EPA-650/2-73-024
July 1972
ENVIRONMENTAL PROTECTION TECHNOLOGY SERIES
-------
EPA-650/2-73-024
MEASUREMENT AND CHARACTERIZATION
OF PARTICLES
IN WET SCRUBBING PROCESS
FOR SOX CONTROL
by
Donald A. Brooks
Walter C. McCrone Associates, Inc.
2820 South Michigan Avenue
Chicago, Illinois 60616
Contract No. EHSD 71-25
Program Element No. 1A2013
EPA Project Officer: D. Bruce Harris
Control Systems Laboratory
National Environmental Research Center
Research Triangle Park, N.C. 27711
Prepared for
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, B.C. 20460
July 1972
-------
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
page
I. Introduction 1
n. Phase I 2
1. Review of Duct Condition at the Scrubber Test Facility 2
2. Evaluation of the Andersen Stack Sampler 9
2.1 Prediction of Particle Collection Efficiency 9
2.2 Sensitivity Analysis of the Particle Collection
Efficiency of Size Selective Particle Sampler 16
2. 3. Experimental Evaluation of the Andersen
Stack Sampler 20
3. Design of a Parallel Cyclone Size Selective Sampler 26
3.1. Conceptual Design of the Sampler and Train 26
3.2. Operation of Sampler 31
3.3. Uses of Data from Size Selective Mass Sampler 32
3.4. Introduction to Cyclone Design 36
3.5. Theory of Cyclone Optimization 45
3.6. Experimental Evaluation of a Sampling Cyclone 53
4. Review of Laboratory Equipment 57
4.1. Wind Tunnel and Lab Prep 58
4.2. ITMC Evaluation 64
5. Field Test: Kansas Power and Light Company 65
5.1. Background 65
5.2. Experimental Plan 65
5.3. Description of KPL Tests 66
5.3.1. Time Priorities 66
5.3.2. Completed Tests 66
5.4. Series Cyclone Efficiency 70
5.5. Probe Deposition 70
6. Field Test: Shawnee Power Plant 71
7. Deposition of Particles in a Horizontal Sampling Tube 76
8. Prediction of Sample Bias Due to Non-Isokenetic Aspiration 82
9. Evaluation of Filters for Parallel Cyclone Samplers 86
HI. Summary of Phase H 89
References ... 91
ill
-------
FIGURES
No. page
1 Particle Size Distribution at the Wet Scrubber
Inlet 8
2 Particle Collection Efficiency of the Andersen
Stack Sampler , Flow Rate = 0.25 cfm 12
3 Particle Collection Efficiency of the Andersen
Stack Sampler, Flow Rate = 0. 5 cfm 13
4 Particle Collection Efficiency of the Andersen
Stack Sampler, Flow Rate = 0. 75 cfm 14
5 Particle Collection Efficiency of the Andersen
Stack Sampler, Flow Rate = 1. 0 cfm 15
6 Comparison Size Distribution Obtained from
Three Andersen Impactor Samples and a
Microscopically Sized Filter Sample at KPL 21
7 Particle Deposits on Stage 6 of the Andersen
Impactor 24
8 Deposition in and Around Jets on Stage 4 24
9 Layout of Size Selective Gravimetric Sampler 27
10 Components of Size Selective Sample Showing One
of the Seven Flow Streams 28
11 Top View of Cyclone Set Showing Tangential
Outline on Cyclone T-3A 30
12 Hypothetical Particle Size Distribution and
Collection Efficiency Curve for a Wet Scrubber 33
13 Schematic Drawing of a Cyclone Showing Relevant
Dimensions and Velocities 37
14 A Comparison of the Ideal and Actual
Collection Efficiency 39
15 The Relationship Between a, A, and b/r3 from
the Measurements of Muschelknautz and Brumer 44
16 Dependence of the Wall Friction Coefficient x on
the Re-Number and the Relative Wall Roughness
49
17 Pressure Loss Coefficients of Sharp- Edged
Outlet Pipes as a Function of the Relative
Inner Peripheral Velocities and the Re-Number
According to Measurements 51
iv
-------
Figures (Continued)
page
18 Overall View of the Wind-Tunnel Test Facility 58
19 Modified Auto Transmission Installed on the
Wind Tunnel 59
20 Climet Light-Scattering Particle Counter with
the New Photomultiplier Power Supply and a
Digital Volt Meter for Setting Precise
Discriminator Levels 63
21 The Pilot Scrubber 72
22 Size Distribution of TVA Samples Collected
with Small Cyclone 75
23 Schematic Arrangement of Equipment for
Particle Deposition Study 78
24 Arrangement of Equipment to Clean Particles
Deposited in the Sampling Tube 79
25 Deposition of Particles in Horizontal Sampling
Tube as a Function of Particle Size and
Sampling Conditions 81
26 Comparison of Predicted Nozzle Bias with the
Same Sampling Time and Nozzle Cross Section 83
27 Predicted Range of Nozzle Bias for 5-^m
Particles Sampled with 1/8 in. Nozzle at
Various Given Velocity Ratios at 90%
Confidence Level 84
28 Predicted Range of Nozzle Bias for 5-/Ltm
