:PA-650/2-74-102
OCTOBER 1974
Environmental Protection Technology Series
PARTICULATE SIZING TECHNIQUES
FOR CONTROL DEVICE EVALUATION
Office of Reseotch ond Development
U S Environmental Ptotection Agency
Woshmgton. DC 20460
-------
EPA-650/2-74-102
PARTICULATE SIZING TECHNIQUES
FOR CONTROL DEVICE EVALUATION
by
W. B . Smith, K. M. Gushing, and J. D. McCain
Southern Research Institute
2000 Ninth Avenue South
Birmingham, Alabama 35205
Contract No. 68-02-0273
ROAP No. 21ADM-011
Program Element No. 1AB012
EPA Project Officer: D . B . Harris
Control Systems Laboratory
National Environmental Research Center
Research Triangle Park, North Carolina 27711
Prepared for
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
October 1974
-------
This report has been reviewed by the Environmental Protection Agency
and approved for publication. Approval does not signify that the
contents necessarily reflect the views and policies of the Agency,
nor does mention of trade names or commercial products constitute
endorsement or recommendation for use.
11
-------
TABLE OF CONTENTS
List of Figures
Sections
I Conclusions 1
II Introduction 3
A. Inertial Methods 4
B. Diffusional Methods 5
C. Optical Particle Counters 6
III Experimental Methods For Measuring Particle
Size Distributions For Control Device
Evaluation 9
A. Impactors 9
B. Optical & Diffusional Methods 14
IV Field & Laboratory Evaluation Of Various
Inertial Particle Sizing Devices 26
A. Field Tests 26
B. Laboratory Tests 29
C. • Cascade Impactor Data Reduction 60
V Appendix A - Impactor Operating Techniques 66
A. Impactor Selection 66
B. Sampling Time 67
C. Collection Substrates & Adhesives 67
D. Multipoint Dampling 71
E. Impactor Orientation 71
F. Heating Impactor 72
G. Probes 72
H. Balance Requirements 72
I. Sampling Configurations 73
J. Preparing the Impactors 77
K. Taking the Sample 79
L. Disassembling the Impactor 80
M. Data Logging 81
N. Cascade Impactor Data Reduction 81
Appendix B - Extractive Sampling 104
Appendix C.- Diffusional Particle Sizing 109
A. Experimental Procedure 109
B. Data Analysis 111
Appendix D - Optical Techniques 116
References 118
iii
-------
LIST OF FIGURES & TABLES
Figure No.
la
Ib
Table I
2
3
8
Title
Parallel Plate Diffusion Battery.
The batteries have 12 or 100 channels,
0.1 x 10 x 48 era
Page No.
Penetration Curves for Monodisperse
Aerosols. (100 channels, 0.1 x 10 x 48 cm).. 7
Size Fractionating Points of Some Commer-
cial Cascade Impactors for Unit Density
Spheres ,
10
Optical and Diffusional Sizing Systems 15
Fractional Loss Rate of Monodisperse
Aerosols by Coagulation (by Haberl &
Fusco1 *) 18
Calibration Curve for Climet Optical Particle
Counter. Polystyrene latex (PSL) spheres
were used as standards 18
Correlation of Optical and Sedimentation
Diameters. Data acquired using fly ash
obtained from a coal-fired boiler 21
Particle Size Distribution At Inlet And
Outlet of Electrostatic Precipitator At
The Effluent of a Coal-Fired Power Boiler.... 25
Schematic Representation of the Vibrating
Orifice Aerosol Generator 30
Ammonium Fluorescein Aerosol Particles
Generated Using The Vibrating Orifice
Generator. The Particle Diameters are
5.4 ym 33
Percentage Wall Loss vs. Particle Size
Andersen Stack Sampler (0.5 acfm, 25°C,
29.60" Hg) f , 36
iv
-------
LIST OF FIGURES & TABLES
Figure No. Title Page No.
10 Percentage Wall Loss vs. Particle
Size. Brink Cascade Irapactor (0.03 acfm,
25°C, 29.60" Hg) 37
11 Percentage Wall Loss vs. Particle Size.
University of Washington (Pilat) Impactor
(0.5 acfm, 25°C, 29.60" Hg) 38
12 Stage Collection Efficiency vs. Particle
Size. First Stage - University of Washington
(Pilat) Impactor (0.5 acfm, 25°C, 29.60" Hg). 39
13 Stage Collection Efficiency vs. Particle
Size. Second Stage - University of Washington
(Pilat) Impactor (0.5 acfm, 25°C, 29.60" Hg). 40
14 Stage Collection Efficiency vs. Particle
Size. First Stage - Andersen Stack Sampler
(0.5 acfm, 25°C, 29.60" Hg) 41
15 Stage Collection Efficiency vs. Particle Size.
Second Stage - Andersen Stack Sampler
(0.5 acfm, 25°C, 29.60" Hg) 42
16 Stage Collection Efficiency vs. Particle Size.
Third Stage - Andersen Stack Sampler
(0.5 acfm, 25°C, 29.60" Hg) 43
17 Stage Collection Efficiency vs. Particle Size.
Fourth Stage - Andersen Stack Sampler
(0.5 acfm, 25<>C, 29.60" Hg) 44
18 Stage Collection Efficiency vs. \V
Andersen Stack Sampler - Glass Fiber Filter
Substrates (0.5 acfm, 25°C, 29.60" Hg) 46
19 Stage Collection Efficiency vs. V*
Brink Cascade Impactor - Bare Collection Plates
(0.03 acfm, 25°C, 29.60" Hg) 47
20 Stage Collection Efficiency vs. Vf
Brink Cascade Impactor - Glass Fiber Filter
Substrates (0.03 acfm, 25°C, 29.60" Hg) 48
-------
LIST OF FIGURES & TABLES
Figure Wo. Title Page No.
21 Stage Collection Efficiency vs. Y¥
Brink C&scade Impactor - Greased Plates
90.03 acfm, 25°C, 29.60" Hg) 49
22 Stage Collection Efficiency vs. v^
University of Washington (Pilat) Impactor
(0.5 acfm, 25°C, 29.60" Hg) 50
23 Stage Collection Efficiency vs. Particle
Size. Brink C2 (Externally Mounted)
Cyclone (0.03 acfm, 25°C, 29.60" Hg) 52
24 Stage Collection Efficiency vs. Particle
Size. Brink C3 (Inline) Cyclone
(0.03 acfm, 25°Cr 29.60" Hg) 53
25 Stage Collection Efficiency vs. Particle
Size. Andersen Cyclone Precollector
(0.5 acfm, 25°C, 29.60" Hg) 54
26 Stage Collection Efficiency vs. Particle
Size. McCrone Cyclone T2A - 0.13* H20 55
27 Stage Collection Efficiency vs. Particle
Size. McCrone Cyclone T2A - 1.00" H20 56
28 Stage Collection Efficiency vs. Particle
Size. McCrone Cyclone T2B - 0.80" H20 57
29 Stage Collection Efficiency vs. Particle
Size. McCrone T2B - 20.0" H20 58
Table II Environmental Research Corporation Tag
Sampler (Greased Plates) 59
30 Stage Collection Efficiency for a Brink
Cascade Impactor with a Pre-collector
Cyclone 61
31 Simulated Impactor Tests: a) Input "true"
distribution, b) Calculated distribution
(Picknett method), c) Calculated distribution
(D50 method) 62
vl
-------
LIST OF FIGURES & TABLES
Figure No. Title Page No.
32 Simulation of a Continuous Particle
Size Distribution 64
33 Impactor Tests Using a Known Particle Size
Distribution 65
Al Nomograph For Selecting Nozzles For Iso-
kinetic Sampling 68
A2 Sampling Time Determination For Total Mass
Collection of 25 Milligrams 69
A3 The Original Brink Impactor System, BMS-11,
Set-up for out of stack sampling 74
A4 An Improved Method of Flow Control for the
Brink Impactor using a calibrated orifice
flowmeter 75
A5 The typical Andersen or Pilat setup using
calibrated orifice and dry gas meter 78
A6 Form used for logging field test data 82
Bl Extraction and Dilution system for out of
stack flue gas sampling 105
B2 Diffusional Adsorption Apparatus for Removal
of H20 From Sample Aerosol 108
Cl Diffusion battery and condensation nuclei
counter lay-out for fine particle sizing 110
C2 Nomograph for determining fine particle size
distribution using diffusion battery data....113
C3 Experimental diffusion battery data to be used
in conjunction with Figure C2 in determining
fine particle size distribution 114
vii
-------
I. CONCLUSIONS
In this study, three techniques were used to measure
particle size distributions in industrial plant environments
under a variety of conditions. Cascade impactors were primarily
used in the particle size range from 0.5 to 20 ym, diffusion
batteries and CN counters in the size range from 0.01 to 0.2 ym,
and optical counters in the range from 0.3 ym to 2 ym. In
addition, laboratory tests were performed on several impactors
and cyclones to measure their stage collection efficiencies.
An Institute-built vibrating orifice aerosol generator was used
in our calibration work.
It was found that when used properly, and with extreme care,
cascade impactors can give reliable particle size information.
No single impactor was versatile enough, however, to make
accurate measurements under the extreme variations in grain
loading which occur between inlets and outlets on control devices,
or from site to site. Low flow rate impactors were found to be
most useful in the case of high grain loadings, while high flow
rate impactors are more convenient in low grain loading situations.
With the exception of the first stage of the University of
Washington Mark III Source Test Cascade Impactor, all of the
impactor stages tested had stage collection efficiencies which
agreed fairly well with the theory of Rantz and Wong. Wall losses
were significant, explaining the fact that cascade impactors
normally give grain loadings about 30% lower than those measured
using the EPA method five. •
Diffusional sizing was found to be practical with diluted
samples, although the procedure is tedious and requires extensive
sampling times. This method is most accurate and useful on
sources where the particle size distribution and number concentra-
tion are relatively stable.
The dilution system which was used with the diffusional and
optical sizing systems was reliable and linear over a wide range
of dilution ratios. Typical dilution factors varied from
-------
-2-
30 to 300 at control device outlets to 200-1000 at the inlets.
Although emphasis was not placed on the use of optical
particle counters during this project, they proved valuable
because their useful size ranges overlapped with that of the
cascade impactors. This allowed for a check on the quality of
the impactor size distribution measurements. In general, the
particle size distributions and fractional efficiencies cal-
culated from optical data and impactor data were in good
agreement.
The system used in this study has proven the feasibility
of measuring particle size distributions in stack gases, and
fractional efficiencies, from 0.01 pro to 15 ym. We believe
that this first generation system represents a viable package
which can be improved and will continue to be useful in making
these extremely important measurements related to the control
of fine particles.
-------
-3-
II. INTRODUCTION
In the last few years, the emphasis on particulate control
has shifted toward the fine particle size range. Fine particles
are defined as particles which have diameters of less than 3 urn.
These particles present special problems because they remain in
the atmosphere for much longer periods of time and because they
follow the gas stream into the depths of the respiratory system
where a large percentage is deposited. Fine particles are more
difficult to control than large particles and contribute more to
the scattering of light in the visible range. In contrast,
large particles are collected with higher efficiency by most
particulate control devices, and those particles which penetrate
the control device tend to settle out of the atmosphere more
quickly. For example, the terminal, or settling, velocity of a
10 ym diameter particle of unit density is 0.3 cm/sec., while
that of a 1 ym diameter particle is 0.003 cm/sec. Although
most of the mass emitted from a particular pollution source may
consist of large particles, in general, the largest number
of particles is in the fine particle range. Thus, high mass
collection efficiency does not always imply high number collection
efficiency.
The work described in this report consists mainly of develop-
ing, improving, and evaluating experimental techniques and
equipment for measuring fine particle size distirbutions and the
fractional efficiency (efficiency as a function of particle
size) of pollution control devices. Laboratory tests have been
performed to evaluate commercial sizing instruments and to
find the optimum operating conditions for these instruments.
A number of field tests were performed which allowed additional
evaluation of the various instruments in harsh industrial
environments and which were extremely useful in developing
practical sampling techniques and procedures for making particle
size distribution measurements.
-------
-4-
In order to determine fractional efficiency of particulate
control devices, measurements must be made of the size and
concentration of particles suspended in the flue gas at both
the inlet and outlet of the devices. In making these measure-
ments, one encounters a wide variety of testing conditions.
Flue gas temperature, pressure, moisture content, and the physical
properties of the particulate vary widely from device to device.
Concentrations vary by orders of magnitude from inlet to outlet,
and from one control device to another. Because of this com-
plexity and range of testing conditions, and the limited useful
size range of individual testing devices, more than one instrument
is required to give complete information on the particle size
distribution, even at a single site.
Instruments which seem to show the most promise for use in
measuring particle size distributions under plant conditions
are: inertial sizing devices, diffusion batteries with con-
densation nuclei counters, optical particle counters, opacity
monitors, and electrical particle counters. The initial goal
of this study was the evaluation of inertial devices, but
diffusional and optical techniques were also used during many
tests and considerable insight was gained into their usefulness
as part of a complete particle sizing package. Thus, this paper
summarizes our work on all three techniques.
A. Inertial Methods
Impactors, impingers, cyclones, and centrifuges have been
used for many years for sampling and estimation of particle
size. Because of its compact arrangement, the cascade impactor
has generally been found to be the most suitable inertial device
for mass distribution measurements of pollution emission sources.
In most cases, the impactors can be inserted directly into the
duct or flue, thus eliminating many condensation and sample
loss problems which occur when probes are used.
-------
-5-
Early work on impactors was done by May1 and published in
1945. A cascade impactor similar to the one described by May has
been manufactured for a number of years in England by C. F.
Casella and Company. During the 1950's, Andersen,2 and Brink3
both developed cascade impactors for special applications and a
considerable amount of theoretical and experimental work with
cascade impactors was done during this period. Much of this
early work was guided by the extensive empirical and theoretical
studies conducted by Rantz and Wong1* and published in 1952.
Pilat5 has provided a coneise background of work on cascade
impactors and a description of the development of modern devices
that are suitable for air pollution research.
Existing impactors will yield information leading to a
particle size distribution over the range in diameter from
0.3 ym to 15 ym. Since the amount of particulate mass collected
depends upon the particulate concentration and volume of gas
sampled, the design flow rate is an important consideration in
selecting an irapactor for a particular test situation.
Cyclones normally are most useful in sizing particles
which are larger than 10 ym and in collecting more total mass
than impactors. For sub-micron work, cyclones are useful as
precollectors to prevent large particles from entering the system.
B. Diffusional Methods
Liu et aJ.6 have developed an electrical sizing device for
sub-micron particles, and Pilat7 is developing a low pressure
impactor for sizing of particles down to 0.02 ym. Neither of
these devices, however, has been used for fractional efficiency
measurements. The only practical means, other than electron
microscopy, for sizing particles in the 0.005 to 0.2 ym diameter
size range is with diffusion batteries. A diffusion battery
consists of a number of long, narrow, parallel channels, or a
cluster of small bore parallel tubes. Variations in length and
number of channels (or tubes) and in the aerosol flow rate are
-------
-6-
used as a means of measuring the number of particles in a
selected size range. As the aerosol moves, in streamline flow,
through the channels, the random Brownian motion of a particle
causes it to be displaced from its original position in the air
flow streamline. The most probable displacement from a stream-
line is zero, but the root mean square displacement is propor-
tioned to the square root of the travel time. Consequently,
some of the particles are displaced sufficiently to reach the
walls of the battery. It is assumed that once a particle contacts
the battery wall, it will adhere. Therefore, only a fraction
of the influent patticles appear at the effluent from the battery.
Because particle diffusivities increase with decreasing particle
size, the extent of the penetration of the battery will depend
on the particle size. The penetration of the battery can be
measured with a condensation-nuclei (CN) counter. Figure la
shows a typical diffusion battery geometry used in the work to be
described in this paper. Figure Ib shows the penetration charac-
teristics. For reasons to be discussed in the next section,
extractive sampling is necessary for diffusional sizing.
C. Optical Particle Counters
Photoelectric or optical particle counters function on
the principle of light scattering. Each particle in a continuous
flowing sample stream is passed through a small illuminated
volume. Light scattered by the particle is imaged on the surface
of a photodetector during the time the particle is illuminated.
The intensity of the scattered light is a function of particle
size and index of refraction. Photoelectric particle counters
will give reliable information if the concentration of particles
is such that the probability of illuminating more than one
particle at a time is low. Typically, this restriction places
-------
-7-
Figure la. Parallel Plate Diffusion Battery. The
batteries have 12 or 100 channels, 0.1 x 10 x 48
cm.
0.01
Particle Diameter,
Figure Ib. Penetration Curves for Monodisperse
Aerosols. (100 channels, 0.1 x 10 x 48 cm)
-------
-8-
an upper limit of about 300 particles/cm3 for instruments
providing size information down to diameters of 0.3 ym, the
practical lowest sensitivity for optical sizing. Thus, optical
particle counters are effective in the same size regime as
inertial sizing devices. Advantages are real-time readout and
particle size data that are directly related to atmospheric
visibility and plume opacity. Disadvantages are the necessity
to dilute the sample to number concentrations less than
300/cm3 and the dependence of the calibration upon the index
of refraction.
By using all three of the techniques mentioned above,
inertial, optical, and diffusional sizing, data can be obtained
over the size range from 0.005 ym to about 20 ym diameter with
some overlap. In the next section, the experimental methods
with results will be described in more detail and the short-
comings of the instrumentation in the present stage of develop-
ment will be discussed. Appendices are included which outline
the actual procedures used for making particle size measurements
in essentially "cook book" fashion. These appendices also
include instruction, charts, and computer programs for data
reduction.
