.0 14.0 14.0 10.0 10.0 12.8
6.40 5.05 6.40 6.40 6.40 8.40 9.50 6.94
19.2 15.2 19.2 19.2 19.2 25.2 28.5 20.8
499
-------
(00
300
- 101
ioo
0
!f
- 100
s»o
100
0
OnIjO
lor clMring «owp»rtmtc«
ifaMMMli * A »•"«•"*•» »"
:
-------
cycle in Figure 2, by certain accuracy (each pressure drop with
±1%, average filtering velocity with ±10% and one cycle cleaning
frequency ±1%), it was concluded that the estimated accuracy of
pressure drop parameter A and B are ±10% and ±40%, respectively.
This means that the estimation of B is more difficult than that
of A. However, increasing the number of measurements will increase
its accuracy and will not require too long a time. Thus this
method is very well applicable to industrial usage. However,
since the estimation of B by this method definitely requires an
accurate measurement of dust concentration in the bag, it will be
expected in the future to have a pressure drop characteristic
measurement and the instantaneous measurement of concentration.
Also since A is the value directly related to the cleaning per-
formance and takes a more important part than B in determination
of bag filter operating condition, this method has the practical
significance in giving a more accurate value of A. Here the dust
concentration of the bag part was assumed to be equal to that
concentration at the entrance of bag filter itself, but when the
dust concentration of the bag part cannot be measured, equation (26)
is the fundamental equation to estimate the product of B and c
(where 6 - 1.0). Table 2 shows the value of Be as a reference.
Conclusion
The estimation methods of pressure drop characteristics A,
B and 6 of mechanical shaking type and reverse collapse type bag
filters from field data were reviewed and the following conclusions
were obtained.
1) The pressure drop parameters A, B and 6 can be estimated
by time variation in partial cleaning cycles of mechanical shaking
type and reverse collapse type multi-compartment bag filters
pressure drops.
2) It was shown that assuming
-------
A = a+b Moq (ii)
BM6 = b(Mo+M)q-b Moq (iii)
The general pressure drop characteristics will have the regions
given by following equations.
30 < a < 1400
4000 < b < 17000 (iv)
0.5 < 9 < 1.5
For the regions of equation (iv) , A, B and 6 were estimated
as a parameter of M by using the least squares method. As a
result, it was determined by B, if b and q are constants, will
have approximately a constant value independent of Mo and 6 will
have a certain relationship with q, as a parameter of Mo, inde-
pendent of a and b as shown in Figure 4. The especially interesting
point is that even if q changes significantly, 6 is always about
1.0. Thus, for the region of equation (iv), the pressure drop
estimation by assuming 6=1.0 and applying the approximate pressure
drop coefficient B will only have errors within i several %.
Therefore the approximation of 6=1.0 is industrially sufficient.
Acknowledgment
We express our sincere appreciation to Shinto Dust Collector
Company and Clean Filter Company concerning important field
pressure drop data for multi-compartment bag filters for this
research.
502
-------
0.8
0.1. 10 0.6
q (-)
Fig. 4 Relationship twiwcm «tod I
503
-------
PERFORMANCE OF FIBROUS POWDER BED FILTER
Koichi linoya, Kazutaka Makino,
Tsutomu Imamura, and Hiroshi Okura
Kyoto University, Chemical Engineering Dept.
The filter collector is relatively efficient and practically
important for eliminating micron and submicron particles. Here,
fibrous bed and filter paper are mainly used for air purifica-
tions, and the estimate of their performance is reasonably
accurate due to the relatively wide fundamental studies of their
behavior.
Although the fabric filter was considered mainly for indus-
trial applications, the method of accumulating fibrous powder on
the filter cloth is being recognized for air purifications.
This is due to the following reasons: 1) It is relatively easy
to evenly accumulate fibrous powder on the filter cloth; 2) as a
consequence of this, the fibrous bed is formed on the filter
cloth surface and the initial collection performance is improved;
3) utilization of fibrous powder has a less pressure drop than
ordinary powder; 4) since particles are mainly collected by
fibrous powder which will be swept off after certain amount of
particle collection, itj. is easy to sweep off the collected tiny
particles and easy for the prevention of filter cloth clogging.
Whereas, there has been hardly any fundamental research for
fibrous powder bed filters, thus this report presents the ex-
perimentally investigated results.
Experimental Apparatus and Method
The_ experimental apparatus is exactly the same as shown in
the previous reports. The fibrous powder bed filter was formed
by accumulating fibrous powder in a stratiform on the filter
paper or cloth. The experimental procedure was to first measure
the collection rate of filter paper or cloth and the overall
collection rate of the entire fibrous powder bed filter under
predetermined conditions. Next, the penetration rate of the
fibrous powder bed was obtained as a ratio of the latter one to
the former one. The stearic-acid-single-diffusion-particles by
Lamer type generator were used as the sample aerosol and the
aerosol concentration was measured by the digital dust counter
(manufactured by Shibata Kagaku Co.) simultaneously at filter
entrance and exit. Table 1 shows the characteristics of the
fibrous powder used in this experiment.
Experimental Results
The penetration rate and the pressure drop across fibrous
powder bed filter;
504
-------
Table 1 Characteristics of test fibrous powder
Fiber length Fiber diameter Specific turface
/ M D, t/i] area S. tmVm']
Precoat (•)
Cellulose ,.,
powder A <*J
Cellulose r- >
powder B (-)
Cellulose , ,-,
powder C CT)
2TO
180
110
80
22
18
17
IS
1.98 *10»
2.03 xlO>
2.20 xlO»
2.43 xlO»
Received on February 22. 1973
•• Keithi linoya (il:.£H),K»iutak« Makino (iE6rtX Tjulomu
Imunur* (73i£H) and lliroshi Okurs (
-------
Figure 1 shows a relationship between penetration rate and
amount of fibrous powder for the fibrous powder bed filter.
According to this, the cellulose powder has a lower penetration
rate than the precoat. This is probably due to the fact that the
true density of the filter material for the former one (pf =1.7
g/cm3) is smaller than the latter one (pf =2.2 g/cm3) and that
the former one has a larger filling volume and collection surface
area for the same fibrous powder amount. The relationship between
the volume fraction and the thickness of fibrous powder bed is
shown in Figure 2 as a reference. Based on this the precoat has
a larger volume fraction and a smaller thickness than the cellulose
powder. On the other hand, both have about the same penetration
rate for the apparent velocity through filter of 1-10 cm/sec.
Also in this region of apparent velocity, the pressure drop AP^
across the fibrous powder bed is given by the following equation
based on the Kozeny-Carman's law.
US 2 (l-ef) u m
Apd - 60 -g;; - ,f» pf (± 10%> (1>
Here, the void fraction £f can be easily estimated in Figure 2.
For example, when mo = 1 Kg/m2 and us = 1 cm/sec. AP^ will be
* 6 mm-water. This value is somewhat smaller than that of
ordinary powder.
Single Fiber Collection Efficiency of Fibrous Powder Bed Filter;
The experimental results were converted to single fiber col-
lection efficency (ne) by the following logarithmic permeable
equation and the relationship between TI£ and (l-£f)'L was investi-
gated.
4(l-ef)L
As a result, in the region of (l-ef) »L^0.15. ne is constant and
is independent with changes of (l-ef) «L. Namely, since (l-ef) •L=m0/pf
in this region and the permeability (1-nd) will decrease as an
exponential function of mo. Also all the data of this experiment
is shown in Figure 3 as a correlation between He a^d l-ef) .
This result can be used as a simple estimation method of collec-
tion efficiency of fibrous powder bed filter. The solid line in
Figure 3 is given by the following equation,
lnne - -1.3(l-ef) -2.9 (3)
However, the applicable region of this equation is for the apparent
velocity across filter of 1~10 cm/sec and particle size of
0.3~0.8y for fibrous powder with diameter of 10~20p and with length
of 100~200y.
506
-------
0123
amount of fibrous powder bad, m.(kq/frf)
Fig. 1
Relation* between penetration (1—*) »nd
unooni of fibrout powder bed m«
507
-------
The symbols aro th?
same is those in
"0.1 23 & 7 1 23 b T K
amount of fibrous powder bed,mf(k
-------
S*~Q07
.i'oos
-*~A
X.
•precoat
A cellulose powder A • cellules* powder B
-—'—L^-'—' "' ' ' ' ' ' ' ' ' '
7v*l * i- _ * . i . - i ii it
oo a05 007 Qffl Oil 013 0.15 017 019 021
volume traction of fibrous powder bed.) -£)(-}
Fig. 3 Relition between collection efficiency of tingle fiber
». «nd Tolume fr«ction (I-,/) of fibrout powder in bed
509
-------
Also, in order to estimate overall collection efficiency of
fibrous powder bed filter, an information concerning collection
efficiency of filter paper or filter cloth itself other than
equation (3) is required. Please refer to Reference (2) and (5),
respectively, for this information. Also, the theoretical
correlation between single fiber collection efficiency obtained
this report and ordinary fibrous filter will require future
evaluation.
This experiment used the powder bed filter with short fiber
(about 20y diameter and lOOy long) previously untested and
measured its performance experimentally. As a result, a simple
estimation equation for collection performance and pressure loss
was obtained.
510
-------
APPENDIX I
UNIVERSITY OP OSAKA
511
-------
REPRINTED
FROM
FUNDAMENTALS
Growth of Aerosol Particles by Condensation
Tetsuo Yoshlda, Yasuo Kousaka, and Klkuo Okuyama
(Reprinted with permission)
Reprinted from I&EC Fundamentals
VOL. 15 NO. 1 FEBRUARY 1976
512
-------
Growth of Aerosol Particles by Condensation
Tetsuo Yoshkla, Yasuo Kousaka/ and Kikuo Okuyama
Department of Chemical Engineering. University of Osaka Prefecture. Sakai, 591. Japan
The rate and the extent of growth of submicron aerosol particles introduced into a supersaturated atmosphere
of water vapor were studied from both theoretical and experimental points of view. It was found that the rate of
growth of aerosol particles undergoing condensation was very rapid, and that the volume-mean diameter of
grown particles was determined by the number concentration of aerosol particles and the initial state of super-
saturation in the surrounding gas. The supersaturation which was necessary to cause condensation of water
vapor around particles was produced by mixing hot saturated air with cold air. The size distribution of grown
particles or grown water droplets was determined by the ultramicroscopic size analysis previously developed
by the authors. The results suggest that particle growth by condensation is one of the most promising precondi-
tioning techniques for industrial dust collection.
Introduction
The cost of removal of submicron dust particles from ex-
haust gas has been considered to be very expensive. If the
growth of such particles into larger ones (to a few microns
or more in diameter) can be easily attained, such precondi-
tioning techniques of dust collection will facilitate air pol-
lution control. Condensation of water vapor on particle sur-
faces has been proposed as one of the most promising tech-
niques to promote particle growth (Fahnoe et al., 1951;
Schauer, 1951; Lapple et al., 1955; Lancaster et al., 1971).
Because of the difficulty in measuring the size distribution
of water droplets smaller than several microns in diameter,
the overall effect of condensation on the rate and extent of
particle growth still remains unknown. In this study the
rate and extent of particle growth by condensation were
studied from both theoretical and experimental points of
view. In order to effect condensation on particle surfaces, a
supersaturated atmosphere of water vapor was produced
by mixing hot saturated air with cold air into which several
kinds of submicron particles, not consisting of soluble sub-
stances, were introduced. The size distribution of grown
water droplets was measured by a new technique previous-
ly developed by the authors (1975).
Rate of Growth of Polydisperse Aerosol Particles
When a water droplet with radius r is put into a super-
saturated atmosphere, the rate of growth of the droplet has
been given as follows (Fuchs, 1959):
£ = -^|Sp9(7\.)-po(To,r)||l +
at rRptT I
2p,
(1)
This equation is based on Maxwell's equation for the sta-
Ind. Eng. Chem., Fundam., Vol. 15, No. 1. 1976
513
-------
'aropitl
Flgnn 1. Illustration of temperature and pnwur* flelds (round a
growing droplet.
tionary evaporation of a ipherical droplet motionlsas rela-
tive to an Infinite uniform medium and, in addition to the
Maxwell equation, the correction for the effect of Stefan
flow ia made. 5 in the equation represents supersaturation,
which ia larger than unity in thia study, but whan S is leas
than unity, the equation represents the diminishing rate of
a droplet by evaporation. p0 represent* the vapor pressure
at the surface of a droplet which is assumed to be equal to
it* equilibrium pressure known ss the Kelvin equation
(2)
The rise in temperature of a condensing droplet is given by
the following equation taking account of Stefan flow:
+ T. (3)
2pt
Thia equation is derived under the assumptions that heat
transfer by convection and radiation is negligible and that
the quantity of heat transferred to the media from the
droplet equals the amount generated in condensation. Fig-
ure 1 illustrates the changes of temperature and pressure
fields around a growing droplet which is put into a closed
and insulated cell initially having a certain degree of super-
saturation. The rate of growth of a single water droplet
with radius r will then be determined by above equations.
In order to apply these equations to aerosol particles in-
stead of water droplets, one must assume that the surface
of each particle is covered by a thin water film at the start
of condensation. This assumption is based upon the insta-
bility of a supersaturated atmosphere, where water vapor
immediately condenses upon any particles as the condensa-
tion nuclei. This will be discussed later. When aerosol par-
ticles are polydispene, the change in sixe distribution of
the growing particles ia then derived from the conservation
of mass
(4)
»n(r, t) m_±\ . dr(t)|
»t »r\ ' dt I
Before computing the above equations, it is necessary to
determine the degree of tupersaturation, 5. Consider a sys-
tem when aerosol particles an steadily introduced into the
supersaturated atmosphen which is produced by contin-
uously mixing hot saturated air with cold air in an insulat-
ed chamber. Such a system may be thought of more simply
as a system where a certain number of aerosol particles an
put uniformly into a dosed and insulated chamber contain-
ing air initially having a certain degree of supersaturation.
In the system the degree of supersaturation will decrease as
condensation increases, and after a sufficient time it ap-
proaches unity. The relation between the degree of super-
saturation and the amount of condensed vapor may be eas-
ily understood with the aid of the humidity chart shown in
M. Eng. Cham., Fundam.. Vol. 15, No. 1.1976
tfmperalur*
Figure 1 Changs in humidity and temperature due to conduits-
tion.
Figure 2, when i indicates the initial state just before aero-
sol particles an introduced. When particles an suddenly
dispersed uniformly into the chamber, condensation upon
the particles occurs, and as a consequence there is a de-
crease in humidity and a simultaneous rise in the tempera-
ture of the air as the condensation progresses. Thus the
change is indicated by the slope of the adiabatic change
shown in Figure 2. Temperature, humidity, and supersatu-
ration during the successive condensation an then given by
the following expressions.
(5)
(6)
(7)
0.24 + 0.45Hi
H-H|--»p.( f" r»n(r, t) dr - f" r»n
-------
99
95
rj90
60
»70
£60
I 50
l'°
= 30
d
3 20
" 10
0.4 0.6 08 1 2
particle diameter
2r
4 6 8 10
02
0.4 0.6 08 1 2 4
particle diameter 2' Cpl
6 8
experimental conditions
particle
key
o
o
e
•
material
tobacco
stearic acid
OOP
carbon black
no
2.66 « 10"
3.31 x 108
3.83 x 107
2.54 « 108
°vi,H
0.85
0.75
0.75
047
dgi
1.34
UO
126
1.42
saturated air
T"A
78
eo
81
61
Vc
21
20
19
26
Rh
0.26
0.27
0.26
0.27
AH
0.0070
0.0078
0.0075
0.0067
change
0«in| Ogm
cal.
1.63
1.61
I.Gb
1.26
1.11
1.0i
1.0S
1.05
uf Oy and a*g
Dvf
*«P.
i.06
3.10
6.20
3.95
cal.
3.81
363
7.30
3.71
rfgt
MP.
1.26
1.17
l.3b
1.23
cal.
1.021
1.006
1.002
1.006
Figure S. Change in particle size distribution with time.
to the saturated vapor pressure, p.(T_), in the present
case, the growth rate of small droplet is generally greater
than that of a large one as shown in eq 1. Thus the differ-
ence of growth rates will make the size distributions nar-
rower in the figures. The final distributions in Figure 3
were taken as those when the growth rates of each droplet
became nearly zero in the numerical calculation. Such dis-
tributions will be, strictly speaking, unstable, since the
evaporation of smaller droplets among polydisperse drop-
lets will occur with a decrease in supersaturation. This ef-
fect of evaporation, however, was so small in the calculation
that the size distributions did not change significantly
within a few seconds.
Estimation of the Extent of Particle Growth to B«
Expected
The analysis described above was on the rate process. In
this section a discussion will originate from another point
of view. While AW represents the quantity of condensable
water vapor per unit mass of dry air as described before,
the following relation must be satisfied when all of the
vapor corresponding to Atf is assumed to condense upon
particle surfaces.
6 810"" 2 46 810"" 2 4 6 810"* 2
Figure 4. Volume mean diameter of grown particle.
•ir
r3n(r, «)dr
- r
Jo
, 0) dr
6
(10)
This equation indicates that the volume mean diameter of
grown particles, Dvf, can be evaluated when the value of
AH and the volume mean diameter of particles before
growth, Dvi, are known. When Dyf3 is large enough com-
pared with D^1 and p. is nearly unity as it may be in most
cases, the volume mean diameter of grown particles, Dvi,
can be written in a simpler form as
(11)
Dv, ~ (6AH/Tno)I/3 (Dvf3 » Dvi3; p. ~ 1)
The straight line in Figure 4 shows this relation.
Figure 5. Schematic diagram of experimental apparatus.
Experimental Apparatus
Figure 5 indicates the schematic diagram of the experi-
mental apparatus. Two kinds of saturated air, one of which
was humidified by contact with hot water and the other
with nonnested water, were continuously mixed in a mixing
chamber to produce a supersaturated atmosphere. The de-
gree of supersaturation or the quantity of condensable
water vapor A// was controlled by changing their mixing
ratio and the combination of their temperatures. Aerosols
were continuously introduced into the supersaturated at-
mosphere at a constant rate. The size distributions of the
aerosol particles used in the experiment were obtained by
an ultramicroscopic size analysis (Yoshida et al., 1975) and
kvl Eng. Cham., Fundanr. Vol. 15. No. 1, 1976
515
-------
ft 80
»70
140-
|30
|20-
u
to
toy QtfOiOl
aroonbao U2
02 OA 06081
particle diameter 2r Cji]
Figure ft, Size distributions of aerosol particles used.
objective*
1 thick bronze lube
lection 7« 7 to
165
outlet
tcondentfete
in mm
Flfui* 7. Observation cell of ultramicroscope.
especially important when the present technique is devel-
oped into industrial applications.
For instance, one may consider exhaust gas which con-
tains dust particles 1 n in volume mean diameter, 2.5 g/cm3
in density, and 500 rng/m3 in concentration. In this case,
the dust particles will grow to 3 n in volume mean diameter
when AH of 0.005 g of condensable vapor per gram of dry
air is established.
Conclusion
The rate and the extent of growth of polydisperse aerosol
particles introduced into a supersaturated atmosphere
were studied, and the following results were obtained. (1)
The rate of growth of aerosol particles was very rapid. The
width of size distributions of grown particles obtained ex-
perimentally was not narrow, while a narrower distribution
was expected from the theoretical analysis. (2) The extent
of particle growth was evaluated in volume mean diameter
and was confirmed by experiments. (3) Even the hydropho-
bic particles grew well.
These results suggested the particle growth by condensa-
tion will be one of the most promising preconditioning
techniques for dust collection. The technique of establish-
ing a supersaturated atmosphere will be important for the
industrial application of these results.
Acknowledgment
F. Nomura, T. Yasumunn, and T. Miyazalci were very
helpful in the experimental work.
are shown in Figure 6. The size distributions of grown par-
ticles were also determined by the same method. In order
to prevent any change in the size of the water droplets due
to evaporation or condensation, the temperature and pres-
sure in the observation cell for the size analysis were kept
the same as those in the mixing chamber. The double-tube
cell shown in Figure 7 was used for this purpose.
Experimental Results and Discussion
As discussed before, the rate of growth was so fast com-
pared with the time scale of measurement that the size dis-
tribution of a growing particle at each stage of the elapsed
time could not be observed. However, observations ob-
tained by varying the residence time of particles in the
mixing chamber suggested that the particle growth was
completed within 1 sec. Some of the experimental results of
the size distribution of grown particles were plotted in Fig-
ure 3. The width of the size distribution of grown particles
did not become narrower as compared to theory. The wider
distribution obtained from the experiments was thought to
be caused by a lack of spatial uniformity in the degree of
supersaturation. The widths of size distributions of grown
particles undergoing condensation were observed, in all
cases including those in Figure 3, to be neither wider nor
narrower.
The estimation of the volume mean diameter of the
grown particles, which would be important for industrial
purposes, was determined by experiments. Figure 4 shows
the comparison of an estimated line with the experimental
results. The abscissa of the figure corresponds to the mean
quantity of condensable vapor per single particle. The fig-
ure indicates that even the hydrophobic particles, such as
carbon black and D.O.P., grow well. The growth of such
particles may be caused by the instability of a supersatu-
rated atmosphere as described previously. This fact will be
tad. Eng. Cham.. Fundam.. Vol. 15. No. 1.1976
Nomenclature
D " diffusion coefficient of vapor, cmVsec
Dyi, Dvra. Dvt * volume mean diameter before, in the mid-
dle of, and after growth, respectively, M or cm
H - absolute hum idity.g of HjO/g of dry air
dH « condensable water vapor, g of HjO/g of dry air
i - enthalpy, cal/g of dry air
K = heat conductivity, cal/cm sec °C
L • latent heat of condensation, cal/g
M " molecular weight nf condensing substance, g/mol
n " particle number concentration, 1/g of dry air
no ™ total particle number concentration, 1/g of dry air
p = vapor pressure, mmHg
R •» gas constant, cm3 mmHg/mol °C
r » radius of particle, M or cm
RI, " mixing ratio, g of dry air of hot saturated air/g of dry
air
S " degree of supersaturation defined by eq 7
T - temperature, "C
T • mean temperature between TO and T_, °C
t • time, sec
Greek Letters
p, " density of condensed liquid, g/cm3
a » surface tension, dyn/cm
, "ft " geometric standard deviation before, in the
mid"
ffrm. "it " geometric standard aeviaiu
middle of and after growth, respectively
Subscript*
f " final state shown in Figure 2
h » hot saturated air
i » initial state
1 « cold saturated air
s • saturated
t - total
w » water
0 « particle surface
<• • far away from particle
516
-------
Literature Cited Vomite. T., Kouuka. V.. Okuyarm, K.. ma. £ng. O»m.. funOtm.. 14, 47
FihKM, F , Un*oo«. A.. AtMtoon. R. J.. ma. £ng. O*m.. 4», 13M (H51).
Fuchfc H. A., "Eviporaton «nd Oroptol »owtt< In Guwxit M«««
mon fnu. London. 185B.
Kowto, A.. Olund, 8.. J. Atm». Set.. H, 1060(1989). Accepted SepUmbtr 5, 1976
Uncntw.B W.,Str»u»», W., M. Eng. Chun.. FunOim., 10,9«a(1971|. . . . . .
Uppl».C. E.K«m»c*.H.J., Oam. fnfl. Proa., 51 (3), 110(1»!5). Pre«nt»d it the Chemic*! Enjinwnng Meetinf in J«p»n, April
Sehiuv. P. J.. M Eng. Chun., «, 1532 (t»S1). 1974.
kid. Eng. Ch»m , Fundam.. Vol. 15. No. 1. 1976
517
-------
STABILITY OF FINE WATER DROPLET CLOUDS
YASUO KOUSAKA, KIKUO OKUYAMA,
KENJI SUMI AND TETSUO YOSHIDA
Department of Chemical Engineering, University
of Osaka Prefecture, Sakai, Japan
518
-------
STABILITY OF FINE WATER DROPLET CLOUDS
Abstract
The stability of fine water droplet clouds was studied for
two standpoints. In the first, the rate of evaporation of
monodisperse water droplets was evaluated by numerically
solving the modified Maxwell's equation assuming the cellular
model for a droplet cloud. In the second, the equilibrated
system, where a water droplet cloud is steadily mixed with
unsaturated air, was analysed on the basis of enthalpy and
material balance of the system to evaluate the total volume
change of the droplets. Some of .these analyses were verified
by the experiment using the ultramicroscopic technique which
is useful for droplet size analysis.
519
-------
1. INTRODUCTION
It is well known that under a certain supersaturation of
water vapor excess water vapor condenses upon aerosol particles
as the condensation nuclei to generate small water droplets.
Such water droplet clouds are usually found when a combustion
gas is cooled, a highly humid gas in high temperature is mixed
with that in low temperature, or steam is injected into a gas.
These droplets often have diameters less than ten microns and
they are thought to be unstable because of their high vapor
pressure at the surface. When these droplets being contained
in an industrial exhaust gas are required to be collected from
the gas, it will be necessary to evaluate the effect of
unstaaility of droplets or the decrease in droplet sizes and
number concentrations in a collector system under various
operating conditions. The stability of droplet clouds is also
important in measuring droplte size distribution and in
evaluating the behavior of atmospheric aerosols.
In discussing about the stability of water droplet clouds,
two kinds of approach to the subject were made in this paper.
In the first, the rate of evaporation of monodisperse pure
water droplets, which are led uniformly into a closed vessel
initially containing air having certain humidity and temperature,
is discussed using cellular model(Fuchs (1959) Zung (1967)).
It is the main purpose in this discussion to evaluate the
lifetime or the time required to be equilibrated of water
droplets in terms of various initial sizes and number
concentrations of droplets as well as various
520
-------
initial air conditions. In the second, the equilibrium state
of droplets in the system where a droplet cloud is steadily
mixed with unsaturated fresh air will be discussed from
enthalpy and material balance of the system for various
conditions.
As the experimental technique, the ultramicroscopic size
analysis previously developed by the authors (1975) was applied.
Because of the difficulty in measuring the unsteady size change
during rapid evaporation, most of the experiment were limited
to verify the analysis of the equilibrium state of droplet
clouds .
2. RATE OF EVAPORATION OF A WATER DROPLET CLOULD
In this section, the rate of evaporation of a water droplet
cloud is discussed using the cellular model (Fuchs (1959) Zung
(1967)). In the cellular model, a droplet cloud, where the
droplets are distributed equidistant ly each other, is assumed,
and the cloud is divided into a number of identical cubic cells
each of which is supposed to contain a single droplet in the
center. The length of the edge of such a cube is then given as:
b = VITn (1)
In this model the evaporation of a droplet cloud may be reduced
to the single droplet evaporation. In the conditions where n'Q
is less than about 106 particles/cm3 and the droplet radius
is less than several microns, the rate of change in droplet
521
-------
radius r' in one cell almost agrees with that of an isolated
droplet according to Fuchs (1959) and Davies (1973), which
can given as t
PM Po
-------
where, nw = n^/vH (6)
In consequence of evaporation of a droplet, there are an
increase in humidity and a simultaneous fall in temperature
of air. These changes are indicated by the slope of the adia-
batic change in the humidity chart as shown in Fig. 1. The
temperature, the humidity and the degree of saturation during
the successive evaporation are given by the following
expressions.
T«2 " T~l " AHeL/{0.24 + 0.45H2) (7)
H2 = Hx - AH£ (8)
S " P«, = H(0.622-»-Hs)/{Hs(0.622+H)J (9)
While the change of the system is unsteady, the unsteady
fields of temperature and pressure are established around each
evaporating droplet. It is difficult, however, to calculate
the change in particle radius strictly taking account of the
unsteady fields. Then a quasi-stationary analysis, where
temperature and pressure fields were considered to be constant
during each step of time, was made. In calculation, the
change in droplet radius was first evaluated by numerically
solving Eq.(2) with Runge-Kutta-Merson method, and then the
consequent change in the state of surrounding air was calculated
by Eq. (5)^ Eq. (9) , and this step was repeated until the
driving force for evaporation became zero. The temperature
dependence of the physical properties appearing in the above
equations was taken into consideration in the computation.
523
-------
Some of the results thus calculated are shown in Figs. 2 and
3. Fig. 2 shows the time dependent change in radius of
evaporating droplets under various particle number concentra-
tions n' in intially saturated air(S=l). It can be seen that
the fine droplets of micron order and having low number concen-
tration tend to evaporate even in saturated air because of the
Kelvin's effect. The broken lines in the figure show the
analytical solutions by Davies (1973) for the evaporation rate
of an isolated droplet(refer to Appendix). With the increase of
number concentration n', the decreasing rate in droplet radius
becomes slow, and at n' larger than about 10 particles/cm
the droplets of 0.5 y in radius seems to be almost stable.
Fig. 3 shows the dependence of evaporation rate of droplets
on the degree of saturation under various conditions. As seen
from Figs. 2 and 3, the stability of water droplet clouds
depends greatly upon the initial degree of saturation S and
number concentration n' Under the low values of S and n',
the quantity of water vapor produced by evaporation of droplets
is not enough to saturate surrounding air and therefore droplets
disappear completely. On the other hand, under the high values
of SQ and n^, droplets continue to evaporate until the equi-
librium vapor pressure as given by Eq. (3) is attained and
because of the sufficient amount of water vapor to be evaporated
in this case,the droplets are stable after a slight change in
size.
3. EVALUATION OF THE CHANGE IN DROPLET SIZE UNDER EQUILIBRIUM
524
-------
STATE BY MIXING OP DROPLET CLOUD WITH UNSATURATED AIR
Since the change in radius of a droplet by evaporation
proceeds in a short time, as calculated in the former section,
the droplet cloud which we can actually observe will be in the
equilibrium state. Therefore there is an importance to analyse
the equilibrium state of a droplet cloud for various situations.
Figure 4 shows a schematic diagram to be analysed in this
section. Such a system, for instance, may be interpreted by
that where an exhaust gas containing a certain amount of small
water droplets encounters unsaturated air resulted from some
leakage in a dust collector system.
As shown in Fig. 4, when air containing droplets is mixed
with unsaturated fresh air at a certain mass ratio, the amount
of water droplet containing in the resultant air is expected
to be decreased. In such a system the following enthalpy and
material balance equations are derived under the assumption
of the existence of droplets in the equilibrium state after
mixing of air, that is, AHf£. 0:
(material balance of water)
R H + (1-R )!! . + (1-R )AH. « H . + AH. (10)
m n * m' si * m' i sf f
(enthalpy balance of the system)
V. + (1-R»)isi + u-VMii - i.f + ^f^f (11)
Another expression of Eq.(11) may also be written as follows:
R fo.24T -H597.1+0.45T )H \ + (l-R ) ) 0.24T .+{597.1+
m nt in mi m i si
525
-------
(l-Rm)AHiTsi « 0.24Tgf+(597.1+0.45Tgf)Hgf
* AHfTsf
where, Hgf - f (Tsf) (12)
AH in the equations represent the mass of water droplets
suspending in air per unit mass of dry air, and R the mixing
ratio of unsaturated air to resultant air in mass basis of
dry air. Since the Kelvin's effect is small enough in the case
of droplets larger than 0.1 y in diameter, it was neglected in
the following analysis.
In the case of AH, = 0 The quantity of water droplets
^^B^"^^^™"^"^^^™*" ^ V^H^MI^^^^W
after mixing of unsaturated air, AH., in Eqs.(10) and (11)
decreases with the increase of the mixing ratio of unsaturated
air, R , and finally AH. becomes zero, that is, all droplets
disappear by evaporation. The relation among each variables
appears in the above equations at such a critical condition
can be calculated by putting AH =0 in the equations. If the
conditions before mixing are known, the values of R , H f and
T are obtainable from Eqs.(10), (11) and (12). Some of the
calculated results are shown in Figs. 5 and 6.
In the case of AH,, > 0 The containable quantity of water
W«_B^_^««_«^H~^B«M^BB^B«^» I ^™«^™^»
droplets after mixing of air, AH kg water per kg dry air,
can be essentially evaluated by Eqs. (10), (11) and (12). The
relation among the variables in the equations, however, is more
complicated in this case. Some of the calculated results are
shown in Figs. 7 and 8.
526
-------
Evaluation of AHf and DU£ If the initial conditions of a
droplet cloud and the state of unsaturated mixing air are given,
the quantity of water droplets after mixing of air, AH., can be
evaluated as described above. The volume mean diameter of the
droplets after mixing of unsaturated air is then evaluated
knowing the number concentration of the droplets, n , as follow.
Dvf
w
f AHf )
- — (13)
l(ir/6)n,p J
. _
ViT 8
4. EXPERIMENTAL APPARATUS AND METHOD
Figure 9 shows the schematic diagram of the experimental
apparatus to examine the analysis in section 2. The water
droplet cloud was steadily generated by mixing hot saturated
air with cold saturated airs**"T'The hot saturated gas contains
by th« wtHors UVTfcl
X& ',
small dust particles having diameters around 0.05 M which are
generated in burning fuel gas and the gas is mixed with cold
saturated air to produce supersaturation which causes droplet
formation on the dust particles as condensation nuclei. Thus
obtained saturated air containing a certain amount of small
water droplets was continuously led into a vinyl chloride pipe
with diameter of 26 mm to make a turbulent flow. The length
of the pipe was 10 m, at inlet and outlet of the pipe the
aerosols were sampled with isokinetic condition. The second
experiment to observe the analysis in section 3 was made by
introducing the saturated air containing droplets
into a mixing chamber instead of the pipe, where unsaturated
527
-------
air was mixed. Then a part of the mixture was drawn out for
observation. Size analysis and determination of concentration
of droplets were made by the same method using an ultramicro-
scope as those previously developed by the authors. In order
to prevent any change in size of droplets due to evaporation
or condensation during observation, the temperature and the
pressure in the observation cell must be kept same as those
in the mixing chamber. A heat exchanger type cell as shown in
Fig. 10 was used for this purpose.
5. EXPERIMENTAL RESULTS AND DISCUSSIONS
Fig. 11 indicates the comparison of the particle size
distribution between at inlet and outlet of the pipe. The
residence time of droplets in the pipe is about 4 second in
this case. If the evaporation theory of an isolated droplet
is applied to this case the droplets smaller than about 1.6 u
disappear by evaporation. As seen from the graph, however, no
appreciable change in the droplet size distribution occurs.
This fact will be reasonable because a droplet cloud having
high number concentration, 10 particles/cm in this experiment,
is expected to be stable from the analysis shown in Fig. 2.
Fig. 12 shows some examples of droplet size distributions
obtained by the second experiment where the droplet cloud is
mixed with unsaturated air in a mixing chamber. Fig. 12(a)
is the size distribution of droplets before mixing of unsatu-
rated air, and Figs. 12(b) and (c) are those after mixing of
528
-------
unsaturated air. Slight difference is found among these three
distributions, whereas fair difference among them in particle
number concentration is found. All of the other experimental
results showed the same tendency as those illustrated in these
figures. Loss of water droplets due to mixing of unsaturated
air shown in Pigs. 12 (b) and (c) will be caused by evaporation
of some of the droplets.
Since the vapor pressure of small droplets is higher than
that of larger ones as expected by the Kelvin's equation, the
droplets once decreased their sizes for a certain reason,
for instance local lack of uniformity of humidity, can easily
evaporate. The smaller the droplets becomes, the quicker
progresses their evaporation, and in consequence smaller
droplets may not exist. It is an interesting phenomena that
droplet sizes do not decrease uniformly but number concentration
only decreases by mixing of unsaturated air, preserving the
/
initial droplet size distribution itself. This phenomena will
not be undesirable for dust collection, because some of the
droplets produced by condensation upon small dust particles
decrease their sizes to those of the former small dust particles
which are difficult to collect.
Fig. 13 shows the comparison of the mixing ratios, Rm/
calculated from Eqs.(lO), (11) and (12) with those observed,
when droplets just disappear by evaporation, that is, AHf = 0.
Good correlation is found between them.
Some examples of the experimental results for AHf>0 are
shown in Pigs. 14 and 15. Fig. 14 shows the effect of the
529
-------
mixing ratio R and the temperature of mixing air T on the
A n
quantity of droplets remaining in air, AH , which was determined
by Eq.(13). This figure suggests that loss of droplets is
significant, when leakage of fresh air having high temperature
into a droplet cloud exists. The deviation of the experimental
results from the calculated curves may be caused by the
experimental error of D f which effects on AHf at third power
as seen in Eg.(13). Fig. 15 shows the relation between the
temperature after mixing of unsaturated air, T ., and the
mixing ratio R . It is obvious that T . decreases with R
m sr m
because the droplet evaporation requires latent heat.
6. CONCLUSION
The stability of pure water droplet clouds was studied
for two standpoints. In the first, the rate of evaporation of
monodisperse water droplets was evaluated by numerically
solving the modified Maxwell's equation assuming the cellualr
model for a droplet cloud. The lifetime of a droplet cloud or
the time required for a cloud to be equilibrated was illus-
trated for some typical conditions for the better understanding
of the phenomena. The experimental verification for the analysis,
especially for the unsteady droplet size change, could not be
made because of the difficulty in measuring the droplet size
undergoing rapid evaporation. However, for the equilibrated
state, which is easily attained in a droplet cloud with high
number concentrations, reasonable experimental results were
530
-------
obtained using the ultramicroscopic technique.