Particles Sampled with 1-in. Nozzle at Various
Given Velocity Ratios at 90% Confidence Level 85
29 Pressure Drop Vs. Flow Rate for Several Filters
in Typical Filter Holders with Teflon Backup
Filter to Reduce the Pressure Drop 88
-------
TABLES
No. page
1 Compositions of Flue Gas in Scrubber Inlets
and Outlets, Based on 0% Excess Limestone
Injection 3
2 Compositions of Flue Gas in Scrubber Inlets
and Outlets Based on 0% Excess Dolomite
Injection 4
3 Composition of Flue Gas in Scrubber Inlets
and Outlets Based on 150% Excess Limestone
Injection 5
4 Compositions of Flue Gas in Scrubber Inlets
and Outlets, Based on 150% Excess Dolomite
Injection 6
5 Ranges of Stream Gas Variables in the Wet
Scrubber Inlets and Outlets 7
6 Particle Size Distribution at the Wet Scrubber
Inlet 7
7 Particle Size Collected with 50% Efficiency by
Each Stage of the Andersen Stack Sampler at
Four Different Flow Rates 10
8 Changes of Air Velocity and Viscosity Due to
a Temperature Variation of 100 °F 19
9 Summary of Andersen Sampler Data from Kansas
Power and Light Company Field Test 22
10 ERC Andersen Stack Sampler Tests--2/um
Mono-Disperse Aerosol (Fluorometric Analysis
of Collected Dye) 25
11A Particulate Sampling Equipment and Cost Summary 29
11B Dimensions and Geometry of a Small Cyclone
Based on U.S. P.H.S. Design 43
12 The Computer Printout for the Optimization
Calculations for Cyclone T-2A 54
13 Cyclone Dimensions 55
14 Parameters for Cyclone T-1A with Q = 1 cfm 55
15 Amount of Water Required to Maintain a Given
Humidity in the Wind Tunnel 60
VI
-------
I. Introduction
This final report is submitted in fulfillment of our contractual agreement
under EHSD 71-25, modification 3. The estimated period of performance for the
Scope of Work under this contract was originally August 15, 1970 through August 15,
1971. However, mod 3 resolved some previous mutual misunderstandings by ex-
tending the contract period to July 7, 1972. This amendment,(3), also deleted the
request for submission of a final report and required that only a draft final report
be submitted. Although this report has been submitted as a draft final report, we
consider it of printable form and quality.
-------
II. Phase I
1. Review of Duct Condition^
Since the main purpose of this project is to develop a size selective par-
ticulate sampler for measuring and characterizing the particles in the wet scrub-
bing process for SOX control, an understanding of duct conditions is necessary
before the sampler is actually designed. The sampler should not only be capable
to fractionate the particles in the desired range, but also be able to tolerate the
physical conditions—such as temperature, pressure, etc., imposed by the gas
being sampled.
The wet scrubbing process for SOX removal consists of two steps—the
injection of pulverized limestone or dolomite into the furnace and the scrubbing
of the flue gas to remove SC>2 and particulate samples. The composition of flue
gas in the scrubber inlet, therefore, would depend on the quantity and type of al-
kali injected into the furnace and that in the outlet would depend on the particle
removal efficiency of the scrubber. Thus, our first step in this project is to ob-
tain information on the process variables of wet scrubbing and the efficiency of
the scrubbers which will be installed in the TVA test station.
1 2
With regard to process variables, Bechtel's Phase I and Phase II Reports '
provide a good information source. Our study of the reports, however, revealed
a lack of some data necessary for the definition of sampling interfacing.
Hoping to elicit this information, we developed individual sets of questions
for Bechtel, the TVA and scrubber manufacturers and.through the assistance of
our project officer, we received much of the information we requested from
Bechtel and the TVA. In regard to questions posed on the scrubbers, the data is
very limited since most of the scrubber manufacturers were unable to answer
our questions either due to their lack of information or unwillingness to disclose
proprietary information. Based on the information presented in Bechtel's Re-
port and what we collected, the flue gas compositions are listed in Tables 1 through
4. The ranges of several important stream gas variables which effect sample
designs were also computed and are listed in Table 5.