-------
-9-
III. EXPERIMENTAL METHODS FOR MEASURING PARTICLE SIZE
DISTRIBUTIONS FOR CONTROL DEVICE EVALUATION
A. Impactors
1. General considerations in selecting the best jmpactor
for a specific test condition -
Table I shows some characteristics of several commercially
available inertial classifiers. It is usually impractical to
use the same impactor at the inlet and outlet of a pollution
control device for efficiency measurements because of the
difference in particle concentration. For example, if a
sampling time of thirty minutes is adequate at the inlet
(control device efficiency of 99% is assumed), for the same
amount of sample to be collected, approximately 3,000 minutes
sampling time would be required at the outlet. Although the
flow rates are somewhat variable, they cannot be adjusted
enough to compensate for this difference without creating other
problems. Extremely high flow rates result in particle bounce
and in scouring of impacted particles from the lower stages
of the impactor where the jet velocities become extremely high.
Short sampling times may result in atypical samples being
obtained as a result of momentary fluctuations in the particle
concentrations or size distribution within the duct. A practical
solution might be to choose, for example, to sample at the
inlet with a Brink impactor and at the outlet with an Andersen.
The sampling times are then adjusted to those times that are
required to collect a weighable sample of small particles on
the fine particle (lower) stages without accumulating so much mass
on the upper stages that overloading and reentrainment occur.
2. Specific techniques used in this study -
The actual extraction of a size-fractionated sample from
a gas stream using cascade impactors is a well established
-------
-10-
Table I
Size Fractionating Points of Some Commercial Cascade
Impactors for Unit Density Spheres
Stage
Cyc
0
1
2
3
4
5
6
7
8
Modified
Brink
0.85 LPM
18.0 ym
11.0
6.29
3.74
2.59
1.41
0.93
0.56
Andersen
Mark III
14 LPM
14.0 ym
8.71
5.92
4.00
2.58
1.29
0.80
0.51
U. of W.
(Pilat)
14 LPM
39.0 ym
15.0
6.5
3.1
1.65
0.80
0.49
E.R.C.
Tag
14 LPM
11.1 ym
7.7
5.5
4.0
2.8
2.0
1.3
0.9
0.6
-------
-11-
procedure at this time,1'5'7 and is summarized in Appendix A.
The following paragraphs deal in general terms, mainly with the
unexpected or non-ideal behavior encountered. Section IV
contains specific information describing our evaluation of
several commercial impactors.
• Wall Losses - Particles are lost within an impactor
by diffusion and impaction on the walls and jets. Lundgren,8
and Gussman et al9 have shown that the losses per stage can
amount to as much material as is collected by the stage. For
fractional efficiency measurements, wall losses might not have
a large effect on the accuracy if the percentage of losses are
the same in the impactors used at the inlet and outlet. With
mass concentration measurements, however, uncorrected errors
of this magnitude could completely void the data. Measurements
of mass concentrations taken with impactors during this study
have consistently produced results about 70% as large as those
obtained by the non-sizing, standard techniques.
• Reentrainment - All the particles that strike a
collection stage do not stick. A technique used to enhance the
retention of particles at the original impaction site which is
gaining wide acceptance is coating the substrates or collection
stages with a suitable viscous grease. In using grease, however,
proper precautions must be taken to prevent loss of mass due to
the evaporation of the volatile evaporents. One approach that
was found to be satisfactory was to make a 10 to 15% suspension
of Dow Corning high vacuum silicone grease in benzene. After
placing a drop of this suspension on the collecting surface
and allowing the benzene to evaporate, the coated substrates
are baked for about an hour at 400°F and desiccated until
actually used. In at least one case, anomolous weight losses
were attributed to the hot grease flowing from the substrates
when the impactor was operated in a horizontal position within
a hot duct. As yet, no coating suitable for use at temperatures
over 400°F has been found.
-------
-12-
Even with greased substrates, significant scouring and loss
of material occurs on the last stages when the jet velocities
are too high. Experimentally, the value of about 65 m/sec has
been established as the maximum velocity obtainable without
reentrainment or grease erosion using greased substrates, and
about 35 m/sec using ungreased substrates. In effect, these
phenomena place upper limits on the flow rates at which an
impactor may be operated to obtain a valid particle size
distribution.
The Andersen Stack Sampler uses glass fiber filters as
impaction substrates. These filters seem to be a satisfactory
alternative to greased substrates in minimizing particle
bounce and reentrainment. Glass fiber substrates have also
been used with the Brink impactor at high temperatures.
• Weighing Accuracy - To achieve an accuracy of about
20% in control device penetration measurements, the mass
loading in the chosen size intervals must be measured to an
accuracy of about 10% at both the inlet and outlet of the control
device. If the mass collected in the large particle size ranges
is kept reasonably low, to avoid reentrainment, this means that
a weighing accuracy of 10 to 30 yg is required for the small
particle size ranges where much less mass is collected.
Techniques which have been used to minimize problems in
weighing accuracy are: to reduce the tare weight of the collection
stages by using light weight inserts made of aluminum foil,
stainless steel shim stock, or glass fiber filter material;
to purposely bias the sample against large particles by pointing
the nozzle downstream; or to use cyclone precollectors to remove
large particles prior to the sample entering the impactor.
The use of lightweight substrates and cyclone precollectors
is becoming more common. This technique gives the maximum
weighing accuracy and still permits isokinetic sampling. The
capture efficiency of the precollector cyclone must be taken
-------
-13-
into account when doing data analysis, however, or the results
will not be representative of the true aerosol for large
particle sizes.
• Isokinetic Sampling - If the velocity of the aerosol
stream entering the impactor nozzle is the same as the velocity
at the stream in the duct, the sampling is said to be "isokinetic."
If the sampling velocity is too high, large particles having
greater inertia cannot follow the streamlines into the nozzle.
In this case, the sample would be biased against large particles.
If the sampling velocity is lower than the stream velocity,
particles with higher inertia will be oversampled, again biasing
the results. Isokinetic sampling is necessary for particles
having aerodynamic diameters greater than above 3 to 5 ym.
A selection of nozzles having different bores is available
for each impactor so that for a unique flow rate, several
velocities may be achieved at the inlet of the impactor. Once
a particular sampling velocity is chosen, however, it cannot
be changed during a test because the fractionation points at
each collection stage would change. Since the velocity is fixed
during a test, any fluctuations in gas velocity with time
or position within the duct will introduce measurement errors
in the data on the larger particles.
• Shape of Particles - The most common parameter used in
describing particle size is the aerodynamic diameter. This is the
diameter of a sphere of unit density which has the same settling
velocity as the particle.
The Stokes or physical diameter may also be used to describe
a particle size. This is the diameter of a sphere having the
same density and terminal velocity as the particle.
• Density of Particles - Collection efficiencies for
cascade impactors are calculated using the equations developed
-------
-14-
by Rantz and Wong.1* The cut points (D$o)* may be determined for
unit spheres (as was done in Table I) and corrected for density
as required, or calculated for spheres having the estimated
density of the aerosol particles. An "average" density can be
calculated from true particle volume-weight data taken with
a helium picnometer for a representative sample. If the chemical
composition of the particles is known, the bulk density may be
used. In some cases, the aerodynamic diameter may be the only
information needed and density is not a factor.
The validity of size information based on an average
density depends upon the uniformity of the density from particle
to particle,. A mixture of particles having very different
densities could cause large errors in size distributions.
B. Optical and Diffusidnal Methods
1. Dilution
Because of the concentration limits for operating optical
counters (300 particles/cm3) and condensation nuclei (CN)
counters (105 particles/cm3), problems with condensation in the
sampling lines, and losses due to agglomeration during the
measurement process, it is necessary to dry and dilute the
sample aerosol before it reaches these devices. The sampling
and dilution problem is simplified; however, in the particle
size range of interest here (0.02 - 2 urn diameter) because
isokinetic sampling is not required.
Figure 2 shows schematically the testing configuration for
optical and diffusional sizing. The sample is introduced at
the apex of a perforated cone and clean dilution air is pumped
*D50 is the particle diameter for which a particular stage
has a collection efficiency of 50%.
-------
-15-
Flowmeters
Cyclone Pump
Process
Exhaust
Line
Charge
Neutral!zer
Cyclone
(Optional)
ll x Flowmeter
Particulate
Sample Line
Orifice
Diffusion
Battery
Dilution
Device
Manometer
Recirculated
Clean Dilution
Air
Filter
Orifice
Manometer
Aerosol
Photometer
Diffusional Dryer
(Optional)
Charge
Neutralizer
Pressure
Balancing
Line
Pump
Bleed
Figure 2. Optical and Diffusional Sizing System
-------
-16-
through the perforations, creating a highly turbulent mixing
zone. In verifying the performance of the diluter in the
laboratory using test aerosols, it was found that the calculated
and measured dilution factors agree to within the uncertainty
in measuring the sample and dilution air flow rates. The
concentration was found to be uniform for sampling points
from wall to wall across the body of the diluter.
Drying is accomplished simultaneously with dilution by
recirculating the filtered dilution air through condensers and
drying tubes. In this way, the concentrations of water and other
condensable vapors in the gas stream are reduced by approximately
the same fraction as the particle concentration. If drying were
not performed, the sample aerosol would be below the dew point
for many applications and condensation would occur in the sampling
lines. When the duct is below atmospheric pressure, and cool
ambient air leaks into the port, it is often necessary to heat
the probe to avoid condensation before the sample reaches the
diluter.
Plugging of the sample metering orifice can be a problem,
even when condensation does not occur. To prevent this, a
cyclone precollector with a D5Q of about 2 ym is used to eliminate
large particles. Typical sample flow rates are from 0.1 1pm to
5 1pm with cyclone flow rates of about 14 1pm.
Strom10 has done experiments to determine the significance
of particle losses in sample lines. Two mechanisms were considered:
1. settling in horizontal lines, and 2. impaction at the bends.
For Strom's particular system (8 m of horizontal and 8 m of
vertical stainless tubing 16.8 mm in diameter, with three 90°
elbows, Reynold's No. >_ 2000), 90% of the particles of 15 ym
diameter were lost, 30-50% of the 8 ym diameter, and almost
none having diameters less than 2 ym. Diffusional losses in the
probes and sample lines used in our work are estimated to be
-------
-17-
about 98% at 0.001 ym diameter, 25% at 0.005 ym. diameter and
12% of the 0.01 ym diameter. Thus, losses in the sampling lines
used in this work are probably not significant for particles
having diameters between about 0.005 ym and 2 ym.
Figure 3, after Haberl and Fusco,11 shows the losses due to
coagulation at various particle sizes and concentrations for
monodisperse aerosols. The loss rate of a particular size, in
this case, is inversely proportional to the concentration.
Because of tiie long residence time in diffusion battery sizing,
significant losses can occur if the concentration is not
reduced by dilution.
2. Optical Sizing
Procedures applicable to measurement of particle sizes
and concentration by light scattering have been reviewed by
Hodkinson,12 and more recently by Berglund.13 In addition,
Berglund compared the actual response of three commercially
available optical particle counters with experiment using
particles with different indices of refraction.
The basic operating principle for optical particle counters
is the scattering of light by individual particles as they pass
through a small viewing volume; the scattered light being measured
by a photomultiplier tube.
A Climet Particle Analyzer, Model No. CI-201 was used in
this study. This instrument uses "near forward" light scattering
to obtain particle size information. The view volume is the
volume shared by the focused light beam and aerosol stream.
The view volume is located at one focus of an elliptical
mirror, and the photomultiplier at the other. The amplitude
of the scattered light pulses is related to the particle size
-------
-18-
0.
)OOI
Figure 3.
0.001 0.01 O.I I 10
Fractional Loss Rate per Second by Coagulation
Fractional Loss Rate of Monodisperse Aerosols by
Coagulation (after Haberl and Fusco11).
1.0
0.2
PANEL METER
Indicated True
Dla. Dlo.
0.3 0.45
0.3 1.0
1.0 1.6
3.0 3.0
0.3
0.5
1.0
2.0 3.0
Particle Diameter, jim
Figure 4. Calibration Curve for Climet Optical Particle
Counter. Polystyrene latex (PSL) spheres were
used as standards.
-------
-19-
and the rate at which the pulses occur is related to the particle
concentration. Thus, a counter of this type gives both size and
number information. Operated in the configuration shown in
Figure 2, with a cyclone precollector, the counter has an
effective upper limit of about 1.5 ym for sizing. Also, the counter
has an inherent lower limit of 0.3 ym. Thus, the counter responds
to a limited size range, but gives information in an important
regime, that where data from the diffusional and impactor
measurements converge. The occurrence of more than one particle
in the viewing volume is interpreted by the counter as a larger
single particle. To avoid this effect, dilution to about
300 particles/cm3 is necessary. This was done in the manner
described in the previous section.
The intensity of the scattered light depends upon the viewing
angle and upon the particle size, index of refraction, absorp-
tivity, and shape. Some generalizations may be made from
Hodkinson's review:* 2
When using white light as a source of illumination, the
response to an assembly of randomly oriented, identical, non-
spherical particles will be the same as that for the spherical
particles of equal mean volume whose polydispersity resembles
the "orientation polydispersity" of these particles.. Since
particles sizes are usually given in terms of some "diameter",
the shape factor is relatively unimportant, but may result in
more apparent polydispersity than actually exists.
If the scattered light is collected in some small angle
about the forward direction, and white light is used, the de-
pendence of the amplitude upon index of refraction can be
minimized. For particles greater than 2 ym in diameter, the
amplitude of the forward scattered light pulse is almost
independent of the index of refraction and is proportional to the
cross-sectional area of the particles. For particles less than
-------
-20-
2 ym in diameter; however/ there is no simple relationship
between particle size and amplitude, and the index of refraction
is more important. Berglund found that measurements deviated
from theory by as much as a factor of two in the size range
of 0.4 urn to 1 ym and, in some cases, the response was multivalued.
Figure 4 shows our laboratory calibration for polystyrene
latex spheres having an index of refraction of 1.6. Because
of the uncertainty in making a theoretical correction for index
of refraction (n) , the optical data was correlated with sedimen-
tation data, a method which is independent of n. Referring again
to Figure 1, if the diffusion batteries are laid on their sides,
so that the long dimension of the slots is horizontal, the most
important mechanism for the removal of large particles will
be sedimentation, with rather high efficiencies being obtained
for micron-sized particles. Concentrations of particles
entering and exiting the sedimentation chambers can be measured
with an optical particle counter. Sizes obtained in this way can
be correlated with impactor or optical pulse height intensity
data. Figure 5 shows data taken using the optical counter in
conjunction with the sedimentation chambers while measuring the
effluent from a coal-fired cyclone steam boiler. The particle
size as indicated by the pulse height correlates very well with
that calculated from sedimentation rates.
3. Diffusional Sizing
General - Diffusional sizing consists of passing an aerosol
through a large number 6f small tubes in parallel, or between
closely spaced parallel plates, and measuring the collection
efficiency of the system (diffusion battery). Diffusional
sizing is most useful in determining particle size distributions
below 0.1 ym diameter.
-------
-21-
• 12 Channel battery data
+ 100 Channel battery data
2
^
0.1 0.5 1.0 5.0
Indicated Optical Diameter (equivalent PSL dia.), microns
10
Figure 5. Correlation of Optical and Sedimentation Diameters.
Data acquired using fly ash obtained from a coal-fired
boiler.
-------
-22-
Fuchs1* has reviewed diffusion battery sizing work up to
1956, while Sinclair," Breslin et al, " Twomey,- and Sansone
and Weyel have reported more recent work, both experimental
and theoretical.
Diffusional measurements are less dependent upon the
aerosol parameters than the other techniques discussed and perhaps
are on a firmer basis theoretically. By changing the flow
rate and diffusion battery geometry, a large number of data
points can be established from which a,particle size distribution
may be calculated. Disadvantages are the bulk of the diffusion
battenes, although advanced technology may eliminate this
problem15'" and peculiar problems introduced by particular
testing situations. For example, when testing emissions with
a high S02 content, on one occasion, particles were
actually "created" within the diffusion batteries by oxidation
of S02 to SO3 and subsequent formation of a sulfuric acid
condensate. As in the case of optical sizing, dilution and
drying of the sample air is necessary to prevent coagulation,
growth of hygroscopic particles, and water condensation.
Theory, - The geometry chosen for this experimental work
was that of parallel plates, partly because of ease of fabri-
cation and availability of suitable materials (Figure la), but
also because sedimentation can be ignored if the slots are vertical
while additional information can be gained through settling, if
the slots are horizontal. Sinclair15 and Breslin et al16
report success with more compact, tube-type arrangements in
laboratory studies. The mathematical expression for the
penetration (n/no) of a rectangular slot or parallel plate
diffusion battery by a monodisperse aerosol was given in series
form by Gormley. '• The coefficients were calculated and tabulated
-------
-23-
by Twomey17 using a computer. The equation is:
n/n° = 0.91e-xmD + O.OSSe-11-37*"10 + 0 .OlSe'33-06xmD + 0.0068e-66-06xinD
+ 0.0037
-------
-24-
rates for each configuration, a large number of data points
can be obtained. The inlet and outlet concentrations may be
measured with a single CN counter, or in cases where the con-
centrations fluctuate rapidly, the inlet and outlet concentra-
tions can be monitored simultaneously using two CN counters.
Flow pulsations caused by the cyclic processes in the CN counters
were minimized by using anti-pulsation devices as described by
Sinclair.l5
Because of the time required to accomplish diffusion sizing,
this technique is most useful where concentrations do not
fluctuate rapidly. Diffusional sizing is independent of density,
index of refraction, and to a large extent, shape. Further,
if the response of the CN counter is linear, diffusional
sizing is independent of errors in calibration of the CN counter.
The calibrations of the instruments used in these studies were
provided by the Environment One Corporation with the exception
that concentrations were kept below 105 particles/cm3 in
accordance with the results of recent work by Liu.21
Figure 6 shows data taken at the inlet and outlet of a
coal fired power boiler using all three techniques discussed
above. The agreement among the methods in the overlap ranges
is quite good.