In the second, the equilibrated system, where a water
droplet cloud is steadily mixed with unsaturated air, was
analysed on the basis of enthalpy and material balance of the
system to evaluate the quantitative change of the total volume
of the droplets. The analysis was verified by experiemtns. As
to the manner of the decrease in total volume of droplets,
the unexpected decrease in droplet number concentration was
observed instead of droplet size change.
ACKNLWLEGHENT
Y. Sakata was very helpful in the experimental work.
531
-------
APPENDIX
Davies (1973) derived the following equations expressing
the decrease in size of an isolated liquid particle due to
evaporation for two situations:
(i) Evaporation of liquid particle into a vapor-saturated
air( S=l ).
dy y+1
- = - D.K' - 5 - (A-l)
dtf r yty^+l
(ii) Evaporation of liquid particle into a vapor-free air
( S=0)
dy y+1
dt' y+1.71y+1.333
where, D. - DMp /RTpeX2, K' = 2Mo/RTp X, y = r' /X
CSS S
532
-------
REFERENCES
Davies, C. N. (1973) Faraday Simp, of Chem. Soc., No.7, 34.
Fuchs, N. A. (1959) Evaporation and droplet growth in gaseous
media, Pergamon Press.
Yoshida, T., Kousaka, Y. and Okuyama, K. (1975) Ind. Eng.
Chem. Fundam. 14, 47.
Yoshida, T., Kousaka, Y. and Okuyama, X. (1976) Ind. Eng.
Chem. Fundam. 15, 37.
Zung, J. T. (1967) J. Chem. Phys. 46, 2064.
533
-------
Nomenclarure
b
O
D
v
Df
H
AH
e
AH
i
Kn
K
K'
L
M
»i
n
R
m
S
T
(cm]
[cm /sec]
length of the edge of a cubic cell
diffusion coefficient of vapor
volume mean diameter of the droplets
diffusion factor used in appendix
absolute humidity
quantity of evaporated water vapor [kg H2O/kg dry air]
quantity of water droplets [kg H2
-------
t' : time [sec]
vu : humid volume [m /kg dry air]
n
y : (-Krf1) [-]
Greek letters
X : mean free path [cm]
p : density of droplet [kg/m ]
a : surface tension [dyne/cm]
o : geometric standard deviation [-]
$ : percentage humidity [%]
Subscripts
f : final state of air after mixing
i : initial state of air before mixing which contains
water droptets
m : unsaturated mixing air
s : saturated
w : water
0 : droplet surface
09 : far away from droplet
Superscript
' : for water
535
-------
Capations of figures
Pig. 1 Change in humidity and temperature due to evaporation
Fig. 2 Evaporation of fine water droplet clouds into
vapour-saturated air
Fig. 3 Evaporation of fine water droplet clouds into air with
various degrees of saturation
Fig. 4 Schematic diagram of the system to be analysed
Fig. 5 Correlation among mixing ratio R , mixing air
temperature T and percentage humidity $ in the case
of AHf = 0
Fig. 6 Correlation among mixing ratio Rffl, mixing air
temperature T , droplet quantity contained in air
m
before mixing AH. and temperature of the same air T .
i si
in the case of AH = 0
Fig. 7 Effects of the quantity of initial droplets on those
after mixing of unsaturated air
Fig. 8 Effects of every variable on containable droplet
quantity, AH , after mixing of unsaturated air
Fig. 9 Schematic diagram of the experimental apparatus
Fig.10 Observation cell
Fig.11 Comaprison of the particle size distribution between
at inlet and outlet of the pipe
Fig.12 Change in size and number concentration of droplets
before and after mixing of unsaturated air
Fig.13 Comparison of mixing ratios, R , calculated and
observed in the case of AH = 0
Fig.14 Effect of mixing ratio R and temperature of mixing
air T on remaining droplet quantity AH
n £
Fig.15 Effect of mixing ratio R and temperature of mixing
air T on air temperature after mixing of unsaturated
n
air' Tsf
536
-------
Cn
Co
E
JC
saturated
line
temperature T(
ao
Fig. 1 Change in humidity and temperature due to evaporation
-------
Ul
OJ
00
' = 3.0u
1—I I I 111
T 1—I i I I I I
i 1—r-j-r
i 111 ' ~ i 1—i | i i 11
nn=1(r pgrticles/cc
initial condition
TQO=20°C.Sb=1.0
• Davies* Eq.
time f
Fig. 2 Evaporation of fine water droplet clouds into vapour-saturated air
-------
VD
1 1 1 I I I I I
initial S =1.0
initial condition
-.1 = 1.0
Davies* Eq
10~3 2
2 5 10-1 2
time t' CsecD
Fig. 3 Evaporation of fine water droplet clouds into air with various degrees of saturation
-------
in saturated air
1-Rm,
sfjwt
droplet cloud equilibrium state
of droplet cloud
Fig. 4 Schematic diagram of the system to be analysed
540
-------
1.01
^^^^ * ^^* * i
-c.Hsi?: 0.0860
0m: persentagt humidity
of unsaturaled mixing air
20
30
40
50
60
m
Pig. 5 Correlation among mixing ratio Rm, mixing air
temperature T and percentage humidity $m in the case
of AH
541
-------
TSi=20tc
TSi=50-c
Fig. 6 Correlation among mixing ratio Rm, mixing air
temperature T , droplet quantity contained in air
n
before mixing AIT and temperature of the same air
in the case of AH, - 0
542
-------
0,01
S 0,01
o
X
•o
o>
JC
en
0.005
TSi«50«c
Tm «30'c
0m =50%
O.S
1,0
Pig. 7 Effects of the quantity of initial droplets on those
after mixing of unsaturated air
543
-------
0.015
T=70*c
Fig. 8 Effects of every variable on containable droplet
quantity, AH , after mixing of unsaturated air
544
-------
thermometer
droplet
formation
humidifier
dia. 200mm
heightlOOOmm
O.Sinch Raschingring
recircu-
latory
water
Ul
*>.
en
t '
1
1 1 f
L— l 1 /
~^»
j>
flow meter
cnuriiu
1
1
1
I
• cooling
J water
i thermometer
| flow
. meter
?
i
c
sampling tap
RV.C.pipe(dia.26mm)jnletside
sampling tap
outlet side
flow meter
-a-
thermometer
mixing
chamber
unsaturated I
air
i
high temp
exhaust gas
saturated air
containg
droplets
dehumidifier
dia.150mm
height 750mm
O.SinchRasching ring
rSflow
LJ meter
observation
cell
I
vaccum
pump
, - i - . nfiow
| ultramicroscopej UJ meter
- i - . 4
TV camera | vaccum
pump
\
})02
OJ
013
T unsaturated
' air
X-Y
recorder
automatic
particle
counter
- -[monitor j
sampling
in mm
first experiment: V, open.V2 shut
second experiment :v, shut.V2 open
(mixing chamber)
Fig. 9
Schematic diagram of the experimental apparatus
to
blower
-------
urethane
8
objective of microscope
_ electro magnetic valve
a * ——
side view
I-A IB
1 thick bronze tube
in mm
Fig.10 Observation cell
-------
2
^
c
^ 1
>s
O
C
0)
D
CT
0)
i_
>4_
0
T=23°C
n0=8.0x105
porticles/cc
rg=1.65p
0^=1.35
-
•
o: inlet
(Osec)
• ; out let
(^ sec)
o
£\
1 *\
A \
i •
/• °\
f^ •
o^ 8
€ ^
i >T .
V
, .°v
.6 .8 1
radius
Fig.11 Comaprison of the particle size distribution between
at inlet and outlet of the pipe
547
-------
en
•u
oo
99.
95
90
n
s« 80
N
| 50
3 40
I 30
a
| 20
u
10
Rm=0 ' ' '(a)"
TSi = 47*c
. Hsi=0.073 kg HJD/kg dry air
particles/kgdryairy
.AHj=0.0059
kgHjO/kgdryair/"
(b)
Tm=30-c
Hm=0.0063 kgH20/kgdryair,.
particles/kg dry air
.AHf =0.0033
kg H20/ kg dry air
/
;
Rm=0.28
(c)
.Hm=0.0063 kgHjO/kgdry ai
"wf=6,3x107
. particles/kg dry air
. Hf=a0026
kg H^O/kg dry air
: /
2 3456 7891 2 3 4 5 5 789 t
particle diameter
i t
3456 789
Fig.12 Change in size and number concentration of droplets
before and after mixing of unsaturated air
-------
HSI=0.031~ 0.073
j =0.0025-0.0070
Hm=0.0058~0.0081
Tm=21-57«c
0.5
Recalculated)
Pig.13 Comparison of mixing ratios, RTO, calculated and
observed in the case of AHf = 0
549
-------
a
>»
i_
TO
en
9,0.005
Ok
X
•«a
Ts j = 47'c.HSj = 0.073
; =0.007. Hm=0.0081
Fig.14 Effect of mixing ratio R and* temperature of mixing
air T on remaining droplet quantity AH
01 ^
550
-------
50
35
30
25
TSi«*7*c,HSj*0.073
4Hi=0,007, Hm«0.0081
0,5
M
1,0
Fig.15 Effect of mixing ratio R and temperature of mixing
m
air T on air temperature after mixing of unsaturated
n
551
-------
BEHAVIOR OF AEROSOLS UNDERGOING BROWNIAN
COAGULATION, BROWNIAN DIFFUSION AND
GRAVITATIONAL SETTLING IN A CLOSED
CHAMBER
KIKUO OKUYAMA, YASUO KOUSAKA
AND TETSUO YOSHIDA
Department of Chemical Engineering. University of Osaka
Prefecture, Sakai, 591
(Reprinted with permission)
Reprinted from
JOURNAL OF
CHEMICAL ENGINEERING
OF
JAPAN
Vol. 9, No. 2 (1976)
Pages UO 146
-------
BEHAVIOR OF AEROSOLS UNDERGOING BROWNIAN
COAGULATION, BROWNIAN DIFFUSION AND
GRAVITATIONAL SETTLING IN A CLOSED
CHAMBER
KIKUO OKUYAMA*. YASUO KOUSAKA
AND TETSUO YOSHIDA
Department of Chemical Engineering, University of Osaka
Prefecture, Sakai, 591
The behavior of aerosols undergoing Brownlan coagulation, Brownian diffusion and gravita-
tional settling in a closed chamber was studied by solving the basic equation, the so-called popula-
tion balance equation, numerically for a polydispene aerosol system and analytically for a mono-
disperse system, and then the results were examined by experiment. In solving the basic equation.
two dimaisionless parameters, which are determined by the initial properties of an aerosol and
the chamber dimension and also characterize the relative effects of Brownian coagulation and
Brownian diffusion to gravitational settling, were introduced in order to generalize the behavior
under arbitrary conditions. The calculated results, the time-dependent changes in particle number
concentration and particle size distribution for a polydispene system, were presented graphically
by using the above two parameters. And further using these parameters, the domains of the
three controlling factors were mapped to snow the extent of each effect of these factors under
various conditions for a monodisperse system. Some of the calculated results were compared
with the experimental results obtained by the ultramicroscopic size analysis previously developed
by the authors.
Introduction
The behavior of aerosols in a closed chamber is
generally characterized by coagulation, diffusion,
sedimentation, thermophoresis, existence of genera-
tion sources of aerosol and so on. Brownian coagula-
tion increases the size of aerosol particles, resulting in
decrease of number concentration. Brownian diffu-
sion decreases the concentration of aerosol particles as
the result of the deposition of small particles at the
walls, while gravitational settling decreases the con-
centration as the result of the deposition of larger
particles at the bottom wall. The effects of these
three factors on aerosol behavior, change in size and
number concentration of aerosol particles in a closed
chamber are discussed in this paper.
Several theoretical1'''''''10' and experimental1-4-9'11'
studies of this behavior have been reported in the field
of nuclear power reactors, where the behavior of
radioactive aerosol generated by an accident in the
reactors is considered to be important from a safety
viewpoint. In theoretical studies, basic equations
considering Brownian coagulation and sedimenta-
Received August 29,1975.
Presented at the 40th Annual Meeting of The Soc. of Chem.
Engrs., Japan at Nagoya, April 3, 1975 (entilted "Behavior of
Aerosols in a Closed Chamber").
tion1'6'101, sedimentation and diffusion81, or three
effects4'*'71'" have been solved numerically on the
assumption that the concentration is uniform except
very close to the walls. Experimental studies, on the
other hand, have been carried out by observing the
change in mass concentration and particle size dis-
tribution of aerosol in the observation chamber.
The agreement between their calculations and ex-
periments has been found to be fairly good4-1".
These results, however, are obtained under particular
conditions and so they seem insufficient to predict the
general behavior under various conditions such as
various particle size distributions, number concentra-
tions and chamber sizes.
In this paper, the basic equation in dimensionless
form taking account of these three effects was solved
numerically under various conditions for polydis-
perse aerosols and analytically for monodisperse
aerosols. In calculations two dimensionless param-
eters which characterize the relative effect of gravita-
tional settling to Brownian coagulation and Brownian
diffusion were introduced. By using these parameters,
the calculated results of the change in particle number
concentration and size distribution with time were
graphed so that the behavior of aerosols under vari-
ous conditions is ready for prediction. Some of the
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
553
-------
calculated results were confirmed by experiment,
using the ultramicroscopic size analysis previously
developed by the authors'".
Theoretical Consideration
Polydisperse aerosol
Suppose an aerosol is dispersed throughout a
chamber, within which aerosol convective currents,
though small, arise due to small temperature vari-
ations. As a result, aerosol concentration in the
chamber is kept uniform except very close to the
walls4'*'". In this case the basic equation expressing
the time dependence of the size distribution of aerosol
particles undergoing Brownian coagulation, Brownian
diffusion and gravitational settling is given as1-4'*1"'1",
when the particle size distribution is a discrete spec-
trum.
l—imln.,.imax
(1)
Eq. (I) is described as a differential equation, but it
can be rewritten as a partial Integra-differential equa-
tion in the case of a continuous spectrum. The left
side of Eq. (I) represents the change in particle num-
ber concentration of size r'( with time. The first
term on the right side represents the rate of formation
of panicles of size r( and the second term the rate of
loss of particles of size r{ due to coagulation.
K(rlp'i) is the coagulation function and in the case
of Brownian coagulation it is given by1"
K(f(, M-KAr'<+rt{Cm(r |)/r{+ CJWUfft,
AT,-2«r/3^ (2)
The third term represents the loss of particles due to
sedimentation with the terminal settling velocity, and
the last term expresses Brownian diffusion onto cham-
ber walls, which is evaluated on the basil of film
theory characterized by the thickness of concentra-
tion boundary layer 9".
Eq. (I) was derived on the following assumptions.
(1) There exist no external forces except gravity.
(2) Particles are spherical and electrically neutral.
(3) Particles collide with each other to form a
single new spherical particle whose mass may be the
same as the combined mass of the two smaller particles.
(4) The aerosol concentration is spatially uniform
except within the boundary layer of thickness 4, and it
changes with linearity within the layer4'.
(5) The chamber has a vertical cylindrical shape.
VOL. 9 HO. 3 1974
The initial particle size spectrum in the chamber was
assumed to be established instantaneously with the
following log-normal form:
"'
«'(rl.0)=
(3)
y '
To generalize the solution the following dimension-
less variables, which have been used in the case of
Brownian coagulation alone1", were chosen:
"(*"<• 0="v<''')/"•> tc=K
-------
»1=0.35 ft and «r,0 = 1.3
spending to these three dimensionless times and their
analytical solutions for monodisperse system are
shown in Table 1 as a function of CG and DG, to-
gether with the time /._0.» when particle number
concentration reduces to half of the initial one.
Calculation Results and Discussion
Particle number concentration
Fig. 1 shows the normalized number concentration,
which is defined as the ratio of the concentration at
any time to the initial one, as a function of the di-
mensionless time based on Brownian coagulation.
The graph also shows the effect of CG,
-------
T»bl« 1 Tiroe-G>pend«it change In particle mmiber concentration for monodlsperse aerosol
Controlling factor and dimen-
sionlcss time
Coagulation
Sedimentation
lc=u,(r't)l'IH
Diffusion
where
Time-dependent change in par-
Bas.c d.mens.onless equation |jc|e num^r concen,r»lion
,„, _df.?*P.<--V<-)_
«/I/tt/C=- —*" —^0" "VI
dn/dto=CD( - 2nl -
CD = CCIDC
ditions by means of curves of cumulative percentage
against particle size like Fig. 2. Generally two
parameters, a geometric mean radius and a geometric
standard deviation, are used to characterize the
particle size and the width of distribution when the
particle size distribution follows log-normal form.
Though there existed some deviation from log-normal
form in this case, a nominal geometric mean radius
and a nominal geometric standard deviation are now
introduced as follows:
(9)
n'<««. ) / "I"/*
£ n'(r!./')In'(r,'/r;) /«'
'<-'!.,. J I
where
n'= L n'(r!.r')
The change of size distribution with time will be dis-
cussed by the aid of r', and a,. The change in the
ratio of r', to the initial one is shown in Fig. 3(a)
using various values of CO. Distinct maxima at CO
larger than 1.0 appear in the figure. The increasing
regions of r',/r',, will be controlled by the particle
growth due to coagulation. The decreasing regions,
on the other hand, are controlled by gravitational
settling, where the enhanced settling velocities of
large particles being grown by coagulation will con-
tribute to the decrease. In both regions, however,
r', of the larger standard deviation changes faster
than that of the smaller one. As seen from Fig.
3(b), which shows the effect of r',,, particles with small
f'it grow much more than those with larger r^. This
dependence on r',, seems to be caused by coagulation
function K(r',.p',)t which increases with decreasing rj,.
Fig. 4 shows the change in a, with time. In spite of
the variation of a,,, the curves seems to converge to
certain values which will be determined by CG and
DC. The effect of r',t on the change of a, with time
is not so large in the range of r^=0.1~1.0^i. The
curves of C(7=100 in both Fig. 3 and Fig. 4 agree
well with those of Brownian coagulation alone"1.
Fig. 4 Variation of nominal geometric standard deviation
with time
COM
Fig. 5 The domaitu of the three controlling ficton
Controlling factor under arbitrary conditions
Fig. 5 shows the domains of the three controlling
factors on CG-DC coordinates. Each domain was
determined essentially from the dependence of r..B.i
for monodisperse aerosol shown in Table I on CG
and DC. In the case where one of the three factors,
coagulation, diffusion and sedimentation, is ignored,
'.-•.I can be determined by a parameter (CG, DC or
CD) consisting of two remaining factors as shown in
VOL. 9 NO.
1976
556
-------
vessel B
monitor TV
Fig. 6 Schematic diagram of experimental appratus
Table 2 Determination of the domains of the three con-
trollliig factors
Negligible ^-o-i.by two
* ST*
Diffusion lim (/c).-o.»
DG-+0 or ro~°r
CD-.00
Equation expressing the
domain where the factor
is ignored
I
A,
lim (f0).-o.»
?g:t"
Sedimen-
tation
CC-»oo or
lim
(11)
Coagulation
CG-Oor
C0-0
(12)
Table 2, and then, thus calculated /..,., agrees with
'•M.I in Table 1 which includes the three factors.
The boundary relations, which express whether one
factor is ignored or not, were obtained as Eqs. (10)
(11) and (12) in Table 2 by regarding one factor as
negligible if the difference between two kinds of /„„„.»
in Table 1 and Table 2 is less than 10% of ',., • in
Table 2.
Experimental Apparatus and Method
\ schematic diagram of the experimental apparatus
it shown in Fig. 6. Aerosols used in this study were
tobacco smoke, stearic acid and DOP. Aerosols of
both stearic acid and DOP were generated by a La
Mer-Sinclair type generator, and tobacco srnake was
generated by a simple apparatus111 by which number
concentration of panicles was controlled from 10* to
10' particles/cc. Aerosols thus generated were intro-
duced promptly throughout the chamber. To main-
tain gentle mixing of the aerosol in the chamber, the
aerosol was mechanically stirred with a small fan set
at the bottom of the chamber. Aerosol sampled at
every given residence time was introduced into the
observation cell installed on the stage of an ultra-
microscope to measure its particle size distribution
arid particle number concentration. This measure-
ment method using an ultramicroscope was developed
previously by the authors1". Experiments were car-
ried out by changing the intitiat particle number
concentration and changing the chamber dimension
to obtain the experimental data for a wide range of
CG and DC.
Experimental Results and Discussion
In a comparison of experimental data with theory,
the boundary layer thickness S in DG must be deter-
mined. Several investigators have experimentally
obtained the values of S, as shown in Table 3, by
combining the deposition rate on the wall per unit
area with the suspended mass concentration. The
values of 8 are scattered in the range of 0=0.01 ~
1.9 p. Figs. 7, 8 and 9 show comparisons of the
change in particle number concentration between
experimental data and theoretical curves taking as
i—0.2 ft and 1.6 ft. As shown in Fig. 7, there is no
difference caused by stirrcr speed or sampling posi-
tion, and these facts suggest that neither turbulent
coagulation nor turbulent diffusion occurs and that
particle number concentration is uniform throughout
the chamber. In every case in the figures experi-
mental data agree with the curves of 8=1.6 ft rather
than those of 6=0.2 ft. Estimating the controlling
factors from Fig. 5, it is found that coagulation in
Fig. 7, coagulation and diffusion in Fig. 8 and coagula-
tion, sedimentation and diffusion in Fig. 9 are con-
trolling. Fig. 10 shows a comparison of the change
in nominal geometric mean radius with time between
experimental data and calculation curves. The agree-
ment is fairly good and it is seen that aerosol having
the larger CG grows much greater than the smaller
one, as expected from theoretical calculations. Fig.
11 is an example of a series of photographs, taken by
a camera directly attached to the ultramicroscope,
which correspond to the data plotted in Fig. 7. It is
seen that particle number concentration decreases
gradually, accompanying particle growth due to
Brownian coagulation.
Conclusion
The behavior of aerosols undergoing Brownian
coagulation, Brownian diffusion and gravitational
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
557
-------
ta
C3
00
296 596
U1
»9' 596
T9 975
477
Big
CM
Hi
\&
13?
477 :0*0|135
7M !05l|l32
3751 9751780 ;05l!l.32
3531975176.0 10.51 |l 32
n.-564.icr-746«10'
k*y
0*0
A A
01
Mintr lftf<3
[rpm]
1950
3060
1950
tamping
top
1
1
1
Fig. 7 Experimental and calculated time-dependent change
in particle number concentration
4 6 BIO' 2 4 6 8)CP Z
t - KoniCmOiJt' C—3
Fig. 8 Experimental and calculated time-dependent change
In particle number concentration
Table 3 Boundary layer thickness
Investigator
Kitani el al.f>
Greenfield el a/.41
Nelson et al'->»
Aerosol
Na,O
NasO
Na,0
Particle radius
w
0.32~1.0
0.29~0.5
1.0~1.5
1
M
O.OI~I.9
(a*. 0.34)
1.75
0.7-1.6
settling in a closed chamber was studied theoretically
and experimentally. To generalize the analysis, two
dimensionless parameters CO and DG, which express
the relative effect of coagulation and of diffusion to
gravitational settling, were introduced in solving the
basic equation. By using these two parameters, the
results of numerical calculation for polydisperse
aerosol having various initial size distributions with
log-normal form were presented graphically, and in
the case of monodisperse aerosol analytical solutions
were obtained. So far as the change in total number
concentration with time was concerned, the calcula-
vou 9 NO. i IWA
,0-1 2 4 6 6,0-' 2 4 6 8100 2 4
Fig. 9 Experimental and calculated time-dependent change
In particle number concentration
10'
Fig. 10 Experimental and calculated variation of nominal
geometric mean radius with time
t'iBmin t':30mtn
tobacco smoke
Fig. 11 Change of particle number with time
tion results of polydisperse aerosol agreed closely
with those of monodisperse aerosol. The manner of
change in particle size distribution with time was
found to depend on the values of CG and DG as shown
in Figs. 2, 3 and 4. As the results of analytical solu-
tion for a monodisperse system, the domains where
each of the three factors becomes controlling were
mapped on CG-DG coordinates, which permitted a
general understanding of the behavior of aerosols
558
-------
under various conditions. Some of the calculated
results were examined experimentally and were found
to be in good agreement when the boundary layer
thickness 8 was taken as 1.6 p.
The results obtained in this paper and in the previ-
ous one10, which showed the effect of Brownian coagu-
lation and diffusion on gravitational settling consider-
ing the variation of particle number concentration in
the direction of sedimentation, provide a basic and
representative concept for predicting the actual be-
havior of aerosols in a closed system.
A,
CD
CO
Cm(ri)
DC
D(r'i)
H
K,
K(r'ti PI)
«',«
r't.r
r't
r i
5
T
('
tr
- constant shown in Table 1 (=(! +DG)/CG) [— J
= dimensionless parameter (-CG/DG) [— ]
- dimensionless parameter defined in Eq. (6)
[— ]
[-1
Cunningham's correction factor of rj
dimensionless parameter defined in Eq. (6)
diffusion coefficient
(-C.(rJ)«r/6it/*ri)
acceleration of gravity
chamber height
coefficient in Eq. (2)
coagulation function for two particle*
of size r i and f\
dimensionless coagulation function
[cm'/sec]
[cm/sec1]
{cm]
[cm'/sec]
[cm'/sec]
l-l
size width between /•; and rj+i [— ]
O - number and dimensionless number
concentration of aerosol particles
(-«'(r5.OM) [partlcles/ccH-J
total and dimensionless total particle
number concentration (-n'/n't)
[partlcles/cc][-]
total particle number concentration
it time zero [particles/eel
particle radius and dimensionless
particle radius (-rJ/rW [cm],M,H
geometric mean radius [cm], [p]
particle radius of monodltperse aerosol [cm], f>]
wall area of chamber [cm1]
absolute temperature [*K]
time [sec]
dlmenilonless time based on coagulation
[-]
- dimensionless time based on diffusion
/a - dimensionless time based on gravitational
settling (-i/,(riV7«) [-]
/..o.i - dimensionless time when « reduces to half [—]
"((''!) ~ terminal settling velocity
(-2CB(ri)0)p-p)/r;ay9^) [cm/sec]
V — chamber volume [cm1]
i - boundary layer thickness [/<] [cm]
<• /"( = particle radius and dimensionless particle
radius in Eqs. (I) (2) and (5) [cm] [—]
I " the number particle size
mln •> minimum
max " maximum
0 — at time zero
Literature Cited
1) Ahn, C. H. and J. W. Gentry: Ind. Eng. Chem. Fundaai.,
11, 483 (1972).
2) Devler, S. E.: /. Colloid Scl., 18. 744 (1963).
3) Devier, S. E.: /. Colloid Interface Scl., 21, 9 (1966).
4) Greenfield, M. A., R. L. Koontz and D. F. Hausknecht:
Ibid., 35,102 (1971).
5) Huang, O. M., M. Kerker and E. Matijevic: ibid., 33, 529
(1970).
6) Takahash), K.: Kuki Seljo, 10,27 (1972).
7) Takahashl, K. and M. Kasahara: Atmos. Environ., 2, 441
(1968).
8) Kitanl, S., H. Matsui, S. Uno, M. Murata and J. Takada:
/. Nuel. Set. Ttchwl., 10, 566 (1973).
9) Langstroth, O. and T. Clllespie: Cand. J. Research, 25B,
455 (1947).
10) Lindauer, O. C. and A. W. Castleman, Jr.: /. Aerosol Set.,
2, 85 (1971).
11) Nelson, C. T., L. Baurmash and R. L, Koontz: Proc. 9th
AEG Air Cleaning Conf. (ACE Report CONF-660904),
454 (1966).
12) Smoluchowskl, M.: ffiyi. Chtm. (Ltlpitg), 92,127 (1917).
13) Yoshida, T., Y. Kousaka and K. Okuyama: Ind. Eng.
Chtm. nmdam., 14,47 (1975).
14) Yoshida, T., Y. Kousaka, K. Okuyama and S. Nlshlo: J.
Chtm. Enf. Japan, 8,137 (1975).
IS) Yoshlda, T., K. Okuyama, Y. Kousaka and Y. Klda:
IbU., 8,317 (1975).
16) Zebel, O.: Kolhtd-Z,, 156,102 (1958).
JOURNAL Or CHEMICAL INOINIMINO OF JAPAN
559
-------
A New Technique of Particle Size Analysis of Aerosols and Fine Powders Using an
Ultramlcroscope
Tetsuo Yoshlda, Yawo Kousaka,* and Klkuo Okuyama
Faculty ot Engineering, University ol Osaka Prefecture. Osaka. Japan
A new technique to determine the particle size distribution of fine powders and aerosols Including those
of fine liquid droplets was developed. The technique Is In principle somewhat similar to the Andreasen
plpet method, but has some distinctive features as follows: (1) sedimentation Is made In air as well as
in water according to the particle size: (2) sedimentation depth is extremely shallow; (3) particle con-
centration at a given depth in a sedimentation cell is detected by an ultramicroscope on the number
basis. The lower limit of the measurable particle size Is several tenths of a micron in diameter and the
upper more than several tens of microns. By sedimentation mostly In air at depths less than a few milli-
meters, very quick measurement was possible even for submlcron particles. It is desirable for the parti-
cle number concentration to be high.
Introduction
In recent years an increasing interest in the size analy-
sis of aerosols has arisen in connection with air pollution
control. In existing ultramicroscopic techniques of particle
size analysis (Richardson, et al., 1956; Mukaibo et a/.,
1962), the settling velocities of hundreds of individual
particles Vnust be observed by an ultramicroscope to de-
termine the size distribution, and thus the techniques re-
quire much time. Trie technique developed ir. this study
is quite different from the existing ones and, above all,
has the distinctive feature of quick analysis of size distri-
bution for particles, including water droplet*, of more
than several tenths of a micron in diameter. The tech-
nique is also applied to the determination of the particle
number concentration of. aerosols and, by some additional
devices, the density of particles. Although the technique
described below is limited to aerosols of small particles
dispersed in air. for the powders of larger particles the
technique can be applied when they are dispersed in
water.
Principle
The principle of the method is almost same as that of
the Andreasen pipet method, but the concentrations of
particles at a given depth as sedimentation progresses are
detected by an ultramicroscope on the number basis. Sed-
imentation in air or in water with shallow sedimentation
lengths makes it possible to analyze smaller particles in a
short time, whereas in the Andreasen pipet method sedi-
mentation is usually made in water with a deep sedimen-
tation length. The sedimentation length is from about 0.5
to a few millimeters, so that, for example, a water droplet
1 it in diameter is measured within several tens of sec-
onds. The detection of particle number concentrations by
ultramicroscope and then the analysis of particle size dis-
tribution is made by the following procedure: aerosol is
introduced into a small cell having valves at both inlet
and outlet sides, the flow of the aerosol is instantaneously
stopped by closing the valves, and then sedimentation is
started. When the focus of the ultramicroecope is prelimi-
narily set at depth H shown in Figure 1, the aerosol parti-
cles existing in the volume vm are recognized because of
their shining at the depth of the focus but they are as yet
unknown in sizes. The particles appearing in the micro-
scope are photographed or recorded by a video recorder as
sedimentation continues until the particles disappear
from sight.
The terminal settling velocity of a particle is represent-
ed by the equation
vt =
18/1
(1)
U\ in this equation is replaced by h/t, where h is the
depth of sedimentation preliminarily set; then the particle
diameter Dp is determined in accordance with the sedi-
mentation time t, that is
Dt =
(2)
The particle with the diameter Dp calculated by eq 2 at
time t is the maximum one which can be recognized at
the depth h; that is, the particles larger than it have al-
ready passed below the depth h. Thus the particle number
in sight at time t after the start of sedimentation is given
by the equation
MO = N(0)fD'l(Dt) AD,
(3)
The integral term of this equation indicates cumulative
fraction undersize, then
F = f\D,) ADt = MO/MO) (4)
N(0), the particle number in the volume Um in Figure 1 at
the beginning of sedimentation, a/id Mr), the numbers at
various elapsed times, are counted by slow video. Then
the size distribution is determined by eq 4. Table I shows
the analysis procedure.
In the procedure, to prevent photophoresis, thermopho-
resis, and thermal convection, the lighting of the micro-
scope must be extinguished during the intervals between
successive observations.
Another method may also be true in principle. If the
sedimentation depth h is varied instantaneously at con-
stant elapsed time, the particle size distribution can be
determined by almost the same procedure described
above. The particle number concentration, no, is given as
follows, when the volumes vm or uh shown in Figure 1 are
predetermined
560
Ind. Eng. Chem.. Fundam.. Vol. 14. No. 1,1975
printed with permission)
-------
beam
/
aerosol out
Figure 1. Illustration of observation cell.
I
eyepiece
U7,xtQ.x15)
\ objective
J0c10,x20.x40)
(Observation celt
Figure 2. Arrangement of uttramicroecope.
Table I. Procedure of Analysis
Nil),
1, number of
time particle
elapsed in (>„
0 MO)
', N(lt)
It Nil,)
r>>,
particle
diameter
F(h/t,}
F(h/lt)
r,
cumulative
undersize
t
Af(?,)/AT(0)
N(/2)/AT(0)
nt =
«o =
(5)
(5')
where Moui is the particle number in the volume v*
which is obtained by counting all the particles passing
through the volume um during sedimentation. Equation 5'
ii especially useful when the' particle number concentra-
tion is mall. In this procedure, however, since the micro-
scope must be kept lighted during the measurement, it in
not mitable to measure the size distribution of particles
iraaller than about 1 M in diameter, where thennophoresis
and thermal convection affect sedimentation.
Irtd. Eng. Ch«m.. Fundim.. Vol. 14. No. 1,1975
\ I (tick ttonrr lub» lij
—:_i.165_i.ii_i— °
valve
outlet
objective of
glow wall
Net
ousel
In mm
Figure 3. (a) Ohmervalinn cell far fog particle, (b) Observation
cell for am all particle.
The estimation of Ah in Figure 1, the depth in focus, is
made as follows. Small numbers of particles to he ana-
lysed are first, deposited on a glass in some way. The glass
is mounted on the stage of the ultramicroscope and the
deposited particles are observed while shifting the stage
up and down. Then &h is given ai the total displacement
of the stage over which thp images of the particles are in
sight with a certain clearness.
When the particle number concentration is thus ob-
tained, and when th* particle weight concentration is sep-
arately determined hy another appropriate way such us
filtering, the particle density is, in principle, obtainable
by means of the above technique of size analysis.
In the principle described here Rrownian coagulation
and diffusion are neglected. When the particle number
concentration exceeds 107 particles per cubic centimeter,
the effect of the Brownian coagulation will not be negligi-
ble (Fuchs, 1964), and when'there ure particles lens trtsn
about 0.4 n in diameter, diffusion influences the gravita-
tional settling (Daviea, 1949). The development of some
devices to remove these effects by forced sedimentation,
such as by thermophoretic or centrifugal forces, or theo-
retical methods to compute the effects will be desirable in
order to extend the application of this technique.
Apparatus
The ultramicros.-vjpe itsed in this study is shown in Fig-
ure 2. It has a 160-W halogen lump and an observation
cell on its stage. The magnification of the objective and
the eyepiece were selected according to the particle size
and the particle number concentration. Two typical types
of the observation cell are shown in Figure 3. The cell
shown in Figure 3a consists of a double cell. The outer cell
U provided for temperature control of the aerosol to be
observed in the inner one. This type is useful for the anal-
ysis of fog particles or "other volatile particles which can
vary their size by evaporation or condensation with
change of temperature. The cell shown in Figure 3b has a
small sectional area so that the effect of convection is re-
moved .
The aerosol particles tested in this study were generated
561
-------
vqlyt
generator (or tobacco a*ro«oi
by-poll
corrprwsor
Figure 4. Arrangement of tobacco nm*ol generator
fotamvtar
filltr
h«t nchangtr
rwonlrars
ttnptraturr
regulator —
itMricacId
i ii Him \
. Sttsrlc icld ifro«ol generilor.
humid Iflef
A£K
N
,- »
r
I
rc
n..
' th«
c
-w—
X*-
|L.
NPump' "
rmom»t»r
tog bai
-
iwotef
rrr-^lonli
1
.»
tobacco at
ascondm*
nuclei
—
ro»oi
31 ion
oir_
'lower
Pleura *. Fo» pirticlf f» nerntor.
ai shown in Figure* 4, S. and R. Figure 4 shown a tobacco
aero«ol K'nerator. The particle number concentration of
tobacco aerosol was controlled from 10* to 10* particle*
per cubic centimeter by air flow rate through tobacco and
the rat« of the by-pnnft. The particle nize was controlled hy
residence time in the aging chamber from 0.7 to 2 n in
geometrical mean diameter. The stearic acid particles
were generated hy a La Mer generator ihown in Figure .r>,
but because of the much control of tempernturr they were
not monodispcree. Figure 6 show* a fog generator. The fog
particles were formed around the tobacco particles as con-
densation nuclei, and the particle number concentration
and the particle size as well were controlled by changing
the number of the condensation nuclei. The controlled
sixe was from 1 to 10 n and the concentration was from
10* to 10* particles/cm'.