-2-
-------
Table 1
Compositions of flue gas in scrubber inlets and outlets, based
on 0% excess limestone injection. (Extracted from Process
Flow Diagram M-101, with totals corrected.)
Inlet
Component
S02
C02
N2
A
02
H20
CaO(s)
CaSO4(s)
Flyash
Inerts
TOTAL
mw
64.06
44.01
28.02
39.95
32.00
18.02
56.08
136.15
—
—
—
mph
7.6
411.8
2,314.7
28.4
142.4
258.9
7.6
2.0
**
**
3,173.4
Ib/hr
487
18,121
64,858
1,135
4,555
4,665
426
265
735
50
95,297
Outlet
mph
0.8
411.1
2,314.7
28.4
140.6
543.5
0.1
0.0
**
**
3,439.2
Ib/hr
49
18,091
64,858
1,135
4,500
9,791
4
3
7
1
98,439
** not included in totals
-3-
-------
Table 2
Compositions of flue gas in scrubber inlets and outlets based on
0% excess dolomite injection, (extracted from Process Flow
Diagram M-104, with totals corrected.)
Inlet
Component
S°2
C°2
N2
A
°2
H2°
CaO(s)
CaSO (s)
MgO(s)
MgS04(s)
flyash
inerts
TOTAL
mw
64.06
44.01
28.02
39.95
32.00
18.02
56.08
136. 15
40.32
120.38
—
—
mph
7.6
411.8
2,314.7
28.4
142.4
258.9
3.8
1.0
3.8
1.0
**
**
3,173.4
Ib/hr
487
18,121
64, 858
1,135
4,555
4,665
213
133
153
118
735
46
95,219
Outlet
mph
0.8
411.4
2,314.7
28.4
140.6
543.6
0.0
0.0
0.0
0.0
**
**
3,439.5
Ib/hr
49
18,106
64, 858
1,135
4,500
9,792
2
1
2
1
7
0
98,453
** not included in totals
-4-
-------
Composition of flue gas in scrubber inlets and outlets based on
Table 3 150% excess limestone injection (extracted from Process Flow
Diagram M-108, with totals corrected).
Inlet
Component
S°2
C02
N2
A
°2
H2°
CaO(s)***
CaSO4(s)
flyash
inerts
TOTALS
mw
64.06
44.01
28.02
39.95
32.00
18.02
56.08
136.15
—
—
—
mph
7.6
426.1
2,314.7
28.4
142.4
258.9
21.9
2.0
**
**
3,202.0
Ib/hr
487
18,752
64, 858
1,135
4,555
4,665
1,230
265
735
126
96,808
Outlet
mph
0.8
424.4
2,314.7
28.4
140.6
543.4
0.2
0.0
**
**
3,452.5
Ib/hr
49
18,678
64, 858
1,135
4,500
9,792
12
3
7
1
99,035
** not included in totals
*** including Ca(OH) as CaO
-5-
-------
Table 4
Compositions of flue gas in scrubber inlets and outlets, based on
150% excess dolomite injection (extracted from Process Flow
Diagram M-115, with totals corrected).
Inlet
Component
SO
2
C02
N2
A
°2
H2°
CaO(s)***
CaS04(s)
MgO(s)
MgS04(s)
flyash
inerts
TOT A 15
mw
64.06
44.01
28.02
39.95
32.00
18.02
56.08
136.15
40.32
120.38
—
—
—
mph
7.6
426.1
2,314.7
28.4
142.4
258.9
10.9
1.0
11.0
1.0
**
**
3,202.0
Ib/hr
487
18,752
64,858
1,135
4,555
4,665
614
133
443
118
735
116
96,611
Outlet
mph
0.8
425.3
2,314.7
28.4
140.6
544.6
0.1
0.0
0.1
0.0
**
**
3,454.6
Ib/hr
49
18,717
64,858
1,135
4,500
9,817
6
1
4
1
7
1
99,096
** not included in totals
*** including Ca(OH) as CaO
I*
-6-
-------
Table 5
Ranges of stream gas variables in the wet scrubber inlets
and outlets
Variables
Temperature (°F)
Pressure (psia)
Velocity (fpm)*
Particle concentration
(grains/ft3)
Air density (grains/cm3)
Air viscosity (poise)
Relative humidity (%)
Particle size distribution
* calculated from
Inlet
300
14.3
3,440
5-10
0.000813
0.00023
1.76
As shown in
Table 6 and
Figure 1
v- 9
Outlet
328 (preheated)
250 (reheated)
13.3
3,440
0.05-0.5
0.000812
0.00022
7.73 (reheated)
90.5 (preheated)
Unknown**
where Q = 30,000 cfm
D = 40 m. = LLQ/12 ft.