-------
E
•v
6
I0'2
10"
I08
107
INLET
OUTLET
a a
a a
a D a a a
a a
O.OOI
O IMPACTOR
A OPTICAL
D DIFFUSIONAL
O.OO5 O.OI 0.05 O.I 0.5
PARTICLE DIAMETER , MICROMETERS
5 10
ro
ui
I
FIGURE 6. PARTICLE SIZE DISTRIBUTION AT INLET AND OUTLET OF ELECTROSTATIC
PRECIPITATOR AT THE EFFLUENT OF A COAL FIRED POWER BOILER.
-------
-26-
IV. FIELD AND LABORATORY EVALUATION OF VARIOUS
INERTIAL PARTICLE SIZING DEVICES
A. Field Tests
Extensive field and laboratory studies have been conducted
during this project to evaluate the performance of commercially
available inertial particle sizing devices. Our early field
test evaluations were reported in Progress Narrative Number 10
(NTIS Pub. No. PB-226-292/lwp). Since that time, we have per-
formed several additional field tests and as a result, have
gained additional insight into the proper use of these devices
to yield valid particle size distributions. The following
paragraphs are a summary of the knowledge gained through our
field experience with cascade impactors.
• Of the inertial classifiers tested, no single device
was found to be useful over the entire range of particle size
and concentration needed for evaluation of a wide variety of
particulate control device. In general, the inertial classi-
fiers designed to operate at low sample flow rates are useful
at the inlet of a control device where the particulate loading
is high; the low flow rates permit reasonably long sampling
times to be used. If these same low-flow-rate classifiers are
used at the outlet of a control device, the sampling time may
be impractically long, so that a high-flow-rate device is usually
a better choice at the outlet.
• One possible solution to the problem of measuring fine
particles in the presence of high concentrations of large
particles is to use one or more small cyclones in front of the
impactor. Modified Brink Impactors and the Andersen Model III
were successfully used in this manner during our field tests.
The Andersen cyclone was supplied by Andersen 2000 Inc.. The
Brink Inline (C3) Cyclone and Brink External (C2) Cyclone were
designed and built by Southern Research Institute.
-------
-27-
• Most of the impactors have collection stages that are
too heavy for obtaining accurate measurements of the weight of
particles collected in a size fraction. It is helpful to
cover the stage with a lightweight collection substrate made
of aluminum foil, teflon, glass fiber filter material, or other
suitable lightweight materials depending on the particular
application. With such an arrangement, it is possible to make
accurate weight determinations of collected samples that are
small enough to prevent or at least minimize reentrainment.
Weighing to a precision of at least 30 yg appears to be required
for impactor size determination for submicron particles and
10 yg precision is desirable.
• When impactors are operated at flow rates higher than
some critical value, particle bounce can lead to incorrect
sizing with non-cohesive particulates. This is especially
true in the lower (fine particulate) stages. In these lower
stages, the high jet velocities cause scouring of the plates
which are quite noticeable and is an almost certain indication
of reentrainment. Experience with the University of Washington
Mark III and the Brink Impactor indicate that the maximum jet
velocities that can be used for impaction on ungreased substrates
without severe reentrainment resulting from particle bounce is
about 40 m/sec. Also, the physical properties of the particles
can effect the maximum loading that can be safely obtained on
an impaction stage. For example, with some emission sources,
a small quantity of cohesive particulate materials will form a
pile that will plug the stage orifice. For other types of emission
sources with dry particulates, reentrainment and particle
bounce problems can be severe, even with low flow rates. The
use of a silicone grease that is stable at high stack temperatures
is helpful for measuring dry particulates. It has been our
experience that even with this grease, erosion and scouring can
-------
-28-
occur on the lower stages with high jet velocities (velocities
greater than 65 m/sec.)* The scouring and reentrainment can be
reduced by reducing the flow rate in the impactor. For example,
if the Brink sample flow is reduced from the recommended value of
2.8 1/min (0.1 cfm), to approximately 1.4 1/min (.05 cfm),
the scouring effect can be eliminated. This change in flow
rate increases the cut point of each size fraction and decreases
the amount of information that can be obtained in the small
particle size range. These comments are appropriate for the
Andersen and other impactors as well as the Brink. However, the
addition of a final stage to the impactor with the proper combi-
nation of \elocity and jet diameter would make it possible to
regain the information lost at the lower flow rate. This has
been done on the Brink by adding a "6th" stage designed and built
at Southern Research Institute.
• Glass fiber "back up" filters have generally been used
after the final stage in our impactor tests. For some emission
sources, the back-up filter frequently showed a weight loss
instead of a weight gain. This result appears to be related to
filter handling problems and poor design of filter retainers
and seats. Care must be taken to avoid tightening of the Andersen
Impactor too severely causing the steel o-rings to tear the filter
substrates and leading to weight losses if pieces of the
filter are lost or blown off.
• Particle size measurements have been made with the
impactors mounted outside the stack. Sampling probes and lines
*Also, under some flue gas conditions, such as high
SO content, we have experienced severe weight loss with the
Ji
high vacuum grease. The reason for this weight loss is still
unknown.
-------
-29-
were used to transport the sample to the impactor. Based on
simultaneous size determination with the instack and out-of-
stack sampling arrangements, it seems almost certain that a
significant quantity of particles are lost in the probe and
sampling lines, particularly in sizes larger than 2 urn.
•A comparison of all of the fractional efficiency
curves shows that measurements with the modified Brink Impactors,
Brink with grease, Brink kit, Andersen Model II,aand U. of W.
Mark III all seem to result in about the same calculated efficiency.
B. Laboratory Tests
Laboratory evaluation of the particle sizing devices
tested during this study involved the use of a vibrating
orifice aerosol generator (VOAG) as a source of monodisperse
aerosols. The VOAG used in this study was designed and built
at Southern Research Institute, although similar devices have
been reported by several authors previously. 22 '2 3 r2 ** Figure 7
is a schematic diagram showing the operating principle of the
VOAG. A solution of known concentration (in our case, a solution
of fluorescein in 0.1N NHi»OH) is forced through a small orifice
(5, 10, 15, or 20 ym diameter). The orifice is attached to a
piezoelectric ceramic which, under electrical stimulation, will
vibrate at a known frequency. This vibration imposes periodic
perturbations on the liquid jet causing it to break up into
uniformly-sized droplets. Knowing the liquid flow rate and the
perturbation frequency, the droplet size can be readily calculated.
The solvent evaporates from the droplets leaving the nonvolatile
solute as a spherical residue. The ultimate dry particle size
is calculated from the droplet size through the known concentration
of the liquid solution.
-------
-30-
Plexigtass Drying
Chamber •
Vibrating
Orifice —
Flow
Meters
V!*1'
\».i i
w.
wrtrri I I
Control
Valves
IT
,Kr85 Charge Neutrallztr
^Signal Generator
7
1
, Membrane
Filter
X
^Syringe
Pump
Filter
X
Dry Air
Figure 7. Schematic Representation of The Vibrating Orifice
Aerosol Generator
-------
-31-
To calculate the dry particle size, the expression
6QC P' is used.
60TTf /
/ volume of solute
C is the solution concentration or volume of solution
Q is the solution flow rate (cm3/min.)/ and
f is the perturbation frequency (Hz).
By going to smaller.orifices, one can obtain much higher
operating frequencies. This yields higher particle number
concentrations and allows shorter running times in the cali-
bration studies. The running time must be sufficiently long
to allow accurate determination of the stage collection effi-
ciencies and wall losses. It was our experience that the
20 urn diameter orifice was consistently easier to use in
particle generation, primarily because of clogging problems.
The orifices were washed in detergent with ultrasonic agitation
and then rinsed several times in distilled water, also with
ultrasonic agitation. After the filter and liquid handling
system was flushed several times with the aerosol solution to
be used, the orifices were placed, still wet with distilled
water, into the crystal holder and the syringe pump turned on.
A jet of air was played over the orifice to keep the surface
clear until enough pressure was built up behind the orifice to
form a jet.
After a stream of particles was generated, a determination
of monodispersity had to be made. There are two methods to
do this. By using a small, well-defined air jet to deflect
the stream of particles, it was possible to tell when the
aerosol was mono- or polydisperse. Depending on particle size,
the stream was deflected by the air at different angles. During
polydispersity, several streams can be seen at one time. By
varying the crystal oscillation frequency, the system could be
-------
-32-
tuned to give only a single deflected particle stream, thus
indicating monodispersity. On several occasions, the aerosol
tended to drift from monodispersity. To protect against this
occurrence, a filter sample was drawn to be viewed under an
optical microscope. This was a good check on the quality of
the aerosol because the final particles were investigated in-
stead of the primary liquid droplet. This served as a check for
proper drying, satellites, proper size, and multiplets. Kr85
beta sources were placed near the air stream as charge neutralizers
to minimize the loss of particles due to electrostatic forces.
A three-foot-high cylinder was placed on the generator and dis-
persion and dilution air turned on to disperse and loft the
particles into a plenum with several sampling ports. It was found
that drying columns had to be put in the dilution air line to
adequately dry the particles before they enter the sizing devices.
(This is not the case with alcohol as a solvent.) During each
test, filter samples were drawn at intervals to insure continual
monodispersity. Because of its nonhygroscopicity and physical25
properties, ammonium fluorescein was used throughout these
studies as our test aerosol, although in theory, any material
that will dissolve readily in an evaporable solvent could be
used. Figure 8 shows a 5.4 pm diameter test aerosol. In general,
we found about 8% by mass of the particles were doublets.
The devices evaluated in this laboratory study were the
Brink Cascade Impactor using bare plates, filter substrates,
and greased plates; The Brink Precollector Cyclones (C2 externally
mounted and the C3 inline cyclone); The Andersen Stack Sampler
with filter substrates; The Andersen Precollector Cyclone;
the University of Washington Source Test Cascade Impactor with
greased plates; the ERG Tag Sampler with greased plates; and the
McCrone Parallel Cyclone.
In all cases, a back-up filter was used to collect material
-------
-33-
Figure 8.
Ammonium Fluorescein Aerosol Particles
Generated Using The Vibrating Orifice
Generator. The Particle Diameters are
5.4 um.
-------
-34-
passing the last irapaction stage. This insured our ability to
calculate the collection efficiency of all stages tested.
After several trials using high vacuum silicone grease,
agar, K-Y Jelly, and Vasoline, it was found that vasoline was
the most convenient material to use as a greased substrate
medium because it goes easily into solution in a small amount of
benzene.
In the initial portion of this laboratory work, the larger
aerosols were generated. During this period, only the upper
impactor stages were prepared since all the aerosol should have
been caught by at least the third or fourth stage in most cases.
In all tests, a back-up filter was used after the last stage.
For the McCrone device, each cyclone was tested separately with
a back-up filter. As smaller particles were generated, more
stages were tested in each device until the entire device was used.
A pump and flow-metering device for each impactor insured
repeatability in flow rate during each test. Most tests ran
30 minutes to 2 hours depending on particle size and the amount
of material to be collected. After several trials, it was
determined that the Brink with bare plates was not successful as
a collection device due to severe bounce and reentrainment.
Accordingly, testing of the Brink with bare plates was discontinued
after sufficient data was taken to prove its unreliability as a
sizing device. Only the Brink with filter substrates and greased
plates were subsequently tested, as well as the C2 & C3 cylones.
At the conclusion of each test, each impactor was carefully
disassembled and each internal surface cleaned using a solution
of 0.1N NH^OH. Using a known amount of solution, each plate
and surface was washed to dissolve and rinse off the ammonium
fluorescein particles. Where vasoline was used, a small amount
of benzene was poured over the greased collection plate in a
small dish which was then placed in an ultrasonic cleaner. A slight
-------
-35-
amount of agitation caused the vaseline to dissolve and the
ammonium fluorescein particles to become well mixed. Adding a
known amount of 0.1N NHi»OH to the mixture with stirring caused
the ammonium fluorescein to dissolve. After the benzene mix-
ture floated on top of the NHi»OH, the ammonium fluorescein
solution was pipetted off.
The mass of material on each surface was determined by
fluorometric analysis. A Beckman Quartz Spectrophotometer,
Model DU, calibrated with solutions of known concentration of
ammonium fluorescein was used to measure the concentration of
ammonium fluorescein in each wash. By knowing the amount of wash
solution, the dilution factor, if any, and the absolute concentra-
tion, the mass of particles on each surface could be calculated.
With the mass on each plate and surface known, the percent of
total mass collected on each surface was then calculated. From
this, the total wall losses and stage collection efficiencies
could be calculated. Figures 9, 10, & 11 show the percent wall
loss for the Andersen, Brink, and Pilat Impactors versus
particle size. Regardless of size, 50% or more of the total
wall loss was due to material in the nozzle and inlet cones
of the impactor. By using the mass collected on each
impaction plate, the collection efficiencies were calculated
for each test. For many of the lower impactor stages,
there is not yet enough data to accurately draw an efficiency
curve because the smallest size particle that has been used
to date has bean 2.0 um. It is instructive, however, to show
examples of collection efficiency curves for some of the
upper stages. Figures 12 and 13 show efficiency versus particle
size for the first two Pilat stages. It can be seen that
the initial Pilat stage in reality does not behave as the theory
predicts. The D5o h*$ shifted to about 14 um from a calculated
value of about 28 um. Figures 14, 15, 16, & 17 show the collection
-------
99
98
96
90
80
7O
•0
50
40
30
20
to
U)
0\
I
I 10
PARTICLE DIAMETER,MICRONS
Figure 9. Percentage Wall Loss vs. Particle Size
Andersen Stack Sampler (0.5 acfm, 25 C, 29.60" Hg)
100
-------
99
95
90
80
70
*
•>
§
90
eo
10
t
U)
i
10
PARTICLE DIAMETER f MICRONS
Figure 10. Percentage Wall Loss vs. Particle Sizg
Brink Cascade Impactor (0.03 acfm, 25 C, 29.60" Hg)
-------
98
95
70
60
50
40
20
10
5
!'
•••
a
i
Figure 11,
10
PARTICLE DIAMETER .MICRONS
Percentage Wall Loss vs. Particle Size
University of Washington (Pilat) Impactor
(0.5 acfm, 25°Cf 29.60" H
-------
EXPERIMENTAL DATA
PARTICLE DIAMETER,MICRONS
Figure 12. Stage Collection Efficiency vs. Particle Size
First Stage - University of Washington (Pilat) Impactor
(0.5 acfm, 25°C, 29.60" Hg)
50.0
-------
EXPERIMENTAL
THEORETICAL
50.0
I
i
O
PARTICLE DIAMETER, MICRONS
Figure 13. Stage Collection Efficiency vs. Particle Size
Second Stage University of Washington (Pilat) Impactor
(0.5 acfm, 25°C, 29.60" Hg)
-------
o - EXPERIMENTAL DATA
THEORETICAL
1.0
50.0
PARTICLE DIAMETER , MICRONS
Figure 14.
Stage Collection Efficiency vs. Particle Size
First Stage - Andersen Stack Sampler
(0.5 acfm, 25°C, 29.60" Hg)
-------
o -EXPERIMENTAL DATA
THEORETICAL
10.0
PARTICAL DIAMETER , MICRONS
50.0
Figure 15. Stage Collection Efficiency vs. Particle Size
Second Stage - Andersen Stack Sampler
(0.5 acfm, 25°C, 29.60" Hg)
-------
IOO
EXPERIMENTAL DATA
THEORETICAL
) 10.0
PARTICAL DIAMETER , MICRONS
Figure 16. Stage Collection Efficiency vs. Particle Size
Third Stage - Andersen Stack Sampler
(0.5 acfm, 25<>C, 29.60" Hg)
50.0
-------
IOO
UJ
o
u.
u
i
I-
o
Ul
3
o
Ul
o
o -EXPERIMENTAL DATA
THEORETICAL
PARTICAL DIAMETER, MICRONS
Figure 17. Stage Collection Efficiency vs. Particle Size
Fourth Stage - Andersen Stack Sampler
(0.5 acfm, 25°C/ 29.60" Hg)
•u
*».
i
-------
-45-
efficiency for the first four Andersen stages. Notice that the
collection efficiency for stages 3 & 4 for large particles is
well below 100%.
A more concise way of presenting all the data is to
graph ^JT vs. EFF as was done by Rantz and Wong.1* In this
way, the efficiency data on all stages can be presented on the
same graph. The dimensionless parameter \ij' is given by
D ± | p F 472 POA C I where
10" y (DC) 3 X(I) PS
Dp = aerosol diameter (urn),
p = aerosol density (gra/cra3),
F = flow rate (cfm),
POA = air pressure at impactor inlet (atm),
C = Cunningham Correction Factor,
y = air viscosity (poise),
DC = Jet diameter (cm),
X(I) = Number of holes per stage, &
PS = Air pressure at stage jet (atm)
It can be seen that for each stage, ^j~^~ is dependent only
on Dp, the aerosol diameter, when all other parameters are held
constant. Figures 18, 19, 20, 21, & 22 show ^JT vs efficiency
for the Andersen, Brink (bare plate), Brink (filter substrate),
Brink (greased plate), and the Pilat Impactors. Separate
symbols are employed to represent data for each stage. For each
symbol, increasing \tj> values indicate increasing aerosol
size. Notice that the collection efficiency does not reach
100% for a large size and stay at that value for all larger
sizes. In fact, if a single symbol is followed from left to right,
it is apparent that the stage reaches a peak in efficiency and
then falls off at larger sizes. This indicates that some large
-------
—t-*-
Figure 18.
Stage Collection Efficiency vs.