Experimental Section
Some characteristics regarding the ultramicroscope
used were predetermined before experiment*. Table II
shnw« the result*. The last column gives a criterion of the
Figure 7. Chinx* of psrticle numh«r with tht Ispur of time (MM-
ric scid leroml): («) ( • 0 oer afirr thr «t»rt of •rriimentstinn:
(b)f • 40 see; (ell -50»eci(d)« -60sec;(») ( -70i««c.
Table II. Characteri»tics of the Ultramicroscope Used
Eve-
plcce
^ nl>-
Jectlve
10 ^ 10
10 *20
10 x 40
&li,u
125
40
15
>'.i
1.9
1.4
1.2
cm'
x 10-t
x 10'''
x 10-B
Visible
pn rile In
di:iin,/i
S2
^0.2
^0.05
cm1! e may be
enlarged by giving a horizontal sweep to the observation
cell. In any case the more dense the particle number con-
centration is, the easier is the analysis in this met hod.
A series of the photos which illustrate the change of the
particle number as time elapses is shown in photos (a) to
(e) in Figure 7. These pictures are only for illustration and
the most of the size analysis were made by the slow video
system because of the more rapid analysis than the photo-
graphic one.
Figure 8 shows the result of the analysis. No differences
in the size distribution are found for various depth* of
sedimentation. This agreement indicates that the aenwol
introduced into the cell is homogeneously dispersed and
also indicates almost no effect of diffusion or thermophor-
esis. which probably appears at the vicinities of the upper
or the lower wall of a cell because of the larger gradient*
Infl Eng. Chem.. Fundam.. Vol. 14. Vlo. 1. 1975
562
-------
95
90
7*>
U70
a60
•«50
•§40
^30
t> 20
>
4rf
{)
c
§5
<
1 ' 1 ' 1
_stearic acid
particle
VIM'
°/
r\,= 2.90x106 *
partlcles/cc , ~
^r-
«-
— • • —
,r
V -
i
1 , 1 ' . 1
key
0
0
0
•
o
0
i 17
hCy3
250
500
750
1000
1250
1500
'0.1 0.2 0.4 0.60.81
Table HI. Representative Experimental Data
Figure 8. Particle «ize distribution of stearic acid aerosol for vari-
ous icdimentation depths.
95
m90
U80
undersize
fiSS 3
I30
|20
0 "*
5
i' '' il / ' I/ ' i
[fog] / /o
n,, = 997x105 / /
_ ptrtfcl«/cc of-
- 1 I ~
* 1 *
/ o/n0=2.95x10-
_ °o / particl«fce_
/ /
1 / / :
/ y . 1 , i
_ / ° L mixing condition
/ ^ hot saturated air
- / on. 60 *c . 16 l/min
' / cold saturated air
III , 20*c . 6 l/min
0.6 OB 1 2 46
Dp/2
Ptfur* I. Particle lite diitribution of fog.
of concentration and temperature, respectively. The
agreement, moreover, indicates that Brownian coagula-
tion, which must progresa with the lapse of time or with
the depth of sedimentation, has no effect on the size anal-
yiis below UP particles/cm1.
Figure 9 shows the result of the analysis of fog particles
or water droplets, which have been thought difficult to
analyie without any disturbance as they are in suspension
in air. In these eased the particle number concentrations
were so small that the sice analyses were made by count-
ing the whole number in the volume un.
Figure 10 shows the two kinds of *ize distribution* of
tobacco aerosols ("Cherry" made by the Japan Monopoly
Corp.). One is that obtained for low concentration and the
lnd.Eng.Chem.. Fundam.. Vol. 14. No. 1.1975
No.
of par-
Time tides in
elapsed, sight,
/, sec N(t)
(Run 1;
0
20
40
50
60
70
80
90
100
110
120
(Run
0
15
30
45
55
70
80
90
100
120
140
160
180
(Run
0
2
6
8
10
16
20
30
40
50
stear 'c
47
46
42
32
24
12
7
8
4
1
1
Par-
ticle
diam
by
eq2,
D,,n
acid,
1.31
0.90
0.80
0.73
0.67
0.62
0.58
0.55
0.51
0.48
Cumu-
lative
undersize
F -
N(I)/NW
i
Remarks
h = 1000 /i in Figure 8)
0.98
0.89
0.68
O.S1
0.26
0.15
0.17
0.085
0.021
0.021
* eyepiece
x objective
x 10 x 20
p. = 0.84g/
cm'
2; tobacco, left side in Figure 10)
145
144
124
110
89
68
52
37
29
13
8
6
2
3; iron
31
31
30
27
21
11
11
9
4
4
1.64
1.12
0.92
0.82
0.72
0.66
0.63
0.59
0.53
0.49
0.45
0.42
oxide
1.74
0.98
0.84
0.74
0.56
0.50
0.40
0.34
0.23
0.99
0.86
0.76
0.61
0.47
0.36
0.26
0.20
0.083
0.05R
0.041
0.014
pigment in
1.00
0.97
0.87
0.68
0.36
0.36
0.29
0.13
0.13
x 10 x 20
p •= 0. 75 R/
'cm5'
/i - 1000 n
Figure 11)
x 10 > 20
p, = 0. 52 g/
cm'
h = 1000 /i
(Run 4;fo(5, right side in Figure 8)
*,„,(/)
*',..«»
Remarks
0-0.5
0.5-1.0
1.0-1.5
1.5-2.0
2.0-2.5
2.5-3.0
3. 0-3. 5
3.5-4.0
4.0-4.5
4.5-5.0
\0-5.5
5.5-6.0
111
107
95
84
61
65
39
37
25
9
6
6
9.5
7.4
6.2
5.5
5.0
4.9
4.3
4.0
3.8
3.6
3.4
0.97
0.86
0.76
0.55
0.59
0.35
0.33
0.23
0.081
0.054
0.054
x 10 x 10
p, = 1.0 g/
cm'
A= 2000 M
NM(I) Is the
sum of 7
observations
other is that for particles grown by Brownian coagulation
after sufficient aging time. The wide size distribution of
the latter indicates a typical feature when coagulation oc-
563
-------
ss
90
^
5*80
tJ
70
I60
to ^^
^/A
C ^*
330
$20
cumulati
en o
1
tobacco /
" aerosol 0
: /.uw /:
0' pariielesfce /°
/ J ~
~ o I ~
— 1 / —
0 /
-1 /
1 •/
/e /n0s2.58xl06~
_ 0 ? portlel«t/ee _
0 /
/ . 4° , t , 1 , 1 i 1
0.2 0.4 0.6 0.8 1
DP/2C,!]
Figure 10. Particle liu distribution of tobacco aerosol.
n
. j —rt*xtron mkroscop*
0.1
02
OA 0.6 0.8 1
C^D
Figure II. Particle SIM distribution of powder of iron oside pig-
ment.
curt. Although the sizes of tobacco particlet reported by
other investigators (Sano, et at.. 1964; Keith and Derrick,
1960; Pontendorfer and Schraub. 1972) are smaller than
those in Figure 10, such difference is thought to be caused
by the existence of water, which ii expected to adhere to
the particle surface.
Figure 11 bhowg one of the results of analysis of Tine
powder*. In this analysis iron oxide pigment powder wait
dispersed into air by a mixer-type disperser. It probably
taken a few daya to measure by the Andreasen pipet
method while only a few minutes was required by thia
method. The result of size analysis by an electron micro
scope for the same powder is also plotted in the figure.
They agree fairly well. Some of the representative data
are shown in Table III.
Conclusion
The technique developed in this study was found to be
useful for the particle size analysis of fine powders and
aerosols, including those of fine water droplets. The tech-
nique has the advantage of sedimentation methods in
which particles are observed while they are suspended.
Moreover, the usual disadvantage of sedimentation proce-
dures, that much time is needed for the fall of small par-
ticles, was overcome by air sedimentation with an ex-
tremely shallow sedimentation depth. As a result, the size
analysis of particles, including water droplets, which are
larger than several tenths micron in diameter, was possi-
ble in a short time. The ultramicroscopic detection of the
particle concentration on a number basis and its recording
by a video system served for quick analysis without dis-
turbing the aerosols during observation. The particle
numlier concentration, however, muxi be rather hii;h, a
tact which is both an advantage and a disadvantage of the
method.
Acknowledgment
T. Miyazaki was very helpful in the experimental work.
Nomenclature
Cm • Cunningham's correction factor
{•" • cumulative fraction underside
D, • effective diameter of microscopic sight shown in
Figure I, cm
U,. • particle diameter, M or cm
f(/),,) • particle size distribution function
h. A/I • values shown in Figure 1, cm
MO » particle number observed by microscope at t sec
after the start of sedimentation
A/iut(t) " particle number observed by microscope during
a certain period of observation at t sec after the start of
sedimentation
Nini.i « total particle number in volume yn
no " particle number concentration, particles/cm*
t * time elapsed, sec
Ui " terminal settling velocity, cm/sec
Vh,Um " volumes shown in Figure 1, cm*
Grvek Letters
it * viscosity, g/(cm sec)
pp * density of particle, g/cms
pi » density of fluid, g/cm1
Literature Cited
Osvws. CM.. Pme. Roy Soe. Str A. 200,100 (1949).
Fucns, N. A.. "Th* M«ch»rUc§ ol Actotols." Pwgamon Prut. 19M.
K»,lh. C H . D«rnck, J C.. J. ColloidScl.. !«. 340 0880)
Mukaibo. T . •( •/. SogoSfiiKtntho Ntmpo. JO. No t (1962)
Por>MAdOrl*r. J . Sehftub. A.. SUuA. 32 (101.33 (1972)
Rlch«td»on, J F.. Wooding. E. R . J Photogr. Set. 4, 75 (19M)
Sano. K . Fuiiya. V . Sakata. S . J. Chum. Sac. Jtp.. 74,664 (1964)
Kterivml f»rretHt-u1 April 1,1974
Accrprrd Septembers, 14(74
Prevented «t the Chemicsl KnKineerinii Meeting in Japan.
Ind. Eng. Cham., Fundam.. Vol. 14, No. 1.197S
564
-------
TURBULENT COAGULATION OF AEROSOLS IN A PIPE FLOW
KIKUO OKUYAMA, YASUO KOUSAKA AND TETSUO YOSHIDA
Department of Chemical Engineering/ University
of Osaka Prefecture/ Sakai 591/ Japan
565
-------
TURBULENT COAGULATION OF AEROSOLS IN A PIPE FLOW
KIKUO OKUYAMA, YASUO KOUSAKA AND TETSUO YOSHIDA
Department of Chemical Engineering, University
of Osaka Prefecture, Sakai 591, Japan
Abstract
Turbulent coagulation of aerosol particles was studied
experimentally by observing the time-dependent changes in
particle number concentration of aerosol suspended in various
turbulent pipe flows, using the ultramicroscopic size analysis.
And these experimental results were confirmed with the calcula-
tion results obtained by numerically solving the population
balance equation, which contained the simultaneous effect of
Brownian coagulation or turbulent deposition in addition to
turbulent coagulation. The effects of Brownian coagulation and
turbulent deposition on 'turbulent coagulation were estimated
using the values of two dimensionless parameters K and T ,
and negligible regions of these effects were indicated.
566
-------
Introduction
For high concentration aerosol particles suspended in a
turbulent flow/ turbulent coagulation is essential for chara-
cterizing the behavior of aerosols, but most of previous studies
on turbulent coagulation have been restricted to theoretical
!,4,b)
ones where the coagulation rate is discussed. Though the changes
in properties of aerosol undergoing turbulent coagulation in
a stirred tank were observed and experimental data were compared
with theoretical calculation results in the previous paper,
experimental study is insufficient to check the property of the
theories. The largest cause of the lack in experimental data is
the difficulty in accurate measurement of the change in proper-
ties of high concentrated aerosol.
In this paper, the changes in particle number concentration
of polydisperse aerosols undergoing turbulent coagulation in a
turbulent pipe flow were observed for various conditions using
the ultramicroscopic size analysis previously develpoed by the
q)
authors. And the observed changes of particle number .concentra-
tions with time were compared with those obtained by numerically
solving the basic equation for polydisperse aerosol. In this
case, these calculations were made for some representative
turbulent coagulation rate constants. Moreover two dimensionless
parameters were introduced to estimate the amount of the effects
of Brownian coagulation and deposition by turbulent diffusion
on turbulent coagulation.
1. Theoretical Consideration
567
-------
The population balance equation describing the time-dependent
change in particle size distribution of polydisperse aerosol under-
going turbulent coagulation alone can be written as"J
M t')
'P*o
In the case of turbulent pipe flow, the time t' corresponds to
the residence time which is approximately the ratio of pipe
length to average flow velocity
t' = L / u (2)
KT(r',p') is the turbulent coagulation function for particles of
radii rf and p', and the representative theoretical equations
denoting the collision rate by the turbulent flow have been
<0 •*)
proposed by Saffman et al. and Levich . Generally collisions
between particles in a turbulent fluid are considered to be
caused by two independent and essentially different mechanisms.
In the first coagulation mechanism particles may collide with
each other as the result of different velocities between parti-
cles because of the spatial non-homogenities characteristic of
turbulent flow. For this mechanism, Saffman et al. has proposed
next equation denoting the turbulent coagulation rate,
KT(r',p') = 1.30(r'+p')3/e0/v (3)
The second coagulation mechanism may be caused by a relative
motion of each particle differing from that of turbulent air
because of its inertia will not be the same as equivalent mass
of air. Saffman et al. also obtained the following equation for
the simultaneous collision rate by first and second coagulation
mechanisms including the collision brought by gravity
568
-------
where t(r') is the particle relaxation time, equal to 2r'2p /9y.
P
This equation reduces to next equation assuming that particles
move with air.
KT(r',p') = 1.67(r'+p')3/^/v" (5)
which is a form similar to Eq.(3). A detailed comparison of Eqs.(3)
and (4) with experimental results, which is one of the most
important purposes of this paper, has not been made at all.
Levich's equation, however, was omitted because it was found that.
this equation overestimated the coagulation rate as shown in the
previous paper
In either mechanism, the coagulation rate depends mainly on
particle size and velocity gradient evaluated from the energy
dissipation rate per unit mass of fluid, because Eqs.(3) and (4)
were derived assuming the theory of isotropic turbulence. In the
case of turbulent pipe flow, the average value of the energy
$)
dissipation rate e has been suggested by Laufer as next equa-
tion
= fu3/D (6)
As the distribution of energy dissipation rate through a pipe
results in coagulation rate distribution, a difference exists
between the coagulation rate based on the average energy dissi-
pation rate e and that based on the local energy dissipation
rate and its distribution through a pipe. In this study, however,
the distribution of energy dissipation rate was neglected and the
average value e was used to evaluate the coagulation rate for
a first approximation.
569.
-------
Integrating Eq. (1) with respect 'to r' from 0 to «, the
total particle number concentration n' of polydisperse aerosol
can be given as
where
n'= Cn'(.r',t)
-------
radius r^ and the partical number concentration of monodisperse
aerosol were assumed to be r' and n' . As seen from the graph,
C tends to increase with a , but the increasement of 5 calculated
by using Eq. (4) as KT(r',p') is much greater than that by US Ing
Eq. (3) for even the small value of e . Prom the value of ? at
Brownian coagulation, which is calculated by introducing the
7)
following Brownian coagulation function K (r',p') in replace of
KT(r',p'} in Eq. (10),
Where K0=
the effect of polydispersity is found to be small in comparison
with those of turbulent coagulation.
1.2 Particle size distribution
Figure 2 shows the time-dependent changes in particle size
distributions for three different initial log-normal forms. The
frequency f (In r') was calculated by the next equation
fUn r')- •n'or'.tOY'/Tlo (13)
The dimensionless time t was expressed in terms of the coagula-
tion time, which is the ratio of the actual elapsed to the half-
life time t , which is given by the next equation from Eq. (9)
In the particle size distribution of o 0=1«1 at t=2.39 the second
mode, which corresponds to the coalescence of two particles hav-
ing the initial geometric mean radius, has begun to appear. In
other cases the particle size distributions broaden rapidly with
the dimensionless time t, and its tendency in calculation results
by Eq. (4) is especially large.
571
-------
1.3 Particle number concentration
Figure 3 shows the normalized number concentration, which
is defined as the ratio of the concentration at any time to the
initial time, versus the dimensionless time based on turbulent
coagulation. There exist very large differences between the
calculation results by Eq.(3) and that by Eq.(4) due to the loss
by second turbulent coagulation mechanism. Even a calculation
results of 0 0~1>1 by Eq. (3), which is nearly monodisperse,
particle number concentration comes to decrease faster than that
of monodisperse aerosol with the elapse of time, because particle
size distributions expand with time as shown in Fig.2.
1.4 Effect of Brownian coagulation
The basic equation expressing the time-dependence of the
size distribution of polydisperse aerosol undergoing simul-
taneous Brownian and turbulent coagulations can be written as,
assuming that both coagulation mechanisms are independent each
other.
P, />)- Krd0r?/0j *'('/&P, t)n(fit)
~ ' !s° L -t- 0 j' i/~ '
When an aerosol is monodisperse Having pailicxfl. raAuts foe
' r«')\n.'z
dt'
Substituting Eqs.(3) and (12), Eq. (16) becomes
5)
As described in the previous paper, the importance of turbulent
coagulation relative to Brownian coagulaiton can be estimated by
the following dimensionless parameter K
572
-------
=. | + 5.2 r«jiv
Normalizing Eq. (17) by the following dimensionless variables,
= .
the basic equation comes to next equation.
dn/dt « -
Solving Eq. (20) under the initial condition that n=l at t=0,
72 =
Figure 4 shows the effect of K on the change in the normalized
number concentration as a function of dimensionless time based
on turbulent coagulation. As the value of K increases the curves
move toward the right and tend to converge to that of KD=
which coincides with that of turbulent coagulation alone. The
graph also shows that the effect of Brownian coagulation can
be almost ignored at the value of KD greater than 10.0. Though
the time-dependent changes for polydisperse aerosol were obtained
by solving Eq. (15) numerically assuming the initial particle
size distribution to be log-normal form, their tendencies by KQ
are found to almost agree with those of monodisperse aerosol.
1.5 Effect of particle deposition
In a turbulent pipe flow of aerosol, particles tend to
deposite by turbulent diffusion and gravitational settling.
The basic equation expressing the time-dependent changes in par-
ticle number concentration of aerosol undergoing turbulent
deposition alone can be given as follows.
573
-------
.
3 It is
St* = ifibO"
where St (r') denotes the local dimensionless deposition velo-
city and is given
; - T/ 0*
*
As this St (r') is different by the effect of gravitational sett-
ling in the case of horizontal pipe flow, St (r') is the average
one around the periphery of the pipe. There exist many theoretical
and experimental researches on evaluating the values of St (r').
Combining the Eqs.(1) and (22), the basic equation for coagu-
lation and deposition can be given as follows.
When an aerosol is monodispeise, Eq. (25) reduces to
Substitution of Eq. (19) into this equation gives the following
equation
dn/dt = - nz - T0ri
where
T0 = *s£ ( rj>) u V ^ -2 r
T is a dimensionless parameter which can be evaluated from
initial aerosol properties and intensity of turbulence. As seen
from the definition, this parameter T means the ratio of depo-
sition rate to turbulent coagulation rate, and therefore T
denotes the relative importance of deposition to turbulent
coagulation. Integrating Eq. (27) under the initial condition
574
-------
n=l at t=0, the time-dependent change in particle number concen-
tration can be given as
Figure 5 shows the dependence of particle number concentration
change on TD« With the increase of TD the particle number decreases
faster due to the larger effect of deposition, but the effect of
deposition can be almost ignored at the value of T less than
0.1 as seen from this graph. Though the discussion described
above is for monodisperse aerosol, it will be essentially valid
for polydisperse one.
2. Experimental Apparatus and Method
The apparatus used in this work consists of the following
parts- blower, aerosol generator, aerosol chamber, P.V.C. pipe
and devices for ultramicroscopic particle size analysis as
shown in Fig. 7. Aerosol used was fog of aqueous ammonium chloride
s
solution. Clean air coming from a blower through a glass fibrous
filter and being regulated the orifice flow meter was first
bubbled into the hydrochloric acid solution and subsequently
bubbled into the aqueous ammonia solution, when ammonia chloride
smoke was produced. During the bubbling of the ammonia chloride
smoke at the final vessel containing water, aerosols were changed
into the fog droplets of aqueous amminium chloride to absorb
water vapor. At the same time, the excess gas, chiefly ammonia
gas, is absorbed through the water. The aerosol generation by
this method was examined by Fujitani in detail. Aerosols thus
generated were continuously led into vinyl chloride pipe with
diameter of 13 or 26 nun to make turbulent flow. The length of
575
-------
pipe was changed to be 20, 52 and 100 m, before and after of
which the aerosol was sampled by sampling tap with isokinetic
sampling condition. Aerosol sampled was introduced into the
observation cell installed on the stage of an ultramicroscope
to measure its particle size distribution and particle number
concentration. The measurement method using as ultramicroscope
Q)
was developed previously by the authors.
Experiments were carried out by changing the initial par-
tical number concentrations and by increasing the particle
radius using particle growth in a ageing chamber. Represen-
tative experimental conditions and properties of aerosols are
shown in Table 1.
3. Experimental Results and Discussion
In a comparison of experimental data with theoretical
calculation results, the residence time of aerosol in a pipe
can be calculated by Eq.(2), and e by Eq. (5). Figure 8 shows
a comparison of the change in particle number concentration
between experimental and theoretical curves. Particle number
concentrations obtained experimentally tend to decrease with
the dimensionless time t indicating some scattering. Theoretical
curves denoted by solid lines, broken lines and one point broken
line show the particle number concentration change of aerosols
undergoing turbulent coagulation alone, which were calculated
by solving Eq.(1) for polydisperse aerosol and Eq.(8) for
monodisperse one. Experimental data decrease faster than the
calculated results of monodisperse aerosol because of the poly-
dispersity, and their tendencies almost agree with those calcu-
576
-------
lated using Eq.(3) as the turbulent coagulation function, but
they do not agree with those calculated using Eq.(4) at all.
Some of experimental data are found to be affected by Brownian
coagulation or turbulent deposition as seen from the values of
KD and TQ, but the deposition by Brownian diffusion can be
ignored in these experimental conditions according to Gromely
et al?} . In the calculation of TQ, Yoshioka at alls10equation
was used as the value of St (r'), which is a relatively simple
form and showed the better fitting with many experimental data
in comparison with other equations. Fig.8(a) shows a comparison
of experimental data(the values of K are less than 10.0) with the
calculation results of Eq. (15) to examine the effect of Brownian
coagulation, and Fig.8(b) a comparison of experimental data (the
values of TQ are larger than 0.1) with the calculation results of
Eq.(25) to examine the effect of deposition. Most of experimen-
tal data almost agree with the tendency of the calculated curves,
and the effect of Brownian coagulation or turbulent deposition
is found to increase as K decrease or T increases.
Fig.9 shows a comparison of experimental data, where the
particle radius is large to be about 0.7 p in comparison with
Fig.8, with the calculation results obtained by solving
Eq.(25). Most of experimental particle number concentrations
tend to decrease faster than those of calculation results, that
is, it is found that these differences can be explained by
neither Brownian coagulation and turbulent deposition. These are
considered to be due to the loss of particles by second turbu-
lent coagulation mechanism which comes to appear with increasing
of the turbulent intensity and particle size, but they can not
be explained by Saffman at alfs Eq. (4).
577
-------
Prom experimental conditions of Figs. (7) and (9), the sec-
ond coagulation mechanism by turbulent flow may be essential when
the values of r'^/T^ is larger than 2.5 x 10"10. These kinds of
experimental data as shown in Figs. (7) and (9) have not been
reported because of the difficulty in accurate measurement of
Changing number'concentration of particles with time.
Conclusion
Turbulent coagulation of aerosol particles was studied
experimentally and theoretically, and the following results
were obtained.
1} At r'0/el less than 2.5 x 10" the time-dependent change in
particle number concentration by turbulent coagulation can
be evaluated by the solution of population balance equation
using Eq.(3) as coagulation function, that is, particles
coagulate by first coagulation mechanism alone in this case.
But at larger than this value the second coagulation mechanism
was found experimentally to be appeared.
2) The effect of Brownian coagulation relative to turbulent
coagulation can be ignored when the dimensionless parameter
K is larger than 10.0.
3} The effect of deposition on turbulent coagulation can be
predicted by the value of the dimensionless parameter TQ.
The deposition can be almost ignored when the value of T0is
less than 0.1.
578
-------
Literature Cited
1) Seal, S. K.: J. Aerosol Sci., 3, 113(1972)
2) Fujitani, Y.: Bull. Chem. Soc. Japan, 30, 683(1957)
3) Gormley, P. G. and M. Kennedy: Proc. Roy. Irish. Acad.,
52-A, 163(1949)
4) Levich, V. G.:"Physicochemical Hydrodynamics", Prentice-
Hall (1962)
5) Okuyaraa, K., Y. Kousaka, Y. Kida and T. Yoshida:
J. Chem. Eng. Japan, 10, 142(1977)
6) Saffman, P. G. and J. S. Turner: J. Fluid Mech., 1, 16(1956)
7) Smoluchowski, M. von Z: Phys. Chem., 92, 129(1917)
8) Uhl, V. W. and J. B. Gray:"Mixing-theory and practice",
Academic Press, 75(1966)
9) Yoshida, T., Y. Kousaka and K. Okuyamailnd. Eng. Chem. Fundam.,
14, 47(1975)
10) Yoshioka, N., C. Kanaoka and H. Emi: Kagaku Kogaku,
36, 1010(1972)
11) Zebel, G.:Kolloid-Z., 156, 102(1958)
579
-------
Table 1 Experimental Conditions and Properties of Aerosols
key Re
2 ° 3
[-] [cm /sec )
n'
0 3
[particles/cm ]
r'
go
Eul
a
go
[-1
K
D
[-1
St* (r'
av v gO
[-1
) T«
' D
[-1
r':
g\
[cm
l/T~
I ° 3/2,
'/sec ' 1
Pipe diameter D = 13 mm
.-11
9
O
4,
o
O
*.
•
®
e
o
5000
m
n
8000
n
n
7400
10000
13000
22400
1.35
4.93
3.98
9.11
1.98
8.38
x
.
n
X
M
H
X
X
X
X
10°
IO6
IO6
IO6
IO7
IO7
3.66-6.06 x 10 0.27-0.32 1.20-1.31
2.61-3.07 x IO6 0.34-0.43 1.24-1.60
3.98-4.64 x IO7 0.35-0.43 1.21-1.30
5.69-6.63 x IO7 0.25-0.27 1.22-1.31
6.71 x IO6 0.36-0.38 1.25-1.32
2.54-3.98 x IO7 0.32-0.41 1.30-1.40
1.01-2.37 x IO7 0.37-0.41 1.21-1.38
1.10-3.17 x IO7 0.30-0.48 1.23-1.27
1.85-2.09 x IO6 0.31-0.34 1.32-1.48
3.62 x IO7 0.30 1.18
2.
3.
ft
2.
5.
n
5.
7.
6.
9.
11
56
33
14
39
64
60
74
8.80
1.49
4.42
8.93
1.06
8.23
5.30
1.35
x
x
M
X
X
n
X
X
X
X
10
io-5
io-6
10"6
io-5
10"6
10"6
io-5
0.
0.
0.
0.
0.
0.
0.
0.
0.
0.
0502
667
0450
0249
160
0329
0669
0459
352
0502
2
4
4
3
1
0
1
0
1
.29-3.
.57-9.
.98-9.
.47-4.
.04-1.
.73-1.
.01-1.
.82-3.
.29-1.
2.
81
24
24
37
22
53
37
34
70
47
x
x
X
X
X
X
X
X
X
X
10 "
io-11
io-11
io-11
io-10
io-10
io-10
io-10
io-10
io-10
Ul
00
o
Pipe diameter D = 26 mm
B 5000 8.47 x IO4 1.22-4.45 x IO7 0.46-0.48 1.25-1.38 2.16 4.33 X 10~5 0.113 2.83-3.22 x 10'11
B " 0.46-4.71 x IO6 0.30-0.44 1.25-1.38 1.54 2.68 x 10~5 1.565 0.79-2.48 x 10*11
• 10000 5.69 x IO5 1.49-5.18 x IO7 0.32-0.45 1.25-1.38 2.66 1.62 x 10~5 0.045 2.47-6.87 x IO"11
• " 0.77-3.74 x IO6 0.35-0.40 1.25-1.38 2.54 1.54 x 10"5 0.665 3.23-4.83 x 10'11
A » 1.18-4.16 x IO6 0.55-0.80 1.36-1.58 10.6 4.94 x 10~5 0.321 1.26-3.86 x 10~10
* 15000 1.74 x IO6 4.60-9.94 x IO5 0.63-0.83 1.37-1.67 21.8 4.03 x 10~5 0.630 3.30-7.54 x 10~
f. -10
A " 1.31-2.19 x IO6 0.55-0.90 1.35-1.51 21.8 " 0.263 2.19-9.62 x 10
-------
Ul
00
Brown ion
coagulation^ -
.0 1.2 1.4 1.6 1.8 2.0
geometric standard deviation (5go
Fig. 1 Polydispersion factor
-------
in
CO
NJ
0.4 0.60.81
2 4 0.4 0.60.81
r'/rgo C-D
Fig. 2 Time-dependent changes in size distribution of particles
undergoing turbulent coagulation
-------
00
I i_L
A 6 8in-1
Fig. 3 Time-dependent changes in number concentration of particles
undergoing turbulent coagulation
-------
i ii
00
4.
dispersion Qgo
4 6 610'
Fig. 4 Time-dependent changes in number concentration of particles
undergoing turbulent and Brownian coagulations
-------
in
oo
in
1.0
0.8
T0.6
o
C
0.4
0.2
0
"^ ** Nk "*»
_
\
V \
N \
-\ 50\ 1C
\ ^
100 \
N X
x x
•X. V
~r n r
x
x x x .
Nx X- -\x
v X v «v
\ \ ->
N N x \s .
. ' \
x\ v-
^ \ \
)\ 5\ 1\0.
\
v\ \
N X
x X
monodisperse
aerosol
calculated by Eq.(29)-
xTD = 0-0.01
^ X
X X
x x
X X
X X
X V v ^ x v
N x x v
x x x ^ *s
•^
6 e,0-i
t = 5.;
6 8100
t' [-)
468
Fig. 5 Time-dependent changes in number concentration of particles
undergoing turbulent coagulation and deposition
-------
V
blower
heat
exchanger
Ol
CO
a\
C
1
k /
oo
00
HCl NH3
orifice
flow meter
ageing
chamber
sampling tap
inlet side
j>
P.V.C. pipe (dia !3mmf 26mm)
filter observation
cell
r
=55. ^-to duct
../sampling tap
outlet side
vacuum
pump
ultramicroscope
T
VTR & moniter]
orifice
flow meter
Fig. 6 Experimental apparatus
-------
in
CO
1.0
0.8
0.6
o
C
0.2
0
D n ' ' ' ' '
/
o
I 1,1,1,
_ i i
A ' fltff) 1 ' '
H c ^^^c^. ^
H 0%_ "*^"_/^"* ^
^\. ^ c
1 cm2/sec3 \
i , j
calculated
curve KjCrif ) dispersion.^
c-— /o\
r
Eq.W) *
mono.
>oly. 6go=1.3
2 4 6 810-2 2
t=5
- \
\
1 ,1,1,
4 6 81
2rrv\V£n/V r
i • i • i • i •
K
>•<:-,
)Q*^:\
»^BV\XN
» Vt.\
^7^- , I , I , I ,
-
-
\
X
s
\:
\
0-1 2 4 6 8lfJ0 2
Vnf C-3
Fig. 7 Comparison of experimental particle number concentration
changes with calculated ones
-------
oo
CO
rl.O
1.0
6 8 10" 2
I ' I ' I'l
calculated by Eq.(25)
calculated by Eq.(15)
3.56 5.39
6 8l 2
6 8-2
Fig. 8 Effect of Brownian coagulation or turbulent deposition
on decrease in particle number concentration
-------
Ul
CO
1.0
0.8
70.6
. o
C
0.2
I ' l ' I '
calculated by Eq.(25)
1
1 .L
I
I . I
6 eio"1
2
C-l
6 6
Fig. 9 Comparison of experimental particle number concentration
changes with calculated ones
-------
EFFECTS OF BROWNIAN COAGULATION AND
BROWNIAN DIFFUSION ON FINE PARTICLE
SIZE ANALYSIS BY SEDIMENTATION
METHOD
KIKUO OKUYAMA, YASUO KOUSAKA,
TOSHIHIRO MIYA2AKI AND TETSUO YOSHIDA
Department of Chemical Engineering, University of Osaka
Prefecture, Sakai 591
Conventional particle size analysis by sedimentation method does not give true size distributions
but only apparent ones, when Brownian coagulation and Brownian diffusion exist. The difference
of true and apparent size distributions was theoretically evaluated by numerically solving the
population balance equation under various conditions. Then some of the theoretical results were
verified experimentally for particles having various sizes and number concentrations which were
obtained under sedimentation in air, in water and in centrifugal field by means of ultramlcroscopic
size analysis. In addition to these analyses, two parameters consisting of given measuring condi-
tions, which were proposed herein, were found to be useful to predict whether or not coagulation
and diffusion effects exist in actually observed size distributions.
Introduction
Sedimentation method for determining particle size
distribution has been widely used because of its con-
venience and its accuracy in size analysis. In this
method, the relation of Stokes-Cunningham equation
is usually applied. However, when the particles once
sufficiently dispersed coagulate each other during
sedimentation, the Stokes-Cunningham equation will
not be satisfactory for size analysis by this method.
Such coagulation will occur in water by using inap-
propriate dispersion agents and also will be unavoid-
able in air having particles in high concentration.
When the particles are small, such as in submicron
ranges, on the other hand, diffusion of particles to
walls and a free surface of a sedimentation cell by
Brownian motion occurs and the Stokes-Cunningham
equation becomes also meaningless. Some studies
on the later cases and none on the former cases have
been made, but they are still insufficient to evaluate
the quantitative effects of those on sedimentation
method'-'-".
The effects of diffusion and coagulation on sedi-
mentation method were theoretically evaluated under
various sizes, concentrations and other various meas-
uring conditions. The limit of the conditions where
no influences of these effects exist was discussed and the
conditions to avoid these influences by means of cen-
Receivvd March 12. 1976. Correspondence concerning this article should
be addressed to K. Okuyanu. T. Miyazaki it with Matsushita Denko. Co.
Ltd.. Kadoma 571.
trifugal force were also discussed. Some of the
theoretical analyses were then compared with experi-
mental results which were obtained by the ultramicro-
scopic technique of particle size analysis".
1. Theoretical Consideration
In the sedimentation method for particle size analy-
sis, cumulative undersize F of particle radius r' is
usually obtained as follows, when the concentration
of particles decreases only by sedimentation.
where JVsis the concentration at the pointy' below the
surface of suspension at every lapse of time, Nt the
initial particle concentration, and r\ is given by Stokes-
Cunningham equation as
for gravitational settling (2)
When the settling depth is sufficiently small compared
with the radius of rotation, r( will be given as follows
for centrifugal settling (3)
When the concentration /Vs decreases to N by particle
deposition due to diffusion to the wall of the sedimenta-
tion cell or by particle coagulation as shown in Fig.
1, such a system can not be solely described by Stokes-
Cunningham relation. Accordingly, the value N/ff,
observed does not give the true cumulative undersize F,
but only the apparent one. Though particles diffuse
\£/ JOURNAL OF CHEMICAL ENGINEERING OFJAPAN
(Reprinted with permission)
590
-------
to both vertical and horizontal walls of the sedimenta-
tion cell, horizontal particle diffusion was ignored
in order to simplify the analysis. The change in parti-
cle concentration in the system where Brownian dif-
fusion, coagulation and gravitational settling are simul-
taneously taken into account should be evaluated by
solving the following equation of population balance
in number concentration'basis".
l_ C.(r{) 3Mr,,j) _ Cm(r{) rJ3«(r(, 0
"
+CG
fl-rimia
•CG Jp" frfa. P<)n(r<, t)n(p,, 0
/=/min ... i max (4)
All the quantities are normalized as follows
t=u,(WzlH, y-y'W, WJr'*
(5)
(6)
DG =
The centrifugal effect z disappears in the case of gravi-
tational sedimentation only in Eqs. (5) and (6). The
dimensionless parameters CG and DC, which are
determined from initial particle properties and physical
conditions, are convenient for predicting the overall
influences of coagulation and diffusion. Eq. (4) was
solved numerically in the previous paper" for various
values of CG and DC assuming initial particle size
distribution to be log-normal form. Vertical particle
diffusion was considered in Eq. (4) and so the calcula-
tion results can be applied directly to evaluate the
effects of coagulation and diffusion on sedimentation
method, when the vertical boundaries are small in area
compared with the horizontal walls in sedimentation
cell. The effect of horizontal particle diffusion will be
discussed in the later section.