** Although the actual particle size distribution in the scrubber outlet is unknown,
it is reported by Bechtel that particles larger than 5 Mm will not get through
the scrubber.
Table 6
Particle size distribution at the wet scrubber inlet
Cumulative percent, by weight 12.5 25 37.5 50 62.5 75
Particle size, Mm 5.3 7.2 8.8 11 13.5 17
87.5 100
23.5 —
—7—
-------
I
00
100
90
80
70
60
50
40
30
20
10
0
0
•_^
•4-1
ct
"a
3
Median value = 11
Std. deviation = 1.93
Particle size
I
_L
I i t
2 3
FIGURE 1
10
20
30
40 50 60 70 8090 100
Particle size distribution at the wet scrubber inlet
(Reproduced from Figures 3-14 in Bechtel's Phase II Report)
-------
2. Evaluation of the Andersen Stack Sampler
Since impactor stages can be used to separate particulate matter from a
dust laden gas, we reviewed the feasibility of using cascade impactors as a size
selective particulate device to collect particles of various size ranges both at
the inlet and outlet of the wet scrubber. One of the most promising samplers
of this type available off-the-shelf is the Andersen Stack Sampler. Although cer-
tain information regarding the performance of this sampler has been provided
3 4
by its manufacturer and others ' , information on its particle collection efficiency
and particle fractionation capability for various stack conditions involved in the
wet scrubbing process for SO control is still lacking. This information is im-
X
portant because it predetermined the possibility of using such a device for this
particular sampling job. We feel, therefore, that a performance evaluation of
the Andersen Stack Sampler is desirable.
Our evaluation was performed in two parts; the theoretical prediction of
the cascade impactor performance and the field test at Kansas Power and Light
Company. Each of these are briefly described below.
2.1. Prediction of particle collection efficiency
The particle collection efficiency is a function of jet diameter of the cas-
cade impactor, gas viscosity, particle density and average jet velocity. Thus,
for a given duct condition, the size of the jet and flow rate determine the collection
efficiency of a single impaction stage (since the flow rate determines the jet vel-
ocity).
In predicting the particle collection efficiency of the Andersen Stack Sampler,
the hole sizes in each stage of the sampler were obtained from its manufacturer.
Based on the size and number of holes in each plate, the jet velocity of each stage
was computed for each of four different flow rates—0. 25, 0.5, 0. 75 and 1. 0 cfm.
Since these flow rates were those originally used by the manufacturer in calibrating
the sampler, they were considered to be the most desirable ranges for the sam-
pler in this project. By applying the equation describing the particle size collected
5 5
with 50% efficiency (X ) and the equation describing the collection efficiency , we
50
were able to obtain the X and collection efficiency curves at the four different
oO
-9-
-------
flow rates for each stage of the Andersen Stack Sampler. The results are tabulated
in Table 7 and shown in Figures 2-5.
. Particle size collected with 50% efficiency by each stage of
the Andersen Stack Sampler at four different flow rates
X of each stage,
ou
Flow rate, cfm
0.25 11.0 6.9 4.7 3.2 2.1 1.1 0.7 0.5
0.50 7.8 4.9 3.3 2.3 1.5 0.8 0.5 0.3
0.75 6.3 4.0 2.7 1.9 1.2 0.6 0.4 0.3
1.00 5.5 3.4 2.3 1.6 1.0 0.5 0.3 0.2
These predictions show that, at a given flow rate, X decreases progressively
50
as the jet diameter decreases. In contrast, for a given impaction stage, the X
uu
decreases progressively as the flow rate increases. According to the design cri-
o
teria of the wet scrubber for SO removal presented in the Bechtel report , about
A
50% by weight of the particles in the scrubber inlet are above 11 Mm, and all par-
ticles in the outlet are below 5 Mm. Thus, the data detailed in Table 7 suggest
that the Andersen Stack Sampler could be a good size-selective sampler at the wet
scrubber outlet for any flow rate from 0. 25 to 1.0 cfm. Within this range, however,
the instrument will collect more than 50% of the total particulate matter sampled
at the inlet on its first stage. Although it might still operate as a size-selective
sampler in this case, the first five stages of the sampler would be quite overloaded
in comparison with the others, and an equal eight-stage size-selective sampling
could hardly be expected. This trouble can be minimized by sampling at a lower
rate and by using several sampling cyclones of appropriate sizes in front of the
impactor to remove the course particles.