Andersen Stack Sampler - Glass Fiber Filter Substrates
(0.5 acfm, 25°C, 29.60" Hg)
-------
— THEORETICAL AFTER
RANTZ AND WON6
0 O.I 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 I.I 12. 1.3 1.4 1.5 1.6
Figure 19. Stage Collection Efficiency vs. V *
Brink Cascade Impactor - Bare Collection Plates
(0.03 acfm, 25°C, 29,60" Eg)
-------
100
0 OL! 0.2 0.9 0.4 0.8 0.6 0.7 04 0.9 1.0 I.I 1.2 L3 1.4 1.5 1.6 1.7
Figure 20.
Stage Collection Efficiency vs. V*P
Brink Cascade Impactor - Glass Fiber Filter Substrates
(0.03 acfm, 25°C, 29.60" Hg)
-------
rrrr THEOfitTlOAL AFTER
O.I 0.8 0.3 0.4 0.5 0.6 0.7 0.6 O.9 1.0 I.I 1.2 1.3 1.4 1.8 1.6 1.7
Figure 21. Stage Collection Efficiency vs.
Brink Cascade Impactor - Greased Plates
90.03 acfm, 25<>c, 29.60" Hg)
-------
*T—i—i
Figure 22. Stage Collection Efficiency vs. V*
University of Washington (Pilat) Impactor
(0.5 acfm, 25°C, 29.60" Hg)
•PHEORETICAl, AFTER
RANTZ AND WONG
I I L
1.3 1.4 1.5 1.6 1.7
i
ui
o
I
-------
-51-
particles impact with enough momentum to bounce and be reen-
trained in the gas stream. Because of this, there may be
difficulty in deciding where a particle will stop. If it is
not caught by stage one or two, for example, there is a high
probability that it will end up on a lower stage or back-up
filter, thus decreasing the reliability of any size distribution
calculated from the data. Figure 19 shows that the Brink Impactor
with bare substrates did not perform well when compared to the
same impactor with glass fiber as greased substrates. The poor
performance of the initial Pilat stage as shown in Figure 22 is
probably due to the failure of the inlet cone to define the
diameter of the air jet. It is assumed in calculating the D50
for the inlet stage, that the gas stream expands to the full cone
diameter. Apparently, this is not the case, and the effective
jet diameter is somewhere between the nozzle inlet diameter and
the cone exit diameter. This will probably not vary in a
predictable fashion with flow rate and may necessitate redesign
of the inlet cone.
Figures 23 and 24 show efficiency versus particle size
for the C2 and C3 cyclone used with the Brink Impactor. Figure 25
presents the Andersen Cyclone data for efficiency. Data for
the four McCrone Cyclones is presented in Figures 26, 27, 28, &
29. In some cases, there have not been enough small particles
generated to indicate the trend of the efficiency at these
sizes.
Because of the small number of tests performed on the E.R.C.
TAG Sampler, collection efficiencies are presented in tabular
form in Table II. Five tests are shown in this table. Each
is indicated at the upper part of the sheet by the particle
diameter and total percent wall loss. The columns marked I
and II are for the percent by stage of the total collected
mass and the stage collection efficiency, respectively. A column
giving the theoretical stage D5o's is presented to aid in under-
standing the stage collection efficiencies.
-------
EXPERIMENTAL DATA
TENTATIVE BEST FIT
Figure 23.
PARTICAL DIAMETER .MICRONS
?fficiency vs- Particle Size
. Hy Mounted) Cyclone
acfm, 25°Cf 29.60" Hg)
-------
5
Ul
EXPERIMENTAL DATA
TENTATIVE BEST FIT
PARTICAL DIAMETER, MICRONS
Figure 24. Stage Collection Efficiency vs. Particle Size
Brink C3 (Inline) Cyclone
(0.03 acfm, 25°C, 29.60" Hg)
-------
EXPERIMENTAL DATA
TENTATIVE BEST FIT
F*RTICAL DIAMETER , MICRONS
Figure 25. Stage Collection Efficiency vs. Particle Size
Andersen Cyclone Precollector
(0.5 acfm, 25°C, 29.60" Hg)
-------
UJ
o
u.
u.
UJ
1
§
A EXPERIMENTAL DATA
TENTATIVE BEST FIT
I Btin
HI I Iiiitiiiiiiiiiiiiiiiiiiimiii
PARTICLE DIAMETER,KMCHONS
Figure 26. Stage Collection Efficiency vs. Particle Size
McCrone Cyclone T2A - 0.13" H20
-------
100
o
UJ
o
u.
Si
i
o
UJ
8
EXPERIMENTAL DATA
TENTATIVE BEST FIT
i
tn
PARTICLE DIAMETER .MICRONS
Figure 27. Stage Collection Efficiency vs. Particle Size
McCrone Cyclone T2A - 1.00" H20
-------
IOO
EXPERIMENTAL DATA
TENTATIVE BEST FIT
01
PARTICLE DIAMETER , MICRONS
Figure 28,
Stage Collection Efficiency vs. Particle Size
McCrone Cyclone T2B - 0.80" H20
-------
too
EXPERIMENTAL DATA
TENTATIVE BEST FIT
Figure 29
PARTICLE DIAMETER , MICRONS
Stage Collection Efficiency vs. Particle Size
McCrone T2B - 20.0" H20
m
CO
t
-------
TABLE II
ENVIRONMENTAL RESEARCH CORPORATION TAG SAMPLER (GREASED PLATES)
Particle Size (microns)
Percent Wall Loss
6.3
32.47
5.0
35.82
5.0
18.06
2.0
4.75
2.0
3.66
Theoretical
D50'8
(microns)
(0.5 CFH)
II
II
II
II
I
Ol
V£>
I
Nozzle
Inlet Cone
O-Rings
E-l
E-2
E-3
E-4
E-5
E-6
E-7
E-8
E-9
E-10
Filter
11,
7,
5.5
4.0
2.8
2.0
1.3
0.9
0.63
26.74
5.73
5.63 8.34
2.56 4.14
1.63 2.75
28.13 48.75
15.28 51.67
10.08
3.41
0.51
0.23
0.02
70.54
80.99
63.75
79.31
33.33
32.58
3.24
1.32 2.06
0.99 1.58
0.92 1.49
1.58 2.59
5.96 10.04
0.04 100.00
47.54
5.38
0.36
0.00
0.00
0.12
89.03
91.81
75.00
9.65
8.41
1.00 1.22
0.26 0.32
0.35 0.43
2.19 2,73
11.24 14.39
56.83
7.34
0.75
0.90
84.96
72.96
27.57
45.68
0.45 42.06
0.62 100.00
0.50
0.89
3.36
0.17
0.00
0.00
0.00
0.65
85.26
8.34
0.71
0.12
0.00
0.18
0.00
0.00
0.00
0.68
90.29
90.95
85.54
100.0
0.64
0.68
2.34
0.35
0.29
0.20
0.25
0.40
72.13
20.13
1.52
0.29
0.05
0.36
0.30
0.21
0.26
0.42
76.05
89.39
63.07
32.58
8.33
0.00
0.55 100.0
-------
-60-
C. Cascade Impactor Data Reduction
Throughout this work on the use of cascade impactors, an
effort has been continually made to improve the accuracy and
convenience of our data reduction schemes. Two methods are
mentioned below. The "D5o" method is used at this time for the
majority of cascade impactor work, but it is an oversimplifica-
tion which can lead to a loss of information. Picknett26 has
introduced a more sophisticated method, yielding maximum in-
formation from the experimental data. This technique, however,
is more difficult to apply and requires the use of a computer.
• COMPARISON OF THE D50. AND PICKNETT DATA REDUCTION TECHNIQUES
USING REAL AND SIMULATED STAGE LOADINGS
In investigating data reduction techniques, it is difficult
to separate inaccuracies in the theory and errors introduced
by reentrainment, bounce, scouring, and poor calibration of the
impactors. This problem can be eliminated by simulating the
capture of a fictitious aerosol using the efficiency curves
shown in Figure 30. Once the stage loadings are calculated,
this data is used to recalculate a particle-size distribution,
using either the D5o or Picknett method, which should ideally be
identical to the fictitious input distribution.
Figure 31 is a test of both the D50 and Picknett techniques
using a fictitious aerosol. The fictitious, or "true", aerosol
consisted of 20% - 1 ym, 20% - 2 urn, 20% - 4 vim, 20% - 8 vim,
and 20% - 12 urn diameter by mass. The zero through fifth stages
were used, and cyclone C3, as shown in Figure 30.
It can be seen that both methods give very nearly correct
values for the mass median diameter,, although neither scheme can
resolve the step functions in the distribution. The Picknett
method is much better at the large particle end, because the D50
-------
Ml 0,04 0,1 O.J OJ 1 I « 111 M N 40 SO M 70 M
90 *5 |« M M.I M.I
Stage Collection Efficiency (%)
Figure 30. Stage Collection Efficiency for a Brink Cascade
Impactor with a Pre-collector Cyclone.
-------
-62-
100
Q)
N
-H
CO
-0
(0
-p
CO
CO
4J
C
0)
O
M
0)
80
60
40
20
-------
-63-
method does not account for the fact that the cyclone collects
some particles which would otherwise reach the upper stages. The
Picknett method does take this overlap in collection efficiencies
into effect, and gives a good "average" of the distribution,
smoothing out the abrupt steps.
Figure 32 is another simulation for which the input, or
"true", distribution, the sum of two log normal distributions
is bimodal and continuous. It can be seen that the Picknett
scheme gives an excellent representation of the true distribution.
Once again the DSO method distorts the distribution at the larger
sizes.
Figure 33 shows true and calculated particle size distributions
using actual experimental stage loadings. An 8.3 ym diameter
ammonium fluorescein aerosol was generated using the vibrating
orifice aerosol generator. About 4% of the particles were found
to be doublets (8% of mass) . This time, we see that neither
method approximates the real distribution accurately, although
the Picknett method does give a good mass median diameter. At
this point, it is impossible to isolate the cause of the
discrepancy, but there was visual evidence of reentrainment.
At the present time, the D50 method is still used for our
data reduction (See Appendix A) . Work is continuing to improve
our computer program based on the Picknett method. We believe
that this method will ultimately be more valuable because it
extracts the maximum amount of particle size information from
the experimental data.
-------
100 -
90
80
70
60
50
40
30
20
10
0.1 i 10
Particle Diameter (ym)
Figure 32. Simulation of a Continuous Particle Size Distribution
100
-------
-65-
100
90
0)
N
•H
W
•0
3
80
70
3 60
CO
to
m
a
50
40
30
20
10
1.0
Figure 33
10.0
Particle Diameter (ym)
Impactor Tests Using a Known Particle Size
Distribution
-------
-66-
APPENDIX A - IMPACTOR OPERATING TECHNIQUES
A. Impactor Selection
In deciding which cascade impactor is the appropriate
device for a sampling program, the main criteria is the mass
loading and its effect on sampling time. In high grain loading
situations such as the inlet to control devices/ a low flow rate
impactor (less than 0.1 acfm) is preferable because it permits
reasonably long process averaging times, although in some cases,
even with low flow rates the sampling time may be limited to
only a few minutes. The use of a high flow rate impactor in
these cases, would be unwise because impractically short sampling
times would be required in order to avoid impactor overloading.
On the other hand, low dust concentrations would result in
excessively long sampling times with the low flow units.
impactors operating at flow rates near 0.5 acfm are normally
used under these conditions to keep sampling times reasonably
short. Even high flow rate impactors frequently require sampling
times in excess of two hours to collect weighable stage loadings,
especially at the outlet of high efficiency collectors.
The Brink Impactor is an example of a low flow rate impactor
which is useful under heavy dust loading conditions, while the
Andersen or Pilat (University of Washington) impactors are the
high flow rate units which were used most often at control
device outlets.
In many instances, the percentage (by weight) of material
with sizes larger than the first impaction stage cut point can
be quite high. In. such cases, precollector cyclones are necessary
to prevent the upper impactor stages fsom overloading. The
Andersen cyclones are available from the manufacturer. The Brink
cyclones used in this study were designed and fabricated by
Southern Research Institute personnel.
-------
-67-
To insure a valid size distribution from impactor measure-
ments, it is imperative that isokinetic sampling be used, if at
all possible. Figure Al depicts the relationship between flue
gas velocity and impactor flow rate for various nozzle sizes.
This graph is used as a convenient means of selecting nozzle
sizes to insure isokinetic sampling when the gas velocity is
known and the appropriate impactor flow rate has been chosen.
Also indicated on the graph are heavy horizontal lines giving
the upper flow rate limits for various impactors to insure that
severe reentrainment does not occur.
B- Sampling_Time
The length of the sampling time is dictated by mass loading
and size distribution. An estimate for initial tests can be
made from Figure A2. Tests subsequent to the first should have
sampling times adjusted such that no single stage, excluding a
cyclone, if one is used, contains more than 10 mg. of mass.
C. Collection Substrates and Adhesives
Due to the necessity for high weighing accuracy and the
restriction of low tare capacity in the case of most field
useable precision balances, foil or glass fiber collection
substrates are used with all impactors. The Brink is generally
used with aluminum foil or glass fiber substrates, the Andersen
utilizes special glass fiber inserts, and the Pilat is used with
aluminum or stainless steel foil substrates. Depending on the
temperature in the sampling duct, silicone vacuum grease or other
high viscosity grease may be used on the foil substrates to aid
in the retention of the impacting particles. It has been found
that a thick (20%) solution or suspension of benzene and grease
-------
-68-
NOTE
HEAVY HORIZONTAL LINES INDICATE MAXIMUM FLOW
RATES WITHOUT REENTRA»NM£NT FOR VARIOUS IMPACTORS.
.5
.4
.3
U.
O
"»•"*
UJ
^ 01
Q- ,\/9
ft 07
O .Wf
.00
.01
.02
OAI
.VI
- _. |. 1 . -
ji Vi'iz
2 '
2 7
7 /
,' ^
j j [ i ' 4 i 2 >
3 J - i < ' 2 - i'Z - ' i
, _ L 0 _ . 1 . , i ! ,.,. .-- •• • C J- . . | . . _
3_Z_ .j_..i...,.i. - - * ..... ... - 2 _.,'.! ^
j ,!.. J ....,'.. .....,!.. .. _.,,.-. i ._
L. _Z._ 3. ..,(.,,,,! ...... .1 ! _-..!.._
X - • • '
^2 2 i ' 2 P' i1 Z
r / ! .- . ! 2 !. . , l! . . E
22 Ji .!. i 2
ZZ,? I' /i! 2
2 it i i i i i '
1 Z 3 4867B9K)
|Z— 8:::iii:;i ::::; !::ji !::: ?i:>5:::
_"__?. . . . i ! j " . . ., _ ^ _ - . ! .
! _. ^ ..,.(., i , , _rf _ _ n ( . .
f. L. 4 ' r f.
A 2 , i 2 c! c
>2 i (' 2 i' i1 2
Z ^ i ' i t'...'. Z
^ / i' \* t / ip
!!rpi::::;!:::: i RiNK-^i •• ( RE/SE :!
2 t ( i - ^
t i p , ^
3. ,_E. ..!.., . ,Z __!..
c «__. _-Z _?_...
,?:;?:: :: BRINK -WITHOI T GREASE':
.1 .__...!.. ...__ ...
i.. i ... ....
r '
2 '
?
?
2 3 4567691
GAS VELOCITY (FT/SEC.)
Figure Al. Nomograph For Selecting Nozzles For Isokinetic
Sampling.
-------
GRAIN LOADING (GRAINS/ACF)
3 456789
.".v.. j tr$: jUXi^" "_"i , - *._ !tij .-.--.rt"'-*-
1.0 0.5 0.4 0.3 O.Z
SELECTED FLOW RATES (ocfm)
O.I 0.05 0.04 0.03 0.02 0.01
Figure A2. Sampling Time Determination For Total Mass Collection
of 25 Milligrams.
-------
-70-
serves this purpose well and is fairly easy to apply. If duct
temperatures are 400°F or lower, this grease-benzene mixture
will hold up satisfactorily under most sampling conditions.
Horizontal operation of the impactors with greased substrates
is not recommended due to possible flow of the grease. The
Brink and Pilat foil substrates are fitted to the shape of the
collection plates. After being shaped, an eye dropper is used to
place the benzene and high vacuum grease solution on the foils.
Four to five drops are placed on the upper stage Brink foils
leaving a residue of about 20 mg. after evaporation of the benzene,
and a single drop is placed in the center of the last two Brink
stage foils. Enough solution is placed on the Pilat substrates
to adequately cover the area under the jets (^30 mg. of grease) .
These greased substrates are then baked at 400 F for 1 to 2
hours. If stage loadings are low, smaller amounts of grease can
be used. When a new batch of grease is used, checks should
be made for weight loss by the grease subsequent to the initial
bakeout. After removal from the oven, the substrates are
conditioned in a desiccator for 12 to 24 hours prior to weighing.
The use of grease with Brink and Pilat Impactors is necessary
because of problems with reentrainment and particle bounce, which
is especially severe at jet velocities over 35 to 40 m/sec and
occurs in some instances at even lower velocities. •High impactor
flow rates result in high jet velocities in the lower stages.
The use of grease allows higher flow rates than would be
useable without grease by improving the particle retention of
the impaction surface, but care must also be taken to insure
that grease is not blown off the substrates. Grease blow off
can occur at jet velocities greater than 65 m/sec.. Therefore,
for each impactor there is a maximum permissible flow rate, the
value of which depends on the type of impaction substrate that
is being used.
The Andersen Impactor uses pre-formed glass fiber
-------
-71-
fliter substrates provided by the manufacturer. Handling
of the Andersen substrates is simplified by cutting aluminum
foil squares slightly larger than the substrates into which the
substrates are folded twice before desiccation and weighing.
After sampling, the substrates are returned to the same foil
wrappers for conditioning and final weighing.