Normalizing Eq. (2) or (3) by Eq. (5), the dimen-
sionless equivalent spherical radius is given by
As Cunningham's correction is not significant in most
liquid sedimentation, then Eq. (7) reduces to
The following dimensionless particle number -con-
centration n, which is obtained by solving Eq. (4)
under a certain sedimentation length y and a certain
time t, is now defined
'(mil
2
,, 0
(9)
Flg. 1 Sedimentation wit
Fig. 2 Time-dependent change in number concentration of
particles undergoing sedimentation and coagulation
Thus defined, n corresponds to N/Na which is actually
observed at y' and /' in sedimentation method under
the influence of coagulation or diffusion, and n never
provides true cumulative undersize defined by Eq. (I)
in this case. The dimensionless equivalent spherical
radius r, determined by Eq. (7) or (8) has a reasonable
physical meaning in gravitational settling only, it is,
however, applied conventionally to the case where
the influence of diffusion or coagulation exists in this
study. Then the relation of r, and NJN, or n, both
determined by observation and by Eq. (4), gives a
true particle size distribution if no influence of dif-
fusion or coagulation exists (CG=Z)G=0). However,
it does not give a true distribution, or it gives only the
apparent distribution, if one of the above influences
exists (CG>0 or £>G>0). The difference of the thus
obtained apparent distributions from a true distri-
bution will be discussed in the following sections.
1.1 Effect of Brownian coagulation
Particle size distribution actually changes as coa-
gulation proceeds in sedimentation ceil, and so it is
necessary to set up a certain distribution to be standard.
In this respect the size distribution at r=0, that is,
initial size distribution, was regarded as the standard
and true distribution. Figure 2, quoted from previous
paper", indicates the time-dependent change in parti-
cle number concentration at every depth y when the
influence of coagulation exists. Two parameters in
the figure were introduced to describe the local effect
VOL 10 NO. I }»7
591
-------
06061
Fig. 3 Apparent particle size distributions due to (he
effect of coagulation
Fig. 4 Variation of 50°,-radius in apparent particle size
distributions by CG y
of coagulation
(10)
Replotting the values of v'v// and n from Fig. 2 to
Fig. 3, apparent particle size distributions defined in the
former section are obtainable. It can be shown that
apparent particle size distributions tend to shift to-
wards larger radii with the values of CG- v. At the
same time the distributions depart from the log-
normal form. The difference between the apparent
distribution and initial (true) one is caused by the de-
crease in particle number by coagulation and at the
same time by the increase in settling velocities of grown
particles. Figure -4 shows the variation in 50%-radius
of apparent size distributions by the dimensionless
parameter CG-y. It can be seen that the curve of
<;,.= 1.5 is slightly larger than that of a,0= 1.2 because
of the effect of polydispersity, and also that, at the
valuesof C6'- v less than 0.04, the effect of coagulation
can be ignored.
02 04 06 08
t-J
02 04 Q6 OB 10
n c-j
(a) (b)
Fig. 5 Time-dependent change in number concentration of
particles undergoing sedimentation and diffusion
1. 2 Effect of Brownian diffusion
True particle size distribution in this case was
defined as that where the influence of diffusion would
be perfectly removed in some way. The effect of
diffusion is very complicated in comparison with that
of coagulation, because of its larger dependence on
particle radius and its irregular dependence on sedi-
mentation depth v. To examine this effect in simple
form, particles are now considered to be monodi>>-
perse. The basic equation in this case, expressing the
timedependent change in concentration of particles
undergoing diffusion and sedimentation, becomes
J"^DG,'''n,- (ID
at ay* ay
where
DGt=D(ri)IHu,(ri)z
Equation(11) was solved analytically by C. N. Davies"
under two kinds of boundary conditions: one is that
both top and bottom walls are absorbing ones, and
another is that the top wall is a free surface where no
particles cross this wall. As the convergence in solu-
tion is very slow for small values of DGa which are
characteristic of larger particles, Eq. (II) was solved
numerically by the same method as the previous
paper" assuming both walls to be absorbing. Figure 5
shows the calculation results of the time-dependent
change in particle number concentration. Particle
number concentration must change with remaining
sharp boundaries, which can be found for such a small
value of DC, as 0.00129 in the figure. The effect of
diffusion, however, is found to become larger with the
values of DC, and the particle number concentration
is found to tend to change with the smooth boundary
JOUtNAL OF CHEMICAL ENGINtfUING OF JAPAN
592
-------
in the figures. The decrease in particle number near
the bottom wall appears at £X70 larger than about 0.07.
Some of these calculation results agreed with Davies'
solutions. It is interesting to find that the solutions
solved under the boundary conditions of free surface
of the upper did not differ significantly from those solv-
ed under absorbing upperwallinthecase of £>G0<0.01.
Figure 6 shows the apparent particle size distribution
obtained from Fig. 5, which is shown by the same
explanation as those in Fig. 3. It is found that in the
case of Z)(?o=0.00129, the results by calculation, as-
suming two absorbing walls almost coincide with the
true monodisperse distribution irrespective of y. The
deviations from the true monodisperse distribution
become larger with increase of DG0, which means that
the plot of actually observed results by means of con-
ventional sedimentation analysis under the effect of
diffusion considerably differs from that of the true
distribution. The apparent size distributions in this
case give those as if they were polydisperse in spite of
the actually monodisperse particles. Figure 7 shows the
region which is determined by putting -<0.95 (12)
When the sedimentation cell is not a shallow one, the
effect of particle diffusion to vertical boundaries must
be taken into account. This effect of vertical walls,
however, can be avoided when the following relation
is satisfied (refer to Appendix).
B>v/50Z>G0W (13)
The discussion above is for monodisperse particles,
but it will be essentially valid for polydisperse particles.
1.3 Dependence of particle radius on the above two
effects
As is found by the definition of CG and DC, the ef-
fects of coagulation and diffusion depend upon particle
properties, fluid properties and dimensions of a sedi-
mentation cell. Among these the remaining values
except particle radius are usually known or can be
suitably chosen in measurement. In this section, the
methods to predict the effects of coagulation and-dif-
fusion will be discussed by two parameters Mr and M„
which do not consist of unknown value of particle
radius.
The parameter CG-y, which appeared in a previous
section, can be rewritten as follows, separating particle
radius:
where
A/r=9 fty'K,n'J2 x \Q-\p,-p)gz
The parameter Mc consists of known valuables.
Figure 8 shows the dependence of 50%-radius (r;)50 of
VOL. 10 NO. i 1977
Fig. 6 Apparent particle size distributions due to the effect
of diffusion
J, 6 810
DG0 C-3
Fig. 7 Region where the effect of diffusion is ignored
Fig. 8 Dependence of (r,')n on M(- and /•„„'
apparent particle size distributions on M, and initial
or true r('0l which was obtained from Fig. 4 by deter-
mining the values of r,'0 and M,-. It can be found that
50%-radius (r',)io agrees with r',, when A/, is less than
0.001 in the case of r;0>0.1 and <7,0 under the effect of coagulation.
In a similar way, a parameter M'„ for diffusion is
defined as follows
593
-------
t. 6
Fig. 9 Dependence of ri an MR and >
Kg. 10 Experimental apparatus
(15)
where
Mn --~ 3*774 x 10- }ir.(!>f
Substituting Mn into Eq. (12), the minimum particle
radius r'0 where the effect of diffusion can be ignored
becomes as follows
In r;>0.333 In M„ -0.868 >•-7.209 (16)
Figure 9 shows the dependence of r'a in Eq. (16) on the
A//,-;- coordinate which are both determined by
measuring conditions. The left-lower regions of each
equi-radius line in the figure indicate that the effect
of diffusion of particles having radius r'^ denoted can
be ignored. For example, the minimum particle
radius r» where the effect of diffusion can be negli-
gible is about 0.05 /A, when A//(=10"' and ^=0.5.
If the observation depth y is shallowed to 0.1, rj
moves to 0.07 ft. Since true particle radius rj, or r'^ is
usually unknown in actual sedimentation tests, it is
unavoidable to use the observed value (r,')so instead of
ri or rj,, for the check whether or not influence of dif-
fusion exists in observed data. If the radius obtained
by Fig. 9 using the known values of M,, and vis smaller
than observed value (r,')50, the effect of diffusion may
be negligible, that is, the observed radius (/•,')«» is
true under the condition of no effect of coagulation or
under enough small value of Mr. If the radius ob-
tained by Fig. 9 is the same order of (/•,'),„ or larger
Fig. 11 Particle size distributions for stearic
acid particles
than (f()M, it is necessary to observe again varying the
measuring conditions to decrease the value of Mn.
The effects of coagulation and diffusion may be mini-
mized under centrifugal sedimentation because the
values of A/c and A/,, become smaller by the centrifugal
effect z.
2. Experimental Apparatus and Method
The ultramicroscopic technique for size analysis
was used to examine the above analyses experimen-
tally, w hich can give accurate particle number concentra-
tion at any given depths. The sedimentation cell for
air sedimentation was the same as those which appear-
ed in the previous paper'-:. The cell for sedimenta-
tion in water is shown in Fig. 10 (a). The cell has a
shallow sedimentation length compared with the
distance between vertical surrounded walls so that Eq.
(13) is satisfied in most cases. The particles used in
this study were two kinds of aerosols and four kinds
of fine powders. The aerosol particles were stearic
acid particles generated by a La Mer-Sinclair type
generator and tobacco smoke generated by a simple
smoking apparatus by which the number concentra-
tion of particles was controlled from 10" to 10" parti-
cles cc. The carbon black and three kinds of iron
oxide particles were provided for the sedimentation
test in water, The carbon black particles were quite
spherical, the iron oxide particles A and B were cubic
with rounded corners, and the iron oxide particles
C were needlelike having a length-iodiameter ratio
of six. A detailed description of the measuring pro-
cedure by gravitational sedimentation for aerosol
particles appeared in the previous paper*'. The
measurement procedure for fine powders in water
sedimentation is as follows. A given amount of
powder was first dispersed uniformly into pure water
JOUINAL OF CHIMICAL (NOINIIRING OP JAPAN
594
-------
Fig. 12 Particle size distributions for tobacco smoke
without any dispersing agents, but with mechanical
stirring to obtain suspension. Then the suspension
was poured gently into the sedimentation cell shown
in Fig. 10 (a) with a syringe, and it was covered with a
thin glass. Putting the sedimentation cell on the
stage of an ultramicroscope as shown in Fig. 10 (c),
particle number at a certain depth in the cell was
observed at every lapse of time by using VTR. The
initial particle number concentration, 7V0(=nJ), was
determined by knowing the initial particle number of
images and the observation volume which corresponds
to the focussed volume of the ultramicroscope pre-
liminarily determined". For centrifugal sedimenta-
tion, on the other hand, the cell was set to a centrifuge
to obtain a given value of centrifugal effect z as shown
in Fig. 10 (b). Experimental conditions and properties
of particles are shown in Table 1.
3. Experimental Results and Discussion
Figures 11 and 12 show the comparisons of apparent
particle size distribution data in air sedimentation
under the effect of coagulation obtained experimentally
with theoretical results. No difference between them
is found irrespective of sedimentation depth y in Fig.
11. The size distribution of highly concentrated
aerosol shown in Fig. 12, on the other hand, seems to
shift to a larger radius. The minimum particle radius
ri where the effect of diffusion can be ignored is ob-
tained to be about 0.2 n from Fig. 9 in those case of
Figs. 11 (A/D=4.64x 10'') and 12 (A/D=5.05x lO'").
This value is sufficiently smaller than those experi-
mentally obtained (/,%, which suggests no effect of
diffusion. The value of Mc of highly concentrated
aerosol in Fig. 12 is found from Fig. 8 to be large
enough to be effected by coagulation. The tendency
of the shift in Fig. 12 is well described by calculation
results of Eq. (4) taking into account the coagulation
effect.
H
390- 6yH«l'on
E microwopr
$70
IN
£ 50-
40
I 30
920
I '0
/<
O wnlrrtu
a •:groviiationol
fiflfl
01
O 0.2* 7.97110* 9.37*'0
239>l6
CO-y 00
JSOilO"4
Fig. 13 Particle size distributions for Iron oxide A
99
5:95(1
k»V
0
0
•
y
nnn
016
006
N,.n«
178. I07
??5>!0'
MC
666HO"'
137. i(T*
B'O.iO"1
MO
I36.IIJ*
J«.IO">
COy
I96«IO"'
3.9««icr1
026
00
J"«IO"*
211,10 '
Fig. 14 Particle size distributions for iron oxide B
Table 1 Experimental conditions
Gravitational sedimentation in air
Aerosols; stearic acid particles, tobacio smoke
^(,=0.3^-0.5^, «r,o= 1.2-1.4
Density of particles; />,=0.85 g/cm» (stearic acid particles)
= 0.78 g/cm' (tobacco smoke)
Concentration of particles; ni= I0'-10' particles/cc
Gravitational and centrifugal sedimentation in water
Powders; iron oxide A, B and C
carbon black
rja=0.09 /i-0.4 ft
ff,o = 1.3-l,6
Density of powders; p, = 5.2 g/cm1 (iron oxide A and B
-4.9 g/cm1 (iron oxide C)
= 1.85 g/cm* (carbon black)
Concentration of powders; «i = IO«-10t particles/cc
Centrifugal effect; 2 = 10-67
/-100 A--1500 ,u (>-=0.08-0.6)
C6=0.07-2.2. £>G=lO-'-10-'
VOL 10 NO. 1 1977
595
-------
-"I
006
«9M<5'
CO,
Fig. 15 Particle size distributions for carbon black
02 04 06
till
kfy
0
0
•
Mud
c»ntnfuoal1i»ldtr.6S3)
gravitational fi»id
y
008
008
N0. nj
tiOi >0<
467* 10«
960« 107
Mc
560.IO"6
1 9K10"'
399.10"'
MO 1 COy
235.10"?1U5.IO""J
ue.io-'i"0"0''!
[ 103
DO
91.0. 10"'
5.84. 10~!
Fig. 16 Particle size distributions for iron oxide C
Figures 13 to 16 show the comparisons of apparent
size distribution data in gravitational and centrifugal
field in water sedimentation obtained experimentally
with calculated results. The effects of coagulation and
diffusionarefoundtobenegligibleinthecase of Fig. 13.
This fact is also readily expected by the values of
Mc, Mr,, y and (r(')j0. The experimental results ob-
tained in gravitational Meld in Figs. 14 and IS are
found to shift towards a larger radius in comparison
with those in centrifugal field. This shift in Fig. 14
is mainly due to the effect of coagulation because of
the larger value of M, • The shift in Fig. 15, on the
other hand, is caused by larger values of MD and small
value of (r,')«o. The minimum radius r', in those cases
is predicted to be above 0.28 n from Fig. 9, while the
apparent radius (r',)M observed is about 0.15 in. Then
it is obvious that the effect of diffusion is thought to
be significant in gravitational sedimentation in Fig. 15.
The calculation results from Eq. (4) taking account of
diffusion effect are found to describe the experimental
results fairly well in the figure.
Some examples of apparent size distributions where
both effects of coagulation and diffusion are significant
under gravitational sedimentation are shown in Fig.
16. The effects of coagulation and diffusion seem to
be negligible by applying centrifugal sedimentation
in those cases, which can be easily understood by
checking Mc, MD, y and (r\'),0. As to the experimental
results obtained by gravitational sedimentation in
Fig. 16 (a), the effect of diffusion obviously exists be-
cause the value of r0' determined by Fig. 9 is 0.17 ^,
while (r;)80=0.12 p. The effect of coagulation in this
case is not significant, the amount of which is shown
by (Sed.-fCoagu.)-curve in the figure. Apparent
particle size distribution in Fig. 16 (b) is influenced
not only by diffusion but by coagulation, because
the particle number concentration is higher than that
of Fig. 16 (a). However, it is difficult to predict the
effect of diffusion under the existence of significant
coagulation such as that of Fig. 16 (b). It is necessary
in such cases to eliminate first the effect of coagula-
tion and then the effect of diffusion must be checked.
Both experimental results obtained by gravitational
sedimentation in Figs. 16 (a) and (b) are found to
agree well with those obtained by directly solving Eq.
(4) under the same condition of the experiment.
The theoretical consideration on coagulation in
this paper is not strictly applicable to solid and non-
spherical particles, because the calculations are based
on the assumption that particles are spherical and par-
ticles collide with each other to form a new spherical
particle whose mass may be the same as the combined
mass of the two smaller particles. It is suggested,
however, that the assumption seems to be fairly ef-
fective even for solid and nonspherical particles, jud-
ging from the fair agreement of theory with experiment
so far as the experimental conditions of this study were
concerned.
Conclusion
The effects of Brownian coagulation and diffusion
on particle size analysis by sedimentation method,
which are of major problem for determining the size of
sub-micron particles, were studied, and the following
results were obtained. I) The change in number
concentration of particles was numerically solved,
when particles exist between two horizontal walls and
are undergoing gravitational sedimentation accom-
panying Brownian coagulation and diffusion. The
calculated results were figured to show the difference of
true size distribution from apparent size distributions
which were numerically obtained by conventional
sedimentation analysis under the effects of coagulation
JOURNAL OP CHEMICAL ENGINEERING OF JAPAN
596
-------
and diffusion. 2) Two parameters, M c for estimating
coagulation effect and MD for diffusion effect, were
proposed to predict whether the influence of coagula-
tion or diffusion exists or not in actually observed size
distribution, that is, whether the size distribution
obtained is true one or apparent one. These param-
eters are useful as a criterion to determine the meas-
uring conditions to avoid the influences of coagula-
tion and diffusion. 3) The above theoretical results
were experimentally examined under various con-
ditions, such as air sedimentation, water sedimentation,
centrifugal sedimentation and various particles, by
means of ultramicroscopic size analysis, and the
theoretical consideration presented in this paper was
found to be valid for prediction and prevention of the
effects of Brownian coagulation and diffusion in a
sedimentation size analysis.
Appendix
The time-dependent change in number concentration of parti-
cles existing in the space surrounded by two vertical walls and
also undergoing Brownian diffusion was solved by N. A. Fuchs"
as follows
(A-l)
(A-2)
The maximum time when the decrease in particle number by
diffusion is almost negligible is given by Eq. (A-2) as B'/SQDM).
Normalized value of this time by Eq. (5) comes to B'/SOH'DG,.
Most particles have been already settling down below //at <«
1.0 in the horizontal walls, consequently the effect of particle
diffusion to vertical walls can be relatively ignored when
The change at the middle of B is given as follows
Nomenclature
B
CG
= width of the sedimentation cell [cm]
= dimensionless parameter defined in Eq. (6)
CJrO
DC
DC,
D(ri)
F
f(r")
g
H
K,
*M, Pi)
*(r,,/»,)
- Cunningham's correction factor of ri
= dimensionless parameter defined in Eq.
= dimensionless parameter defined in Eq.
( - 3c TI4x(pf - p)gHri,'z)
«• diffusion coefficient of ri
(=*CJirl)xT/6iiitri)
= cumulative undersize
~ particle size distribution function
= acceleration of gravity
= height of the sedimentation cell
= coefficient in Eq. (5) (=2*773 n)
— coagulation function for two particles
of sizes ri and p>
— dimensionless coagulation function
[-J
(6)
(II)
[-1
[cm'/sec]
H
[-]
[cm/sec1]
[cm]
[cm'/sec]
(cm'/sec)
Mc = parameter for coagulation defined in Eq. (14)
( = 9 nK*iy'H x \0-t(pf-p)g2) [cm!]
MD = parameter for diffusion defined in Eq. (IS)
•< = 3«774xlO-l'm>,-/»)fWz) [cm']
N.Nt.N, = measured particle concentration [particles/cc]
n'(ri,i'),n(rt.t) — number concentration and dimen-
sionless number concentration of
particles ( = n'(r't ,»')/«,) [particles/cc] [—]
n = dimensionless total number concentra-
tion of particles ( = En(r,.t)) [— ]
ni = total number concentration of particles at
time zero [particles/cc]
r' » particle radius for continuous spectrum [cm]
fir, = particle radius and dimensionless
particle radius (=ri/r;0) [/'][cm)[— ]
ri = particle radius of monodisperse
particles [/<] [cm]
(r<)« = 50%-radius in apparent particle size
distribution [;<]
r;0 = geometric mean radius at time zero [/*] [cm]
T = absolute temperature l°K]
»',/ = time and dimensionless time
[sec] [-]
[cm/sec]
y'.y
terminal settling velocity of ri
Pi. Pi
P.P,
= horizontal distance from one wall of the
sedimentation cell [cm]
= vertical and dimensionless vertical
distance from top of the sedimentation
cell (=//«) [cm] [—]
= centrifugal effect [—]
= geometric standard deviation of apparent
particle size distribution [—1
= geometric standard deviation at time zero [—]
= Boltzman's constant (-1.38X10-14) [erg/°K]
= viscosity of fluid [g/cnvsec]
= particle radius and dimensionless
particle radius (- p't/r^t)
•= fluid and particle density [g/cm*]
i = refers to the number of particle size
max = maximum
min = minimum
Literature Cited
1) Davies, C. N.: Proc. Roy. Soc., A200. 100 (1949).
2) Fuchs, N. A.: "The Mechanics of Aerosols", Pergamon
Press (1964).
3) Irani, R. R. and C. F. Callis: "Particle Size; Measurement,
Interpretation and Application", Willey (1963).
4) Jelinek, Z.: "Particle Size Analysis". Willey (1974).
5) Moore, D. W. and C. Orr, Jr. : Powder Techno!., 8, 1 3
(1973).
6) Yoshida,T., Y.Kousaka and K. Okuyama: lad. Eng. Chem.,
Fundam., 14, 47 (1975).
7) Yoshida, T., Y. Kousaka, K. Okuyama and S. Nishio: J.
Chem: Eng. Japan, 8, 137 (1975).
(Presented at Hokkaido Meeting of The Soc. of Chem.
Engrs., Japan at Muroran, July 24, 1975 and 9th Autumn
Meeting of The Soc. of Chem. Engrs., Japan at Fukuoka, Oct.
17. 1975.)
I-}
VOl. 10 NO. 1 t»77
597
-------
CONSTANT PRESSURE FILTRATION OF POWER-
LAW NON-NEWTONIAN FLUIDS
MOMPH SHIRATO, TSUTOMU ARAGAKI, Em IRITANI,
MASAAKI WAKIMOTO, SATOSHI FUJIYOSHI,
AND SH&I NANDA
Department of Chemical Engineering, Nagoya University,
Nagoya 464
A filtration theory for the power-law non-Newtonian fluids b developed on the basis of the ex-
Rabbxmibcn-Mooiiey equation, and several definitions of non-Newtonian nitration charac-
teristics are defined. The equations presented in this paper may be considered as equations
applicable both for non-Newtonian filtration and for conventional Newtonian filtration. Methods
for evaluating the overall characteristics of non-Newtonian filtration are suggested by using the
compression permeability data. In order to confirm the validity of the theory, constant pressure
filtration experiments are carried out under various conditions of the Bow behavior index N ranging
from 0.404 to 0.504 and the filtration pressure p from 1000 to 3000 G/cm', and it is shown that
the methods presented in thb paper are valid. It is also shown that the average specific filtration
resistance varies considerably with change in the /V-value of the power-law, and the cakes formed
from non-Newtonian filtration of pseudo-plastic fluid are denser than those from usual Newtonian
filtration.
Introduction
In spite of the basic importance of non-Newtonian
filtration in broad fields of petrochemical and food
processing industries etc., very little has been studied in
theories and experiments, while W. Kozicki ft a/.7'"
have made valuable contributions to the nitration
theory of power-law non-Newtonian fluids.
In this paper, the conventional filtration theory of
Newtonian fluids at constant pressure is reexamined in
view of the power-law for flow of non-Newtonian
fluids. In order to provide a useful mathematical
tool of simplified form to industrial filtration, a genera-
lized theory which is applicable to both non-Newtonian
and conventional Newtonian filtration is presented
in this paper. It will also be demonstrated that the
non-Newtonian filtration behaviors can be calculated
on the basis of compression-permeability cell measure-
ments and the estimated results are compared with
constant pressure filtration experiments of pseudo-
plastic non-Newtonian fluids.
1. Experimental Equipment and Procedures
The experimental filter, shown in Fig. 1, essentially
consists of a plexiglass cylinder of 130mm inside
diameter, a brass upper plate with a connection for
applying air pressure and a stainless-steel bottom
Received August 12, 1976. Correspondence concerning thii irtkU ibould
be addretMd to M. Shinto. M. Wakimoto it with Noritake Co.. Ltd.. Nt-
goya 451, S. Fujiyoihi ii with Shin-Nippon Iron * Sue! Co.. Ltd., Tokai
476 mod S. Nindi i> with Japan Catalytic Chem. Ind. Co., Ltd.. Hiraeji
671-12.
plate which supports a perforated plexiglass plate
with a filter paper on it.
In order to conduct non-Newtonian filtration ex-
periments, the selection of solid materials is important
and it is essential that addition of non-Newtonian
liquid to suspensions may not substantially affect
particle flocculation, and that the viscous character-
istics of the filtrate do not vary during filtration pro-
cess"'. Several kinds of slurry materials (i.e., Gai-
rome clay, Korean kaolin, calcium carbonate, fine
silica sand, Filter-Cel, Standard-Supercel, Hyflo-
Supercel and Radiolite) have been examined. The
last two materials have proven to be appropriate in
the above mentioned views and Radiolite (# 1100)* is
used for the experiments attempted in this study.
The non-Newtonian fluids used in this study are
(rest of paper is missing)
PLEXIGLASS
CYLINDER
\ FILTER (WER ft
PERFORATED
FILTRATE PLATE
Fig. 1 Schematic diagram of experimental apparatus
* Diatomaceous filter aid, Showa Kagaku Kogyo Co., Ltd.
JOUtNAL OF CHEMICAL ENOINfERINO OF JAPAN
598
-------
EXPERIMENTAL STUDY OF THERMOPHORESIS OF
AEROSOL PARTICLES
YASUO KOUSAKA, KIKUO OKUYAMA,
SHIGERU NISHIO AND TETSUO YOSHIDA
Department of Chemical Engineering, University of Osaka
Prefecture, Sakai, 591
(Reprinted with permission)
599
Reprinted from
JOURNAL OF
CHEMICAL ENGINEERING
© OF
JAPAN
Vol. 9, No. 2 (1976)
Pagei 147 -150
-------
EXPERIMENTAL STUDY OF THERMOPHORESIS OF
AEROSOL PARTICLES
YASUO KOUSAKA*, KIKUO OKUYAMA.
SHIGERU NISHIO AND TETSDO YOSHIDA
Department of Chemical Engineering, University of Osaka
Prefecture, Sakai, 591
The velocity of thermophoresis of aerosol particles in the slip flow region, about which no con-
clusion has yet been drawn from the many results of experimental and theoretical work, was studied
from the experimental point of view. A new experimental method using an ultramicroscope
was developed to meet most of the necessary conditions to obtain reliable data on thermophoresis,
such as accurate observation of velocity under an accurately known temperature gradient and
prevention of the action of any forces except thermal force. The experimental results were
compared with some of the most representative theories, and were found in good agreement with
Derjaguin's theory.
Introduction
Because of the practical interest in particle deposi-
tion on heat exchanger tubes and in particle collection
by scrubbers and thermal precipitators, as well as the
theoretical interest in evaluating the phenomenon,
extensive experimental investigations of thermo-
phoresis have been made".
In the large-Knudsen number region /fi>>l, theory
and experiment are found to be in satisfactory agree-
ment. In the smaller region or the slip flow region,
Kn<\, though various theories have been proposed,
sufficient reliable experimental data to verify them
have not been obtained because of the difficulty in
accurate measurement of the velocity of thermo-
phoresis.
This paper presents experimental data on the veloc-
ity of thermophoresis in the slip flow region obtained
by a new technique developed to determine the ac-
curate velocity of thermophoresis. The data are
then compared with some representative theories of
thermophoresis1"'•'•"'"".
Experimental Method
Several experimental methods to determine the
velocity of thermophoresis have been developed.
Derjaguin examined the available experimental meth-
ods which had been reported and classified them into
four types". Each of them, however, seems to have
some unavoidable faults. In accurate determination
of the velocity of thermophoresis, the following con-
Received September 19, 1975.
Presented at the Tokushima Meeting (at Tokushima, July
1975) of The Sex;, of Chem. Engrs., Japan.
VOL. 9 NO. 2 1974
ditions should be essentially satisfied: 1) to know the
accurate temperature' gradient where the velocity of
particles is just observed; 2) to prevent the action of
any non-thermal forces, such as fluid drag due to
fluid flow, photophoretic and electric forces; 3) to
avoid convective flow of aerosol induced by the tem-
perature difference"; 4) to know accurately the diam-
eter of spherical particles: and 5) to observe the
velocity itself directly under the above conditions.
Considerable spread in experimental data obtained by
different authors is thought to be caused by lack of
some of the above conditions. The experimental
method presented in this paper was developed so that
the above conditions were satisfied as much as pos-
sible.
The experimental technique applied in this study is
in principle much the same as that previously devel-
oped by the authors for size analysis of aerosol
particles'". The only difference between them lies
in the observation cells.
Fig. 1 shows the experimental apparatus. The ob-
servation cell fixed on the stage of an ultrumicroscope
has a water jacket into which cooling water controlled
in temperature ranging 0C'C to room temperature is
circulated to cool the bottom wall of the cell. The
bottom wall was made from brass plate which has a
large heat capacity. The upper wall of the observa-
tion cell consists of a glass plate through which the
particles suspended in the cell were observed by the
ultramicroscope. The side walls of the cell were
made from polyvinyl chloride for thermal insulation.
A temperature gradient was formed between the
upper glass wall and the bottom one, and its extent
was controlled by changing the temperature of the
600
-------
ultromicroscope
UldW by
cyonooiry&tt bindir>g
~ i plan view o(
thermocouple
stageofurlramkroscop* H O.lmm Cu-Con&tantan
Sectional thermocouw.
in mm
Fig. 1 Experimental apparatus
bottom wall, the upper wall being left at room tem-
perature. The ratio of cell width to height was
selected as ten so that convective flow induced by
temperature gradient in the cell could be avoided".
The aerosol to be observed is cooled by a heat
exchanger to the mean temperature in the cell and is
introduced into the cell. After several seconds of
admitting the aerosol, the flow is instantaneously
stopped by closing the electric valves shown in Fig. 1.
Then a linear temperature field is formed within a
short time throughout the cell except in the vicinity of
the side walls. The value of the time interval, r,
needed for the aerosol in the cell to warm up is given
by"
r=«js>/C,/A, (1)
where Cf is the specific heat of the aerosol, pf density,
K, thermal conductivity of the aerosol and ht half of
the cell depth. In the present case r comes to 0.05
seconds, which is negligibly small compared with the
observation period /,/, described later. Exact solu-
tion of this problem can be obtained by an analogical
method in solving the establishment of Couette flow,
and it gives a still smaller value. Thus the field may
be regarded as at steady state.
The focus of the ultramicroscope was set at a given
depth h from (he inner surface of the upper glass wall
by adjusting the height of the stage, on which the cell
was fixed, up and down. Thus the particle numbers
at various depths was observed. The temperature
gradient in the cell, on the other hand, was also pre-
liminarily measured by a small thermocouple in the
cell shown in Fig. 1, the depth of which was also ad-
justed by displacement of the stage. The stage dis-
placement in these measurements was determined by
the height gauge installed in the microscope.
Aerosol particles in the cell start to settle just after
closing the valves of the cell under the influence of
gravity and thermophoresis. The particles appearing
in sight of the microscope, which is focused at a
certain depth of the cell, h, are recorded by a video
recorder as sedimentation progresses until the particle
disappears from sight. Knowing the depth h and the
time fi/j at which half of the initial particles disappear
from sight of the microscope, the settling velocity of a
particle having median diameter DPao, though it is
resultant velocity shown below, is determined:
l/e(D,,.) -r Ur(Dpis) = */*,.,
(2)
£/c(/?fH) represents the gravitational settling velocity
and is easily obtainable by measurement where no
temperature gradient is formed in the cell. Further-
more, Uc(Dritl) can be converted into the median
diameter of the particles, Dfw by using the Stokes-
Cunningham equation. Thus the values of both
Ue(DTia) and Z>,,so are accurately evaluated. In con-
sequence the velocity of thermophoresis of a particle
of Dp,, in diameter, (J,-(Dfia), can be determined by
observing ti/t under existence of temperature gradient.
Eq. (2) is valid when resultant velocity of UG and VT
increases monotonously with particle diameter. As it
is well known that the dependence of particle size on
the velocity of thermophoresis is small this condition
will be satisfied in most cases unless there exists an
extremely large temperature gradient.
Aerosol particles used in this study were tobacco
smoke, stearic acid and OOP. Aerosols of both
stearic acid and DOP were generated by a La Mer-
Sinclair type generator and tobacco smoke was
generated by a simple smoking apparatus" !. Aerosols
thus generated were cooled by a heat exchanger and
were then observed.
The size distributions of aerosol partides are shown
in Fig. 2. They were obtained by the ultramicro-
scopic method1" using the same cell as shown in
Fig. 1 but having no temperature gradient in it.
Experimental Results and Discussion
Figs. 3 and 4 show the experimental results. The
temperature gradients .measured in the cell are shown
at the right side of the figures. They seem to be
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
601
-------
Fig. 3 (a) Relation between /, „ and A
(b) Temperature gradient
OA 0.60.61.0
Dp t>n
Fig. 2 Size distribution of aerosol particles
Fig. 4 (a) Relation between r,,2 and h
(b) Temperature gradient
linear. The abscissa of the figures, y, indicates the
depth from an arbitrary position which roughly cor-
responds to 0.15mm in actual depth h from the
upper wall. In the left-side figure, the relation be-
tween tl/t in Eq. (2), at which half of the initial particles
disappear from sight of the microscope, and h, the
depth from the inner surface of the upper wall of the
cell, were plotted. To compare these experimental
values with theoretical ones, the following two re-
presentative theoretical equations proposed by Derja-
guin" and Brock" were adopted,
(3)
(4)
grad T
l + 2Cm(2*ID,)
U?(D,)= \ x Uf(D,)
VOL. * NO. 1 1974
Since the gravitational settling velocity, U^D,^),
and accordingly the median diameter of the particles,
Z)P80, have been already determined by experiment,
substitution of U,,(DrJ and l'ADfJ which can be
calculated by the above equations knowing DPJI) and
other experimental constants into Eq. (2) gives the
theoretical relation between t,-, and h. The solid
and dotted lines in Figs. 3 and 4 are those thus pre-
dicted by Derjaguin and Brock, respectively It can
be found that good agreement exists between the
solid lines calculated on the basis of Derjaguin's
theory and the experimental values, except only the
case of large temperature gradient, grad 7"---~74.8"C/
cm. The experimental data of grad 7"^-74.8'C/cm
were obtained under an undesirable condition where
hot air was blown onto the surface of the upper wall
of the cell to obtain a large temperature gradient.
The deviation from Derjaguin's theory in this case is
602
-------
Aerosol
Table 1 Comparison of experimental results win theories
grad T Ur(DfK) Icm/sec]
[°C/cm] Exp. Eq."(3) Eq. (4)
n°"Ce
OOP
Stearic acid
Tobacco smoke
-9.2
-19.2
-38.0
-74.8
-13.5
-14.4
-18.3
2.72x10-
5.83x10-
9.90x10-
1.16x10-
3.39x10-
3.86x10-
4.93x10-
2.64x10-
5.51 xlO-
1.09x10-
2.14x10-
3.83x10-
4.08x10-
1 4.82x10-
1.33x10-
2.75x10-
5.45x10-
l.Olx 10-
1.91x10-
2.04x10-
2.41 x 10-
Jf.=5.9xlO-»[ca!/cm
*(=3.0xlO-«[cal/cm
convective flow occurs
oil-l°0410"4"
i if t AV \c\~ * 5I n
r\i^=j.\j*' t\J , fj)M
sec-°KJ"
•sec-'K]"
=0.94
caused by poor temperature control of the upper wall
of the cell and also by convective flow of aerosol in
the cell owing to the large temperature gradient.
The velocities of thermophoresis in various experi-
mental conditions were determined by the slope of
Figs. 3 and 4, subtracting those at zero temperature
gradient. The results are shown in Table 1. Good
agreement is also found between the experimental
results and the values calculated by Eq. (3).
In ullramicroscopic observation of particle numbers
at a certain depth of the cell, h, it was noteworthy
that the particles disappeared suddenly from sight of
the microscope at the time tlft while they disappeared
gradually under zero temperature gradient. This
sudden disappearance shows the small dependence
of particle diameter on the velocity of thermophoresis,
as expected from Eqs. (3) and (4).