-10-
-------
On the basis of the theoretical prediction, it is concluded that the Andersen
Stack Sampler is suitable for size-selective particle sampling both at the inlet and
outlet of the wet-scrubber for SO control if a proper sampling flow rate is used.
A
A flow rate of 0.75 cfm, as originally recommended by the manufacturer of the
Andersen Stack Sampler, is reasonable to use.
-11-
-------
1.0
0.9
1
to
0.8
% 0.7
rf
O*
g 0.6
0*
0
| 0-5
Q
1 0.4
D
0
39 0.3
•
•
•
-
•
-
0.2 .
0.1 .
345 6 7 8 9 10 11 12 13 14
FIGURE 2 Particle collection efficiency of the Andersen Stack Sampler
Flow rate =0.25 cfm
-------
1.0
0.9 -
0. 8 _
0.1 .
2? 0.7
r-r-
cT
o 0.6
o*
o
, ? °'5
1-1 3
CO O
1 **>
o 0.4
n
a
o
f\ n
-
<
-<
m
i
ol o
Q'
i
I
4
•a
'
•
1 0
to
5
1
1
l
11
FIGURE 3 Particle collection efficiency of the Andersen Stack Sampler
Flow rate = 0. 5 cfm
-------
1.0
0.9
0.8,.
0.7
o
a 0.6
0.4
0.3
0.2
0.1
FIGURE 4 Particle collection efficiency of the Andersen Stack Sampler
Flow rate = 0. 75 cf m
-------
C71
I
FIGURE 5 Particle collection efficiency of the Andersen Stack Sampler
Flow rate =1.0 cfm
-------
2. 2. Sensitivity analysis of particle collection efficiency of size selective
particle samplers
The particle collection efficiency of a size selective sampler, whether it
is a cyclone, an impactor, or an elutriator, is dependent upon the mass flow rate,
gas velocity and density, particle density and so forth. For a given sampler, the
relation can be expressed as follows :
x50=f(Q, l.H.p) Eq. (1)
where
X_ = particle size collected with 50% efficiency;
50
Q = mass flow rate;
t, = gas density;
M = gas viscosity; and
p = particle density.
The gas density and viscosity is dependent upon gas temperature. Thus,
equation (1) can be expressed as follows :
T,/>) Eq. (2)
where
T = gas temperature.
This equation indicates that in order to maintain a constant particle fractioning
ability of a sampler, the flow rate, gas temperature and particle density must be
kept constant. However, in actual stack or duct conditions, these three variables
can vary to some extent. We therefore considered it necessary to examine sensi-
tivity of the impactors, cyclones and elutriators to small changes in these three
variables .
-16-
-------
The analysis was made independently to examine the effects of small variations
in the density of aerosol, temperature and flow rate to the variation of X or Z , and
o U o 0
the aerodynamic particle size collected with 50% collection efficiency. An example
is given here to illustrate the sensitivity of the process.
The X of the impactor can be expressed as follows
50
and simplified to:
where
D = the impactor jet diameter as slit width;
U = the gas jet velocity;
o
V- = the gas jet viscosity;
P = the particle density; and
B = a constant of proportionality
Eq. (3)
Eq. (4)
Using equation (4) we can express the changes in X corresponding to small
o u
changes in D, A*, p and U as follows:
o
8X
50
:50 . 3X50 . 3X50
AU
( » f = I BD }"•*
I I AD * I —77- I AM
L\
DpU
BDAt
V0.5
^/ -^» -•
O/ \ 0
Eq. (5)
-17-
-------
and dividing both sides by X , we have:
50
Eq. (6)
"50
To obtain the percent change in X we multiply this expression by 100:
ou
AU
Equation (4) can be rearranged to obtain Z .
50
This is as follows :
substitutes into equation (4).
We obtain
and
% change in ZgQ = 50 \^- - ^ - -^- Eq. (10)
I o J
Equations (7) and (10) are convenient for analyzing the effects of small
changes in variables.
First we will examine the effect of aerosol density variations. Assuming
all other variables are constant, we have:
Ap
% change in X = -50 -75- Eq (11)
oU
The density of flyash particles, however, may range from 1 gm/cc to over
6 gm/cc. Since this is not a small change, equation (11) cannot be used since Ap
must be small enough so that the second order terms are neglible.