Back-up filters are used on all impactors to collect the
material that passes the last impaction stage. Binderless
glass fiber filter material such as Gelman Type A Glass Fiber
Filter Web is used for this purpose. For the Brink, 1" diameter
circular discs are placed under the last spring in the outlet
stage of the impactor. The filter is protected by a teflon
washer and a second filter disc placed behind the actual filter,
which acts as a support. The Andersen uses 2%" diameter filter
discs placed above the final "F" stage. The Pilat Impactor
uses 47 mm filter discs in an integral filter holder.
D. Multipoint Sampling
Although it is desirable to sample at several points across
a duct to insure that the sample collected is representative of
the flue gas, the impactors should be operated isokinetically
at a constant flow rate for each point sample. To accomplish
a traverse, the impactor is operated at several discrete points
across the duct, with properly selected nozzles and flow rates
for isokinetic sampling, and the results averaged to give
an "average" dust loading.
E. Impactor Orientation
Whenever possible, the impactors are oriented vertically to
minimize gravitational effects such as flow of grease or falloff of
collected particles. Horizontal placement is necessary at
-------
-72-
times and extra care must be taken on such to not bump the
impactor against the port during removal operations.
F. Heating Impactor
All condensible vapors must be in a gaseous state until they
exit from the impactor unless a condensate is the prime aerosol
being measured. In streams above 350°F, auxiliary heating is not
usually required. Below 350°F, the exit temperature of the
impactor is maintained at least 20°F above the process temperature,
A thermocouple is normally used to monitor the temperature of the
exit gas from the impactor.
Whether the impactor is being heated in the duct or exter-
nally, with heater tape/ etc., an allowance of 30 minutes
warm-up time is allowed as a minimum to insure that the impactor
has been heated to duct or operating temperature.
G. Probes
Sampling probes to an impactor outside the duct are used only
if there is no other way. Probes are kept to a minimum length
and contain the fewest possible bends. A precollector cyclone
is mounted at the end of the probe to remove the large particles
and thus reduce line losses.
H. Balance Requirements
For accurate weighing of collected material, a balance
with a sensitivity of at least 0.01 milligrams is required.
This is especially true for the lower stages of the Brink
Impactor where collection of 0.3 mg. or less is not uncommon.
The balance must also be insensitive to vibration if it is to be
-------
-73-
used in the field. During this study, a Cahn Model G-2
Electrobalance was used and found satisfactory.
I. Sampling Configurations
Several sampling train arrangements are available depending
on the impactor used, flow, sampling time, etc.. The original
Brink design called for a system similar to that shown in
Figure A3. The flow rate is determined by the pressure drop
across the impactor. When low flow rates are required, it is
very difficult to accurately measure the pressure drop across
the Brink. The low flow rates also make a dry gas meter useless.
In performing most of our tests, we have used an in-stack
system which incorporates a calibrated orifice as shown in
Figure A4.
To insure proper measurement by the orifices and dry gas
meter and to protect the vacuum pumps from damage resulting
from condensable vapors, it is necessary to cool and dry the
sampled gases immediately after they leave the impactor. For
long sampling times or in cases where there is high water
content, a series of condensers in an ice bath is useful in
removing the water. A drying column is normally used after the
condensers for further protection of the gas meter. Different
size orifices such as 0.03, 0.06, 0.09 inches allow for a wide
range of flow rates with reasonable pressure drops using a water
manometer. Normally a Hg manometer is placed upstream of the
orifice to determine the pressure at the inlet to the orifice.
The use of a calibrated orifice to monitor the impactor
flow rate involves the following equation giving the pressure
drop across the orifice water manometer.
* (i-rB,e)
o c s
-------
-74-
BACK
FRONT
TO DUCT
CYCLONE
IMPACTOR
CATCH
BOTTLE
1
PUMP
J
I
o
F
1
m^
•••
—
VJ
•LC
TO
Q° OPEN
^=±/ TO AIR
HQ MANOMETERS
C
!
k?
1WMETER DUST
[SSUREDftOP PRESS!
'?
JRE
Figure A3. The Original Brink Impactor System,
BMS-11,
Set-up for out of stack sampling.
-------
COOLING COIL
IMPACTOR
c
Hg MANOMETER
PUMP
Ul
r
MANOMETER
Figure A4
An improved Method of Flow Control for the Brink
Spacer using a calibrated orifice flowmeter.
-------
-76-
MM = Mean Molecular Weight of Flue Gas
MA = Mean Molecular Weight of Air
AP = Calibrated orifice pressure drop "H20
AP = Pressure drop at which orifice calibrated "HzO
o f
QI = Impactor flow rate chosen for isokinetic sampling ACFM
Q = Calibration flow rate for orifice ACFM
c
F., n = Volume fraction of water in the flue gas
11 2 U
P = Ambient stack pressure P = P +• AP "Hg
s r s a s
P = Pressure upstream of orifice referred to ambient "Hg
P = Ambient pressure when orifice calibrated "Hg
C _
T = Temperature of the orifice R
T = Temperature of the orifice when calibrated R
T = Stack temperature °R
To monitor impactor flow rate with a dry gas meter, either
individually or in conjunction with a calibrated orifice, the
following equation pertains:
s a
Q = Flow rate indicated by the dry gas meter ACFM
rt
T = Temperature of metered air R
a
T = Flue gas temperature R
P = Ambient pressure upstream of the meter "Hg
P = Ambient stack pressure "Hg
S
F« n = Volume fraction of water in flue gas
Hj U
In-stack sampling is recommended in all cases where practical;
however, in certain cases, out-of -stack sampling is the only
solution.
To apply equation Al, monitoring of the orifice or meter
inlet pressures is necessary. We have found that the total
-------
-77-
volume of gas sampled, as measured by the calibrated orifice
and by the dry gas meter, agree within a fraction of one percent.
Thus, the gas meter may be left out of the system if it is
inconvenient or difficult to use at a particular location.
For most cases, with the Andersen and Pilat Impactors, where large
gas volumes are to be sampled, the system shown in Figure A5
can be used depending primarily on the ability of the gas meter
to withstand the large pressure differentials in the system.
A sufficiently long piece of pipe is attached to the impactor
to insure proper positioning and traverse capabilities in the
duct and to insure that the impactor is not cooled by heat
transfer along the probe if external heating of the impactor is
not used.
Negative duct pressures can cause problems resulting from
backflow through the impactor causing material to be blown off
the collection substrates onto the underside of the jet plate
after conclusion of sampling. Thus, care was taken to insure
that no gas flow through the impactor takes place except when
sampling.
J. Preparing _th_e Impactors
The impactor is carefully loaded with the preweighed stage
substrates. The Andersen requires that extra attention be paid
to the alignment of the substrate to the stage and stage to
stage insuring that the jets are not blocked by the substrate
and the jets of one stage are above the collection surface of
the next stage. After all stages are loaded, the entrance
cap is placed on the Brink and the shells are placed on the
Andersen and Pilat. The Brink is tightened with wrenches to
make certain the asbestos gaskets are seated. Handtightening
is sufficient for the Andersen or Pilat Impactors. Over tightening
will cause the stainless seals to cut into the Andersen substrates.
After assembly, the appropriate nozzle is added. (See Figure Al.)
-------
CONDENSERS
PROBE
PORT
ANDERSEN OR PILAT
MR FLOW--
Hg MANOMETER
DRYING
COLUMN
COOLING COIL
1
/i/ve MFTFR
PUMP
oo
i
MANOMETER
Figure A5. The typical Andersen or Pilat setup using calibrated orifice and
dry gas meter.
-------
-79-
If supplemental heating is required, the heating tapes,
insulation and temperature monitors need to be added. A
thermocouple mounted in the gas flow immediately after the
impactor is best for controlling heating. This also yields the
temperature needed for calculating impactor cut points. A
heating tape of sufficient heating capacity is wrapped around
the impactor. Fiber glass tape is used for holding the heating
tape. Insulation such as asbestos tape is then wound around the
impactor. Fiber glass tape is again added to hold the asbestos
in place as insulation. An easily removed wrapping of aluminum
foil is sometimes used to keep the impactor clean while in the
duct.
The impactor is then mounted on the appropriate probe,
taken to the sampling position and installed in the sampling
system.
K. Taking the Sample
The impactor should be preheated for at least 30 minutes
before sampling. If supplemental heat is being used, the impactor
should be brought up to temperature outside the duct and then
allowed some time to equilibrate after insertion. The nozzle
should not point upstream into the gas flow during this phase.
The flow rate must be maintained at the predetermined level
during testing to assure stable cut points. Any attempt to
modulate flow to provide isokinetic sampling could destroy the
validity of the data. The correct flow rate should be established
quickly, especially for the short sampling times typically
found at inlets.
-------
-80-
L. Disassembling the Impactor
The post test procedure is very important in obtaining
useful results. The crucial part is to make sure the collected
material stays where it originally impacted. After the test,
the impactor should be carefully removed from the duct without
jarring, removed from the probe, and allowed to cool. Disassembly
is quite tricky in some cases. It is good to have a pair of
fine tweezers and a balance brush. Careful disassembly of a
Brink is a necessity for obtaining good stage weights. If a
precollector cyclone has been used, all material from nozzle to
the outlet of the cyclone is included with the cyclone catch.
All of this material is brushed onto a small tared 1" x 1"
aluminum foil square to be saved for weighing. Cleaning the
nozzle well is also important, especially if it is a small bore
nozzle. All material between the cyclone outlet and second stage
nozzle is generally included with material collected on the first
collection substrate. All appropriate walls are brushed off
as well as around the underside of the nozzle where a halo
frequently occurs on the upper Brink stages. All material
between the second stage nozzle and third stage nozzle is
generally included with that on the second collection substrate.
This process should be continued down to the last collection
substrate. Care is necessary in taking out the filter. A
good pair of sharp fine tweezers is essential in removing the
foil substrates from the plates without losing grease and
collected material.
Cleaning an Andersen is a demanding chore. A foil to hold
the Stage 1 substrate is laid out. Next, the nozzle and entrance
cone are brushed out and onto the foil. Then the material on
Stage 0 is brushed off. Next, any material on the top o-ring
and bottom of Stage 0 is brushed onto the foil. Finally, the
-------
-81-
Stage 1 filter substrate material is placed on the foil and last,
the top of the Stage 1 plate o-ring and cross piece is brushed
off. Depending on how tight the impactor was screwed shut,
some filter material may stick to the stainless steel edge
contacting the substrate. This is carefully brushed onto the
appropriate foil. This process is continued through the lower
stages. Finally, the filter is carefully removed. Again, all
material is desiccated 12 to 24 hours before final weighing.
Pilat clean up is similar to the Brink. Some problems
have been noted with o-rings sticking and care must be exercised
not to dislodge the sample while trying to separate the stages.
M. Data Logging
Records should be kept in notebooks for all phases of the
sampling program. It was generally found convenient to keep
three notebooks/ one for recording weight, which remains in the
lab with the balance and one each for recording the details on
inlet and outlet sampling runs. A form similar to that shown
in Figure A6 was used for recording the necessary information
while sampling.
N. Cascade Impactor Data Reduction
After an impactor runs, it is necessary to obtain a
particle size distribution from the mass loadings on each staao.
The conditions at which the impactor was run determine the stage
D50 cut points. These can be calculated by an iterative solution
of the following two equations:
'so
= 1.43 x 10'
P X(I)
c s
C 472.0
-------
-82-
IMPACTOR IDENT.
OPERATOR
PORT NO.
AMB. PRES.
IN STACK/OUT-OF-STACK
PITOT DELTA P
NOZZLE DIA.
GAS METER END
GAS METER START
METERING ORIFICE TEMP.
DATE
PROBE DEPTH
STACK TEMP.
GAS VELOCITY
IMPACTOR DELTA P.
RUN CODE NO.
SAMPLING LOCATION
STACK PRES.
IMPACTOR TEMP.
SCALPING CYCLONE(S)
IMP. FLOW RATE
START TIME
TOTAL SAMPLING TIME
METERING ORIFICE AP
PRESS. AT METERING ORIFICE
REMARKS:
Figure A6. Form used for logging field test data.
-------
-83-
C = 1 +
2L
D50 x 10"
where
1.23 + 0.41 EXP
(-0.44 D50)/L x ICT")
D50 is the stage cut point (ym) ,
U = gas viscosity (poise),
DC = stage jet diameter (cm),
P = local pressure at stage jet (atm),
s
p = particle density (gm/cm3),
QI = impactor flow rate (cfm),
P = ambient pressure at impactor inlet (atm),
C = Cunningham Correction Factor,
L = Gas mean free path (cm), and
X(I)= Number of rhbles aper stage.
The most feasible way to calculate these cut points is to
write a computer program. Otherwise, it is a slow and tedious
process. The size parameter reported can be either aerodynamic
diameter, that is, diameter based on the behavior of unit
density particles, or approximate physical diameter, based on
the estimate of the true particle density. In either case,
the particles are assumed to be spherical.
After calculation of the stage D5o's, there are two possible
methods of presenting the data. The "Dso" method is used at
this time for the majority of cascade impactor work, but it is
an oversimplification which can cause loss of information.
Picknett26 has developed a method which takes into account the
non-ideal nature of cascade impactor collection efficiencies,
specifically, the fact that a single particle can be collected
by two or more stages is incorporated into the theory. In the
Picknett method, discrete particle sizes are arbitrarily chosen
usingihe D50 as an initial set, and their concentration calculated
-------
-84-
such that the measured stage loadings would be obtained if the
real particle size distribution had consisted of these chosen
sizes. A large number of particle sizes may be chosen, giving
a very good representation of the real distribution.
The particle size distribution may be presented on a
differential or cumulative basis.
• Differential Particle Size Distributions - It is assumed
for the purpose of analysis that all of the mass caught upon an
impaction stage consists of material having aerodynamic diameters
equal to, or greater than the D5o for that stage, and less than
the D50 for the next higher stage. For the first stage (or
cyclone) , it is assumed that all the material caught has
aerodynamic diameters greater than, or equal to, the DSO for
that stage (or cyclone) , but less than some arbitrarily large
value, say 100 um.
If the true particle size distribution constituted a
continuum, the amount of material having diameters between D
and D+dD could be represented by dM. Then the integral
dD
Fi
would yield the total mass having diameters between Dj and D2.
Because the intervals between the stage D50's are
logarithmically related, and to minimize scaling problems, the
differential particle size distributions are plotted on log-log
or semi-log paper with
dM
d (log D)
as the ordinate and Log D as the abscissa. (D is the geometric
geo geo 3
mean of DI and D2.) The mass on stage "n" is designated by dM .
-------
-85-
The d(log D) associated with dM is log (D50) ,, - log (D50) .
The total mass having diamel
to the area under the curve
The total mass having diameters between (D50) and (D50) is equal
Mass =
• I
dM
log (D50)t+1 ~ log (D50)t
log (D50)t+1-log (D5o)t
or
Mass = / dM . d (log D)
d (log D)
'm
for a near continuum.
The procedure outlined above describes the construction
of a histogram. In practice, a smooth curve is drawn through
the points, yielding an approximation to the real particle
siz» distribution. Such a curve is needed to calculate fraction
efficiencies of control devices if the D50's differ between
inlet and outlet measurements. The accuracy of the approximation
is limited by the number of points, and by the basic inaccuracy
of neglecting the non-ideal behavior of the impactors, especially
overlapping collection efficiencies for adjacent stages.
• Cumulative Particle Size Distributions - The data may be
presented on a cumulative basis by summing the mass on all the
collection stages and back-up filter, and plotting the fraction
of the mass below a given size, versus size. This is frequently
done on special log-probability paper. This paper is especially
convenient for log normal distributions, but semi-log paper may
be preferable for interpretation, especially if the distribution
is not log normal. In general, cumulative distributions are
more difficult to interpret than differential plots. The abscissa
is the logarithm of the particle diameter, and the ordinate is the
percentage smaller than this size. The value of the ordinate
-------
-86-
II
12
at a given (D50) would be t = n
n 1
IS
-j 16
Percent less than stated size = - t " ° - x 100% 19
IK
15
t = o |2
13
where "*
ij5
t = o corresponds to the filter, and 1)6
t = N corresponds to the coarsest jet or cyclone. }J
to
Alternately, an analytical curve might be fitted to the 19
cumulative distribution obtained above, and values of dM/d (log D) !?
obtained by differentiation of the analytical expression. U
P»
The curve fitting approach is especially useful when using the jj
Picknett method because the method results in the development 15
of a "best fit" cumulative distribution providing a large number |*
of discrete points on the cumulative distribution curve. IB
On the following pages is our computer program for ^
10
impactor data reduction based on the D50 scheme. The program Si
presented is especially for the Brink Cascade Impactor. It can be ®
adapted for the Andersen or Pilat by altering the number of 14
stages, the number of holes per stage, the jet diameter, the '5
jet to plate spacing, the initial D50 estimate, and the fraction 37
of the pressure drop at each stage. 58
19
10
11
»E
U
14
15
it
IT
18
19
SO
H
*
Is
44
$5
ft
17
18
19
-------
-87-
C
C MODIFIED BRINK CASCADE IMPACTOR
C
C THIS IS A FORTRAN IV PRHGKAM pOR CALCULATING STAGE CUT POINTS
C (050'S) AMD THE PARTICLE SIZE DISTRIBUTION OF MATERIAL COL-
C LECTED BY A MODIFIED BRI^K CASCADE IMPACTOR , COMMENT CARDS
C DESCRIBE THE INPUT AND OUTPUT DATA AND THE IMPORTANT CALCULA-
C TIONS,
C
C
INTEGER xc?)
REAL MASS<9),JSPA(7),MM,MU,LC7)
REAL IMASS(9),ICUMM(9)
DIMENSION PRCU(9)
COMMON MA5S,JSPA,MM,MU,L,X,POA,DPA,TTK,RHn,DUR,RA,F,CCUMM,MC,MCC
COMMON MO,M*1»H,DMAX,DELPC7),FGC5),PCC7),SUBC7),PE»CU(),CUMM(9}
COMMON GRNA,GRNS,GRNAM,GRNSM,CYC2,CYC3,DPC(7),RfcYNH7),RF_YN2(7)
COMMON FD(7)»PS(7),AMASS<9)
COMMON IMASS,MC2,MC3,MOO,MS
C
C IF EsO, READ A NEW SET OF DATA, IF E = l, STOP.