The particle number concentration of aerosols in
experiment was about 4xlO'~8xl06, which cor-
responds to about 50~100 particles in sight of the
microscope. At these concentrations almost no effect
of Brownian coagulation on the change in particle
number concentration occurs1". The effect of photo-
phoresis by illumination of the ultramicroscope on
settling velocity was completely avoided by intermit-
tent lighting.
Conclusion
The velocity of thermophoresis in the slip flow
region was studied experimentally. The experimental
method presented herein was developed to meet most
of the necessary conditions for accurate measurement
of thermophoresis, and it gives very reliable data on
thermophoresis compared with those so far reported.
The results were compared with the theories proposed
by Derjaguin and by Brock, and were found in good
agreement with Derjaguin's theory rather than Brock's.
Adnmrledgmeat
I. Nishioka was very helpful in the experimental work.
Nomenclature
C.
C,
D,
h
K..K,
T
P
9*
= tangential momentum first-order slip
coefficient" « 1.23 [—]
= temperature jump first-order slip
coefficient" =2.16 [_]
= diameter of particle [cm], [/i]
= depth from inner surface of the upper
wall of a cell [cm], [mm]
= thermal conductivity of gas and
particle, respectively [cal/cnvsec-°K]
= temperature of gas [°C], [°K]
= the time when half of the initial number
disappear [sec]
= velocity of gravitational settling [cm/secj
= velocity of thermophoresis [cm/sec]
= depth from an arbitrary position in
a cell [mm]
= mean free path of gas molecules [cm]
= viscosity of fluid [g/cm-sec]
= density of fluid [g/cm1]
= density of particle [g/cm'J
— for median diameter
D = for Derjaguin
B = for Brock
Literature Cited
1) Brock, J, R.: /. Colloid Sci., 17, 768 (1962).
2) Derjaguin, B. V. and Y. Yalamov: J. Colloid Sci., 20, 555
(1965).
3) Derjaguin, B. V. and Y. Yalamov: / Colloid and Interface
Sci., 22, 195 (1966).
4) Gudzinowicz, B. J.: J. Chem. Eng. Data, 9, 79 (1964).
5) Hidy, G. M. and J. R. Brock: "Topics in Current Aerosol
Research (Part 2)", Chapter 5, Pergamon Press Inc. (1972).
6) Jacobsen, S. and J. R. Brock: /. Colloid Sci., 20, 544 (1965).
7) Paranjpe, M. K.: Proc. Indian Acad. Sci., 4a, 423 (1936).
8) Phillips, W. F.: Physics of Fluids, 18, No. 2. 144 (1975).
9) Sone, Y., K. Aoki and Y. Onishi: J. Japan Soc. Aero,
Space Sci,, 23, 568 (1975).
10) Sone, Y. and K. Aoki: ibid., 23, 575 (1975).
II) Yoshida, T., Y. Kousaka and K. Okuyama: Ind. Eng.
Chem.. Fundam., 14, 47 (1975).
12) Yoshida, T., Y. Kousaka, K. Okuyama and S. Nishio: /
Chem. Eng. Japan, 8, 137 (1975).
JOURNAL OF CHEMICAL ENGINEEKIN6 OP JAPAN
603
-------
TURBULENT COAGULATION OF AEROSOLS IN A
STIRRED TANK
KIKUO OKUYAMA, YASUO KOUSAKA,
YOSHINORI KIDA AND TETSUO YOSHIDA
Department of Chemical Engineering, University of
Osaka Prefecture, Safari 591
(Reprinted with permission)
Reprinted from
JOURNAL OF
CHEMICAL ENGINEERING
OF
JAPAN
Vol. 10, No. 2 (1977)
Pagts 142—U7
-------
TURBULENT COAGULATION OF AEROSOLS IN A
STIRRED TANK
KIKUO OKU YAM A, YASUO KOUSAKA,
YOSHINORI KIDA AND TETSUO YOSHIDA
Department of Chemical Engineering, University of
Osaka Prefecture. Sakai 591
Turbulent coagulation of aerosol particles was studied experimentally by observing the time-
dependent changes in particle number concentration and size distribution of aerosol for various
intensities of stirring in a stirred tank, using the ultramicroscopic size analysis. From the ob-
served data on the decrease in particle number concentration of aerosol, the rate constants of tur-
bulent coagulation were evaluated and compared with some of tlie most representative theories, and
were found in good agreement with SafTman and Turner's theory. Further, the changes in particle
size distribution observed were confirmed by those obtained by numerically solving the equation of
coagulation for polydisperse aerosol, the so-called population balance equation.
Introduction
The rate of coagulation of aerosol particles depends
on Brownian motion of particles and turbulent motion
of the fluid in which particles are entrained, when
particles are not electrically charged. Brownian co-
agulation controls the rate of coagulation under small
particle sizes and small turbulence, while the effect of
turbulent coagulation begins to exceed Brownain co-
agulation with the increase of particle radius and tur-
bulent intensity. The behavior of polydisperse aero-
sols undergoing Brownian coagulation has been studi-
ed by many researchers' 6I"-1S| and a general under-
standing of the behavior under various conditions has
been almost obtained. Previous studies1'7'9' on tur-
bulent coagulation, on the other hand, have been limit-
ed to theoretical ones where coagulation rate is dis-
cussed, but few experimental data are available to
verify the theories because of the difficulty in accurate
measurement of the change in properties of highly
concentrated aerosol".
In this paper, the changes in particle number and size
of aerosols undergoing turbulent coagulation in a
stirred tank were observed for various intensities of
stirring, using the ultramicroscopic size analysis pre-
viously developed by the authors'". Then the rate
constants of turbulent coagulation were evaluated from
the observed data on the decrease in particle number of
aerosols, and they were compared with some repre-
sentative theories on turbulent coagulation. Further-
more, the changes in particle size distribution observed
were compared with those obtained by numerically
Received September I, 1976. Correspondence concerning this article
ihould be addressed to K. Okuyama. Y. Kida is with Kurnho Co., Ltd.,
Hink*t»S7J.
solving the equation of coagulation for polydisperse
aerosols.
1. Experimental Apparatus und Method
A schematic diagram of (he experimental apparatus
is shown in Fig. 1. The vessel used was mack- of acry-
lic resin, and equipped with four vertical batik's each
of which has a width of one tenth of the tank diameter.
The stirrer was six llat-bladcd turbine. The vessel and
stirrer dimensions are standard ones as shown in Fig.
I. Aerosol used in this study was tobacco smoke
generated by a simple smoking apparatus11-, by which
number concentration of particles was controlled from
10" to 10s particles/cc. Aerosols thus generated were
introduced promptly throughout the chamber and was
mechanically stirred for a short period with extremely
low revolution to make the aerosol uniform. Then
the revolution of the stirrer was raised to the desired
speed. The stirrer speed was checked by a photo tran-
|VTR > monitor |-t><-
*[r?
U T J
photo tronstiiof
E-JT.'- ' 1
JtL
lomp
-'- 1
%i
Jii
^ _. T!. _..,.?,
-
i
-t>
-
^xl
[tlffOIOl
<1 -" Qif COI
- 4 batti* f
...L
un.rj
I
npiPiSOrJ
kili*s
in tm
%iirrtd tonfcl
tlirrtd tankl
H
100
705.
_!L..
100
190
_0r
89
om,
WT
10
i 8
IT
2 •;
2 ^
H
It/10
Ti/To"
Fig. 1 Experimental apparatus
JOURNAL OF CHEMICAL ENGINEERING Of JAPAN
605
-------
Table 1 Experimental conditions and properties of aerosols
N
Stirred
600
1800
3780
5000
9000
Stirred
1440
3000
•
«
5400
Re
tank I
1.81 v
5.44 •
1.14x
l.Slx
2.72x
tank II
1.28x
2.66-
•
.
4.78x
I01 1
10s :
10* :
10«
10«
10'
10*
10'
"o
.00;- 10'
.00 • 107
..00 10'
.75.- 107
.50x10'
.30A10'
.80x 10"
.00x10'
5.25xlO«
.00x10'
r>o
0.47
0.42
0.45
0.45
0.45
0.47
0.40
0.46
0.64
0.47
"*>
.34
.31
.38
.40
.40
,34
.33
.28
.48
.46
'0
3.00X104
8.10x10*
7.45x10"
1.74x10'
1.01 xlO"
8.10x10*
7.45x10*
i
t
4.35x10'
K
1.02x10-
1.28x10-
6.44x10"
1.25x10-
3.76x10-
2.11x10-
3.66x10-
5.65x10-
1.13x10-
1.30x10-
2
1
6
6
1
1
1
7
1
3.
P
32xiQ-
09 x 10-
37x10-
91x10-
78x10-
14x10-
89x10-
33 x 10-
68 x 10-
68x10-
KD
2.9
3.6
18.2
35.2
106
6.0
10.1
16.0
33.5
37.0
Pn'jKn?
1.14x10-
4.26x10-
5.23x10-
3.16x10-
3.16x10-
4.16x10-
2.86x10-
1.30x10-
2.38x10-
2.38x10-
[sec]
stirred tankH
3000 r.p.m. 5400 r.p.m.
Fig. 2 Photographs on time-dependent changes in aerosol
sistor and was varied from 300 to 9000 rpm. Aerosol
sampled at any given residence time was introduced
into the observation cell installed on the stage of an
ultramicroscope to measure its particle size distribu-
tion and particle number concentration1". Experi-
ments were carried out by changing initial particle
number concentrations, particle sizes, stirrer speeds
and sampling positions. Representative experimental
conditions and properties of aerosols are shown in
Table 1.
2. Experimental Results and Discussion
No difference in experimental results was found in
changing sampling positions, which indicates that
the aerosol is uniform throughout the tank. Figure 2
is an example of a series of photographs, taken by a
camera directly attached to the ultramicroscope. It
is se*n that particle number decreases rapidly with
VOL. 10 NO. 2 1977
2 4 6 BIO1
/Browrvan
/ coagulation olon«
-. n,,10-10' -
45 /"
6 BIO1
(1 i
Fig. 3 Effect of stirrcr speed on decrease In particle
number concentration of aerosol
time, and the decrease is more rapid at higher revolu-
tion of the stirrer. The particle growth due to turbu-
lent coagulation is also found in the photographs.
2.1 Particle number concentration
Figure 3 shows experimental relations between the
ratio of particle number concentrations at any time
with those at initial time and stirring time in a tank.
It can be seen that particle number concentrations in
all cases decrease faster than the estimated ones from
the equation of Smoluchowski"" for monodisperse
aerosols undergoing Brownian coagulation, which is
given by
l/ji'-l/Hi=2«iC.(rrt)f', Ar,=2/cr/3^ (I)
These differences increase with the intensity of stirring
when initial aerosol properties are of the same order.
606
-------
In the case of same stirrer speed, the decrease in parti*
cle number depends on the initial particle number con-
centrations and initial particle sizes. As these experi-
mental results depend not only on coagulation but on
deposition loss of particles to the walls, it is impossible
to compare directly these results with the theory of
coagulation alone. According to Gillespie and
Langstroth", the effects of coagulation and deposi-
tion on the decrease in particle number concentration
can be seperated quantitatively by introducing the fol-
lowing equation
dn'ldt' Kn'*-pn' (2)
where K denotes the coagulation rate constant and j9
the deposition rate constant.
As both K and /8 may be considered to depend to
some extent on particle size distributions which are
subject to change during the ageing of an aerosol,
they will be a function of time. Since it is very com-
plicate to introduce the change of particle size into Eq.
(2), time-dependences of K and ft were disregarded
here. Integration of Eq. (2) gives
The coagulation rate constant K and deposition rate
constant ft were determined by fitting the experimental
data to Eq. (3) using the nonlinear squares method.
The curves in Fig. 3 are the fitting curves thus obtained.
The values of A'and /3 are shown in Table 1 for various
conditions. As seen from the table, values of K
ranged from 10~' to 4xlO-'cm'/sec, while ft from
10-' to 2 x 10-' I/sec. The values of K and ft seem to
increase with the stirrer speed and initial particle size.
2.2 Coagulation rate
The values of K evaluated from experimental results
include both effects of Brownian and turbulent coagu-
lation, and turbulent coagulation is first discussed in
this section.
Turbulence can affect coagulation by two different
mechanisms. In the first mechanism, since the tur-
bulent flow brings spatial non-homogeneities, different
velocities in neighbouring particles appear and, as a
result, particles collide with each other by a mechanism
analogous to the mechanism of laminar shearing flow.
A second coagulation mechanism is caused by the rel-
ative motion of each particle differing from that of the
turbulent air, because its inertia will not be the same as
an equivalent mass of air". This second mechanism
may be neglected" when
(1) the sum of colliding particle radii is small com-
pared to the smallest eddies in the fluid, and
(2) the particles follow the fluid motion completely.
In either mechanism, the coagulation rate depends
mainly on particle size and velocity gradient evaluated
from the energy dissipation rate per unit mass of fluid.
In the case of stirred tank, the average value of the
energy dissipation rate «<, is taken to be equal to the
power consumption rate per unit mass of mixing
fluid. Some investigators including Schwartzberg and
Treyball" give e, for the standard stirred tank used in
this experiment as, from the data for water
«o=7.9JVJ/yr/r?// at /te>5000 (4)
The distribution of energy dissipation rate through a
tank results in coagulation rate distribution. That is,
a difference exists between the coagulation rate based
on the average energy dissipation rate «„ and that
based on the local energy dissipation rate and its distri-
bution through a tank. Kuboi, Komasawa and Otake,
however, suggested that the effect of energy dissipation
rate distribution through a tank on the mass transfer
coefficient is relatively small. For a first approxima-
tion, the coagulation rate experimentally obtained was
connected with average value of «0 neglecting the effect
of local values of energy dissipation rate in this study.
The values of e0 are shown in Table 1. According to
the theory of isotropic turbulence proposed by Kolmo-
goroff, the micro-scale of the turbulence >J0 is given by
the next following equation from e0T)
^(v'Ao)1" (5)
Under the present experimental conditions, the mini-
mum value of I, is about 25 ft, which is sufficiently
large compared with the particle radius. As the
relaxation time is sufficiently small for sub-micron
panicles, the particles will follow fluid motion com-
pletely.
From these discussions it can be concluded that,
in the present case, the above second mechanism of
turbulent coagulation due to the inertia of the particles
may be ignored and the first mechanism is important.
The representative theoretical equations denoting the
collision rate by the first coagulation mechanism have
been proposed by
Saffman and Turner91 ;
r',)= \ .300-,' +r;)»
Levich";
(6)
(7)
The value d which appears in the equation is given as
about 0.25 by Fuchs". Equation (7) reduces to next
equation when ri+rj is less than /i0
The comparison of these equations with experimental
results has not been made so far. When an aerosol
is monodisperse, coagulation rate Kr is given as
*r(r,')=*r(r,'.r,')/2 (9)
The values of AT evaluated from experimental results
including the effects of both Brownian and turbulent
coagulation as described before, the following value
JOURNAL OP CHEMICAL ENOINJKtNO OP JAPAN
607
-------
KD was next introduced to obtain the importance of
turbulent coagulation relative to Brownian coagulation
y Yf ¥ (» ' \~ / f t tf ' \ i Vfr'\\f&fff'\ l\ f\\
•A.J) — "f**B\'tQJ — \**-T\'ffQJ~i •"•Bx*Bfl'//**J?V'flO/ x^-V
where Ka(r^)—2K0Cm(r',0), which can be easily deter-
mined. The values of KD are shown in Table 1.
Figure 4 shows the variation of KD against the values
of rjo Veo, together with the theoretical curves of
KD from Eq. (6) by Saffman and Turner and Eq. (8)
by Levich. It will be seen that the experimental data
agree with the curve of Saffman and Turner's equation
rather than that of Levich. Turbulent coagulation can
be ignored at r',l V«o less than 2 x !0~", while Brownian
coagulation can be ignored at r^Vso larger than 2x
10"10. When particle size increases significantly by
turbulent coagulation, comparison of experimentally
obtained KD with the theoretical one by using the ini-
tial geometric mean radius will be erroneous.
2.3 Deposition rate
Figure 5 shows the dependence of the deposition rate
/S obtained experimentally on the energy dissipation
rate £«, together with the theoretical values calculated
by next equation
8 ,„„,,„.„, w,(r,') _..,.( ffH<(rd) )
01)
Equation (Jl) was derived by Takahashi and Kasa-
hara1" for a cylindrical vessel in the same way as
Corner and Pendlebury" for a cubic vessel, which
gives the deposition rate due to Brownian and turbu-
lent diffusion accompanying the gravitational settling.
In Eq. (11), K. is equal to ic'du/dx, where *' is the
Karman's constant (~0.4) and dujdx is given by
Saffman and Turner81 as follows
du/dx=(2s,/l5t>Y11
(12)
Figure 5 suggests that experimental values agree ap-
proximately with the theoretical ones and that the
values of stirred tank I are larger than those of stirred
tank II, because the tank I has a larger ratio of wall
surface to the volume.
When the values ofpnyKn? in Table 1, the ratio of
deposition rate to coagulation rate at the initial ageing
stage, are less than about O.I, it is found that the
effect of deposition on the decrease of particle number
is relatively small. Figure 6 shows the comparison of
actual time-dependent change in particle number con-
centration with calculated curves obtained by ignoring
the one of the values of K and /3 in Eq. (3). It is
seen that experimental results agree with the curve of
coagulation alone. Figure 7 shows the effect of initial
particle number concentration n'a on decrease in par-
ticle number concentration. Particle number of highly
concentrated aerosol decreases faster than that of lower
101
E
A
"i. I
"to'
1
6
1
10s
l
.
^S"
, , ,
/.
//
/S
1 '//'''
^ i,
's* • %•;
• • ''
,-,'•' • tli'iK
;' O ttint
(Soltmon a
cottulalM t>»
(Itvith
#•£' '-
V
.
1 tank 1 ~
1 lank 1 ~
EqlO'EqdO)
nd Tumrr)
E5(»-Eposition alone
-K -6 tt ilfl'em^sec K«0
:—:—z— •* A_C ITvinJ ««/*'
J u_JL
i .i.l-
v p
p=637xl03sec-'
\
10'
,« •» / c «•«' 2 i 6 8IO! 2 46 8101
4 6 810
f CseO
Fig. 6 Effects of deposition and coagulation on decrease
in particle number concentration
one, which indicates that the effect of deposition is not
significant since the effect of deposition must be in-
dependent of n't. These resulcs suggest that the behavior
of aerosol for one micron order in a stirred tank is
dominated by coagulation when particle number con-
centration is larger than 5x 10* particles/cc.
2.4 Particle size distribution
The basic equation for the time-dependent change in
particle size distribution of polydisperse aerosols
undergoing Brownian and turbulent coagulation can
be written as1"
VOJ- 10 NO.'2 1977
608
-------
be established instantaneously with the following
log-normal form
4 6 8 I01
Fig. 7 Effect of Initial particle number concentration on
decrease of particle number concentration
V0.4 0.6Q8I 2 4 0.4 06 081 2 4
r'/'v> [-]
Fig. 8 Change in particle size distribution with time
3n'(r',t')ldt'
7^', p')+KT( Wr-^p*, p')
-{f~~{KB(r',p')+KT(r',p')}n'(r',t')n'(p',t')dp'
Jf-*0
(13)
KB(r',p') is the Brownian coagulation function and is
given101 by
KB(r',p')=K,(r'+p'){Cm(r')lr' + Cm(p')lp'} (14)
Saffman and Turner's Eq. (6) was used here as KT(r',
p'), the turbulent coagulation function, which showed
better fitting with experimental results.
The left side of Eq. (13) is the change in particle
number concentration of size r' with time. The first
term on the right side represents the rate of formation
of particles of size r' due to coagulation of two particles
smaller than size r' and the second term the rate of
loss of particles of size r' due to their coagulation with
particles of other sizes including r'.
The initial particle size distribution was assumed to
21n1o-,,
(15)
Since Eq. (13) cannot be solved analytically, the Runge-
Kutta-Merson method was employed to solve it1".
As a numerical check the total mass of aerosol was cal-
culated every few time steps and compared with the ini-
tial value. Figure 8 shows the comparison of time-de-
pendent change in particle size distribution between
calculated and experimental ones. The frequency
/(In r') was calculated by the next equation
/(In r')=/iV. '>'/«; (16)
The manner of the change by calculation seems to de-
pend on the value of KD which expresses the relative
importance between Brownian and turbulent coagula-
tion. When the value of K,, is 1 .28, as shown on the
left side of the figure, where Brownian coagulation is
controlling, the particle size distributions shift towards
the larger radius with time. When the value of KD
is large, as shown on the right side of the figure, where
turbulent coagulation is controlling, the mode radius
in the particle size distribution does not tend to move to
the larger radius. These tendencies will be caused by
the difference in the dependences of the coagulation
functions on particle sizes, that is, the dependence of
Brownian coagulation function on particle radius is
not very large for r0'>0. 1 p, while turbulent coagula-
tion function is propotional to the cube of particle size.
The decreasing rate of total particle number concentra-
tion by turbulent coagulation is found to be very large
in comparison with that by Brownian coagulation in
Fig. 8. The figure indicates that the calculation re-
sults agree approximately with experimental results
including the small effect of deposition.
Conclusion
Turbulent coagulation of aerosols was studied
experimentally by observing the time-dependent
change in particle number concentration and size
distribution of aerosols in a stirred tank, and the coagu-
lation rate and the deposition rate were determined.
The turbulent coagulation rate experimentally obtained
was compared with the theories proposed by Saffman
and Turner, and by Levich, and were found to be in
good agreement with that proposed by Saffman and
Turner rather than that by Levich. The deposition
rate agreed approximately with the theory proposed
by Takahashi and Kasahara. The time-dependent
changes in particle size distributions were evaluated by
numerically solving the coagulation equation for poly-
disperse aerosols using the turbulent coagulation func-
tion proposed by Saffman and Turner, and then they
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
609
-------
were compared with experimental ones. The change
in size distribution by turbulent coagulation obtained
by both calculation and experiment give a distinctive
feature which is different from those by Brownian
coagulation. It was found that at i$ \/«o less than
2x 10~" turbulent coagulation is not effective and at
itf \/t« larger than 2 x 10~ 10 Brownain coagulation can
be ignored.
Acknowlcdgnient
N. Hishlo was very helpful in the experimental work.
NomendataK
C«(rj) — Cunningham's correction factor of radius r't
D(f\)
DT
g
H
K
AT,
KD
K.
N.N,
n'
n'(r',t')
n,
Re
= Diffusion coefficienct (
Cw(r|)«776it/«r;)
[cm'/sec]
[cm]
[cm/sec*]
[cm]
[cm'/sec]
[cm'/sec]
= stirrer diameter
=• acceleration of gravity
- height of a stirred tank
- overall coagulation rate constant
- coefficient in Eq. (1)
" Brownian coagulation rate constant for
monodisperse aerosol [cm'/sec]
— Brownian coagulation function for two
particles of size r\ and r'/ [cm'/sec]
- defined by Eq. (10) [— ]
- t'duldx [I/sec]
— turbulent coagulation rate constant for
monodisperse aerosol [cm'/sec]
— turbulent coagulation function for two
particles of size- r\ and r'f [cm'/sec]
« stirrer speed [l/min][l/sec]
= total particle number concentration at any
time [particles/cm*]
- particle size distribution [particles/cm1 -em]
- total panicle number concentration at
time zero [panicles/cm1]
— Reynolds number based on stirrer tip
velocity (-/),Ar.O}./rt [-1
— particle radius [cm]
~ geometric mean radius [cm]M
- particle radius for monodisperse aerosol
T — absolute temperature
TT — diameter of a stirred tank
I' - time
u •» average velocity in a 'stirred tank
ut(r') = terminal settling velocity
(=2Cm(r')(p,-pf)or'*l9n)
x — arbitrary direction in a stirred tank
ft - deposition rate constant
.to = average energy dissipation rate
i = Boltzman's constant ( =1.38x10' '•)
t' = Karman's constant («0.4)
PT — fluid and particle (tensity
{°K]
[cm]
[sec]
[cm/sec]
[cm/sec]
[cm]
[I/sec]
[cm'/sec1]
[erg/°K]
[— ]
[cm][H
[g/cm-sec]
[cm1 /sec]
[cm]
[g/cm']
Ofi = geometric standard deviation at time zero [ — ]
Literatim Cited
1) Beal, S. K.: /. Aerosol Set., 3, 113 (1972).
2) Corner, J. and E. D. Pendlebury: Proc. Phys. Sac., BM,
645(1951).
3) Gillespie, T. and C. O. Langstroth: Canad. J. Chem., 30,
1003 (1952).
4) Huang, C. M.. M. Kerker and E. Matijevic: /. Colloid
Interface Set.. 33. 529 (1970).
5) Kuboi, R., I. Komasawa and T. Otake: Chem. Eng. Scl.,
29, 65 (1974).
6) Lai, F. S.. S. K. Friedlander, J. Pich and G. M. Hidy: J.
Colloid Interface Scl., 39, 395 (1972).
7) Levich, V. O.: "Physicochemical Hydrodynamics", Prentice-
Hall, London (1962).
8) Saffman, P. G. and J. S. Turner: J. Fluid Mech., 1, 16 (1956).
9) Schwartzberg, H. G. and R. E. Treybal: Ind. Eng. Chem.,
Fundam.. 7, 1 (1968).
10) Smoluchowski. M. von Z. : Phys. Chem., 92, 129 (1917).
11) Takahashi, K. and M. Kasahara: Atmos Environ., 2, 441
(1968).
12) Yoshida.T., Y.Kousakaand K.Okuyama: Ind. Eng. Chem.,
Fundam., 14, 47 (1975).
13) Yoshida, T., K. Okuyama, Y. Kousaka and Y. Kida: J.
Chem. Eng. Japan, 8, 317 (1975).
14) Zebel, O.: Kolhld-Z., 156, 102 (1958).
VOL 10 NO. I If77
610
-------
THE EFFECT OF NEIGHBOURING FIBERS ON THE
SINGLE FIBER INERTIA-INTERCEPTION
EFFICIENCY OF AEROSOLS
HITOSHI EMI, KIKUO OKUYAMA
AND MOTOAKI ADACHI
Department of Chemical Engineering, Kanazawa University,
Kanazawa 920
The effect of the volume fraction of the fibers in fibrous ah- filters OB the collection efficiency of
• tingle fiber has been examined both theoretically and experimentally In the Inertia predominant
region.
Numerical solutions have been obtained for the flow around a circular cylinder In a cell deter-
mined by the volume fraction of the fibers for a Reynolds number of 10 and the potential flow.
Particle trajectories have been calculated in the cell by introducing the above numerical values
tato the equations of particle motion. As a result, the inertia-interception efficiency was evaluated
through four parameters; inertia parameter, interception parameter, Reynolds number and volume
friction.
Experimental data on model filters, which are made of a uniform parallel arrangement of wires
orientated at right angles to the flow direction have shown good agreement with the calculation
results m the Intermediate Reynolds numbers region.
Irtnttoction
In the filtration of aerosols by a high-porosity fibrous
filter (e>0.98), the collection efficiency due to inertial
impaction, Brownian diffusion, gravitational settling
and direct interception can be predicted using an isolat-
ed fiber model. In the previous paper41, some appro-
priate expressions on collection efficiency of an isolated
fiber were shown and a method was proposed to pre-
dict the efficiency readily under arbitrary operating
conditions. When the porosity becomes lower,
however, the efficiency will deviate from that of an
isolated fiber. This deviation from an isolated fiber
is known as the interference effect between neigh-
bouring fibers, and in classical filtration, the empirical
correction had been made by Chen1' and others1'8'111,
without any certain theoretical background. A mo-
dern filtration theory on the interference effect was
developed by Kirsch, Stechkina and Fuchs*-"' on the
basis of Kuwabara's flow field1" transverse to a
random assembly of parallel cylinders. After the
Kuwabara's flow had been confirmed valid experi-
mentally at low Reynolds number, they showed a
method to estimate the combined diffusion and inter-
ception efficiency which correlated very well with ex-
perimental results by model filters and by real ones.
In the inertia predominant region they also calculated
Received June 1, 1976. Correspondence concerning this article should be
addressed to H. Emi. K. Okuysmi is it Dept of Chemical Engineering,
Uttivaflily of Osaka Prefecture. Sakti 591. M. Adachi ii at Radiation
Center of Osaka Prefecture. Sakti 593.
the collection efficiency for Sf< 1 by an analytical pro-
cedure1" using Kuwabara's stream function. Sten-
house calculated the particle trajectory by a stepwise
method'*1 using Happel's flow field and gave the
inertia-interception efficiency as the function of the
volume fraction of the filter. These theoretical treat-
ments have not taken into account the effect of Reyn-
olds number, which is not negligible especially in the
inertia predominant region.
In the present work, the collection efficiencies due to
inertia and interception have been calculated numeri-
cally from the flow pattern around a circular cylinder
on the basis of a cell model. Calculation results
are compared with the experimental ones obtained
by model filters.
1. Flow Pattern in Fibrous Filter
Most fibrous filters are built up from fiber layers,
in which individual fibers are arranged nearly perpendi-
cular to the flow and keep a proper distance from each
other corresponding to the volume fraction of the
fibers. The flow pattern around a circular cylinder,
therefore, is influenced by its neighbours, and a filter
may be considered to consist of a number of cells,
each of which comprises a single fiber surrounded by
a concentric envelope of air. Though Kuwabara1"
and Happel" solved the viscous flow equation in one
cell for a small Reynolds number as a function of only
the volume fraction of the fibers, the flow in the cell
should be determined by both volume fraction and
JOURNAL OF CHEMICAL ENGINEERING OP JAPAN
611
-------
REPRINTED
FROM
PROCESS DESIGN
AND DEVELOPMENT
Pressure Drop and Collection Efficiency of an Irrigated Bag Filter
Tetsuo Yoshlda, Yasuo Kousaka, SMgeo Inake, and Shigcyukl Nakai
Reprinted from I&EC Process Design and Development
Volume 14, April 1975, Page 101 (c)
Copyright 1975 by the American Chemical Society and reprinted by permission of the copyright owner
612
-------
Pressure Drop and Collection Efficiency of an Irrigated Bag Filter
Tetsuo Yoshida, Yasuo Kousaka,* Shlgeo Inake, and Shlgeyuki Nakal
Faculty ol Engineering, University ol Osaka Prefecture. Osaka. Japan
An irrigated bag filter has been developed to improve performances of existing dry bag filters. Irrigation to
filter surface by spraying or overflowing water prevents filter media from firing in handling hot gas and
makes it possible to wash away the precipitated dusts from filter surfaces. Some characteristics regarding
pressure drop and dust collection of an irrigated filter which were quite different from dry ones were
studied, and then basic mechanisms of them were discussed. A series of studies suggested that this kind of
collector will be useful in certain industrial fields.
Introduction
An irrigated bag filter described here is quite different
from existing dry bag filters, because the surface of filter
media is covered by water. It has been reported that an ir-
rigated bag filter can treat high-temperature and highly
humid gas, and that sweepage procedures of dust cake
necessary for a dry filter are not needed because of water
falling along the filter surface (Minami, et a/., 1969). It
has been also reported that the relation between pressure
drop and gas flow rate is peculiar compared with that of
dry filters (Muhlrad, 1970) and that collection efficiency
is fairly high (Minami, et al., 1969).
In this paper, pressure drop and collection efficiency of
inifated bag filters were tested and.their basic mecha-
nisms were studied by using nets of standard wire meshes
instead of bag cloths.
Experimental Section
One of the experimental apparatuses used in this study
is shown in Figure 1. Water is supplied along the inside of
a ring dam to the top of the bag cloth. Gas flows out from
the inside of the bag cloth just contacting with falling
water in the manner of crossflow. Superficial filtering gas
velocities were varied within 20 cm/sec and water rates
were from 2 to 20 l./min. The dust particle used was
CaCOs having a median diameter of 3.6 M (in weight base)
and concentrations at the inlet were from 2 to 8 g/m9.
Some physical properties of bag cloths are shown in Table
I.
Pressure Drop. Figure 2 shows the comparison of the
pressure drop of irrigated bag filters and that of dry ones
when they are clean. As is shown in Figure 2 the charac-
teristics of pressure drop considerably differ from each
other. Figure 3 indicates the same comparison but with
dust loads. Because of washing action against deposited
dusts by the down stream of water, almost no pressure
rite occurred in irrigated bag filters. For certain dusts,
however, which contain some tar substances, the pressure
drop increased with operation period.
Collection Efficiency. Figure 4 indicates the collection
efficiency of irrigated bag filters. The collection efficien-
cies seem to be correlated to the pressure drops as shown
in Figure 3. Although other experimental conditions of
various water rates ranged from 2 to 10 l./min and those
of superficial gas velocities from 1.5 to 8 cm/sec were also
examined, almost no differences among them were found.
Discussion
In this section, some basic mechanisms of pressure drop
and dust collection of an irrigated filter are studied by
using nets of standard wire meshes instead of bag cloths.
Pressure Drop. Figure 5a indicates a model of the
mesh over which a water film covers. The equilibrium of
force is given as follows when pressure difference exists
between the two sides of the film.
cos
or
AP = 4ff.«/Dldrc
(1)
(2)
The pressure difference AP gives the critical one at which
the film is just broken. In existing wire meshes, because of
a three-dimensional structure shown in Figure 5b, the di-
rection of the force of surface tension varies with positions
of a mesh. Then this factor was included in { in eq 2. The
coefficient, f, however, must be constant when the mate-
rial of the meshes, the manner of weaving, and liquid, re-
spectively, are the same. Equation 2 indicates that the
pressure drop to break a film is inversely proportional to
the opening size of a mesh.
An irrigated net of wire meshes whose openings have a
size distribution is next discussed. When the pressure dif-
ference between both sides of the net is gradually raised,
the film covering over a mesh with the maximum opening
size, in this case, will be broken first because of the mini-
mum pressure to break it as shown in eq 2. Subsequently,
with a slight pressure rise, the film over a mesh with the
next larger opening size is then broken. Thus films are
broken in order of their opening sizes as pressure rises.
When the film over the mesh having a hydraulic diameter
of Dm of the ith size is just broken at the pressure of
AP, = 4o.«;/£>H*?c > *P\ <3)
gas must flow out through the opening with the velocity of
ui to keep pressure drop in AP/. The pressure required to
break the t'th film against the force of surface tension is
caused by the resistance of gas flow through the openings
over which films are already broken. Then
This equation indicates the flow resistance on the ith
opening, and on other openings there must be the fol-
lowing relations
Ind. Eng. Chem., Process Oes. Dev., Vol. 14, No. 2, 1975
613
-------
Table I. Properties of Filters Used
Resistance Hydraulic
Filter cloth Fabric coefficient, t/m mean radius, cm
A Teviron
B Tetoron
C Tetoron
Filament, plain 0.432 x iOr 76.9 x 1Q-4
Spun, plain 2.10 x 101 48.5 x IQ-*
Spun, plain 1.48 x 10' 65.8 x 10'4
D Kanekalocv Spun, plain 2.6* x 1C7 64.1 x 10-*
E Saran
FLOW 3
METER T
I
t
W4.TE*
INLE1
f
.J
•pn
~~
<
Filament, satin — —
/"
r /
t \ f
J|«o"»
\ \
It
,'BAd
5 J FUER
r
iiJ-
J -f
Lui^i
«M
3Z-
VC
Xr-
J*C
JT^J
,LLB«n
V MS J*0
v
_. £
0.160
S
o
S 120-
3
•
f to-
JJ 40
-i 6*5
r*" •
f
• 0
1*
Porosity
6.07 x 10-J
4.46 x 10-'
7.67 x ID'1
4.47X10-1
—
• FILTER *
• C
• 0
• E
OR*
IRRIGATED
»
i; SUPERFICIAL GAS-
f VELOCITV 6crrV»#c
i
k
i^ » °~ "..-••""
• --•"""'"
A_> *"
Figuie 1. Experimental apparatus.
200
i « n is 20
SUPERFICIAL CAS VELOCITY u cmfs*C
Figure 2. Comparison of pressure drop of irritated and dry
filters with no dust load.
T£TV^ (5)
When meshes with a narrow size distribution, such as
standard wire meshes, are concerned, each cat velocity
through openings is expressed aa
r, = »
OUST LOAD
20
Figure 4. Comparison of collection efficiency of irrigated and dry
»>•* beg filters.
When all of openings may be regarded as almost square,
At is represented as
AI f
'HI
2 _
(8)
(7)
The above equations give the relation between superficial
gas velocity and pressure drop, when the size distribution
of meshes, two-dimensional porosity of a mesh, surface
tension of liquid, and its correction factor £ are known.
When A, is first assumed, then CHI, AP,, v,, and 0, are
given respectively by using eq 8, 3, 4, and 7. Repeating
the same procedure for various A,, the correlation of AP
and 17is obtainable.
The experimental apparatus to test the above analysis
was essentially similar to that of Figure 1, but a net of
standard wire meshes was installed instead of a bag cloth
and the sice of the apparatus was about half of that of
Figure 1. Table n shows some physical properties of wire
meshes used in the experiment. The dust particle used in
Ind. Eng. Chem.. Process Oes. Dev., Vol. 14. No. 2. 1975
614
-------
Table II. Properties of Standard Wire Meshes
Wire mesh
JIS 500
M60
JIS 149
JIS 74
Hydraulic jnean
Fabric diameter, DH, mm Porosity,
1150
o
olOO
T 50
a.
a
**~aa o o o o o
^» JIS 7*
» :
JIS U9 '
M 60
^^" O O O w
. " 0 E»PERi"ENTAL .
f^ — CAlCUUTEO
« O
./o" • • J. f ^ .^ .
; / / / <'' **''„•''* is soo
SUPERFICIAL
VE.OC.TV
Q cnwttc
Flfura 7. Comparison of experimental and calculated niulti of
piettun drop.
10
a
%
6
•4
•
: 2
n
rf
"hi
as 0.75 1 i» «i10-j
4t CsecJ
Figure 8. Interval of water film being kept broken (JIS 500),
speed camera, is shown in Figure 8. This sudden opening
and closing action of an orifice, that is shutter action of
an orifice, may be considered as another mechanism of
collection.
Collection by an Orifice in an Infinite Plane. The col-
lection efficiency by an orifice in an infinite plane can be
obtained when the stream line around the orifice and then
the trajectory of a particle in the stream are calculated. In
order to estimate the collection efficiency, potential flow
for stream line and Stoke's law for drag of a particle an
hen assumed. The stnam function of ideal gas around an
orifice in an infinite plane is given by (Lamb, 1932)
(9)
The gat velocitiet are also Riven as
' " jb Jf c'*>i
v, m t- §ln (to) dfr (10')
When the stream is thus given, trajectories of a particle in
ind, Ing, Chenv, Process 0»s Dev,, Vol. 14, No 9, 1976
615
-------
the stream are computed as follows by assuming Stokes'
law and no disturbance in the stream by the panicle. The
dimensionless equations of motion of a particle are ex-
pressed as follows when any external forces including
gravity may be ignored
particle
(12)
where
f
~ a
x_
a
At a point far away from the orifice, the velocity of a par-
ticle is assumed to be equal to that of fluid, so the initial
conditions are
= 0;
(13)
(13')
The trajectories of a particle for various values of * are
obtained by performing numerical calculation of above
equations. Figure 9 illustrates the relation of a stream line
and a trajectory of a particle. The collection efficiency of
a particle of dp in diameter is consequently given as the
ratio of the volumetric gas flow rate Qp to the total flow
rate Q as shown in Figure 9. Figure 10 indicates the result
of calculation of the efficiency defined above. The figure
suggests that the interceptional collection is important in
this case, and so collection efficiency cannot be expected
to be too great unless the interceptional parameter is suf-
ficiently large.
Collection by Shutter Action of an Orifice. One may
suppose a case where a shutter which covers an orifice is
suddenly opened and then gas flows out through the ori-
fice. In this case, if dust particles are contained in gas,
dust-free gas only may flow out at first while it takes some
instants to accelerate the dust particles. When the orifice
is recovered by a shutter in the next instant, the particles
go straight ahead to be caught to the shutter. If the inter-
val of the shutter being kept open is short enough, a fair
contribution to dust collection is expected.
The analysis of this mechanism of collection may be ac-
complished by calculating the unsteady particle motion in
an unsteady velocity field of fluid around the orifice. It
will be difficult, however, to estimate the unsteady veloci-
ty profile of gas around the orifice which is confronted
with the sudden opening and closing action. It is assumed
here that a steady velocity profile of gas may be instanta-
neously built up at opening and that the flow may also be
instantaneously stopped at closing. This assumption may
be valid only for rough estimation of the extent of the col-
lection efficiency described above. Under the assumption,
the collection efficiency of sudden closing of the orifice
may be defined as follows (see Appendix)
n. =
Q,t,
(14)
In the equation o>top represents the volume surrounded by
a stopping distance shown in Figure 11. The stopping dis-
tance of a particle with dp in diameter and pp in density
is given by
_ <*.'*>."»
(15)
Flffure 9. Illustration of collection efficiency of an orifice in an in-
finite plane.
1.0
o
'C
u.
,
-------
o.
fu
5
8 .
PARTICLE OIA : I micron
DENSITY' 1.1 gxn/
OPENINO INTERVAL
ORIFICE 01*- 0»'
"0 10 20 30 40
ACTUAL VELOCITY THROUGH ORIFlC HVHC
Figure 12. Collection efficiency when shutter being just closed.
RANGE OF ACTUAL VELOCITY
THROUGH ORIFICE
lOi-^jl.5 mine
FILTER JIS 900
SO 100 ISO
SUPERFICIAL GAS VELOCITY
200
cnuwc
Figure IS. Collection efficiency obtained by standard wire mesh
experiment.
Conclusion
The manner of change in pressure drop with gas veloci-
ties of irrigated bag filters was first tested and was found
to be very peculiar compared with the dry ones. The
mechanism of pressure drop was then analyzed from the
equilibrium of forces such as surface tension and static
pressure of gas, and the result of the analysis was found to
agree well with that of experiments using a net of stan-
dard wire meshes. The slight change in pressure drop over
the wide range of gas velocities seemed to be one of the
interesting characteristics of this type of filter. It was also
found that almost no pressure rise occurred after a long
operation for dusts not containing tar substances.
The collection efficiency of an irrigated bag filter was
found to be fairly high. It was impossible to give a full ex-
planation of the experimental results because of the com-
plexity of collection mechanisms. However, two mecha-
nisms of collection, one by an orifice in an infinite plane
and another by shutter action of an orifice, were pointed
out by a simplified analysis in the case of standard wire
meshes being irrigated.
Acknowledgment
S. Magono, K. Yamadaki, T. Yasumune, and Y. Adachi
were very helpful in the experimental work.
Appendix
The material balance before and behind the collector
with orifices having shutter action is expressed as
where ci, Co, and cm represent respectively the dust con-
centration at inlet, outlet, and at collection chamber
shown in Figure 14. Q represents the volumetric gas flow
rate, and Q»top the total volume of rg = force by surface tension shown in Figure 5, kg/
sec8
-------
(Reprinted with permission)
[CJ JOURNAL OP CHEMICAL WGWERING OP JAPAN
(0«Mi) ft*
/ '
licatioj^* Utter to the Editor
(«>*•>
GROWTH OF AEROSOL PARTICLES BY STEAM INJECTION
*»«•«
x-f *•
jt tt« <«*. ***>
TETSUO YOSHIDA, YASUO KOUSAKA, KIKUO OKUYAMA
AND FUMINORI NOMURA
Department of Chemical Engineering, University of
Osaka Prefecture, Sakai 591
Kmivd
^rmpoadtnc* eonecmlnf UUt mMe *ould be xidrawd to f. Kousaka.
F. Nomura is now with Kanebo Co., Ltd., Osaka
(ft*)
» M
TKU
IN "»
V6L . No.
(19 )
i! A)
(ft*M
«
M«
1
ii
ft
R*
* Jt
(^fM
7
A
Tikta «
« 0 »
« >
1 )
( »
• )
'J
3
i
• M
rn» •
1 • 7 « 7 »
I I
( 1
- 1 >
• ( 1
M
V «
n» N
10 »
« . »
i »
i »
< >
#
K
• 7»
( i
l i
I l
( l
1
« I; "» *»-!>•
'.a.B'-^r.f,,
( » t-K
( » «•-*»
( )«-v
( Jt-i*
• 4
618
-------
Introduction
It has been reported that steam injection into scrubber
systems improves dust collection performance1"*5*. Two
mechanisms have been suggested for this improvement : (1)
condensation of water vapor upon dust particles, which
increases particle size to improve inertial dust collection
at scrubbers, (2) deposition of particles on condensing
surface by the Stefan flow. The first mechanism, which
remains unevaluated by theory so far, was first studied
theoretically in this paper. The analysis was then examined
by the experiment where the ultramicroscopic technique for
droplet size analysis previously developed by the authors
was applied.
1. Estimation of Particle Growth
When steam is injected into dust free air, a certain
degree of supersaturation will be produced according to the
condition and the quantity of both steam and air. Figure 1
illustrates the change of air initially in the stage of
"g" to a supersaturated state "i" by steam injection on a
humidity chart. If some particles as condensation nuclei
were introduced into the supersaturated air, condensation
of water vapor upon the particles occurs to decrease
supersaturation. This change was illustrated in "i-f line
in the figure. In actual case where some condensation nuclei
619
-------
or aerosol particle*exist in air before steam injection, the
change like "i-f" will not take place because condensation
occurs before the point "i" in Fig. 1 is attained. Thus the
point "i" is imaginary, but even in such a case the amount
of the condensable water vapor per unit mass of dry air will
be coincide with the value, AH, shown in the figure if the
process is adiabatic. The value of AH and the equilibrium
temperature T _ of the system after steam injection will be
determined by the following enthalpy and material balances
on dry air basis:
v Hstl* A:t + (l - *> ^tl- v + vi - * > *if + * ^(1)
Hg + Hst x " Hsf( Tsf > + m (2)
The fianl state of air was regarded saturated in the above
equations because the vapor pressure at the surface of grown
particles having diameter over 0.1 micron, as is predicted
by the Kelvin's equation, is nearly equal to that of
saturation.
The calculated value of AH against the quantity of
steam injection per unit mass of dry air was shown in Fig. 2.
It will be noteworthy that AH considerably decreases with
the increase of air temperature. The temperature at the
equilibrium state after steam injection, T0f, was also shown
in Fig. 3. The steam condition in these calculations was
taken as that of 100°C in temperature, 1 atm. abs. in
pressure and 1.0 in dryness fraction. When the dryness
fraction x is less than unity, the value of AH decreases
620
-------
with the decrease of x shown in Fig. 4. When supersaturated
steam is injected into air, AH will decrease with the degree
of superheat of the steam.
Whlie AH represents the quantity of condensable water
vapor per unit mass of dry air as described before, the
following relation must be satisfied when all of the vapor
corresponding to the amount of AH are assumed to condense
upon particles contained in air
H-7< Dvf -Dvfi>V.
H7DvfnO for Dvf»Dvi and "s^1 (3)
where n_ represents the particle number concentration of
aerosol on dry air basis. D . and D . represent the volume
mean diameters of the particles before and after steam
injection respectively, or before and after growth of parti-
cles. The increase in size of aerosol particles by steam
injection can then be evaluated in volume mean diameter Dyf
knowing the value AH from Eqs. (1) and (2), and the value n .
The line in Fig. 5 shows the relation of Eg. (3).
2. Experimental Method
Figure 6 shows the schematic diagram of the experiemntal
method. Air having T in temerature and H in humidity was
9 9
introduced into an 1 inch insulated pipe and then the aerosol
particles, tobacco smoke in this study, were dispersed into
the air stream. Steam having 100°C in temperature, 1 atm abs. in
621
-------
pressure and dryness fraction of nearly unity was injected
at the point of 2 meter down stream from that of the particle
dispersion point. The air flow rate was about 180 1/min and
the quantity of steam injection was ranged from 0.05 to 0.5
gram steam/gram dry air. The particle number concentration
and size distribution of the aerosol particles at the point
of 1.5 meter down stream from the steam injection point were
observed by the ultramicroscopic technique . The observation
cell used in this study was the same as that used in the
previous paper , which was composed of a double tube to
prevent the change in size of grown particles or water
droplets due to the temperature change. When the temperature
of the aerosol to be observed was high, the cell was further
surrounded by a cover into which air with controlled
temperature was blown.
3. Experimental Results and Discussions
Figure 7 shows one of the experimental results of
particle size distribution of the grown particles together
with the initial size distribution of tobacco aerosol
particles. It will be found that the considerable increase
in size occurs by steam injection. The width of size distri-
bution seems to become neither narrower nor wider. The
similar results were also found in the previous work where
the particle growth was promoted by mixing hot saturated
air with cold saturated air7). Furthermore, as found in the
622
-------
previous work, the rapid growth rate was also suggested in
this case judging from the fact that the residence time of
aerosols in the pipe was very short.
The volume mean diameters of grown particles obtained
by experiment were plotted in Fig. 5. In spite of the
difficulty in measuring the size of grown particles or
water droplets in high temperature, the experimental results
agree with the estimation line.
Conclusion
The effect of steam injection into dust-laden air on
the increase in the size of dust particles was studied. The
procedure of the estimation of AH, the amount of condensable
water vapor per unit mass of dry air by steam injection,
which is effective for particle growth, was made clear for
various conditions of air and steam injected. By using the
value of AH, the increase in size of aerosol particles
having number concentration of n^ was then evaluated in
volume mean diameter. These analysis was verified by
directly measuring the size of grown particles using the
ultramicroscopic technique previously developed by the
authors.
The results suggested that the particle growth by
•team injection will be one of the most promising pre-
conditioning technique to improve •crubber performance,
•ipeciallv for an exhauit gas having low temperature.
€23
-------
Acknowledgment
K. Yamadaki was very helpful in the experimental work
Nomenclature
D ., D .
VI Vf
H
Hst
AH
st
^f
no
T
x
p
Subscripts
f
g
i
s
sf
st
= particle diameter
= volume mean diameter before and after growth
[cm][p]
= absolute humidity
= quantity of steam injection
= condensable water vapor
= enthalpy
= enthalpy
= enthalpy
= particle number concentration
= temperature
= dryness fraction of steam
= density of condensed liquid
= final state
= initial state of air
= initial state
= saturated
= saturated air in final state
= steam
[g H20/g dry air]
[g steam/g dry air]
[g H20/g dry air]
[cal/g dry air]
[cal/g steam]
[cal/g water]
[1/g dry air]
[-1
[g/cm3]
624
-------
wf = water in final state
Superscripts
- dry
1 = wet
Literature Cited
1) Calvert, S. and N. C. Jhaveri : J. Air Pollu. Control
Assoc., 24, 946(1974)
2) Fahnoe, P., A. Lindroos and R. J. Abelson : Ind. Eng. Chem.,
43, 1336(1951)
3) Lancaster, S. W. and W. Strauss : Ind. Eng. Chem. Fundam.,
10, 362(1971)
4) Lapple, C. E. and H. J. Kamack : Chem. Eng. Progr., 51,
No. 3, 110(1953)
5) Schauer, P. J. : Ind. Eng. Chem., 43, 1532(1951)
6) Yoshida, T., Y. Kousaka and K. Okuyama : Ind. Eng. Chem.
Fundam., 14, 47(1975)
7) Yoshida, T., Y. Kousaka and K. Okuyama : Ind. Eng. Chem.
Fundam., 15, 37(1976)
625
-------
Captions of Figures
Figure 1. Change in humidity and temperature by steam injection
Figure 2. Condensable water vapor AH against quantity of steam
injection H
Figure 3. Temperature T f at the equilibrium state after steam
injection
Figure 4. Effect of dryness fraction x of steam on condensable
water vapor AH
Figure 5. Relation between volume mean diameter of grown
particle D . and parameter AH/n , condensable
water vapor per single particle
Figure 6. Experimental method
Figure 7. Change in size distribution by steam injection
626
-------
to
3
-C
O
(A
temperature
Figure 1. Change in humidity and temperature by steam injection
-------
0.020
0.015
•o
CT
O
(M
tfO.010
0.005
relatve
humidity
100 '/.
50%
0.1 0.2 0.3 0.4 0.5
HstCgsteam/gdryairD
Figure 2. Condensable water vapor AH against quantity of steam
injection Hgt
628
-------
80
Tg=50*C
relative
humidity
100 */.
0 */.
• saturated point
I
0.1 0.2 0.3 0.4 0.5
H$tCg steam/g dry air
Figure 3. Temperature Tgf at the equilibrium state after steam
injection
629
-------
0.020
0.015
O
CM
50.010
0.005
relative humidity 100*/•
x =1.0
x = 0.75
x =0.5
0.1 0.2 0.3 0.4
HstCgsteam/g dry air D
Figure 4. Effect of dryness fraction x of steam on condensable
water vapor AH
630
-------
o\
u>
10
6
S-4
«*-
Q 2
I 'I
6 810"
tobacco smoke
j , 1 . I . I
key
n0C1/g dryaiO
5x!07~108
4 6 810"1U 2
rgH20/particleD
» '
4 6
Figure 5. Relation between volume mean diameter of grown particle
Dvf and parameter AH/ng, condensable water vapor per
single particle
-------
CTl
OJ
NJ
/Hst=0.05-0.5 g steam/gdry air *
|steamWc.1atm.abs,x* 1 9
.thermometer
air (insulated)
to blower
180 1/hiin
observation
celt
aerosol
partkles
ultramicroscope]
vacuum
pump [ VTR& monitor)
Figure 6. Experimental method
-------
95
u80
a,70
•Seo
S50
3 30
>20
310
I 5
11' i 'i i 'nr
4H=0.0107 CgH2(Vtldryalr3
cVgdryair]
-0
1)
£/
tobacco
smoke
0.6 1
Dp
46
Figure 7. Change in size distribution by steam injection
633
-------
EFFECT OF BROWNIAN COAGULATION AND
BROWNIAN DIFFUSION ON GRAVITATIONAL
SETTLING OF POLYDISPERSE AEROSOLS
TETSUO YOSHIDA, YASUO KOUSAKA, KIKUO OKUYAMA
AND SHIOERU NISHIO
Department of Chemical Engineering, University of Osaka Prefec-
ture, Sakat, 591
(Reprinted with permission)
Reprinted from
JOURNAL OF
CHEMICAL ENGINEERING
OF
JAPAN
Vol. 8, No. 2 (1975)
Paflei 137—142
-------
EFFECT OF BROWNIAN COAGULATION AND
BROWNIAN DIFFUSION ON GRAVITATIONAL
SETTLING OF POLYDISPERSE AEROSOLS
TETSUO YOSHIDA, YASUO KOUSAKA, KIKUO OKUYAMA*
AND SHICERU NISHIO
Department of Chemical Engineering, University of Osaka Prefec-
ture, Sakai, 591
The behavior of aerosol particles of sub-micron diameter undergoing Brownlan coagulatkm,
Brownian diffusion and gravitational settling was studied by numerically solving the equation of
population balance and by experimentally observing the change of aerosol properties with time.
In calculation, two dlmenslooless parameters, which are determined by the initial properties of
aerosols and vessel dimension In which aerosols are suspended, were introduced to evaluate the
magnitude of influence of Brownian coagulation and Brownlan diffusion on gravitational settling.
The results of numerical calculation were indicated by graphical representation, which is
usable for quantitative estimation of the magnitude of these two effects. These computational
results were found to be in good agreement with the experimental results obtained by an ultra-
microscopic technique of particle size analysis.
Introduction
The behavior of aerosols is influenced by coagula-
tion, diffusion, sedimentation, the rate of generation
of particles, particle characteristics and the dynamics
of the fluid in which particles are suspended. For
aerosol particles of sub-micron diameter, coagulation
by Brownian motion and deposition by Brownian
diffusion as well as gravitational settling are essential
for characterizing the behavior of aerosols41.
As basic research in the size distribution which
changes with time undergoing Brownian coagulation
and gravitational settling, a few analytical solutions'-"
and numerical solutions1''1'1" have been obtained
under some simplified or particular conditions. As
one of the representative studies G. C. Lindauer et a/."
and C. H. Ahn et a/.11 showed the effect of Brownian
coagulation on gravitational settling by a numerical
method, but they included the unrealistic assumption
that particle concentrations are always uniform in the
direction of settling.
The purpose of this paper is to discuss the effect of
Brownian coagulation and Brownian diffusion on
gravitational settling considering the variation of
particle concentrations in the direction of sedimenta-
tion. The equation including these three effects has
been solved numerically on the assumption that the
Received July 6, 1974.
Presented at the 39th Annual Meeting of The Soc. of Chem.
Engrs., Japan, at Kobe, April 4 1974 (entitled "Effect of
Brownian Coagulation on Gravitational Settling of Poly-
dbperse Aerosols".)
VOL 8 NO. 2 1975
initial particle size spectrum is of log-normal form.
In calculation, two dimensionless parameters were
introduced to estimate the amount of the effect of
Brownian coagulation and of Brownian diffusion on
gravitational settling. The calculated results were
arranged in graphs to predict these two effects quanti-
tatively, and some of them were compared with ex-
perimental results obtained by an ultramicroscopic
technique11'.
Theoretical Calculation
Basic equation
Consider an aerosol located between parallel hori-
zontal walls as shown in Fig. 1 on which aerosol
particles are deposited by Brownian diffusion and by
gravitational settling accompanying Brownian coagu-
lation. The basic equation expressing the time-
dependent variation in particle number concentration
of aerosols can be given by the population balance of
the element Ay' (refer to Appendix)9*1". The basic
ry'y ///// .
o Brownian dHfution 6
0 °
t *t . »
fr 9 "I"
production »; Owpolion _,_ ]
o—o—~0~ o-—-Q f
Brownian coagulation }
O gravitational stilling
~> I
Fig. 1 Behavior of aerosols between two horizontal walls
635
-------
equation written in dimensionless form becomes
3t
- _
* C.(r,0)
+ CG
f=pf, 0
, 0
/=/min ..... /max (1)
where
<, t)=n'(ri r')/no,
t'/H, y=y'/H,
(2)
nr_
3* r
2D(r,,K\ 7 M((r,0)n0 \
//' A' "H )
(3)
^•'i ,°<) is the coagulation function of particles, and
in the case of Brownian coagulation K(r-, p'J is given
by101
r-)/r- + Cm(p't)lpt],
(4)
In Eq. (1), the first term accounts for Brownian dif-
fusion, the second for gravitational settling, and the
third represents the number of newly formed particles
with radius r< by collision of two particles, while the
last term represents the decrease in number of par-
ticles with radius r, by collision with other particles.
CG and DC are dimensionless parameters which
can be evaluated from initial aerosol properties and
physical conditions. As seen from the definition, the
parameter CG means the ratio of Brownian coagula-
tion rate to deposition rate by gravitational settling,
Another parameter DG, which was proposed by
C. N. Davies" to evaluate the concentration change
of monodisperse aerosols undergoing diffusion and
gravitational settling in the non-coagulation field,
describes the deposition rate ratio of Brownian dif-
fusion to gravitational settling. Though the param-
eters CG and DG are based on an initial geometric
mean radius, they are convenient parameters to
predict the influence of Brownian coagulation or
Brownian diffusion relative to gravitational settling.
In the derivation of Eq. (1), the following assump-
tions were made:
(1) There exist no external forces except gravity and
the medium is in stationary state with no convection.
(2) Particles are spherical and electrically neutral.
(3) Panicles collide with each other to form single
new spherical particle whose mass may be the same
as the combined mass of two smaller particles.
(4) AH particles colliding with two horizontal walls
are caught by them.
The initial particle size spectrum is assumed to be
of log-normal distribution and to be generated in-
stantaneously with spatial uniformity; then
forO<^0 (6)
If the coefficient Cm(r<')/CB(r,0) on the right-hand
side of Eq. (1) is normalized about r,0, solutions de-
pend on <7,0 and two dimensionless parameters CG
and DG.
Finite difference approximation and calculation method
Since Eq. (1) is a nonlinear differential equation, it
cannot be solved analytically. It is approximated by
using central difference formulae accurate to second
order, and then the finite difference equation is given
as follows:
j «fr«.^b-"(r
( Ay
(rj/^i _ _
S KVri-ri.pMVrl-p^y.t)
n^'imln
(f. \* '(mi*
3/rS-'-;r)- E *<>«,/>()
V^i—Pi / fi=rimtu
xn(r4,y,t)n(p(,y,t){, /=/min ..... /max (7)
Calculation was started by assigning initial values of
particle number concentration by Eq. (5) at every in-
terior point between two walls. Being stepped by
dt, new values of particle number concentration were
calculated by Eq. (7), and this step was repeated until
particle number came to zero.
Calculation Results and Discussion
It can be considered that the main factors which
influence the time dependence of particle size distri-
bution are initial properties of particles such as geo-
metric mean radius, number concentration and geo-
metric standard deviation. Therefore solutions were
obtained for various values of them. The results are
shown in Table 1, together with the values of CG,
DG and computational parameters.
The effect of Brownian diffusion
According to the definition the parameter DG in-
creases as particle radius decreases, and consequently
the effect of Brownian diffusion becomes large. Fig. 2
shows the time dependence on dimensionless number
JOURNAL OF CHEMICAL ENGINEERING OF JAFAN
636
-------
Table 1 Coodltloof for calculation
fit
M
0.1
0.2
0.3
0.5
1.0
ft
1.2
1.2
1.2
1.2
1.5
1.2
no
[particles/cc]
10*
4.0x10*
10*
9.0x10*
10'
10"
10*
10'°
10'
10'
10"
10*
10"
10*
10'
10"
10*
10"
CG
[-1
3.10x10-'
3.10x10-'
3.44xlO-«
3.10x10-'
3.44x10-'
3.44x10°
3.44x10'
3.44xlO»
1.24x10-'
1.24x10-'
1,24x10°
1.24x10'
1. 24xlO«
3.10x10-'
3.10x10-'
3,10x10-'
3,10x10°
3.10x10'
DC
l-l
3.94x10-'
4,92x10-'
1.46x10-'
3,15x10-'
3.94x10-'
Jlnr—0.077, 4y=0.01, 0.02, J/=0.01, 0.002, 0.0004
//=0.25cm
Fig. 2 Distribution of aerosol concentration undergoing
Brownian coagulation, Brownian diffusion and gravitational
settling
concentration of paricles n at each height under con-
stant CG. When coagulation and diffusion occur,
that is CG>0 and DG>0, more rapid decrease of
particle concentrations is seen in the figure in com-
parison with those of gravitational settling (CG=0
and DG=0). When the values of DC are small
enough, the difference from gravitational settling
depends mainly on Brownian coagulation, and then
particle concentrations are determined by the values
of CG and o-|0. It may be seen from Fig. 2 that in
the case of />G<0.004 the effect of Brownian diffusion
VOL. 8 NO. 2 1975
(a)
(b)
Fig. 3 Effect of Brownian coagulation on
gravitational settling
seems to be negligible, which almost agrees with
Davies' analytical solution" obtained for monodis-
perse aerosols undergoing Brownian diffusion and
gravitational settling. So far as aerosol particles of
more than several tenths micron in diameter arc con-
cerned, the effect of Brownian diffusion may be
negligible. In the following section, the effect of
Brownian coagulation on settling aerosols in the
absence of Brownian diffusion will be discussed.
The effect of Brownian coagulation
The curves of concentration change with time for
various values of CG are shown in Figs. 3 (a) and (b).
As the values of CG increase the decrease of particle
concentration proceeds more rapidly because of the
loss of particles by coagulation and because of en-
hanced settling velocity due to the growth of particles
by coagulation. Moreover, for a constant value of
CG, the discrepancy in concentration change from
that of gravitational settling (CC=0) increases with
time and settling depth. This is also due to the loss
and the growth of particles by coagulation which
637
-------
objective ol
microscope |n
Fig. 4 Dependence of aerosol concentration on settling
depth
3 468 10 J 4 6 B 10 2 4 6 « 10° 2 4 6 «
(b) --
y
W=l-
(8)
Thus the concentration change of aerosols is arranged
as shown in Figs. 5 (a) and (b). It may be seen from
these graphs that at values of CG-y below 0.02 the
effect of coagulation may be ignored and the change
in concentration of particles depends mainly on
gravitational settling. With the values of CG-y, the
curves have more gentle slopes than that of gravita-
tional settling only. At values of CG-y above 20.0
the concentration change is dominated only by co-
agulation, almost regardless of the initial standard
deviation. Figs. 5 (a) and (b) for two initial geo-
metric standard deviations enabled one to predict the
concentration change of aerosols having various
initial size distributions.
Experimental Apparatus and Method
In this study it is necessary to measure accurately
the variation in number concentration of aerosols
over short periods of time. For this purpose, an
ultramicroscopic technique"1 was used. The ob-
servation cell installed on the stage of an ultramicro-
scope, as shown in Fig. 6, has a small sectional area
to prevent the effect of thermal convection. Aerosols
used in this study were stearic acid particles and tobac-
co smoke. Stearic acid aerosols were generated by a
La Mer-Sinclair type generator. Tobacco smoke was
generated by a simple apparatus'" by which number
concentration was controlled from 10' to 10'par-
ticles/cc. The experimental procedure is as follows.
After the focus of the ultramicroscope is preliminarily
set at the depth h shown in Fig. 7, aerosol is intro-
duced into the observation cell and the flow of aerosol
is instantaneously stopped by closing the valves.
From that moment aerosol particles existing in the
volume v*, which are recognized because of their
JOURNAL OP CHEMICAL ENGINEERING OF JAPAN
638
-------
TaWe } Experimental condition.
Klndi of aeroioli: itMric acid parllclM, tobacco imoke
Initial geometric mean radlut: r,o-0.3 ^-0.5 /<
Initial geometric tUndard deviation: «,o-l.2~l.4
Denilty of particles: p,-0.85g/cm' (iteailc add partlcl«)
•0,78 g/cm1 (tobacco imoke)
Concentration of aeroioli: m-10*~ 10* p»nlclei/cc
A: 200/<~1500^
v. : 1.4x10-•cm* (eyepiece x objective. 10x20)
1.2xlO*«cm' ( » , 10x40)
CO; 0.03-1.7
DC: 4.7xlO-«~ 1.7x10-'
shining, are recorded by a video recorder, until all
particles disappear in the sight. The number of
particles at any given time are counted by reproduc-
ing the recorder.
Initial number concentration of particles, n0i >s
given as follows by the initial particle number of
images N (0) and the observation volume vm:
n,=N(Q)lvm (9)
The particle size spectrum was also determined by
the ultramicroscopic technique developed previously
by the authors'", whose principle is almost the same
as that of the Andreasen-Pipette method. Experi-
mental conditions and physical properties of aerosols
are shown in Table 2.
Experimental Remits and Discussion
Fig. 8 shows a comparison of the relative con-
centration change with time between experimental
data and theoretical curves. In this comparison the
decreases of the experimental values seem to be slightly
slower than those of the predicted ones. However,
overall agreement is good within experimental error.
As seen from the values of CG and DC, the effect of
Brownian coagulation and diffusion on particle con-
centration can be almost ignored. This agreement
indicates that convection, photophoresis and thermo-
phoresis do not occur in the observation cell.
Fig. 9 shows the experimental data of the relative
concentration change on tobacco smoke and stearic
acid particles, together with corresponding theoretical
curves. Experimental data are in good agreement
with the tendency of the calculated curves, and the
effect of Brownian coagulation increases as the values
of CG-y increase. This kind of experimental data
has not been reported because of the difficulty in ac-
curate measurement of changing number concentra-
tion of particles with time. In this experiment the
values of CG-y were rather small, but the effect of
Brownian coagulation on gravitational settling was
clearly found to be characterized. These experi-
mental results might indicate, in a certain sense, that
the conception used in deriving the basic equation,
above all the assumption that particles stick together
upon impact and will not repel, are basically correct.
VOL I NO. 2 1975
260
Fig. I Comparison of
concentration with calculated one*
Fig. 9 Comparison of experimental aerocol
concentration with calculated ones
One more sensitive test of the coagulation theory
could be made by the change of concentration of each
species of particle size distribution.
Conclusion
The behavior of aerosols undergoing Brownian
coagulation, diffusion and gravitational settling be-
tween two horizontal walls was studied theoretically
and experimentally.
The basic equation considering the above three
effects and the variation of particle concentration in
the direction of sedimentation was solved numeri-
cally for various initial particle size spectrums. The
overall effect of diffusion or coagulation on particle
concentration of settling aerosol particles was esti-
mated using the values of two dimensionless param-
eters DC and CG respectively. At values of CG
less than 0.02, the effect of coagulation was then
found to be negligible, and in the regions of DG<,
0.004 the effect of diffusion could be ignored. In
the absence of Brownian diffusion the change of
particle concentration with time at every height could
639
-------
be estimated by introducing a new dimensionless
parameter CG-y. Some of these theoretical results
were examined by a technique of ultramicroscopic
size analysis and were found to be in good agreement
with experimental ones.
The calculation results presented in this paper are
an elementary step in providing a theoretical approach
to predict the behavior of aerosols in a closed vessel
and also will be useful to estimate the influence of
Brownian coagulation or diffusion on particle size
analysis of aerosols by sedimentation methods.
Appendix
The basic equation expressing the lime dependence of the size
distribution of aerosols undergoing Brownian coagulation,
Brownian diffusion and gravitational settling is given4-*-'" by:
7*1
r'*t
^J:
r'*-p'*,n
r*r-,'» )'<">'
-\ K(.r',P-)n'(r',i-)n(p'.t')dp-
where D(r') is the diffusion coefficient and u,(r') is the terminal
settling velocity.
Acknowledgment
T. Miyazaki and Y. Kida were very helpful ia the experimental
work.
Nomenclature
CC
C.(ri')
D.
DC
g
H
h
K,
K(r<', pt'
k(f i, Pi)
Jlnr
= dimensionless parameter defined in Eq. (3)
= Cunningham's correction factor of n' [—]
-= effective diameter of microscopic sight
shown in Fig. 7 [cm]
= dimensionless parameter defined in Eq. (3)
= diffusion coefficient
— acceleration of gravity [cm/sec1]
= height between two horizontal walls [cm]
= values shown in Fig. 7 H[cm]
= coefficient in Eq. (4) (cm'/sec]
— coagulation function for two particles of
sizes n' and pi' [cm'/sec]
= dimensionless coagulation function
n'(n't
no
JV(0)
r', T
r,o
T
i', l
Jr
ut(r\)
f.
y', y
I1), n(ri, t) = number and dimensionless number
of aerosol particles (=/i'(iV, r')/«o)
[particles/cc][— ]
— dimensionless total number of particles
(-ZXK.O/no) [-]
= total number of particles at time
zero [particles/cc]
= number of aerosol particles in Vm at
time zero [particles]
= particle radius and dimensionless
particle radius [;<][ — ]
= geometric mean radius at time zero [/<]
= absolute temperature (°K]
= time and dimensionless time [sec][ — ]
= dimensionless time step [ — ]
= terminal settling velocity
[cm/sec]
[cm'J
[cm][ — ]
= volume shown in Fig. 7
= vertical and dimensionless vertical
distance from top of the cell
mesh width and dimensionless mesh width
along the vertical direction [cm][ — ]
geometric standard deviation at time zero [ — 1
f =
P, PP
fti
i
min =
max =
Boltzman's constant (=^-1.38 x 10"")
viscosity of fluid
fluid and particle density
dimensionless particle radius
refers to the number particle size
minimum
maximum
[erg/°K]
[g/cm-sec]
[g/cmj]
dimensionless size width between n and r<+i [ — ]
Literature Cited
1) Ahn, C. H. and J. W. Gentry: Ind. Eng. Chem., Fundam.,
11,483(1972).
2) Davies, C. N.: Proc. Roy. Sac., A200, 100 (1949).
3) Friedlander, S. K. and C. S. Wang.: /. Coll. Int. Sri., 22,
126 (1966).
4) Fuchs, N. A.: "The Mechanics of Aerosols," Pergamon
Press, Oxford, (1964).
5) Greenfield, M. A., R. L. Koontz and D. F. Hausknecht:
J. Coll. Int. Sci., 35, 102 (1971).
6) Hauck, H.: Slaub-Reinhalt. Luft., 33, 1233 (1973).
7) Langstroth, G. and T. Giilespie: Cane/. /. Research, 25B,
455(1947).
8) Lindauer, G. C. and A. W. Castlcman, Jr.: Aerosol Science,
2. 85 (1971).
9) Mori, Y. and A. Yoshizawa: Kagaku Kogaku, 32, 1233
(1968).
10) Smoluchwoski. M.: fhys. Chem. (Leipzig), 92, 129 (1917).
11) Yoshida, T., Y. Kousaka and K. Okuyama: Ind. Eng.
Chem. Fundam.. 14, 47 (1975).
12) Zebel, G.: Kolhid-Z, 156, 102 (1958).
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
640
-------
CHANGE IN PARTICLE SIZE DISTRIBUTIONS OF
POLYDISPERSE AEROSOLS UNDERGOING
BROWNIAN COAGULATION
TETSUO YOSHIDA, KIKUO OKUYAMA, YASUO KOUSAKA
AND YOSHINORI KIDA
Department of Chemical Engineering, University of Osaka
Prefecture, Sakai. 591
(Reprinted with permission)
641
Reprinted from
JOURNAL OF
CHEMICAL ENGINEERING
OF
JAPAN
Vol. 8, No. 4 (1975)
Pages 317—322
-------
CHANGE IN PARTICLE SIZE DISTRIBUTIONS OF
POLYDISPERSE AEROSOLS UNDERGOING
BROWNIAN COAGULATION
TETSUO YOSHIDA, KIKUO OKUYAMA*. YASUO KOUSAKA
AND YOSHINORI KIDA
Department of Chemical Engineering, University of Osaka
Prefecture, Sakai, 591
The time-dependent change ID particle size distributions of highly concentrated polydisperse
idergoug Brownian coagulation was studied by numerically solving the basic equation
of coagnlation for various size distributions Initially having log-normal form. The results were
plotted la the forms of the change with tune In cumulative size distribution* and the changes in
nominal geometric mean radius, as well as standard deviation for various initial distributions of
aerosol*. These figures showed that size distributions approached certain asymptotic ones,
which might correspond to SPDF (self-preserving distribution function), almost independently
of Initial distributions as coagulation proceeded. The process of the approach to asymptotic
distributions was also made dear by the graphs. Some of these results were verified by ex-
perimental results obtained by the uftramlcroscoplc size analysis previously developed by the
authors.