C
12 REAO(2,99) E
IF(E) 11,11,93
C
C PO *t GAS PRESSURE AT IMPACTOR INLET. INCHES OF MERCURY,
C DP *•> PRESSURE DROP ACROSS THE IMPACTOR, INCHES OF MERCURY.
C T »- TEMPERATURE OF IMPACTOR, DEGREES FAHRENHIET.
C RMO -- PARTICLE DENSITY, GRAMS/CUBIC CENTIMETER,
C OUR *- DURATION OF IMPACTOR SAMPLING, MINUTES.
c DMAX — MAXIMUM DIAMETER OF MATERIAL COLLECTED, MICRONS,
C IF C2 USED, MC2=1> OTHERWISE, MC2=0,
C IF C3 USED, MC3ei> OTHERWISE, MC3*0.
C IF SO USED, MOOslf OTHERWISE, M00=0,
C IF LAST STAGE IS S5(S6), MSs5(6).
C IF BACK»UP FILTER USED, MFsl? OTHERWISE, MFsQ.
C G —» IF G«l, WRITE REYNOLD'S NUMBERS AND JET VELOCITIES,
C IF GsO, DO NOT WRITE THEM
C
11 READ(?,300) PO,DP,T,RHO,DUR,DMAX,MC2,MC3,MOO,MS,MF,G
C
C READ IN GAS COMPOSITION IN THIS ORDER—C02(DRY) CO(DRY)
C N2CDRY) 02CDRY) H20
C
READ(2,102) (FG(I),I»l/5)
C
C READ IN 5TAGE COLLECTIONS IN MILLIGRAMS IN THIS ORDERS
C FILTER STAGESC6,5,«,3,2,1,0) C3 OR C?
C
READ(2,106) (MASS(I),I«lf9)
DO 299 I«l,9
MAS3(I)*MASS(I)/1000tO
299 CONTINUE
C
c READ IN IMPACTOR SAMPLING FLOW RATE IN ACFM.
c
READ(2,310) F
C
C READ IN TEST INFORMATIONCDATE,TIME,ETC,) BETWEEN COLUMNS
-------
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
30
SI
52
33
C
C
C
C
C
C
C
C
C
C
C
J7
J8
19
10
>t
>5
»7
'8
•9
0
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
C
C
C
C
C
C
C
C
C
C
r88-
P TO 51, PUT A 1 IN COLU*Kr 1,
13 READ<2,200)
CHANGE DRY GAS COMPOSITION TO
DO 251 1=1,a
FG(I)=FG(I)*(1.0-FG(5))
251 CONTINUE
DEl,P(I) IS THE FRACTION OF IMPACTOR PRESSURE DROP AT EACH
DELP(l)sOtOO
DELP(2)»0,004
DELP(3)sO,006
DELP(4)«0,014
DFLP(5)*0,045
DELP(6)BO,143
DELP(7)»1.000
DC(I) IS THE JET DIAMETER AT EACH STAGE IN CENTIMETERS.
DC(l)eO,361
DC(2)sO,249
OC(3)sO,1775
DC(4)a0.1396
DC(5>sOt0946
DCC6)eO,0731
DC(7)sO,056
SUB(I) IS AN INITIAL ESTIMATE OF DSO'S
VALUES OF SAME. MICRONS.
SUB(1)66.00
SUB(2)=3,15
USED IN CALCULATING FINAL
,
SUB(5)«0,57
SUB(6)*0,33
SUB(7)30'.20
X(I) IS THE NUMBER OF HOLES PER STAGE.
XC6>«1
JSPA(I) IS THE JET TO PLATE SPACING IN CENTIMETERS.
JSPA(l)s|t016
JSPA(2)»0.7
-------
-89-
C CHANGE PO TO ATMOSPHERES.
C
POA=PO/29.92
C
C CHANGE DP TO ATMOSPHERES,
C
DPAsDP/29,92
C
C RA IS THE AVERAGE MOLECULAR WEIGHT OF AIR,
C
RAe28,97
C
C MM 18 THE AVERAGE MOLECULAR WEIGHT OF THE FLUE GAS.
C
MMami,10*FG
-------
-90-
180 00 30 1=1,9
161 KsIABS(J«-10)
162 IMASS(K)sMASSfl)
183 ICUMM(K)=CUMM(I)+OfQOGQ05
iea pRcu02 C
>03 *RITEC3,203) FG(U,FGC2) ,FGC3),FG(,FG(5)
>oa c
?05 C THIS STATEMENT WRITES THE GRAINS PER ACTUAL CUBIC FOOT, THE
}06 C GRAINS PER STANDARD DRV CUBIC FOOT, AND THE IMPACTOR PRESSURE
|07 C DROP,
i08 C
J09 WRITEf3,20«) G»NA,GRNS,DP
J10 C
,U C THIS STATEMENT WRITES THE MILLIGRAMS PER ACTUAL CUBIC METER,
,12 c THE MILLIGRAMS PER STANDARD DRY CUBIC METER,
4 3 C
1« NRITE(3»214> GRNAM,GRNSM
,15 C
,16 C THIS STATEMENT WRITES A HEADING,
47 C
A» WRiTE(3,205)
jl9 C
20 C THIS LOOP WRITES THE STAGE D50'S, THE STAGE COLLECTED MASS,
21 C THE CUMMULATIVE MASS, AND THE CUMMULATIVE PERCENT,
22 C
»3 1F(MC2) «0,aO,50
»4 50 WRXTE(3,206) CVC2,IMASS(1),ICUMM(i),PRCUf2)
>5 40 1F(MC3) 60r<>0,70
>6 70 WRITEC3,207) CYC3,IMASS(1),ICUMM(1),PRCU(2)
>7 60 IF(MOO) 60,60,90
!8 90 WRXTE<3,206> OPC(t),IMASS(?),ICUMM(2),PRCU(3)
(9 60 DO 100 1=1,MS
tO WRJTE(3,209) I,PPC(I+i),IMASS(I«2),XCUMM(I+2),PRCU(I+3)
H 100 CONTINUE
12 IF(MF) 110,110,120
•3 120 WRITE(3,210) IMASSC9),XCUMM(9)
a no CONTINUE
5 C
6 C XF G«l THIS GROUP WRITES THE REYNOLD'S NUMBERS AND THE LINEAR
7 C VELOCITY AT EACH STAGE,
8 C
9 97 IF(G) 95,95,94
-------
HO 9 REYNl(I>,REYNpCl),FDn)
*4 303 CONTINUE
!«5 95 CONTINUE
46 C
IT C THIS SUBROUTINE CALCULATES THE SIZE DISTRIBUTION ON A MASS BASIS
&fl C
w CALL DMOLOB
50 98 60 TO 12
Si 93 STOP
52 99 FORMATCI1)
53 300 FORMAT(F5t2»F6.3fF6tl,F«t2,F5
IT 213 FORHAT(lHO^XrF8,2,9X,F8,2,10X,F9,2)
IB 2l<» FORMAT(1HO,2X,»CALC, GRAIN LOADING - %F9,«,* MGM/ACM»,&X,F9,«,
f9 i' MGM/SOCM*)
10 END
-------
-92-
001 SUBROUTINE STAGES
002 C
003 C
000 C THIS SUBROUTINE CALCULATES THE PRFSSURE AT EACH STAGE,
005 C
006 C
007 INTEGER xt7>
mDCTAi M A c G / o % f c o A / *? % IM&J AJ 11 i tTF\
~ w ™ l» ™r*^W\'/0wVrf*i rJj™*
-------
-93-
H SUBROUTINE VISB
iz c
tt C
34 C THIS SUBROUTINE CALCULATES THE VISCOSITY OF THE CAS USING
js c A METHOD PRESENTED BY c, ». WILKE IN A PAPER ENTITLED
|6 C "A VISCOSITY EQUATION FOR GAS MIXTURES" IN THE .JOURNAL OF
J7 C CHEMICAL PHYSICS VOLUME 8, NUMBER U, APRIL 1950, PAGE 51
18 C
W C
10 INTEGER X(7)
U REAL MASS(9),JSPA(7)fHM,MU,L(7J
|2 DIMENSION WT(5),VSr5)
15 COMMON MASSjJSPAfMM^M
ia COMMON MO,MM1,H,DMAX,OELPC7),FG(5>,OCC7),SUB(7),PERCUC9),CUMM(9)
15 COMMON GRNA,GRNS,GRNAMfaRNSM,CYC2,CYC3,OPCmrREYNl(7).REYN2m
16 COMMON FD(7),PS(7>»AMASSC9)
IT TT*TTK«273,0
18 C
19 C VSCI) ARE THE PURE GAS VISCOSITIES OF C02,CO,N2,02,H20.
!0 C
jl VSU)sl38, 494+0. «99*TT*0,267E-02*TT*TTfO,972E»05*TT*TT*TT
{£ VS(?JB165. 763+0 1«42*TT«-0. 21 3E-03*TT*TT
15
J5 VS(S)sfl7, 800*0 ,
16 C
n C WT(J) ARE THE MOLECULAR ^EIGHTS OF C02,CO,N2,02,H20,
» C
[9 WT(U«««.lO
Ifl WT(2)B28,01
It WTC3)=28.02
14 MUsO.O
15 DO 200 1 = 1,5
!b JF(FG(I)"0,0) 200,199,200
17 199 FG(I)st,OE»aO
!8 200 CONTINUE
19 DO 300 I = l»5
10 XPHEEUO.O
II XPHEEBO.O
« PMEE«0,0
!J DO 400 JB}fS
14 XPH£E»(CifO+(SORT(VS(n/VSM)))*((WT(J)/wTU))**Ot25))**2,0)/'((«.
* XPHEEl«FC(J)*XPHEE
17 JF(J*n 399,400,399
II 399 PH£E«PHEE+XPHEE1
19 400 CONTINUE
10 PHEEBPHEE/FGCIHl.O
il MU8MU*V8(I)/PHEE
!2 300 CONTINUE
iS C
!4 C THE FINAL VISCOSITV MU IS IN POISE.
15 C
16 MU*MUM,OE*06
S7 RETURN
II END
-------
-94-
001
002
005
006
007
008
009
010
Oil
012
013
014
015
016
017
016
019
020
C
C
C
C
C
30
SUBROUTINE
THIS SUBROUTINE CALCULATES THE MOLECULAR MEAN FREE PATH AT tACH
STAGE JET IM CENTIMETERS,
INTEGER xc?)
REAL K
REAL MASS(9),JSPAC7),MM,MU,L(7) ar M_r
COMMON MASS,JSPA,MM,MU,L,X,POA,DPA,TTK.RHO,DUR,RA,F,CLUMM,MC,HCC
COMMON MO,MMi,H,OMAX,DELP(7),FG(5),OC(7),SUB(7),PERCUt9),CUMr-iC^
COMMON GRNA,GRNS.GRNAM,GRNSM,CYCa,CYC3,DPCC7>,REYNK7),RrYN2C7)
COMMON FD(7)fPSm.AMASS(9)
RETURN
END
-------
-95-
001 SUBROUTINE CUTS
002 C
003 C
804 C THIS SUBROUTINE CALCULATES THE STAGE CUT POINTS OR DSO'S BASED
005 C ON EQUATIONS DEVELOPED BY RAN? AND WONG GIVEN IN "IMPACTION
006 C OP DUST AND SMOKE PARTICLES ON SURFACE AND BODY COLLECTORS",
007 C INDUSTRIAL AND ENGINEERING CHEMISTRY, 1952.
008 C
009 C
010 INTEGER X(7)
9U REAL MASS(9),JSPAC7),MM,MU,L<7)
012 COMMON MASS,JSPA,MMfMU,L,X,POA,DPA,TTK,RHO,DUR,RA,F,CCUMM,MC,MCC
013 COMMON MO,MMl,H,DMAX,DrLPC7),FGCS),DC(7),SUBC7),PERCU(95,CUMM(9)
014 COMMON GRNA,GRNS,GRNAM,GRNSM,CYC2,CYC3»DPC(7),REYN1<7),R£YN2(7)
915 COMMON FD(7),PSC7),AMAS$<9)
016 C
017 C THIS ITERATIVE LOOP CONTINUES UNTIL CONVERGENCE WITHIN 0,1%.
016 C
019 DO 30 Iei,7
020 C«0,0
021 DPC(I)«3UBCn
022 4 DPCI«DPCCI>
823 5 CM,0+(2,OH.m/(DPCCI)MffE»O3*Ci«23+0,ai*EXPC~t44*DPC(X)*ltE«4
024 J/LCI)))
025 DPCmBl.03E04*CSQRTCMU*xm*CDCCI)**3)*PSm/(RMO*F*472.0*POA*C):
026 1)
(27 XFCABS(l,0*CDPC(n/DPCI))»O.QOn 30,30,4
(28 30 CONTINUE
»29 CYC2«J55.6*SORT(MU/(RHO*F))
950 CYCS*232,2*SQRTCMU/(RHO*F))
m RETURN
J32 END
-------
-96-
001 SUBROUTINE CUMB
002 C
003 C
004 C THIS SUBROUTINE CALCULATES THE CUMMULATIVE MASS AND CUMMULATW
005 C PERCENT DISTRIBUTION AT EACH STAGE.
006 C
007 C
oos INTEGER xc?)
009 REAL MASS(95f JSPAC7) ,MM,MU,L(7)
Dio COMMON MASS,JSPA,MM,MU,L,X,POA,DPA,TTK,RHO,DUR,RA,F,CCUMM,MC,MCC
•H COMMON MO,MM!,H,DMAX,DELP(7),FGC5),DCC7),SUBC7),PERCU(9)ICUM»U9)
912 COMMON GRNA,GRNS,GRNAM,GRN5M,CYC2,CYC3,DPC(7),REYNim,REYN2C7)
B$S COMMON FD(7)»PS<7>,AMA8S(9)
B14 SUM«0,0
H5 DO 50 I = t, 9
H6 SUNeSUH+MASSCI)
lift 50 CONTINUE
)19 DO 60 1=1,9
>20 PERCU(I)«(CUMM(I)/8UM)*100.0
)21 60 CONTINUE
>aa c
>23 C GRNA IS THE GRAINS PER ACTUAL CUBIC FOOT.
)24 C
)26 C
>27 C GRN8 18 THE GRAINS PER STANDARD DRY CUBIC FOOT.
>28 C
>29 GRNS(s((SUM*lSt432«)/C(F*OUR*298fO*POA)/(TTK*ltO»)/(l,Oi-FG(5))
ISO C
'31 C GRNAM IS THE MILLIGRAMS PER ACTUAL CUBIC METER.
132 C
133 GRNAMcGRNA*2288,34
134 C
135 C GRNSM IS THE MILLIGRAMS PER STANDARD DRY CUBIC METER.
>36 C
>37
138 RETURN
>S9 END
-------
-97-
191 SUBROUTINE REYNNB
102 C
»3 C
P« C THIS SUBROUTINE CALCULATES THE LI^FAR VELOCITY OF THE GAS AT
105 C EACH STAGE JET IN CM/SEC, AND THE REYNOLD'S NUMBERS BASED ON
(06 C JET TO PLATE SPACING AND JET DIAMETER.
197 C
108 C
109 INTEGER X(7)
HO REAL MASS(9),JSPA(7),MM,MU,LC7)
111 DIMENSION PPflO),RHOD<10)
112 COMMON
113 COMMON MQlMMi,«,DMAX,DELP(7),FG(5),DC(7),3UB(7),PE»CU(9),CUMM(9)
HA COMMON GRNA,GRNS,GRNAM,GRNSM,CYC2,CYC3,OPC(7),REYNim,REYN2m
115 COMMON FD(7),P8(7),AMASS(9)
lib DO 10! Xsi,7
"17 PP
-------
-98-
SUBROUTINE DMDLDB
C
C
C THIS SUBROUTINE CALCULATES THE SIZE DISTRIBUTION ON A MASS BASIS.
C
C
INTEGER X(7)
REAL IHASSC9)
REAL MASS(9),JSPA(7),MM,MU,L(7)
DIMENSION GGRN5C9),DIFF(9),DMDID(9),GEOMDC9)
COMMON MASS,JSPA,MM,MU,L,X,POA,DPA,TTK,RHO,DUR,RA,F,CCUMM,MC,MCC
COMMON MO,MMi,H,OMAX,OELPC7),FG(5),DC(7),3UB(7),PERCU<9),CUMM(9)
COMMON GRNA,GRNS,GRNAM,GRNSM,CYC2,CYC3,DPC(7),REYN1<7),REYN2C7>
COMMON FD(7),PS<7),AMASS(9)
COMMON IMASS,MC2,MC3,M00,MS
C ' -
C THIS LOOP CALCULATES THE GRAINS PER STANDARD DRY CUBIC FOOT PER
C STAGE.
C
DO 505 1*1,9
GGRN*(I)«((IMAS8G(DMAX)-ALOGCCYC2»
DIFF(2)sO.«3«29<»*{ALOG(CYC2)^ALOG(DPC(2)))
DIFF(3)sO,43))
DIFF(4)80.a3«294*(ALOG(DPC(3))»ALOG(DPC(a)))
DIFF(5)80.434294*(ALOG(DPC(4))«ALOG(DPC(5)))
DIFF(6)80.434294*(ALOG(DPC(5))*ALOG(DPC(6)))
DIFF(7)«0.434294^(ALOGCDPC(6))»ALOG(O.OB))
C
C DMDLO ZS A DIFFERENTIAL SIZE DISTRIBUTION ON A MASS BASIS.