Introduction
In the previous paper1", the change in particle
number concentration of polydisperse aerosols under-
going gravitational settling, Brownian diffusion and
Brownian coagulation was discussed, and Brownian
coagulation was found to be important when high
concentration aerosols were concerned. Analytical
solutions1-111, asymptotic solutions8'''8'1" and nu-
merical solutions*'"' have been reported for the
change in pariclc size distribution of polydisperse
aerosols with time undergoing Brownian coagulation.
These solutions, however, have been obtained under
some simplified or specialized conditions, and seem
unsatisfactory for understanding the general aspects
of time-dependent change in particle size distribu-
tions of polydisperse aerosols. In experimental
studies, because of the difficulty in accurate measure-
ment of particle size distribution of highly concen-
trated aerosol, sufficient amounts of available data
have not been reported'-"1.
In this paper, the change in particle size distribu-
tion of highly concentrated polydisperse aerosols was
studied by numerically solving the basic equation for
Brownian coagulation with various initial log-normal
size distributions. The results of the calculation were
Received November 6,1974.
Presented at the 8th Autumn Meeting of The Soc. of Chem.
Engrs., Japan at Tokyo, Oct. 8, 1974. {"Particle Growth of
For/disperse Aerosols by Brownian Coagulation")
VOL. • NO. 4 197J
graphically presented to show the effect of polydis-
persion on the change in size distribution of aerosols,
and then were confirmed by experimental results
obtained by the technique of ultramicroscopic size
analysis previously developed by the authors"'. The
study presented here gives one of the basic proper-
ties of polydisperse aerosols, and is useful, for instance,
for estimation of the extent of the contribution of
particle growth to industrial precipitators.
Theoretical Calculation
The basic equation for the time-dependent size
distribution change of aerosols undergoing Brownian
coagulation has been reported by Miiller*"' and its
dimensiohless form can be written as
r\
i=imin... .imax (1)
where,
K(r'(,p',) is the coagulation function of particles and in
the case of Brownian coagulation it is given by111
Wt,iQ=KJri+pMCJir't\lr't + CJipi)lp:} (3)
The left side of Eq. (1) is the change in particle con-
centration of size r, with time. The first term on the
642
-------
Ofl
Q6
f
nr
o?
1C
99
395
"90
• 60
£70
$60
350
**o
I*
|20
§10
5
'(
""""*'SS^
1 1 1 1 — i
^k
'"•fe
.
tobacco
k£V ritTpnrtirlcBf cl
steoric acid particles .'X
key rtffaiamia roo
-o- 3.55 * 10* a
cju Ogoc :
32 1.28
OOP particles
key rtCpariKMKd Ifco
* 1.99 «10« 0
<• 2.14 K 10« 0.
r2 2 «s
rSo=Q35u
ffgoC— 3
1.50
; "1
CJ13 C%0 C-3
37 122
37 ).22 i
• 769«H3*
e 1.05 > K)'
o 4.16 > 10'
» 7.38* 107
» 84 1 >10'
' «
T9»W
- 0.1
c« 0.3S
V, l.Q
810H 2 4 6 8IOC
t C-3
I'M •"//'}"
fjf'i •
I l''/T'l !
/////' / < / *
Wl
: ^////
:;w/
.' t/llfl / ' /I .'1 11
J.2 0.4 06 1
r
4
9
1
t
I
I
1
(
/
t
t
1
1
j
-
rA * 6 1°
2 4
(re
an
em
va
chi
of
me
r'
wa
ch<
a
cal
foi
smofat
^otf«
0.36
3.375
J51
>.64
3.465
<%OC-J
1.23
1.32
1.53
1.71
1.36
olculoted
dgot-
1.30
1 SO
Fig. 1 Experimental awl calculated
Dumber concentration
6 8 101 2
Equation (1) is approximately normalized on r^
fer to Appendix). Since Eq. (1) cannot be solved
alytically, the Runge-Kutta-Merson method was
ployed to solve it. To compare the effect of
rious initial sizes and their distributions on their
inges with time, calculations were made for values
r(0 from 0.1 /t to 1.0 o and cr(0 from 1.3 to 2.0. In
>st computations r'4m0.1 ft the effect of r,0 on the con-
centration change seems to be negligible in the figure.
Particle size distribution
Figure 2 indicates the calculated results for the
change of particle size distribution with time for
three different initial size distributions. Size distri-
butions of aerosols tend not to remain log-normal but
asymptotically approach equilibrium states as time
proceeds. The shapes of asymptotic distributions
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
643
-------
o.i 0.2 0.4 0.6 as i
Fig. 3 Comparison of self-preserving size distributions
after long periods of time resemble each other ir-
respective of initial geometric standard deviations.
These asymptotic distributions at equilibrium state
correspond to self-preserving distribution function
(SPDF) which was derived by Friedlander1" using the
similarity theory. To compare asymptotic size dis-
tributions numerically obtained in this study with
SPDF given by Friedlander the following common
variables used among most investigators are intro-
duced1-1"
I0o_ LlJS .._
OOP Darticl«»
(6)
where
The correlation of 0(i?,) and 7, after sufficiently long
periods of coagulation is shown in Fig. 3 together
with SPDF given by Wang and Friedlander1" for
Brownian coagulation without the Stokes-Cunning-
ham correction. Though the agreement among
them is fairly good, ^(j?r) numerically obtained in
this study seems to depend slightly on initial rrt and
a#. This dependence on initial r,, and
-------
Fig. 5 Experimental and calculated
variation of nominal geometric standard
deviation with time
4 6 6 10*
Fig. 6 Schematic diagram of experimental apparatus
with flow rates ranging from 0.25 to 0.5 //min, and at
these flow rates the flow in the pipe was observed to
be plug flow rather than laminar flow. The loss effect
by Brownian diffusion was evaluated to be negligi-
ble". Aerosols sampled at every given residence
time were introduced into the observation cell installed
on the stage of an ultramicroscope to measure their
particle size distributions and particle number con-
centration. The measurement method using an
ultramicroscope was developed previously by the
authors141, and the procedure is as follows: After the
focus of the ultramicroscope is preliminarily set at a
given depth of the observation cell, the flow of intro-
duced aerosol is instantaneously stopped by closing
valves. From that moment particles existing in the
field of vision are recorded until all particles disap-
pear from sight, and the number of particles at any
given time are counted by reproducing the recorder.
From these data, size distributions are obtained by a
method almost same as the Andreasen Pipette method,
and initial number concentration, na, is given by the
initial particle number of images JV(0) and the ob-
servation volume vm:
(8)
Experimental Results and Discussion
Figure 1 shows a comparison of the relative con-
centration change with time between experimental
data and theoretical curves. Agreement between
them for r90>0.3 p and a,^<\.l is fairly good. The
effect of r,0 and a,, on the concentration change is
found to be negligible, as discussed in theory. This
fact shows, in a sence, that probability of contact
may be regarded as unity, or that no repulsion may
occur when they collide. A detailed comparison of
the theoretical calculations with experiments shown
in Fig. 1 will be made by turning the point to the
change in particle size distribution with time.
In Fig. 7, changes of size distribution of tobacco
smoke are shown together with corresponding theo-
retical curves for comparison. The agreement is
fairly good. A comparison of the change of nominal
geometric mean radius with time between experi-
mental data and calculation curves is shown in Fig. 4.
Figure 5 shows a comparison of the change of experi-
mental nominal geometric standard deviations with
theoretical ones. Although some scatter is found in
experimental results, the effect of polydispersion on
the change of particle size distribution with time is
approximately confirmed for a,a less than 1.6. Theo-
retical curves for larger a,a, however, cannot be
confirmed by experiment. This is due to the diffi-
culty of constantly generating aerosols having larger
geometric standard deviation. Keith and Derrick"
performed experiments to obtain data on the change
of size distribution. In their experiments tobacco
smoke produced by a burning cigarette was used,
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
645
-------
99
04 0.60.8 1
20
Fig. 7 Comparison of experimental particle size distri-
bution changes with calculated ones
and size distribution as well as concentration was
measured by a conifuge. Figure 8 shows a comparison
of Keith and Derrick's data with calculation results
in this study. Good agreement is found again. It
is interesting that particle size distribution is found
to grow wider with lapse of time in Fig. 7, while they
grow narrower in Fig. 8. These facts suggest that an
equilibrium size distribution may exist.
In these experiments no difference between solid
particles such as stearic acid particles and liquid
particles such as DOP particles and tobacco smoke
(semi-liquid) is found.
Figure 9 shows the half-life of particle number
concentration and also shows the time required to
grow to a geometric mean radius twice the initial
one for various initial number concentrations and for
various initial geometric standard deviations. It can
be understood that at «<, above 107 particles/cc par-
ticles grow in a short time.
Conclusion
The change in size distribution of highly concen-
trated polydisperse aerosols undergoing Brownian
coagulation was studied. The results of numerical
calculation for aerosols having various initial size
distributions which followed log-normal form were
presented graphically. So far as the change in total
number concentration of polydisperse aerosols with
time was concerned, the results of calculation almost
coincided with that of monodisperse aerosols, which
was simply predicted from Smoluchowski's theory.
This was also confirmed experimentally. The man-
ner of change in size distribution with time, on the
other hand, was found to be different among initial
VOL. 8 NO. 4 1975
Keith et al'sdata
tobacco smoke
6 810
Fig. 8 Comparison nf calculated particle size distribution
changes with Keith er al. 's experimental data
- id —
10°
106 2 4 6 810' 2 4 6 BifjB
n0 [particles/cc)
Fig. 9 Dependence of values of r.-n
on initial particle number concentration
and tr,~\
size distributions, which can not be evaluated from
Smoluchowski's theory. Geometric standard devi-
ation decreased with lapse of time when initial devi-
ations were larger than about 1.5, and increased
when initial deviations were smaller than about 1.3.
Some of these results were also ascertained by ex-
periment within experimental error, and the existence
of SPDF was suggested.
The graphs presented in this paper are useful for
industrial purpose where the extent or the rate of
particle growth of highly concentrated polydisperse
aerosols is important.
Appendix
From Eqs. (2) and (3), it follows that
ff- + -£=fcg ) (A-l)
»(rj
-------
\V V
«—»:th)« rang* is"
chosen such that
90'/. aerosols exist
610*
Fig. A Depeadew* of values of C*(r< 0/C.(r,,) oo Initial
particle slie distribution
Eq. (1) can be normalized on r,0. Fig. A shows the depend-
ence of C.CrD/C.tr,,) for various r,0 against dimensionless
particle radius. So far as small a,t or large r,0 is concerned,
values of C.(ri)/C.(r,0) are approximately the same, Eq. (1)
thus can be considered to be almost normalized on r,0.
AcbMwfedgmeBt
S. Nishio was very helpful in the experimental work.
- Cunningham's correction factor of radius r< [— ]
- coefficient in Eq. (2) (-2*773/0 [cm'/sec]
— dimensionless coagulation function
CJf.fi)
JT»
*v"V> P<)
*W, /><)
M
«'W,
*', ir
Hi
MO)
r",r
coagulation function for two particles of
sizes r\ and />< [cm'/sec]
— dimensionless size width between n and ri+i
t-]
-= number of divisions in radius { — ]
''), "(ft, 0 — number and dimensionless number
concentration of aerosol particles
(-fl'W, /-)/».) [particles/eel [— ]
= total and dimensionless total number
concentration (-ZXW. '")/»»>
[particles/cc] [— ]
= total particle number concentration at
time zero [particles/cc]
= number of aerosol particles in v« at
time zero [particles]
— particle radius and dimensionless particle
radius (=/"/r«o) [cm] [/i] [— ]
= particle size width between rj and ri+i [cm]
= geometric mean radius [ft]
i?r
T » absolute temperature [*K]
f ', ' - lime and dimensionlesi time
(-mKtCJirrft') [MC] [-]
«« •« observation volume of ultremicroscope [cm1]
« size scaling function Tor self-preserving
function [— ]
- Boltzman's constant (-1.38x10-") Ierj/°X]
- viscosity of fluid [g/cm-sec]
— another dimeruiontess particle radius in
Eq.(l) [-]
*• geometric standard deviation [— ]
— total volume of particles per unit volume
[cm«/ccj
=° self-preserving number density for aerosol
size distribution [ — ]
/ — refers to the number of particle size
min
max
0
= minimum
" maximum
—at time zero
Literature Cited
1) Clark, W. and K. Whitby: /. Atmos. Sci., 24. 677 (1967).
2) Cohen, E. C. and E. V. Vaughan: /. Colloid Int. Sci., 35,
612 (1971).
3) Friedlander, S. K. and C. S. Wang: /. Colloid Set., 22,
126 (1966).
4) Gormley, P. O. and M. Kennedy: Proc. Ray. Irish. Acad.,
52-A, 163(1949).
5) Hidy, G. M.: J. Colloid Sci., 20, 123 (1965).
6) Huang, C M., M. Kerker and E. Matijevic: /. Colloid
M. Set., 33, 529 (1970).
7) Keith, C. H. and J. C. Derrick: J. Colloid Sci., IS, 340
(1960).
8) Lai, F. S., S. K. Friedlander, J. Pich and O. M. Hidy: J.
Colloid Int. Sci., 39, 395 (1972).
,9) MOller, H.: Kolloidbeihefle, 27, 233 (1928).
10) Nicolaon, G., M. Kerker, O. D. Cooke and E. Malijevic:
/. Colloid Int. Sci., 38, 460 (1972).
H) Smoluchowski, M von Z: Phys. Chem., 92, 129 (1917).
12) Takahashi, K. and M. Kasahara: Atmos. Environ., 2, 441
(1968).
13) Wang, C. S. and S. K. Friedlander.: J. Colloid Int. Sci.,
24, 170 (1967).
14) Yoshida, T., Y. Kousaka and K. Okuyama: lad. Eng.
Chem. Fundam., 14, 47 (1975).
15) Yoshida, T., Y. Kousaka, K. Okuyama and S. Nishio: J.
Chem. Eng. Japan, 8, 137 (1975).
16) Zebel, G.: Kolhid-Z, 156, 102(1958).
JOURNAL OF CHEMICAL ENGINEERING OF JAPAN
647
-------
Application of Particle Enlargement by Condensation
to Industrial Dust Collection
Tetsuo Yoshida, Yasuo Kousaka,
Kikuo Okuyama, and Fuminori Nomura
Chemical Engineering Department,
University of Osaka Prefecture,
Sakai, 591, Japan
648
-------
Application of Particle Enlargement by Condensation
to Industrial Dust Collection
Tetsuo Yoshida, Yasuo Kousaka,
Kikuo Okuyama, and Fuminori Nomura
Chemical Engineering Department,
University of Osaka Prefecture,
Sakai, 591, Japan
( Abstract )
Application of the phenomena of particle growth by condensation
to industrial dust collection was studied. The analysis to
evaluate the extent of size enlargement was first introduced
for the two essential and industrially useful methods, one of
which is that by mixing hot saturated air with cold one and the
other by injection of steam into air. Four typical processes were
proposed in their effective application to industrial exhaust gas
and the procedure of utilization of these processes was illus-
trated according to the various conditions of exhaust gas. The
technique for size enlargement of aerosol particles by conden-
sation was found to be essentially applicable to any industrial
exhaust gas which contains submicron dust particles in low number
concentration, when the appropriate process shown herein was
selected.
649
-------
Introduction
Fundamental analysis and experiment for growth of aerosol
particles by condensation were made in our previous papers (1976;
1977), where it was suggested that the particle growth by con-
densation would be one of the most promising preconditioning
techniques for the collection of submicron dust particles. In-
dustrial application of the fundamental results was developed
in this paper.
Condensation of water vapor on aerosol particles, not con-
sisting of soluble substances, will essentially occur wherever
a certain degree of supersaturation is produced around the parti-
cles. The fundamental analysis of two essential methods to pro-
duce supersaturation, one of which is that by mixing hot saturated
air with cold air (Yoshida et al.,1976) and the other by injection
of steam into air (Yoshida et al.,1977), was first briefly intro-
duced. The establishment of economic processes to produce super-
saturation using these methods will be important from the industri-
al point of view. Processes for this purpose were discussed and
developed under consideration of various conditions, such as temp-
erature and humidity, of industrial exhaust gas. Some processes
were proposed according to various conditions of exhaust gas and
they were arranged into several charts for the facility of the
design of a preconditioner of industrial dust collection.
Basic Consideration
In this section the fundamental aspects for particle growth
by condensation, which were made clear in the previous papers
(1976; 1977), will be briefly introduced.
The point "i" on the humidity chart of Figure 1 indicates a
state of supersaturation of air. When this point is attained in
an insulated chamber by some methods and a certain amount of
650
-------
aerosol particles is introduced into the chamber as condensation
nuclei, condensation upon the particles will occur. And as the
result the state of air changes along the adiabatic line to the
point "f" which almost coincides with the saturated state. Then
the value of AH shown in the figure which represents the quantity
of condensed water vapor per unit mass of dry air will be pro-
duced. In order to attain point "i"ora supersaturated state,
two methods were suggested to be effective: one of which is that
by mixing high temperature saturated air with low temperature
saturated air and the other by steam injection into air. Existence
of aerosol particles, the value of AH in both methods is given
by the following enthalpy and material balances on dry air mass
basis:
(mixing method)
i8f + AHiwf
Hsf + AH (2)
(steam injection method)
ig + Qst { xisfc + (l-x)ijt} =isf +Qstd-x)iwf+ AHiwf (3)
Hg + Qstx = Hsf + AH (4)
Some of the calculated results are shown in Figures 2 and 3.
When all of the vapor corresponding to the amount of AH are
assumed to condense upon particles which are introduced into
supersaturated air, the following relations must be satisfied:
AH = -i-Trpsf fr3n(r,»)dr- [r3n (r,0)drl = - < Dyf " Dvi } nOps <5>
3 l'o J0 J 6
Dvf = (6AH/Trn0)1/3 when D^f^-D3^ and pg = 1 (6)
AQ represents the particle number concentration of aerosol on dry
air mass basis. Dyi and Dyf represent the volume mean diameters
of the particles before and after growth respectively. The increase
in size of aerosol particles undergoing condensation can then be
651
-------
evaluated in volume mean diameter Dvf knowing the values AH and
ng in most cases.
The growing rate of particles undergoing condensation was
found to be very rapid in the previous paper (1976), so the above
analysis in equilibrium state only will be essential in developing
this technique to industrial application.
Typical Operations to Obtain High Temperature and Low Temperature
Saturated Air
(High temperature saturated air)
(a) adiabatic humidification
When exhaust gas lias high temperature or moderate temperature
with high humidity, direct contact of the gas with recirculating
water in a humidifier will be effective. This operation is simple
and any heat source is unnecessary. The operation is illustrated
as line (a) on the humidity chart of Figure 4.
(b) humidification by contacting with heated water
When the exhaust gas has moderate temperature and humidity,
contact of it with heated water will be effective. The line (b)
in Figure 4 illustrates this operation.
(c) humidification by steam injection
This operation will be essentiallly effective for gas having
any temperature or humidity, but this is especially effective
for low temperature gas, the reason of which appears later. The
line (c) in the figure illustrates this operation.
(Low temperature saturated air)
(d) adiabatic humidification
When exhaust gas has low temperature or low humidity, the
same method as (a) is applicable. This is shown as the line (d)
in Figure 4
(e) dehumidification by contacting with cooling water
652
-------
When the operation (d) is not available, the direct contact
of gas with cooling water will be inevitable. This is shown as
the line (e) in the figure.
These typical operations are applied to industrial exhaust
gas according to temperature and humidity of the gas in the
following section.
Methods to Produce AH in Various Industrial Exhaust Gas
The value of AH, which represents the quantity of condensable
water vapor per unit mass of dry air described before, is im-
portant, while the size of grown particles is determined by this
value under a given particle number concentration of exhaust gas
as shown in eq(6). Some of the representative industrial processes
to produce AH are discussed according to some classified gas
conditions in this section. The property of exhaust gas was re-
garded the same as that of air in the following discussions. The
notations (a)~-(e) appears in the following figures indicate the
above classification for typical operations.
(High temperature and high humidity exhaust gas)
This is probably the most profitable case to apply the method
of particle growth by condensation when cooling water is obtainable.
The flow sheet of this process is shown in Figure 5. Gas after
adiabatic humidifier is divided into two parts. One part is de-
humidified by contacting with cooling water, and is then mixed
with the other to produce AH or to enlarge the particles in gas
at the mixing chamber. This process is named "A". The value of AH
in this process depends on the temperature T and humidity Hg
of exhaust gas, and on the temperature of cooling water which
is available. The correlation among them was calculated by eqs(l)
and (2), and was illustrated on the right upper side of the
humidity chart of Figures 6«-10 for various AH. The mixing ratio
653
-------
Rh was chosen in any cases as the optimal value shown in Figure 2
in calculation. If one may require the value of AH=0.006 and
if the cooling water temperature of 20"C is available, then the
exhaust gas conditions must exist at least on the line of 20°C
of cooling water in Figure 8, for instance Tg = 1000°C and H =
0.07gH2O/g dry air, T_ = 100°C and Hg = 0.32 g H2O/g dry air, and
so forth. The temperature of saturated air or the equilibrium
temperature after adiabatic humidification, in this case, comes
to Te = 72.5°C. It is matter of course that the larger values than
AH =0.006 can be obtained if the gas conditions exist in the
upper side of the line of 20°C of cooling water in Figure 8.
(Low temperature exhaust gas)
In this case steam injection method is effective, but this
method has a fault that it requires saturated steam as a heat
and water vapor source. The process is very simple as shown in
Figure 11, and this process is named "B". The value of AH in this
process depends on the steam quantity Qst and on exhaust gas
conditions such as T and H_. The correlation among them was
calculated by eqs(3) and (4) and was illustrated on the left side
of the humidity chart of Figures 6 ~10. As the property of steam,
100°C, 1 ata and unity in dryness fraction were assumed in the
calculation. If the same value of AH as the above example is
required and if the steam quantity of Qgt=0.1 kg steam/kg dry air
is available, then the exhaust gas conditions must exist at
least on the line of Qst = 0.1 in Figure 8. The larger values
than AH=0.006 can be obtained if the gas conditions exist in
the left side of this line. The temperature rise in air was
also illustrated in the figure as T@, in this example Te = 37.9°C.
If the saturated air after steam injection has high temperature,
the process "A" may be applied successively after steam injection.
This is one of the advantages of this process.
654
-------
(Gas having intermediate conditions)
When exhaust gas has temperature below about 200°C and low
humidity,ateam injection can be applied after application of
adiabatic humidification. This process is shown in Figure 12,
and is named process "B"1 because of similarity to process "B".
The calculated results in this case were illustrated on the
humidity chart of Figures 7~10 .
When both hot water and cooling water are available, the
following process may be applicable. One part of gas is humidified
and at the same time heated by hot water to obtain high temp-
erature saturated air, and the remainder is cooled down by con-
tacting with cooling water to obtain low temperature saturated
air. Then they are mixed to produce AH. The process is shown in
Figure 13, and is named process "A1". The calculated results can
not be shown in this case in Figures 6/w. 10 because of one more
additional parameter of heated water temperature. The value of
AH, however, is obtainable for every given condition since the
temperatures of points "2" and "3" in Figure 13 can be evaluated
from given gas conditions.
Figure 14 illustrates the rough domain in applying the above
processes to various exhaust gas conditions.
Utilization of the Above Results
The steps of procedure to utilize the above results for in-
dustrial purpose are as follows. The appropriate process is first
selected according to the given temperature and humidity of ex-
haust gas referring to Figures 6~10 and 14. The value of AH is
next evaluated from Figures 6 ~10 or from calculation by setting
up the quantity of steam, or the temperature of cooling water
which is obtainable. Then the volume mean diameter of grown parti-
cles, Dvf, can be evaluated from eq(6), using the value of AH
and knowing the particle number concentration of the gas, n0.
655
-------
A dust collector after the preconditioned gas should be designed
using the value D f thus obtained. If the value D ~, on the other
hand, is first given from the point of performance of a collector
installed after particle enlargement, the value AH should be
first determined from eq(6) knowing the value nQ. Consequently
the steam quantity Qst or cooling water temperature will be
determined from the figures using the known value AH and knowing
the gas conditions.
(Example)
Tg = 30°C, H =0.01 g H20/g dry air and nQ = 108 particles/g
dry air (roughly corresponds to 10 particles/cm gas) are given.
It is required to enlarge submicron particles in the gas to 5
microns in Dvf.
In this case the value of AH which is required is found to
be 0.006 g H20/g dry air from eq(6). The appropriate process in
this case is found to be process "B" because of low temperature
gas. The required steam quantity Qst (100°C, lata, x =1) is
then found to be 0.1 g steam/g dry air from the point of Tq=30°C
and H =0.01 in Figure 8.
Experimental
The experimental results on the processes "A"1 and "B" were
reported in the previous papers (1976; 1977), and so the experi-
ment on the process "A" and "B1" were briefly shown in this
section. Figure 15 shows the schematic diagram of the experimental
method. The apparatus can be operated in both processes "A" and
"B" by opening and closing valves. The mixture of air with com-
bustion gas from gas burner was supplied as a high temperature
gas from the bottom of the adiabatic humidifier. The temperature
of the gas was several hundreds centigrade degrees for the process
"A" and about 200°C for the process "B1". Tobacco smoke(Dvi=0.35
micron) and dust particles contained in combustion gas(Dvi<0.1
656
-------
micron) were used as submicron dust particles. Total gas flow
rate was 180 1/min. In the experiment of process "A", the valves
of v, and v2 are adequately opened to obtain a certain mixing
ratio R. keeping v, close. In the experiment of process "B1",
the valve of v_ is closed and v. is moderately opened to inject
a certain amount of steam into the gas. The technique of size
measurement of grown particles was the same as that appeared
in the previous papers(1975; 1976; 1977). Experimental results
were plotted in the same graph as those of the previous papers
(1976; 1977), which was shown in Figure 16. The results obtained
in the previous works were also collectively plotted in the
figure. It will be found in Figure 16 that the all processes
"A", "A1", "B" and "B"1 are useful for size enlargement of
submicron dust particles by condensation.
Conclusion
Application of the phenomena of particle growth by conden-
sation to industrial dust collection was developed. Four typical
processes were proposed for the effective application of the
phenomena and the procedure of their utilization was shown ac-
cording to various exhaust gas conditions. The technique for
size enlargement of aerosol particles by condensation was found
to be essentially applicable to any industrial exhaust gas which
contains submicron dust particles in low number concentration,
when the appropriate process shown in this paper was selected.
The results suggested that the exhaust gas having high temperature
and high humidity is especially profitable to apply this technique
because any heat source but cooling water is unnecessary. The
exhaust gas having low temperature such as 30°C or below, on the
other hand, is advantageous in very simple apparatus but is dis-
advantageous in necessity of steam.
657
-------
Nomenclature
Dvi' Dvf*vo^ume mean diameter before and after growth, re-
spectively, cm
H - absolute humidity, g H20/g dry air
AH«condensable water vapor, g H20/g dry air
i»enthalpy, cal/g dry air or g steam
n0-particle number concentration, 1/g dry air
n(r,0), n(r,») - particle number having particle radius r before
and after growth, particles/g dry air
Qst«quantity of steam injection, g steara/g dry air
r* radius of particle, cm
Rn»mixing ratio, g dry air of high temperature saturated air/
g total dry air
T =temperature, °C
x «dryness fraction of steam
Greek Letter
Ps= density of condensed liquid, g/cm
Subscripts
e = equilibrium state of air
f = final state shown in Figure 1
g = initial state of air
i = initial state
s = saturated
sf « saturated air in final state
sh • high temperature saturated air
si »low temperature saturated air
st »steam
wf *water in final state
Superscripts
1 =wet
" « dry
658
-------
References
Yoshida, T., Kousaka, Y. and Okuyama, K.f"A New Technique
of Particle Size Analysis of Aerosols and Fine Powders
Using an Ultramicroscope", Ind. Eng. Chem., Fundam.,
14, No.l, 47(1975)
Yoshida, T., Kousaka, Y. and Okuyama, K.,"Growth of
Aerosol Particles by Condensation", Ind. Eng. Chem.,
Fundam., 15, No.l, 37(1976)
Yoshida, T., Kousaka, Y., Okuyama, K. and Nomura, F.,
"Growth of Aerosol Particles by Steam Injection",
J. Chem. Eng. Japan(in contribution)
659
-------
Captions of figures
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11
Figure 12
Figure 13
Figure 14
Figure 15
Figure 16
Change in humidity and temperature due to condensation
Condensable water vapor AH at various mixing conditions
Condensable water vapor AH against quantity of steam
injection Qst
Typical operations to obtain high and low temperature
saturated gases
Illustration of process "A"
Application of processes "A" and "B" to various exhaust
gas conditions( AH=0.002)
Application of processes "A", "B" and "B"1 to various
exhaust gas conditions(AH =0.004)
Application of processes "A", "B" and "B1" to various
exhaust gas conditions(AH=0.006)
Application of processes "A", "B" and "B"' to various
exhaust gas conditions{AH=0.008)
Application of processes "A", "B" and "B"1 to various
exhaust gas conditions(AH=0.010)
Illustration of process "B"
Illustration of process "BIH
Illustration of process "A1"
Effective application of process "A", "A"', "B" and "B"
to various exhaust gas conditions
Schematic diagram of experimental apparatus
Relation between grown particle diameter DVf and
condensable water vapor per single particle
660
-------
en
temperature
Figure 1 Change in humidity and temperature due to condensation
-------
0.005
cr,
Ch
ro
:Tsh=60'c
r~:Tsh=50'c
0 0.2 0.4 0.6 0.8 1.0
0 0.2 OA 0.6 0.8 1.0
Rh C-3
Figure 2 Condensable water vapor AH at various mixing conditions
-------
0.020
relative
humidity
100 •/•
0.1 0.2 0.3 0.4 0.5
Q$t Tgsteam/gdry atrl
Figure 3 Condensable water vapor AH against quantity of steam
injection Qgt
663
-------
CTl
•U
;
6
£.
air temperature
Figure 4 Typical operations to obtain high and low temperature
saturated gases
-------
adiabatic
high temp, humidifier
exhaust gas
Tg, Hg
mixing chamber
to collector
recirc.
water
cooling water
dehumidifier
en
cr>
temp.
process "A*
Figure 5 Illustration of process "A"
-------
tf\
m
ON
steam injection
st=0.2
5 10 20 40 60 100 200 400600 1000 2000
Ta fa
Figure 6 Application of processes "A" and "B" to various exhaust
yas conditions ( AH = 0.002)
-------
steam injection
= 0.20
0.15
0.10
0.05
0.004
10 20 4060 100 200 400 600 1000 2000
Tg
Figure 7 Application of processes "A", "B" and "B"' to various
exhaust gas conditions (AH = 0.004)
-------
T.=88.7'C
process'A
37.9
26.8*0-
steam injection
CLt= 0.20
* 0.15
0.10
0.05
10 20 40 60 100 200 400 600 1000 2000
exhaust gas temp. Tg C*C3
Figure 8 Application of processes "A", "B" and "B'" to various
exhaust gas conditions(AH - 0.006)
668
-------
O.OOA
60 100 200 400600 1000 2000
Tg C-C3
Figure 9 Application of processes "A", "B" and "B'H to various
exhaust gas conditions(AH »0.008)
-------
0.004
5 10 20 40 60 100 200 400600 1000 2000
Tg C'C]
Figure 10 Application of processes "A", "B" and "B1" to various
exhaust gas conditions(AH=0.010)
-------
low temp.
exhaust gas
Tg.Hg
steam
high temp.
Figure 11 Illustration of process MB'
-------
Intermediate
temp, exhaust c
g. Mg
i
a\
vj
>, Hs
6
£.
Hsl
Hg
adiabatic steam
humidifier
)°s 1 ^1 12 ' 3 *° c
f ^X |"j
I ,
4^
*/7
y
recirc. water
n
-SLi
to process A
temp.
process** B*"
Figure 12 Illustration of process "B*
-------
U)
pump heater
.
humidifier 'heated water
intermediate
temp, exhaust gas j
Tg. H
4 to collector
1*.
--- cooling water
dehumidifier
temp.
* *• *
process A
Figure
13 Illustration of process "A1
-------
3
C.
air temperature
Figure 14 Effective application of process "A", "A1", "B" and "B1"
to various exhaust gas conditions
-------
w
steam M 1 thermometer-
100'C.1ata.x=l v. ' \\
humidifier
did. 200mm
height 1000mm
0.5inch Raschig rings
i recirculatory
A water
t
i
i
i
i
pump
*v» • i
\>
exhaust gas
mix ing chamber
^ capacity
60 -600cm3
filter
orifice
flow
meter
_L
[ultramicroscope}
vacuum ;
pump
JVTR & monitorl
dehumidifrer
dia. 150mm
height 750mm
0.5 inch Raschig rings
IBOVmin
olower
Figure 15 Schematic diagram of experimental apparatus
-------
.2 6
"O
a\
-*±
en
c 2
a
'1
1—' I''"
I I I • I I I
(example) ff .-process "A"*, DO
and n^lO7-11-1
6 810"10 2
CgH20/particle3
10
Figure 16 Relation between grown particle diameter Dvf and
condensable water vapor per single particle
-------
APPENDIX J
MITSUBISHI HEAVY INDUSTRIES, LTD,
677
-------
The Latest Dust Collecting Technique
October 1976
A MITSUBISHI HEAVY INDUSTRIES. LTD.
5-1, Marunouchi 2-chom», Chiyoda-ku, Tokyo 100, Japan
)by Mitsubishi Heavy Industries. Ltd.
jproduced by permission) 678
-------
The Latest Dust Collecting Technique
Yasuo Saito*
Naoji Tachibana**
Koji Tashiro*
Kazuo Matsui**
Shigenori Komura*
Katsutoshi Yada**
The electrostatic precipitator is well known as the most efficient and most economical dust collector for a large amount of
gas volume and high performance and its role is becoming more and more important with the aggravation of environmental
pollution.
The precipitator is now confronted with various problems requiring immediate solution:
(I) Difficulty of collecting dutt of submicron and/or high-resistivity.
(2) Capability of maintaining stable perofrinance over a long period.
(3) Structural reliability against a large quantity of gas treated.
{4) Applicability to desulfurizing and denitraling processes.
This paper introduces our company's views on these problems.
\. Introduction
Mitsubishi Heavy Industries has supplied more than
1000 units of dust collectors to many industrial fields
since this item was added to their lines of production in
1960. These dust collectors have treated a total of some
200 million Nm3/h of flue gases till now. At the begin-
ning, the dust collectors were used as attached facilities in
works and plants but gradually came to be regarded as
indispensable facilities for industries. On the other hand,
users' requirements also became severer, calling for the
following performances:
(1) To be able to effect designed performance surely.
(2) To be capable of maintaining designed performance
for a long time.
(3) To have a stable performance against changes in
conditions.
(4) To have long intervals between servicings for easy
inspect and repair.
In other words, they came to demand the same or
even higher performances than those of the master plants.
Our company is attaching the greatest importance to the
electric precipitator (abbreviated as EP hereinafter) for
the time being and most of the precipitators manufac-
tured and supplied by our company hitherto are EPs.
EPs are generally superior to other types of dust
collectors in that they have a very high dust collecting
efficiency and that they require less power charge,
maintenance and depreciation costs. On the other hand,
EPs, which treat dust-containing gases electrically, are
susceptible to the effect of the physical and chemical
properties of treated gases and dust and therefore, involve
much of an empirical engineering element. That is to say,
the element which governs dust collecting performance is
what is generally called the intensity variable having a
"dispersion" which cannot be specified with a simple
scale, and the resultant dust collecting efficiency cannot
be free from probability variation and their performance
guaranty can only be made by a certain reliability.
With the increasing demand for high dust collecting
efficiency in recent years, high-performance EPs which
can collect dust to such a high degree that flue gases
become invisible have been commonly equipped. Such
high-performance EPs, however, are susceptible to the
effect of probability variation as mentioned above in
many cases and this problem is drawing a particular atten-
tion of the circles concerned. A simple increase of their
sizes cannot be said to be a practical method of solution
if equipment economy is taken into account and it is
necessary to study, by making analysis of the results of
actual operation, a method of grasping the element which
governs the dust collecting efficiency of EP accurately. It
has also been made clear theoretically that the larger is
the volume of past data, the higher performance relia-
bility is obtainable with a smaller risk*. One of the tech-
nical aims of our company will also be placed on this
point for the time being.
Note: «-S. Masudu et al.. Probability Design Method and
Its Application to the Guaranty of Performance of Electric
Precipitators, Trans. I.K.E.J. 8/73.
In this paper, the authors discuss, in relation to this
point, the collection of fine-particle dust contained in flue
gas which is rather significant for the small quantity of
dust contained and the collection of high-resistivity dust
which is comparatively difficult to collect.
With the increase in the size of equipment and its
efficiency, any trouble with the equipment came to give a
greater effect than ever. As a result, lasting performance
and stable continuous operation came to be severely
demanded in recent years. EPs with dust-collecting areas
of several tens of thousand square meters per unit and
several tens of thousand discharging electrodes and at-
tached mechanism have made debut recently. For such
large-sized EPs, structural reliability and easy servicea-
•Takas»go Technical Institute. Technical HeadquarUri
••Kobe Shipyard & Engine Works
679
-------
bilily, in particular, are essential conditions for the main-
tenance of stable performance, for which perfect design
which enables quality control to be made thoroughly,
effectively and accurately is demanded. In order to ensure
the outlet dust burden close to the minimum stably and
for a long time besides improving dust collectability on
the collecting electrodes, the major function of EP, the
improvement of the conventional mechanism for rapping
collected dust on the electrodes and the conventional
mechanism for preventing reentrainmcnt at the time of
rapping are attracting a strong concern of the circles
concerned as indispensable factors for the attainment of
high performance. In this paper, the authors also touch
upon these mechanisms.
On the other hand, flue gas desulfurization or denitra-
tion plants are installed as a means to prevent air pollu-
tion. An EP is used as equipment for pretreatment, inter-
mediate treatment and after-treatment for such plant and
plays a significent role for the attainment of purifying
performance and economic design. Consequently, an EP
has become indispensable for a total system. In this
paper, the authors also discuss its relation with such
equipment as a technique for utilizing EP.
2. Measures to realize high grade dust collecting tech-
nique
2.1 Grasping of problematical points
As the recent problems of dust collecting technique,
measures for fine-particle dust and high-resisti. ty dust
collection ire proposed. These two problems have aspects
which cannot be considered separately. Taking fine-
particle dust collection for example, it is neceuary to
foresee or investigate beforehand electric resistivity in
addition to dust burden, particle distribution, bulk den-
sity and corona discharge characteristics.
Fine-particle dust is apt to scatter again by rapping or
other treatment even after it has been collected. There-
fore, it is neceuary to take into full consideration that
this kind of dust cannot always be collected efficiently.
High-resistivity dust is contained in gases emitted from
low-sulfur coal-fired boilers, sintering machines in steel
mills, cement manufacturing boiler* and CO boilers. For
these gases, dust collectors fail to attain a high collecting
efficiency in many cases due to the back corona action of
dust deposited on the collecting electrode. Concerning the
back corona phenomenon, realities are being made clear
gradually. It ii considered that conditioning, Improvement
of structure, and high-temperature treatment are effective
for the phenomenon. Our company is also conducting
research on this phenomenon by use of our research
facilities shown in Fig. 1.
2.2 Collection of fine-particle dust
In conducting fine-particle dust collection, it is neces-
sary to foresee corona discharge characteristics. Thit is
because there is a fear of current shortage occurring due
to the effect of the so-called space charge. Current short-
age occurs in the fore chamberof EP, in particular. Our
company is coping with the fault by employing various
directional barbed wires selectively.
Fig. I Dust collection research laboratory
Fig. 2 Relition between the performance of EP
and that of filter
What is necessary for the determination of the scale of
EP is to grasp the relation between flying grain size and
dust collecting rate quantitatively. It is needless to say to
point out the importance of recognizing the size of a
grain as that of a grain in flight. For example, if it is
compared with the size of a grain collected by EP, they
do not always agree with each other partly due to the
effect of coagulation in the equipment. For this reason,
our company developed a special filtering device (patent
pending) and conducted measurement at site repeatedly.
As the result of statistical analysis of the results of meas-
urement we could obtain the relation between the per-
formance of EP and that of filter (for actual EP and
model EP) as shown in Fig. 2. By utilijing this achieve-
ment as well, we can increase the reliability of fine-
particle dust collection.
What should not he overlooked in fine-particle dust
collection it the re-entrainment phenomenon of duit as
mentioned above. In case a very high dust collecting
performance is demanded, even an instantaneous increase
of dust burden due to re-entrsinment by rapping becomes
a problem. Re-entrainment i« a phenomenon that dust
once deposited on the collecting electrode comes off due.
to various reasons and is carried away by gas flow. The
680
-------
main causes of re-entrainment are as follows:
(1) Rapping
(2) Electric factor
(3) Gas now
(4) Ionic wind
(1) Rapping produces the largest re-entrainment quantity.
As a measure to prevent re-«ntrainment of dust, our
company adopted the damper system as mentioned
below, with satisfactory result. (2) Re-entrainment due to
electric factor occurs in case of low-resistivity dust with
an electric resistivity lower than 104fi-cm. Dust collected
on the electrode loses its electric charge as soon as it is
collected and at the same time, obtains the same electric
charge as the collecting electrode by electrostatic induc-
tion and flies into the field space isactionally. Dust which
flied out into the field space is charged again by the ion
emitted from the discharging electrode and goes toward
the collecting electrode. This phenomenon is repeated and
in the meantime, a portion of the dust escapes out of the
equipment without being collected by it. As a counter-
measure, it is necessary to design so as to make gas speed
low and conceive proper electrode construction and
arrangement. (3) Re-entrainment by gas flow occurs in
relation to gas speed in case adhesive force between elec-
trode and grain or between grains is small. Re-entrainment
in this case is simple separation and conveyance phenome-
non by fluid. To prevent this phenomenon, it is necessary
to take the following measures: (a) to make gas speed in
the equipment low, (b) to make gas speed distribution
uniform, (c) to provide the form of the collecting elec-
trode with a pocket characteristic, (d) to maintain the
quantity of dust sticking to the collecting electrode at a
small level. (4) Re-entrainment is caused by ionic wind
(also called corona wind or electric wind) which is gener-
ated when ion generated by corona discharge share* its
kinetic energy to neutral molecules as it strikes with them
on the way of its movement toward the opposite elec-
trode by the action of the electric field, and the ion and
neutral molecules move toward the opposite electrode.
This phenomenon is shown in Fig. 3 (a) and (b). This
ionic wind has a function to carry dust to the opposite
electrode but, on the other hand, produces eddy current
on the surface of the opposite elctrode, causing collected
dust to come off. Particularly, in case of dust with a
small apparent specific gravity and weak adhesive power
such as the dust discharged from naphtha-fired boilers, it
seems that dust collecting efficiency is reduced under the
effect of ionic wind. As countermeasures to this, it is
necessary to pay consideration to the following points:
(a) to reduce gas speed as much as possible, (b) to select
discharge electrodes and electrode spacing with considera-
tion to proper ionic wind and (c) proper rapping.
2.3 Collection of high-resistivity dust
High electric resistivity-dust collection must be con-
ducted carefully. This is because when electric resistivity
exceeds 1011 I2f2-cm, back corona occurs. Back corona
is «h abnormal discharge phenomenon which occurs in the
interior or on the surface of dust layer on the collecting
electrode in case the strength of the electric field formed
(by courtesy of Dr. Adachi, Asiii. Professor
of Yimaguchi University)
Fig. 3 (a) Generation of ionic wind
X
"X
N«|it/v» ionic *nd
fx
FrtN MI* Vdi vi««
Fig. 3 (b) Flow pattern of Ionic wind
(by courtesy of Dr. Maiudj, Professor of Tokyo
University)
Fl|. 4 ApjMannce of back corona (32 kV, 10jM/cm')
on the dust layer exceeds the electric breakdown field
strength due to corona current passing the dust layer. Fig.
4 ihowt the phenomenon. This phenomenon reduces dust
collecting efficiency for the following reasons:
(1) Inverse ion is generated from the. collecting electrode
side to the corona discharge electrode side, and in
consequence, streamer or abnormal current increase
occurs easily, causing a sharp decline of applied volt-
age.
(2) At the same time inve.ie-polarity ionic wind occurs
from the collecting electrode side and fiuidly prevents
the normal movement cf dust to be collected.
681
-------
GJI
(X)
Fig. 5 Relation between gas temperature and
current density (fly ash)
(u occvrred it « low temperature)
c* .
I
(as occurred at • high temperature)
Fig. 6 effect of gas temperature on back corona
(3) Inverse-polarity ion from the collecting electrode side
reduces or neutralizes the electric charge of dust which
was charged by the corona discharge of the discharge
electrode and disturbs dust collection by coulomb
power electrically.
U
Vrktft rrpon
Auunpfeon lor ipptutXM
ol Ut'l B»ory
birabit ripen
B, Rmjelnun WK«
H*»v> oi
OKttic turntnu
SnOnn| connrtlf
Fig. 7 Permeability based on Mie'» theory on scattered
light and visible region distribution
As effective methods of collecting high-resistivity dust,
various methods are conceivable, one of which is high-
temperature dust collecting method. Fig. 5 shows the
relation between the gas temperature of high-resistivity fly
ash arising from low-sulfur coal-fired boiler and its dis-
charge characteristic. Fig. 6 shows the patterns of back
corona at low and high temperatures. As can be under-
stood from Figs. 5 and 6, back corona disappears when
temperature is high.
2.4 Smoke density and color
2.4.1 Dust
The color of smoke represents the scattered light of
grains visible to the eyes of observers as "smoke". Its
thickness varies by the dust content and increases as the
sum of the surface areas of grains becomes larger, or in
other words, the grain size is smaller. Fig. 7 shows, by
applying Mie's theory, isodensity lines corresponding to
the Ringelman densities 1, 2 and 3. Also shown in the
figure for reference is the visible region which was esti-
mated on the basis of the results of our actual operation.
It can be understood from the "figure that smoke color
varies by grain size even if density is the same. However,
as the grain size from the same process is not always the
same, special care must be taken-for the application of
the above data.
2.4.2 Steam and S(>3 mist
As typical visible smokes, steam and SOj mist can be
mentioned besides dust. Steam and SO3 mist generate
682
-------
Fig. 8 Comparison of densities
of smoke emitted from
heavy oil fired boiler
Operating condition:
Pbnl output 25GMW. NH,. magnesium hydroxide injection.
Ft* in operation. SO, at economizer oullet-l Sppm ,
Dust at EP outlet Smg/Nm'
2SOMW
Equipped with EH
With NH, injection
white plume when they are emitted from stacks and
cooled by atmospheric air. In this case too, their condi-
tion is affected by the meteorological condition near the
exhaust ports of the stacks. That is to say, in case the
atmospheric condition is in low-temperature high-humidity
condition, steam generates a large volume of white plume.
To prevent this, it is necessary to raise exhaust gas tem-
perature, particularly in case atmospheric temperature is
low.
Fig. 8 shows a difference in the colors of smokes
emitted from SOj mist arising from a heavy oil-fired boiler,
whether NHj is injected into it or nor. From the figures
it can be understood that the color of smoke is governed
by SOj mist rather than dust density.
3. Recent dust collecting techniques-and trend of applica-
tion
3.1 Boilers for utility thermal power plants
3.1.1 EPs for heavy oil (crude oil) fired boilers.
Mitsubishi Heavy Industries completed the world's first
fullscale NH3 injection type EP in 1966 (the injection
method is patented by MHI). During the 10 years since,
the EP has spread at a rapid pace and almost all of the
heavy oil fired boilers in our country are equipped with
EPs. At the beginning, EPs with an efficiency of about 80
per cent were adopted as the standard. At present, in
response to the severe requirement of less than 10-20
mg/Nm3 for outlet dust burden, manufacturers are taking
such methods as the remodelling of the damper type EP
as described below for higher efficiency and the adoption
of low-sulfur fuels.
As a recent tendency, the adoption of high-sulfur
heavy oil coupled with a full-scale flue gas desulfurization
equipment is increasing. In this case, the quantity of dust
is larger and moreover, reaction product of NHj-SOj
line occupies a greater ratio of its components, giving rise
10 problems relating to the performance and operation of
EP, and making it necessary to improve the NHj injecting
device, EP and ash treating device in some cases. Study
for diversified use of EP is under study. That is to say,
for desulfurization and denitration purposes, besides, the
conventional low temperature EPs installed at the air
prelieater outlet, liie high-temperature EPs to be installed
before preheaters and wet type EPs to be used after
desulfurization are under study.
3.1.2 EPs lor coal-fired boilers
Coal-fired boilers are under review. In the past, many
coal-fired boilers were used, but for use in new plants,
they have many difficulties which cannot be covered by
past experience only. The reasons are as follows:
(1) The capacity of one plant is several times as large as
that in the past.
(2) It is scheduled to use imported coals and domestic
coals jointly. Imported coals are more difficult for EP
to collect dust than domestic coals.
(3) Smaller quantity of dust than ever before and a high
efficiency over 99 per cent are demanded.
As countermeasuros to these, measures adapted to the
characteristics of each plant, such as the increase of the
size of the conventional EP, the adoption of high tem-
perature EP and the injection of additive are under study.
3.2 Boilers for private power plants and industries
These boilers are mostly fired by heavy oil. With the
strengthening of dust exhaust standard in 1971 and the
enforcement of the environmental pollution control agree-
ment concluded later, various kinds of dust collectors
have come to be installed at a rapid pace. Among various
kinds of dust collectors, high-efficiency EPs are showing a
steady increase, though there are plants where multiclone
(MC) or flue gas -desulfurization equipment is used as
substitute.
Difference of the EPs for private power plants and
industries from those for utility is that because of the
high temperature of exhaust gases NHj injection for
preventing the corrosion of EP is not necessary and most
of them are not injected with NHs- In case NHj is not
injected, a greater part of the components of heavy oil
dust is carbon, and it has been considered that because of
683
-------
its low electric resistance, it is difficult to collect with EP.
According to the recent results of operation, it has been
made clear that this is not true but rather it is easier to
collect than NH3 injected dust, as only unburnt portion
of comparatively large grain sizes remains. Even if NHj
injection aimed at the prevention of corrosion of EP is
not necessary, SO} cannot be removed unless NHj is
injected. For this reason, there are cases where NHj is.
injected as a countermeasure to acid smut or to make
smokes transparent.
Some of the boilers for private power plants are not
yet equipped with EP. There is a movement, however, to
impose duty to install EP capable of collecting fine parti-
cles, or dust collector having the same or higher perform-
ance as equipment standard for the purpose of reducing
suspended particulates from the viewpoint of environ-
mental protection. It is expected that sooner or later EPs
will be adopted for all boilers for private power plants as
they have been adopted for boilers for utility power
plants. It is also presumed that new uses of EPa will be
developed In relation to desulfuriaation and denltration.
3.3 Main exhaust gas arising from sintering machines
In steel mills
3.3.1 Problematics! points in equipment planning
Sintering machine Is one of the facilities discharging
the largest quantity of exhaust gas in steel mills. Dust is
mainly composed of iron content and has a density of
around 1 g/Nms generally. Gas contains a considerable
•mount of SOj. EPs wen adopted comparatively recently
in place of. the conventional cyclones which were con-
sidered insufficient for the prevention of environmental
pollution.
Mitsubishi Heavy Industries has turned out 38 units of
sintering machines since completing its No. I unit in
1967. Its performance and equipment planning involves
considerable difficulties and requires sufficient considera-
tion. The important points in equipment planning are as
follows:
(1)' Working voltage is low because of high electric resist-
ance of dust (performance).
(2) Dust collection is comparatively difficult as the grain
six* of dust is small (performance).
(3) Then is t fear of dust burning, to it is necessary to
take • measure to prevent burning (equipment).
(4) Powerful rapping to necessary as • large quantity of
dust sticks (equipment).
(5) For largMited equipment, in particular, study of
thermal expansion- and deformation-proof structure is
essential (equipment).
3.3.2 Features of our company's planning
Dust contains considerable amounts of metallic and
nonmetallic fine particles. Its electric resistivity is
10IJn*cm or higher at normal temperatures (around
ISO*) and gas moisture content (6-8 volfc). The high
resistivity is psrtly due to the resistivity of the abovemen-
tioned fine particles themselves and partly to the fact that
SOj in gai cannot perform its role as surface resistance
reducing agent as it forms gypsum by combining with
CaO contained in dust. On the other hand, this gypsum
hardens densely by suffering a hysteresis of low tempera-
ture - water drop spray - high temperature during the
operation of the equipment. This also serves as a factor to
increase resistivity.
In case electric resistivity is high, a back corona phe-
nomenon occurs and in consequence, the working voltage
drops as already described in 2.'Countermea$ures to this
are as follows:
(1) To reduce resistivity.
(2) To adopt a sufficiently effective dust collecting struc-
ture based on new idea even in case electric resistivity
is high.
For item (1), various methods are available. However,
it is difficult ot find out a practical method for general
use because of restrictions from sintering operation and
expenses.
For item (2), formation of a strong electric'field or a
new electric fluid field by utilizing electrode shape and
electrode spaced loading system independently or in
combination was undertaken. However, its effectiveness is
considered not definite in comparison with the con-
ventional EP. At present stage, our company determines
the capacity of equipment on the basis of the statistical
results of actual operations in the past and by assuming
the working voltage on the basis of the characteristic
electrode shape and spacing, rapping strength and mecha-
nism and power equipment control system and by use of
gas temperature as a parameter, obtaining expected results
stably.
In the above, we mainly referred to fine-particle dust.
It is clear from the internal condition of the equipment
that the fine particles are subject to the effect of re-
scattering in the equipment and dust is collected sele-
ctively. These fine psrticles, coupled with the effect of
high electric resistance, cause the comparatively low dust
collecting performance.
Perceiving the behavior of fine particles in fluid, our
company determines the capacity of equipment on the
basis of the characteristic mechanism and system and by
use of gas temperature and a certain component of dust
as parameters. Particularly, it is well known that rapping
of attached dust has a great effect on dust collecting
efficiency.
Problematical points in equipment planning were
enumerated in 3.3.1. They will be discussed later in this
paper.
3.3.3 Relation between dcsulfurixation / denltn-
tion processes and EP
Then is a movement to equip sintering machines with
desulfurication and denitration equipment. Here, EPs are
also playing a. very important role and giving serious
effect on the technical and economical aspects of the
equipment, which is described in detail in 3.4.
3.4 How to utilise EP for desulfurlsation and denitri.
tion processes
3,4.! Utilltation of EP for desulfuritation process
Taking heavy oil fired boilers for example, various
combinations are conceivable for desulfuritation equip-
ment and EP as shown in Table 1. In the past, combina-
684
-------
Table 1 Various combinations of desulfurization equipment
with EP for heavy oil fired boiler
C...
will EP
(EPwthoiilNHilniocHon)
E?
(Dry EP tHor disullurluHon)
NHi
«Wlh-L(D^EPl
(Wit tP)
ProM
kwfficiml oWr
*V»lhritii|
Formtioi of >eM IML
Gmrllioi ot wluto *|MO
orSOj.
bcrtMi cl Mdtl int kordo*
in le roictioi product cirriod
o»or Item dtiuUuriioHoi KOCOM
to Ci«« 1 to 3).
biuflitiwl doMlhritiif.
Btrhrimoci.
ForMlio* ol icid iwrt.
Gmritloo of wUtt »!»•
b»SOi
Pollution it EP
ojr iimporilod nisi.
S0t it Iko ouUn ot diitltiirliitio*
09!,
oo II CM ko colloctoJ o| EP.
tlons 1-3 were most popular. It is desirable for the
improvement of the quality of the by-product of the
desulfurization process, and soot separation and waste
water treating processes to remove dust beforehand by
Installing a dry type EP before the desulfurization equip-
ment. Apart from this point, it is a basic problem for the
determination of the necessity of EP to investigate to
what extent desulfurization equipment can play the role
of removing dust and SO3 which is the role of the con-
ventional EP.
3.4.1.1 Dust removing capacit'y of desulfurization
equipment
The gas cooling section and gas absorbing section of
the wot typo desulfurization equipment widely in use now
have * function of a kind of wet type mechanical dust
collector. Therefore, tho desulfurization equipment dis-
plays a considerable performance for dust of compara-
tively largo grain sizes by giving proper pressure loss to it
and it is considered that in case it is applied to heavy
oil fired boilers, it can remove dust down to 50 mg/Nnr
in outlet dust burden. However, in case a higher dust
removing performance than this is required, it is necessary
to increase pressure loss to a large extent and for such a
caw it is rather economical in many cases to install EP if
the running expense is taken into account.
On tho other hand, in the case an EP is installed
before the desulfurization equipment, dust at the inlet of
tho desulfuricailon equipment can be collected compara-
tively well by the effect of charging and coagulation in
the EP and therefore, it is possible to make an economi-
cal design by combining desulfuri«flon equipment and
EP properly. Table 2 compares 'the performance of
disulfurlzatlon equipment between a case it is equipped
with an EP and mother eau without EP. The dust remov-
Table 2 Desulfurizing and dust-removing performance in case
EP and desulfurization equipment are combined
r~s
r~1 MOT
1 : — i
Howy ol S*I.9K'
, Al Ik tw«
EP tflkMwr
•VIM
Ill
NHi ,,
1 1 | f"~ iH'C 132'C M
L.^
Totol «j,ui!ir ,( Jwi (m,/ymJ)
6C
AOMM of (NH,). SO, in du»t («g/.\m>) 4S
Dwl eolloeliof
rito ojr J«inlf.
Tolil du«l iliniird
(NH,).S04 tonltnt
Totol ) 72
D*«l collMliif
rito li 4owU.
OOJllp«MI IK)
Toltl duu iiudtrd
(NH.).SO, COM..I
57
55
4»
25
ing performance of the desulfurization equipment at the
time the EP was completely stopped is very low as
compared with that at the time the EP was in operation,
and particularly, ammonium sulfate content [(NH«)
mS04) collecting efficiency was very low. This is due to
the fact that ammonium sulfate content which was
created by NH3 injection became a fine particle of
around lp in grain size.
3.4.1.2 SOj removing capacity of desulfurization
equipment
For exhaust gases containing SOa such as those
emitted from heavy oil fired boilers, it was a general
practice to solidify SO) in the form of ammonium sulfate
content by injecting NHj into the gas before removing
with EP and thereafter, remove the ammonium sulfate
with EP for the purpose of preventing low-temperature
corrosion and acid smut formation. As the result, smoke
color caused by SOj mist became thinner.
In contrast, in case desulfurization equipment only was
used (example 1 in Table 1), there was formation of acid
smut and even after reheating of the gas with after-
burner, white plume (violet) remaining phenomenon was
observed, in consequence of which doubts arose a* to the
SOj removing performance of desulfuritatlon equipment,
It wis confirmed as the result of actual measurement that
the SO3 removing rate of deiulfurizatlon equipment to
not so high M Us SO} removing rate.
The reason for the low SOj removing rate of desulfuri-
nit ion equipment is presumed to bo that when |is is
cooled in desulfurization equipment, the greatest portion
of SO) turns into sulfuric acid mist (HjSQ*) in the
temperature range of 140-120*C and tho mist is fino
particle smaller than \n. While S03 gas is diffused tnd
absorbed, this fine H2SC>4 mist can not be collected in tho
desulfurization equipment except by tho removing action
by impact alike in the case of dust. Because of this rela-
tion, in case it is necessary to keep tho discharged quan-
tity of dust at a sufficiently low level Or to remove SOj,
it is essential to us6 EP jointly. For this reason, case 2 in
Table l can be said to be the most desirable method as a
countermeasure to SOj.
685
-------
A rtethod to collect and remove mist and salt gener-
ated by dcsulfuri/ation and denitration equipment by
iitttalling a wet type EP after desulfurization (case No. 5)
is under study. The features of wet type EP are that it is
capable of reducing the density of discharged dust to such
a low level that is difficult for dry type EP to attain and
that it is effective for coupling with denitration equip-
ment described below, making after-burner unnecessary or
reducing operating cost.
3.4.1.3 Relation between denitration process and EP
In the selective catalytic reduction process with am-
monia in the dry denitration process, which is in the most
advanced stage of technical development, some catalyst
is used in a high temperature state of 250 to 400°C gen-
erally. In case exhaust gas contains dust or SOx, it is
often the case that dust attaches to the catalyst, causing
an increase of pressure loss or spoiling of the catalyst, and
countermeasure to this is one of the important problems
of dry desulfurization process.
Concerning this problem, various studies have been
made hitherto, of which the most desirable method is to *
remove dust completely by installing a dust collector
before denitration equipment. This case also involves
various technical and economical problems, depending on
the operating conditions of the dust collector and it is
necessary to investigate these problems for the whole
plant.
In order to enable a dust collector to give full play to
its functions for pretreatment to desulfurization, it is
considered necessary that the dust collector has a high
performance enough to reduce dust burden to a low level
about 10 mg/Nm3 lower than the outlet dust burden
specified on the basis of the exhaust standard or visible
limit. As such high-performance dust collector of this
class, EP is most suitable. For this reason, our company is
also pushing forward the development of dust collectors
for denitration purpose, centering on EP.
Various methods of combination are conceivable for
desulfurization equipment or denitration equipment with
EP. Basically, they can be squeezed to the following three
cases in Table 3. The main problematical points of EP in
each case and the present state of our company's develop-
ment activities are briefly introduced below.
Table 3 Combination of desulfurization equipment, denitrition
equipment with EP for heavy oil fired boilers
ClM
i
NHi
EP
Out it Mtltt
tkfera
UBMin
wlfilt)
Table 4 Combination of desulfurlzatioiv equipment, denitration
equipment and EP (example) for sintering machines
Afttr'burning
686
-------
furization and denitration equipment with EP to be used
for exhaust gas from sintering machines in a steel mill.
This combination type is used for removing dust arising
from solid content remaining after mist eliminator treat-
ment to less than 3 mg/Nm3. The merit of the installa-
tion of wet type EP is that it is capable of high-perform-
ance dust removing as mentioned above, and thereby
increases the effect of the following denitration equip-
ment, and moreover, it can collect mist almost com-
pletely, reducing reheating cost remarkably.
Our company conducted pilot tests to grasp the
properties of gas after various kinds of desulfurization
processes and to confirm the dust collecting character-
istics of EP and furthermore, improved the material
quality, the methods of cleaning, water treatment and
electrode structure. This EP already entered the stage of
practical use and actual units are under production for
heavy oil fired boilers and sintering machines.
4. Improvement of EP performance — EP with damper
In the case of dry EP, attention must be paid to the
re-scattering of dust in the equipment. Concerning rescat-
tering, we referred to in 2.2. Of scattering, rapping scat-
tering is violent. In order to improve dust collecting effi-
ciency by preventing rapping scattering, we developed an
EP with damper (patent pending) as described below,
which is obtaining excellent result.
Fig. 9 shows a rough sketch of the EP with damper.
The interior of the equipment is divided into four sec-
tions, for example (A, B, C, D) and each section is
independent for gas passage, electrode, charging, and
electrode rapping. For dust collection, all sections are
made full use of for almost all time zones (all dampers
are kept open to the full). During this period, electrode
rapping does not take place. With the lapse of preset time
depending on the density and property of dust, the
dampers at the inlet and exit of section A, for example,
are closed to generate electrode rapping only in this
section for dust discharge. Besides, as necessity demands,
the section A works complementary operations such as
charge stopping or quick rapping for the efficiency of
dust discharging.
The time of operation of section A is determined
beforehand. After finish of the operation of section A,
sections B, C and D are operated one after another in the
same manner, completing one cycle. These operations are
Piftitiw Mil
conducted automatically in accordance with a prepared
program.
The locations of the dampers are determined according
to dust collecting performance required. For EPs for
heavy oil fired boilers, they are installed at the inlet and
exit in case outlet dust burden is less than 10 mg/Nm3
and the partition accuracy of the partition structure of
each section is raised.
In case of planning for outlet dust burden exceeding
10 mg/Nm3, a damper is installed at the exit only. As for
the partition of each section, it is enough to partition the
L part only. This system makes it possible to perform
quick rapping after stopping of charging as mentioned
above, and moreover, has a merit to increase the cleanness
of electrodes to a great extent and if this merit is taken
into account, it can be said to be a noteworthy operation
system. Fig. 10 shows the performance of the EP with
damper in comparision with that of the EP without
damper.
Fig. 9 EP with damper
Fig. 10 Relation between dust collecting efficiency
and size of EP
5. Reliability of EP
EPs are the best among various kinds of dust collectors
now in existence both in performance and a wide range
of application. This equipment is always coupled with
production facilities directly or indirectly and their work-
ing ratios and ease of maintenance are matters of great
concern for user. For this reason, it is necessary to pay
proper consideration to each condition, in the design of
the equipment. With the recent progress toward larger air
now rate and" higher performance, the capacities of EPs
have greatly increased (not a few of our products have a
capacity exceeding 15000 m3 per unit), and moreover,
687
-------
they have been improved in heat resistance (up to 400°C)
and pressure resistance (up to 2 SOO mmAq). On the
other hand, EPs have a defect that it is difficult to pre-
vent or control damages as their internal condition cannot
be thoroughly inspected during operation and in most
cases, damages can only be known through changes in
situation.
From this viewpoint, we take up the reliability of this
equipment and briefly introduce the main aim of our
design of the elements of the equipment and actual
products.
S.I EP casing
The casing of the EP with a capacity exceeding 15 000
m3 has a surface area as large as 4000 m3. As it must
have strength and rigidity enough to withstand heat and
pressure besides various kinds of load including the weight
of their internal furnishings, earthquake and wind pres-
sure, it must be built to be a reasonable structural body
backed by very strict strength calculation, since it is ot
comparatively thin steel sheet construction. Furthermore,
as thermal expansion and contraction cannot be disre-
garded, the distance between struts becomes larger than
110 mm maximum in the above example.
Our company entered the actual computerization stage
for structural design several years ago and has succeeded
in establishing, by using structural analysis programs
jointly, reasonable wasteless safe structure designing. For
example, a center pillar structure is adopted for large-type
high-pressure EPs for higher safety. Thermal expansion
and contraction between the casing and the supporting
frame is absorbed by a special pin joint or a plain bearing.
If the strength of the casing is insufficient, a considerable
local deformation will occur (though entire fracture may
not occur), causing changes in the arrangement of internal
furnishings, particularly electrodes. With this point in
mind, our company is making efforts for securing safety
and accuracy.
5.2 Collecting electrode
For functional purpose, a collecting electrode is made*
of thin steel plate and it is necessary to prevent internal
temperature difference and deformation caused by
impacts due to rapping. The collecting electrode devel-
oped by our company is composed of narrow rigid mold-
ings which are combined mutually independently and is
free from individual or overall deformation nor irregular
interval between each molding.
5.3 Discharging electrode
Generally, this is the part most susceptible to damages
and the damage of this part leads to stoppage of the
equipment. The discharging electrode developed by our
company adopts a' short span construction and is stably
charged, almost free from chord vibration and disconnec-
tion. Consequently, it is not necessary to divide the charg-
ing section finely to save the increase in equipment
expenses.
5.4 Rapping
The features of our products-lie in (1) collecting and
discharging electrodes and their supporting mechanisms,
(2) method of fixing striking lever and electrodes, (3)
striking points and (4) individual rapping and even in case
the surface area of the electrode is larger than 70m2, it
can obtain an effective vibration acceleration of about
2000 g near the striking point and several hundred g at
the remotest point. Therefore, even adhesive dust comes
off from the whole surface easily. The striking hammer is
comparatively light in weight and is safe against damage.
Striking noise is small, being 65 phons at a distance of
1 m from the equipment.
5.5 Hopper
The hopper should be designed with emphasis on the
prevention of dust bridge and unavailable gas flow. Our
company adopted, for the former, an angle of repose,
comer plate, downward local cone, through shape, agitator
and vibrator according to the property of dust and arranged
a large-sized gas cut-off plate for the latter. Recently, we
developed a special device (patent pending) to be installed
in the hopper for discharging very light, fine-grained
adhesive dust. This special device is already in actual use
with good results.
5.6 High-voltage insulator
Employing electric dehumidification and blast cleaning
.systems, this high-voltage insulator is free from damage
due to dielectric breakdown, and can be held without
damage as it is supported free of thermal expansion and
contraction.
5.7 Corrosion and wear
As corrosion preventing measures, we enforce heat
retaining and local heating and make severe selection of
materials for dry type EP. For sources of corrosion, the
best efforts are made for the prevention of outbreak of
corrosive atmosphere and for avoiding operation under
corrosive atmosphere. For the wet type whose feed water
system adopts a closed system recently, various kinds of
synthetic nonmetallic materials are used properly to
prevent corrosion by feed water.
The largest wear is the wear of the inlet flue by fluid.
As a countermeasure to this, we improved the shape of
the flue and installed a special current control plate which
obtained satisfactory result for the protection of the wall
surfaces, also serving for the reduction of ventilation loss.
5.8 Explosion and combusion
The major combustible gas which is liable to explode
may be CO. Explosion is a sudden combustion and
combustion easily occurs when (1) combustible gas, (2)
proper oxygen density and atmospheric temperature and
(3) ignition source are present. Our company pays much
attention to the following points and takes the following
safety measures, which have prevented occurrence of
extraordinary accidents in our products.
(1) To check. whether oxygen density is lower than the
safe oxygen density which is determined by the kind
and density of combustible gas contained in mixed gas.
(2) To keep atmospheric temperature off the explosive
limits.
(3) To control the combustion of source so as to reduce
the maximum value of combustible gas density.
(4) In order to check ignition energy on EP side: to
adopt power source equipment which does not gener-
688
-------
ate high-frequency sparks, glow and arc by controlling
voltage by spark detection for each operation, an4 also
of an interlock of CO and Oj densities and power
source equipment. That is, practice of low voltage
operation, always, not accompanied by spark.
(5) The special safe structure enabling the largest-surface
part of the body of EP to open instantaneously.
Generally, the interlocking operation, which detects
combustible gas density and injects inert gas, is apt to be
segregated to density and the detected density does not
always indicate the maximum density. Besides, in relation
to time lag, the interlocking operation cannot be said to
be a sufficient preventive measure.
Conditions for dust combustion are the same as those
for gas. The neighbourhood of specific dust which emits
component .gas with the lowest ignition point ignites first
and its heat energy burns dust in other parts one after
another. Considering that the quantity of dust sticking to
the surfaces of the electrodes which are the most impor-
tant but the weakest parts of EP is responsible for overall
combustion, our company enforced electrode damage
preventing measures (patent pending) including the
above-mentioned measures for preventing explosion. As its
result, we could prevent serious accidents till the present.
6. Conclusion
In this paper, the authors took up EP, particularly dry
type EP preponderantly and described, concerning the
performance and reliability of the equipment, the results
of our research and practical application and the tech-
nique of combining EP with desulfurization and denitra-
tion equipment. It is considered that many technical
problema which have greater effect for our company and
for a number of researchers on dust collecting techniques
from practical engineering viewpoint are still left for
solution.
New needs will be developed for EP in future at the
request of the nation. We feel great responsibility in our
study and solicit continued guidance of the circles con-
cerned.
689
©by Mitsubishi Heavy Industries, Ltd.
(Reproduced by permission)
-------
TECHNICAL REPORT DATA
(Please read Inunctions on the reverse before completing/
REPORT NO.
EPA-600/7-78-110a
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Electrostatic Precipitator Technology Assessment:
Visits in Japan, November 1977
5. REPORT DATE
June 1978
6. PERFORMING ORGANIZATION CODE
7. AUTHORIS)
Grady B. Nichols
8. PERFORMING ORGANIZATION REPORT NO.
Project 3858-5
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Southern Research Institute
2000 Ninth Avenue, South
Birmingham, Alabama 35205
10. PROGRAM ELEMENT NO.
EHE624
11. CONTRACT/GRANT NO.
68-02-2610, W.A. 5
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final: 11/77-4/78
14. SPONSORING AGENCY CODE
EPA/600/13
is.SUPPLEMENTARY NOTES T£RL-RTP project officer is James H. Abbott, Mail Drop 61,
919/541-2925. EPA-600/7-78-110 is the basic report.
i6. ABSTRACT
report gives results of a particulate control technology assessment
visit to Japan by a team of U.S. investigators. The visit included discussions with
personnel from universities , industries , and other major installations involved with
particulate control. Significant research activities were noted in both the academic
and industrial sectors related to particulate control and measurements . The report
summarizes results of the individual discussions, observations during the tour, and
discussions of technical papers. Many valuable technical papers supplied to the U.S.
team are reproduced the the Appendix of the report.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
Air Pollution
Electrostatic Precipitation
Dust
Measurement
Air Pollution Control
Stationary Sources
Japan
Particulates
13B
13H
11G
14B
3. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (ThisReport)
Unclassified
21. NO. OF PAGES
697
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (»-73) gQQ
AU.S.GOVERNMENTPRINTINGOFFICE:l979-6i+0-013' it 2 39 REGION NO. 4
------- |