C
DMDLDC1)aGGRN8(1)/OIFF(1)
DMDU><2)»GGRN8(3)/DIFF(2)
DMOtD(3)BQGRN8(4)/DIFF(3)
DMDLD(4)a6GRN8C5)/DIFF(4)
DMDLD(5)«GGRNS(6)/DIFF(5)
DMDLD(6)»GGRNS(7)/DIFF(6)
.. DMDLD(7)«GGRNS(9)/DIPFm
C
C GEOMD(I) 18 THE GEOMETRIC MEAN OF THE STAGE DSO'S,
C
GEOMDU)*EXP(2,302585*(0,434294*ALOG(CYC2)+0,5*DXFF(l)n
GEOMD(2)sEXP(a,302585*(0,43«294*ALOG(DPC(2))*0.5*DlFF{2)))
GEOMo(3)«EXP(2.3025B5*tO.«34294*ALOG(DPC(3))+0.5*DIFF(3)))
GEOMD(4)BEXP(?t302585*(Ol43a294*ALOG(DPC(4»'fO,5*DlFF(4)))
GEOMD(5)»EXP(a,302585*(0,434294*ALOG(DPC(5))*0,5*DIFF(5)))
-------
-99-
GEOMD(6)=EXP{2,302585*(0.434294*ALOG(DPC(6mO,5*DlFF(6)))
WRITE (3, 506) GGRNS(3),DMDLD<2),GEOMD(2)
WRITE (3, 506) GGRNSf4),DMOtD(3),GEOMDm
WRITE(3,506) GGRNS(5),DMOtD(4),GEOMD(4)
WRITE (3, 506) GGRNS(6),DMDLDC5),GEOMD(5)
WRXTE(3,506) GGRN517) ,&MOLD(63 ,GEOMD(6)
WRITE (3, 566) GGRNS(V) ,OHOLO(7) ,GEOMD(7)
GO TO J50
CONTINUE
DIFF(2)80.434294*(ALOG(CYC2)oALQG(DPC(2)))
DIFFC3)«Ot434294*C(3n)
OIFFC6)"0,tt3fl29«*(ALOGfDPCf5))-ALOG(DPC(6)))
OXFF(7)=0,a3«29fl«CALOG(l>PC(6))»ALOG(OPC(7)))
DXFP(6)«0,434294*CAt.OGCOPC<7))-AL(tGCOt06))
DMDI,Dn)»CORNS(l)/OJPF(l)
DMOLD(3)BGGRN3(«)/DIFF(5)
DMDi.DC4)aGfiRNSC5)/DIFF(4)
DMDlO(5)eGGRNS(6)/DZFPC5)
OMOLb(6)«CGRN8C7)/OIFF(6)
f>MDLD(8)eGGRNS(9)/DlFF(S)
GEOHD(i)BEXP(2,302S85*(0,434294*AtOG(CYC2)40,5*DXFFCl)))
CEOMD(2)»EXP(2t302585*(Ot434294*AlOG(DPC(2))+0J5*DIFF(2)))
GEOMDf3)sEXP(2.302585*(Ot434294*ALOG(OPC(3))*0,5*OlFF(S)J)
.302565* <0,434294*ALOG(OPC( 4) )*0,5*DIFF( 4)))
,302585* CO,434294*ALOG(OPC( 5) )*0,5*DIFP (5)))
GEOMD(6)BEXP(2,302585*(Ot434294*ALOG(OPC(6))+0,5*DlFF(6)))
GEOMD(6)cEXP(2,302565*C0.434294*ALOG(0,OB)+0,5*DXFF(e))3
&EOMO(7)»EXP<2, 302585* (0,434294*ALOG(DPC (7) )*0.5*OIFF(7)))
WRITE (3, 506) GGRN3(n^OMOLOCl)«GEOMD(t)
MRXTE(3»506) GGRN5<3)«DMDL&C2)«GEOMDC2)
WRITE(3»506) GGRN3(4)»DMOtO(3),GEOMO(3)
WRITE(3,50fc) GGRN8C5)*DMDLD(4),GEOMD(4)
WRITE (3,506) G6RN8(6),DMDID{5),GEOMD(5>
WRITE(3«506) GGRNS(7J ,OMDLDC6) ,GEOMO(6)
WRITE (3, 506) GGRH8(8) ,OMDLDC7) ,GEOMO(7)
MRXTE(3»506) GGRN5m*DMDLDC6),GEOMD<8)
GO TO 150
K XF(MC3) 50«50r60
60 IF(M5«5) 70,70,80
70 CONTINUE
DIFFCn-0. 03429*1* CALOGAI.OG(DPC(2)))
DXFF(4)BQ,434294*CALOG(OPC(2n*ALOG(DPC(3)))
DXFF(5)B09434294*eALOG(DPC(3))*ALOG(OPCC4)))
DXFPC6)80I434294*(ALOG(OPC(4))»ALOG(OPC(5)3)
DXFF(7)BO,43429«*(At.06(OPC(5))pALOG(DPC(6)))
DXFF(B)ii0.434294*eALOG(OPC(6))«ALOG(OtOe))
DMDIDCI )«GGRN8n )/DX?F Ci )
DHDLD(2)*GGRN$(2)/DXFFC2)
DMOCO(5)«6GRN3(5J/OIFF(5)
-------
-100-
120 DMr>LD(6)=GGRNS(6)/DIFF(6)
122 DMDLD(8)=GGRNS(9)/DIFF(8)
123
12«
125
I?6 GEOMD(4)=EXP(2.302585*(0,434294*ALOG(DPCC3))*0,5*DIf-F(a}) )
127 GEOMD(5)=EXP(?f 302585* (0 ,434294* At_OG(DPC(
-------
180 110 CONTINUE -101-
181 DlFF(l)sot434294*CAlOG(DMAX)*ALOGIFF<6)sot434294*(Al.OGCDPC(5))-ALOG(DPC<6)))
187 DIFF(7)eot434294*(ALOG(DPC(6))-AUOG(0,Ofl))
188 DMDLD(1)=GGRNS(2>/DIFFC1)
189 DMDUD(2)*GGRNS(3)/DIFF(2)
190 DMDLDC3)eGGRNS(4)/DIFF(3)
191 DWDi.D(4)3GGRNS(5)/DIFF(4)
192 DMDl.DC5)sGGRNS(6)/DXFF(5)
193 DKDUDC6)sGGRNS(7)/DIFFC6)
194 OMOLD(7)sGGRNS(9)/oiFF(7)
195 GEOMDn>sEXP(2,302585* (0.434294* ALOGCDPC(i))*0,5*DlFFU)))
196 GEOMOC2)«EXP<2,302585*{0,434294*ALOGCDPC(2))+0,5*DIFF(2)M
197 . GEOHDC3)«EXP(2,302585*(Ot434294*AlQGCDPCC3)) + 0.5*DIFF{3M)
198 GEOMDC«5»eXP(2f302565*(0.434294*At.OGCDPC(4))*0,5*DIFFC4)))
199 GEOMO(5)«EXP(2t302585*CO,434294*AtOG(OPC(5))tO,5*DIFF(5))J
200 CEOMDC6)«EXP(2,302585*(Of434294*ALOGCOPC(6))-fOt5*DIFF(6)))
201 GEO«DC7)«EXf»(2,302565*C0.434294*AUOGC0.08) + 0
202 MRITEC3«506) GGRNS<2),DMDLD(1),GEOMD(l)
203 WR!TE(3,506) GGRNSC3),OMOLD(2),GEOMD(2)
204 NRITE(3,506) GGRNS(fl),OMOLO(3),GEOMD(3)
205 WRITEC3,506) GGRNS(S),OMOLO(4),GEOMDC4)
206 WRITE(3,506) GGRNS(6),DMOLO(S),GEOMOC5)
207 WRJTEC3,506) GGRNSC7),DMOtD(6),GEOHDC6)
208 WRITEC3,506) GGRNS(9),DMOLO(7),GEOMD(7)
209 GO TO 150
210 120 CONTINUE
211 DIFF(1)BO, 434294* (ALOG(OHAX)f.ALOG(OPC(l)))
212 DlFF(2)eO,434294*(ALOGCOPC(1))*ALOG(OPC(2)))
213 DIFF(3)«0.434294*(ALOG(OPC(2))»ALOG(OPC(3)))
214 DIFF(4)aO,434294*CALOGCDPC(3>)-ALOG(DPC(4)))
215 DIFF (5) «0,434294* (ALOG(DPCf4n-ALOG(DPC(S»)
216 DIFF(6)BO,434294*(ALOG(DPC(5))»ALOG(OPC(6)))
217 OIFF(7)»0.434294*(AUOG(OPC(6))-AL06(DPC(7)))
216 DIFF(8)»0,434294*(ALOGCDPCC7))«ALOGCO,08))
219 DMDlDCn»GGRNSC2)/DIFF(l)
220 DMDLO(2)*GGRN8(3)/OIFF(2)
22S DMDLO(3)»GGRN8(a)/OIFF(i)
222 OHOtD(«)»GGRNS(5)/DIFF(4)
223 DHDUD(5)«GGRNS(6)/OIFF(5)
224 DMDUDC6)«GGRNS(7)/DIFFC6)
225 DMDUD(7)«G6RNS(8)/OIFF(7)
226 OMOLD(8)aGGRNS(9)/OIFF(8)
227
226
CC TF W^V'^^^^^i"*^7 ^••p*VV*««'^^^'~'*Vp-««^-T^V-Vvvp-i||pW^«^i •*«•»« ^-^WyMrr^vAr f + ^ f f f
230 GEOHD(4)*EXP(2(302585*CO«434294*AtOG(DPC(4))^0,5*DZFF(4)))
231 GEOMDC$)*EXP(2,30258S*(0V434294*AIPOGCDPC(5))*0.5*DIFF(5)))
232 GEOMO(6)«EXP(2,302565*(0,434294*ALOGCDPC(6))«0,5*DIFF(6)))
233 GEOttDe7)»EXP(2,302565*CO>434294*ALOGCDPC(7))+0,5*DlFF(7)))
234 GEOHD(6)BEXP(2,302585*(0,434294*ALOGCO,08)f0.5*OIFF(8)))
235 WRITE(3,50b) GGRN8(2),OMOLO(1),GEOMD(1)
236 WRITE(3,506) GGRNSC3),DMDLD(2),GEOMDC2)
Z37 WRlTp:(3,506) GGRN8<4) ,DMOLDC3),GEOMO(3)
238 «RITE(3,506) GGRN8(5),OMDUD(4),GEQMD(4)
239 WRITE(3,506) GGRNS(6),DMDLO(5)fGEOMOC5)
-------
-102-
240 HRXTE(3,506) GGRN5C7) ,DMDLt><6) ,GEOMD(6)
241 WRITE(3,506) GGRN5(8),DMDUO(7),GEOMO(7)
242 WRITE(3,506) GGRNS(9),DMDLD(8),GEOMD(8)
243 GO TO 150
244 90 IF(MS*5) 130,130,140
245 130 CONTINUE
246 DIFF(l)sOt434294*(ALOG(DMAX)-ALOG(DPC(2)))
2«7 DJFF(2)sO,434294*(ALOG(OPC(2))-AUOG(DPC(3)))
248 DIFF(3)=Of4S4294*(ALQG(OPC(3))-ALOG(DPC(«)))
2^9 DJFF(4)BOe434294*(ALOG(DPC(4))»ALOr»(DPC(5)))
250 DIFF(5)s0.43429a*(ALOG(DPC(5»*ALOG(DPC(6)))
251 DIFF(6)so,434294*(AlpGCDPC(6))-ALOG(0,08))
252 DMDLD(i)sGGRNS(3)/OXFF(l)
254 DMDIO(3)8GGRNS(5)/DIFF(3)
256 OMDLD(5)8GGRNS(7)/DXFF(5)
257 DMOL.P(6)aGGRN8(9)/DXFF(6)
258 GEOMO(l)iiEXP(2,302585*(0,434294*ALOG(OPC(2))fO,5*DlFF(l)))
259 6EOMO(2)8EXP(2,302585*(0,434294*ALOG(OPC(3))tO,5*OIFF(2)))
260 GEOHO(3)8EXP(2,302585*(0,434294*ALOG(DPC(4))tO,5*OXFF(3)))
261 GEOMD(4)8EXP(2,302585*(0,434294*ALOG(DPC(S))40,5*OIFF(4)))
262 GEOMD(5)aEXP(2,302585*(0,434294*AlOG(DPC(6))+0,5*DXFF(5)))
263 GEOHD(6)8EXP(2,302585*(0,434294*AUOG(0,08)t0.5*OIFF(6)))
264 WR!TE(3,506) GGRNS(3),OHOID(1),GEOMD(1)
265 HRIT£(3,506) GGR*N8(4),DHDLD(2),GEOMD(2)
266 WRXTE(3,506) GGRN8(5),DHOID(3),GEOMD(3)
267 MRXTE(3,506) GGRN$(6),DHDLD(4),GEOHO(4)
268 WRITE(3,506) GGRN8(7),PMOLD(5),GEOWO(5)
269 WRXTE(3,506) GGRN8(9),OMOLD(6),GEOMOC6)
270 GO TO 150
271 140 CONTINUE
272 OIFF(l)so,434294*(ALOG(OMAX)».ALOG(OPC(2)))
273 DXFF(2)sOf434294*(ALOG(PPC(2))wALOG(DPC(3))
274 PXFF(3)80,fl34294*(AlOG(DPC(3))»ALOG(DPC<4))
275 OIFF(4)s04434294*(AUOG(DPC(4))»ALOG(OPC(5))
276 DIFF(5)80t434294*(AL06(OPC(5))«ALOG(OPC(6))
277 DIFF(6)80,434294*(AtOG(OPC(6))»ALOG(OPC(7))
278 DXFF(7)aO,434294*(AiOC(PPC(7))-ALOG(0,08))
279 PMOLO(l)aGORN8(3)/OXFF(l)
280 PMDLO(2)»6GRN8(4)/DXFF(2)
281 OMDLD(3)aCGRNS(5)/DIFF(3)
282 DMOLD(4)866RN8C6)/OXFF(4)
283 D«OLOf5)8GORN8(7)/OXFF(5)
284 DMDLO(6)aGGRN8(8)/OXFF(6)
285 P«OtP(7)aGGRN8(9)/OXFF(7)
286 6EOMO(1)8EXP(2,302585*(0,434294*ALOG(DPC(2))«0,S*DIFFC1))
287 6EOMO(2)aEXP(2,302585«(0,434294*ALOG(DPC(3))t0.5*OXFFC2))
288 GEOMD(3)aEXP(2,302585*(0,434294*Al.OG(OPC(4))*0.5*OIFF(3))
289 6EOMOe4)aEXP(2,302585*(0,434294*AtOG(DPC(5))*0,5*OlFF(4))
290 GEOHD(5)aEXP(2,30258S*(0,434294*ALOG(DPC(6))«0,5*DXFF(5))
2^1 GEOMO(6)aEXP(2i302585*(0^434?94*ALOG(DPC(7))*0,5*DXFF(6)))
292 GEOMD(7)8eXP(2.3025B5*(0,434294*Al06(0.08)+0,5*DXFF<7)))
293 HRXTE(3,506) GQRN8(3),DHDLO(1),GEOMO(1)
294 HRITE(3,506) GGRN8(4),PMDLO(2),GEOMO(2)
295 WRXTE(3»506) GGRN3(5),pMOUO(3),GEOHO(3)
296 WRXTE(3,506) GGRNS(6),OMOUD(4),GEOMD(4)
297 VIRXTE(3,506) GGRNSC7),0*010(5),GEOMD(5)
298 WRXTE(3,506) GGRN8(8),OMDLO(6),GEOMD(6)
299 WRXTE(3,506) GGRNS(9),DMOUOC7),GEOMD(7)
-------
JOO 150 CONTINUE -103-
301 RETURN
502 END
PROG > 4K
-------
-104-
APPENDIX B - EXTRACTIVE SAMPLING
Figure Bl shows our experimental setup for extracting,
diluting, and drying representative aerosol samples from the
duct. As shown in the schematic, a precollector cyclone is used
to prevent particles from clogging the sample metering orifice.
At the flow rate normally used, the cyclone has a cut point
or D50 of about 5 jam and is oriented with the inlet nozzle
downstream. Thus, this sampling method is anisokinetic and the
sample withdrawn from the duct is representative of the flue
gas only at diameters of about 2 ym and less.
The main probe and*.cyclone support is a 24 inch long
section of 3/16 inch pipe. The sampling line is a length of
k" copper tubing. The diluter body is a hollow aluminum cylinder
approximately 5 inches in diameter, and 18 inches long. A
perforated cone is located at one end within the diluter cavity
and dilution air is?forced through a number of jets arranged
at various angles on the cone so that a highly turbulent mixing
zone is established. The saaple aerosol is drawn in at the apex
of this eone.
Several taps are available on the diluter body where inlet
lines to the optical and CN counters are connected. Other taps
are located downstream on the diluter for return lines from
these instruments. Return lines are sometimes necessary for
proper operation of the aerosol sampling instruments when the
diluter interior is significantly above or below atmospheric
pressure because the sample flow rates will fluctuate and
pneumatic valves will not function if the inlet and outlet
pressures differ by more than a few inches of water.
Fine adjustments of the dilution ratio are made by opening
or closing a bleed valve located on the dilution air pump.
With this valve closed and return lines connected for all the
aerosol sampling instruments, the air flow system is a closed
-------
-105-
f
CYCLONE
(OPTIONAL)
FLOWMETER
FLUE
GAS
CYCLONE
PUMP
SAMPLE
METERING
MANOMETER
DILUTION
DEVICE
CHARGE
NEUTRALIZER
DILUTION AIR
METERING
ORIFICE
MANOMETER
BLEED
Figure Bl.
Extraction and Dilution system for out of stack
flue gas sampling.
-------
-106-
loop and no sample is drawn into the diluter. If, however, the
bleed valve is opened, some air escapes and the sample air
enters as makeup air. Using the bleed valve, the dilution isatio
can be varied in a predictable fashion by about a factor of
ten. For large changes in dilution, say on going from an inlet
situation to an outlet, the "sampling orifice is replacear"with a
venturi. Water manometers were adequate to monitor the orifice
differential pressures.
For €he sampling orifice,
AP T
QSO - cso oso
For the venturi,
= Cv
pv
And, for the dilution air metering orifice,
where
Q0_ = flow rate through sample metering orifice,
oU
CSQ = calibration constant for sample metering orifice,
APg = ppessune- differential across the sample metering orifice,
TSO = temperature of the sample metering orifice, &
P__ = the absolute pressure upstream from the sample
metering orifice.
-------
-107-
Subscripts of v and D denote these same parameters for the
sample metering venturi and dilution air metering orifice,
respectively.
The absolute pressure does not vary significantly from the
dilution air metering orifice to the sample metering orifice,
so that the dilution ratio can be calculated from:
D.F. = = APDTD
80*80
D F . = O = • D / AP T
D -ff /DP , depending on which sample metering
Qv v |/ ApvTv
arrangement is used.
The fraction of diluent air that is returned from the
instruments is passed through glass condensers in an ice bath
and dried. In some cases, additional drying is needed to avoid
condensation in the sampling lines. Diffusional drying cylinders
were used in the sampling lines to the instruments on one occasion
where the flue gas contained 40% water. A diffusional dryer
is shown in Figure B2. The sample aerosol is passed through the
cylindrical cavity bounded by the 100 mesh stainless screen.
Water molecules, having much higher mobilities than particulate,
diffuse through the screen and are absorbed by the drierite.
We hope that a similar device can be used to remove SO which
H
also presents condensation problems when sampling. This depends
upon selecting a material which will adsorb or absorb SOX readily.
-------
-108-
DRJ
GLASS CYLINDER
100 MESH STAINLESS
SCREEN
SAMPLE AEROSOL
Figure B2. Diffuaional Adsorption Apparatus for Removal of
HaO From Sample Aerosol.
-------
-109-
APPENDIX C - DIFFUSIONAL PARTICLE SIZING
A. Experimental Procedure
Figure Cl is a block diagram showing the experimental set
up for making particle size distribution measurements in the
ultrafine range. One condensation nuclei (CN) counter is
used to constantly monitor the diluted sample so that variations
in particle concentration due to process fluctuations can be
accounted for in the data reduction. A second CN counter, with
a variable flow rate, is used to measure the particle concentration
in the aerosol after passing through the diffusion batteries.
The plumbing is arranged so that this CN counter reads the
number concentration at the diluter outlet, or after penetrating
one, two, three, or four diffusion batteries in series. The
exhaust air from the CN counters is dried, filtered, and then
returned to the diluter. The use of return lines minimizes
pressure differentials between the inlet and exhaust section of
the CN counters which might otherwise prevent proper operation.
The antipulsation device shown in Figure Cl consists of a
0.005" thick surgical glove stretched over a 4V plastic
funnel as described by Sinclair.15 These are necessary to damp
out flow pulsations caused by the valving action in the CN
counters and to insure laminar flow through the diffusion batteries,
When calculating — , the fraction of the aerosol which penetrates
o
a series of diffusion batteries, the transport time through the
diffusion batteries must be taken into account. This is about
3%-5 minutes for each of the 100 channel diffusion batteries.
Thus, with four large diffusion batteries in series, (n) at a
particular time, would be related to nQ at a point 16-20 minutes
earlier in time on a chart recording. Making a complete measure-
ment of a particle size distribution takes 2-4 hours; thus,
diffusional measurements are most useful on stable sources where
the particle size distribution is constant in time.
-------
-110-
ANTI-PULSATION
OCVtCE
SAMPLE
DILfltlft
ANTI-
CN COUNTER
•t MODEL
RETURN TO
WLUTER
CN COUNTER
El MOOCL RICH KX)
D.B. 2
D.B. 3
D.B. 4
D.B. 5
RETURN
TO WLUTCR
Figure Cl. Diffusion battery and condensation nuclei
counter ley-out for fi«e particle sizing.
-------
-111-
Although the tests are normally run at a constant
dilution ratio, the bleed value on the dilution air pump should
be adjusted periodically to obtain a new dilution ratio and check
the linearity of the diluter. Agreement between the calculated
change in dilution (calculated from flow rates as measured
using the calibrated orifice) and the response of the CN counters
indicates that the diluter and CN counters are functioning
properly .
By varying the number of diffusion batteries in series and
the flow rate, it is possible to measure penetrations under a
variety of conditions. In most cases, an effort is made to
measure the penetration jp£ four-100 channel diffusion batteries,
one to four batteries in series and the 12 channel diffusion
battery, all at three different flow rates. This yields fifteen
data points from which the particle size distribution (0.01-0.2 ym
diameter) can be reconstructed.
B. Data Analysis
The penetration of parallel plate diffusion batteries by
a monodisperse aerosol is given by:17
no
+ 0. OSle'33'06^ + ........ (Cl)
where
x - 3.77 bl/aQ,;
b = height of each rectangular slot (cm) ,
1 = length of each rectangular slot (cm) ,
2 a = width of each rectangular slot (cm) ,
Q = volumetric flow rate (cm3/sec) ,
n = outlet concentration (No. /cm3) ,
n = inlet concentration (No. /cm3) ,
o
-------
-112-
D = diffusion coefficient of particles (cm3/sec) , and
m = number of rectangular channels in each diffusion battery.
Three methods were used to calculate particle size
distributions from raw diffusional data. One method was to
calculate a set of penetration curves similar to those in
Figure Ib, and to apply the Dgo technique which is commonly
used with cascade impactors. This method was described in
Appendix A. A second technique is that introduced by Fuchs et al20
which assumes a log normal distribution. We regard this assumption
as a loss of generality and this technique was not often used.
The third technique/ and the one which we normally use, was
suggested by Sinclair.15 Looking at equation Cl, it can be
seen that for a particular particle size (or diffusion coefficient) ,
a diffusional sampling geometry can be described by the parameter
xm, which involves the diffusion battery geometry and aerosol
flow rate. In order to facilitate1 the data reduction/ a nomograph
is prepared/ plotting °- versus xm for a number of particle
"o
sizes. Our nomograph is shown in Figure C2. The experimental
data points will constitute a single curve similar to that shown
in Figure C3. There is, however, some ambiguity associated
with forming this experimental graph. Theoretically, the pene-
tration of two or more diffusion batteries in series is given by
the product of the penetrations of each individual battery. If
all the terms in equation Cl are used, there is no unique value
of the parameter xm, representing an equivalent diffusion
battery geometry which is valid for all particle sizes. For the
purpose of plotting the experimental data (n/n vs. xm) , we
have approximated the penetration by the first term in this
equation. The error introduced by this approximation is less
than two percent in all cases. Specifically, for one diffusion
battery, is plotted versus tin; for two similar batteries,
-------
M
too
2.06
xm,
4.12 6.18 8.24 I0.3O
CMFFUSION BATTERY PARAMETER (SEC/CM8* IO
12.36
14.42
0) US
rH T3
O
•H >1
-P H
M
-P C
(U-H
tJ n
0 S
»M O
•H
^J 4J
H-H
tn ^
o -P
•
CN
0
i»
g.
-------
100
UJ
Q.
O
Figure C3. Experimental diffusion battery data to be used in
conjunction with Figure C2 in determining fine particle
size distribution.
14.42
-------
-115-
n/nQ is plotted versus 2xm; for three similar batteries
is plotted versus 3xm; etc..
(0.91)2
The graphical stripping procedure consists of placing the
experimental curve on top of the nomograph (C2) so that the
ordinate axes lie along the same line. The experimental curve
is then shifted vertically until it is slightly above a matching
curve on the nomograph. The experimental curve should approach,
but not cross, the theoretical (nomograph) curve at the extreme
right hand side of the nomograph. The value on the ordinate
on the experimental curve where the monodisperse penetration
curve crosses (y intercept) is the fraction of the aerosol
which is contained in that size (theo. curvef and larger. The
theoretical curve is subtracted from the experimental curve and
the result is treated as a new experimental curve. From this
point, the process is repeated until a cumulative particle
size distribution is obtained over the desired range.
-------
-116-
APPENDIX D - OPTICAL TECHNIQUES
Insofar as this project is concerned, the study of aerosol
size distributions by optical techniques has taken second place
to impactor and diffusional sizing methods. It has been used
mainly to correlate inertial and diffusional data and has not
undergone as extensive a study as these other sizing techniques.
These measurements of particle size and concentration by
light scattering involved the use of a Climet Particle Size
Analyzer, Model CI-201. The theory of operation has been
described in an earlier section. In this case, no correction was
made for aerosol refractive index. The particle counter was
calibrated in our laboratory using polystyrene latex spheres.
The calibration curve is shown in Figure 4.
Data was obtained in two ways. The Climet has a front
panel meter showing cumulative particles per cubic foot larger
than an indicated size, selected as either 0.3, 0.5, 1.0, 3.0,
5.0 or 10.0 microns. Full scale on the meter is 106 particles/cu.ft. f
In normal field use, sample stream dilution was required to keep
the concentration below this full scale value. The dilution
System is described in the Appendix on Extractive Sampling.
A size distribution was obtained by taking the differences in
particle concentrations at adjacent size thresholds and plotting
these values at the geometric mean of these particle sizes.
The particular Climet Analyzer received by Southern Research
Institute from the Environmental Protection Agency had an Ortec
digital ratemeter and an Ortec digital voltage spectrum scanner
incorporated with the size analyzer. This allowed more resolution
and a greater number of data points. The ratemeter also allowed
for time averaging from 10 seconds to 10 minutes. This was
useful at some testing locations where the concentration of large
particles was quite low. Also, the scanning cycle would repeat
automatically, allowing for uninterrupted data accumulation
over several hours. This data was handled in the same fashion
-------
-117-
as that acquired using the panel meter; i.e., by taking concen-
tration differences and plotting at the particle size geometric
mean.
Figure 6, found in the Experimental Methods Section,
shows data acquired by the optical particle counter at a
coal-fired power boiler. It correlates quite well with the
impactor and diffusional data.
B i rmi ngham, Alabama
July 12, 1974
2923-XXI
(10:1:2:15) mlm
Mr. D. B. Harris, Project Officer - 10 copies
Mr. M. P. Huneycutt, Contracting Officer - 1 copy
Mr. Robert Lorentz - 2 copies
-------
-118-
REFERENCES
1. May, K. R., "The Cascade Impactor: An Instrument for Sampling
Coarse Aerosols", J. Sci. Instr. 22 (October, 1945).
2. Andersen, A. A., "New Sampler for the Collection, Sizing,
and Enumeration of Viable Airborne Particles", J. of
Bacteriology, 76 (1958).
3. Brink, J. A., Jr., "Cascade Impactor for Adiabatic Measure-
ments", Ind. and Eng. Chem., 44, No. 6 (June, 1952).
4. Rantz, W. E. and Wong, J. B. , "Impaction of Dust and Smoke
Particles", Ind. and Eng. Chem., 50, No. 4 (April, 1958).
5. Pilat, M. J., Ensor, D.S., and Busch, J. C. , "Cascade Impactor
for Sizing Particulates in Emission Sources", Am. Ind.
Hygiene Assoc. J., 32, No. 8, (August, 1971).
6. Liu, B. Y. H., Whitby, K. T., and Pui, D. Y. H. , "A Portable
Electrical Aerosol Analyzer for Size Distribution Measure-
ment of Submicron Aerosols", Presented at the 66th Annual
Meeting of the Air Pollution Control Association, Paper No.
73-283, (June, 19730.
7. Cohen, J. J. and Montan, D. N., "Theoretical Considerations,
Design, and Evaluation of a Cascade Impactor," Am. Ind.
Hygiene Assoc. J., (March-April, 1967) .
8. Lundgren, D. A., "An Aerosol Sampler for Determination of
Particle Concentration as a Function of Size and Time",
J. Air Pol. Con. Aasoc., 17, No. 4, (April, 1967).
9. Gussman, R. A., Sacca, A. M., and McMahon, N. M. , "Design
and Calibration of a High Volume Cascade Impactor", J. Air
Pol. Con. Assoc. , 23, No. 9, (September, 1973) .
10. Strom, L., "Transmission Efficiency of Aerosol Sampling Lines",
Atmos. Env., 6, (1972).
11. Haberl, J. B. and Fusco, S. J., "Condensation Nuclei Counters:
Theory and Principles of Operation", General Electric
Technical Information Series, No. 70-POD 12 (1970).
12. Hodkinson, J. R. , "The Optical Measurement of Aerosols",
Aerosol Science, Academic Press (1966) ed. by C. N. Davies.
13. Berglund, R. N., "Basic Aerosol Standards and Optical Measure-
ments of Aerosol Particles," Ph.D. dissertation, Mech.
Eng. Dept., U. of Minnesota, (1972).
-------
-119-
14. Fuchs, N. A., The Mechanics of Aerosols, The MacMillan Co.,
60-5th Avenue, N. Y. (1964).
15. Sinclair, D., "A Portable Diffusion Battery", Am. Inst.
Hygiene Assoc. J., (November, 1972).
16. Breslin, A. J. , Guggenheim, S. P., and George, A. C.,
"Compact High Efficiency Diffusion Batteries", Staub-
Reinhalt, Luft, Vol. 31, No. 8 (August, 1971).
17. Twomey, S., "The Determination of Aerosol Size Distributions
from Diffusional Decay Measurements", J. F. I. (February, 1963)
18. Sansone, E. B. and Weyel, D. A., "A Note on the Penetration
of a Circular Tube by an Aerosol with a Log-normal Size
Distribution", Aerosol Science, Vol. 2 (1971).
19. Gormley, P. G. and Kennedy, M., Proc. Roy. Irish Acad.,
Vol. 52A (1949).
20. Fuchs, N. A., Stechkina, I. B. , and Starasselskii, V. I.,
COn the Determination of Particle Size Distribution on
Polydisperse Aerosols by the Diffusion Method", Brit. J. Appl.
Phys., 13, pp. 280-281 (1962).
21. Liu, B. Y. H., Mechanical Eng. Dept., University of Minnesota,
private communication.
22. Berglund, R. N. and Liu, B. Y. H. , "Generation of Monodisperse
Aerosol Standards", Environmental Science and Technology,
Vol. 6, No. 2, 1973.
23. Lindblad, N. R. and Schneider, J. M. , "Production of Uniform-
Sized Liquid Droplets", J. Sci. Instr., Vol. 42, 1965.
24. Strom. L. , "The Generation of Monodisperse Aerosols by Means
of a Disintegrated Jet of Liquid", Rev. Sci. Instr. , Vol. 40,
No. 6, 1969.
25. Stober, W. and Flachsbart, H. , "An Evaluation of Ammonium
Flporescein as a Laboratory Aerosol", Atmos. Environ.,
Vol. 7, 1973.
26. Picknett, R. G. , "A New Method of Determining Aerosol Size
Distributions From Multistage Sampler Data", Aerosol
Science, 1972.
-------
-120-
TECHNICAL REPORT DATA
(Pleat read Imitncttons on the nvene before completing)
1. REPORT NO.
EPA-650/2-74-102
2.
3. RECIPIENT'S ACCESSIONED.
4. TITLE AND SUBTITLE
Particulate Sizing Techniques for Control Device
Evaluation
5. REPORT DATE
October 1974
6. PERFORMING ORGANIZATION CODE
7. AUTHORS) Wallace B. Smith, Kenneth M. Gushing, and
Joseph D. McCain
8. PERFORMING ORGANIZATION REPORT NO.
SORI-E AS-74-138
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Southern Research Institute
2000 Ninth Avenue, South
Birmingham, Alabama 35205
1O. PROGRAM ELEMENT NO.
1AB012; ROAP 21ADM-011
11. CONTRACT/GRANT NOT
68-02-0273
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
NERC-RTP, Control Systems Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Summary: 7/73-6/74
14. SPONSORING AGENCY CODE
18. SUPPLEMENTARY NOTES
ipne report gives results of a study that has proven the feasibility of meas-
uring particle size distributions in stack gases (and fractional efficiencies) from
0.01 to 15 urn. It describes a first-generation system that represents a viable pack-
age that can be improved and will continue to be useful in making these extremely
important measurements related to the control of fine particles. Three techniques
were used to measure particle size distributions in industrial plant environments
under a variety of conditions: cascade impactors, in the particle size range 0.5-20
/im; diffusion batteries and CN counters, 0.01-0.2 /im; and optical counters, 0.3-2
jLtm. The stage collection efficiencies of several impactors and cyclones were also
measured in the laboratory. A vibrating-orifice aerosol generator was used for cali-
bration. With careful use, cascade impactors can give reliable particle size infor-
mation; however, no single impactor was versatile enough to measure accurately
under the extreme grain loading variations found between control device inlets and
outlets. Stage collection efficiencies of all impactor stages tested (except one)
agreed fairly well with the theory of Rantz and Wong. Diffusional sizing was practi-
cal with diluted samples; although it is tedious and time consuming: it is most
accurate for sources with relatively stable particle size distribution and concentration
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Croup
Air Pollution
Measurement
Flue Gases
Particle Size Dis-
tribution
Industrial Plants
Optical Measurement
Cyclone Separators
Air Pollution Control
Stationary Sources
Fine Particulate
Cascade Impactors
Diffusion Batteries
CN Counters
13B
21B
14B
13H
07A
S, DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
127
20. SECURITY CLASS (TM, page)
Unclassified
22. PRICE
EPA Form 2220-1
------- |