-18-
-------
To determine the percent of change in X corresponding to the change in
*5U
p from 1 gm/cc to 6 gm/cc, we use equation (4).
% change in X = lOo Eq. (12)
X50(P=1)
= 100 (!-- = 60
Therefore, if p increases from 1 gm/cc to 6 gm/cc, X decreases 60%
o(J
in comparison with its value at p = 1 gm/cc.
Now, suppose we want to examine the effect of a 100 T temperature increase
relative to the collection efficiency of the sampler. We can use equation (10) by
holding the slit width and the mass flow rate constant. Hence, we have:
r AU 1
% change in ZgQ = 50 -^- - -^ Eq. (13)
I o J
The changes of air velocity and viscosity due to a 100°F temperature increase
are given in Table 8 below:
_, , Changes of air velocity and viscosity due to a temperature
1 able 8 variation of 100 T
Item
Temperature (T)
Viscosity, M, (lb /ft/sec)
Air velocity, U , (ft/sec)
o
Original
Condition
300
1. 16 x 10"5
100
Final
Condition
400
1.75 x 10"5
113
The percent change in Zcn is therefore,
ou
% change in Z5Q = 50 ^- =-2.15
-19-
-------
Thus, the change is about -0.02% per °F increase in temperature.
Based on the results using the technique illustrated above, it was determined,
for cyclones and impactors, that if the particle density is increased from 1 gm/cc
to 6 gm/cc with a constant flow rate and temperature, X decreases 60% in com-
50
parison with its value at 1 gm/cc. At constant mass flow rate, variation of tem-
perature within ±100°F would introduce ±2% variation in Z , the aerodynamic
ou
particle size collected with 50% efficiency. For elutriators, variation of tempera-
ture within ±20°F will introduce the same percent of variation in Z . For all
50
types of samples, precise control of the flow rate is more important than temperature
control. On the basis of our analysis, the following has been concluded:
(1) Sampling cyclones, impactors and elutriators are not suitable
for actual size analysis on aerosols containing a wide range of
particle densities.
(2) Precise aerodynamic particle size analysis is possible and is
mainly dependent upon precise mass flow rate and temperature
control, with precise mass flow rate control of particular concern.
These conclusions indicate that presice flow rate and temperature control
are necessary in particle sampling. Therefore, during sampling development,
these ideas were seriously considered.
2.3. Experimental evaluation of the Andersen Stack Sampler
Three tests were made with the Andersen sampler during our Kansas Power
and Light Company field test. The procedures used during these tests and the analysis
of the results are reported in another section. The three tests were performed using
sampling times of 10, 4 and 1 minute. If significant particle re-entrainment occurs
as the dust builds up on the collection stages, the three indicated size distributions would
differ considerably by being shifted toward smaller particle sizes with longer sampling
time. For comparison purposes, a microscopically determined size distribution was
also obtained from a high flow rate filter sample. The four distributions are shown in
Figure 6, but their comparison is complicated by another factor besides re-entrainment.
The calculated grain loadings from the three Andersen sampler tests indicate that dust
concentrations in the stack changed considerably during the three tests. The Andersen
-20-
-------
9.0
8.0
7.0
6..0
5.0
y
Hj
r^
O*
04.0
-•
g
o
ff
>-s
^3.0
3
2.0
1.0
Andersen Impactor
1 min. sample
Microscope filter sample
Andersen Impactor
4 min. sample
Andersen Impactor
10 min. sample
0.1 0.2 0.5 1
5 10 20 30 40 50 60 70 80
Percent cumulative mass less than size
90
95
98 99
99.8 99.9 99.99
FIGURE 6 Comparison size distribution obtained from three Andersen impactor samples
and a microscopically sized filter sample at KPL
-------
sampler tests were performed with the 10 minute run first, the 1 minute run last. On
the day of the testing, no soot-blowing operations were performed in the boiler since
these would interfere with the emission tests being performed by York Research. By
the late evening, while performing our Andersen sampler test, soot build-up in the
boiler became a problem and it is likely that this led to successively higher grain
loadings in the stack.
These conditions help explain the size distribution and grain loading data
shown in Table 9. A much greater percentage of large particles were present
during the 1 minute run.
. Summary of Andersen Sampler data from Kansas Power
and Light Company field test
Item
Sampling time (min)
Total sample weight (mg)
Flow rate through sampler
Run 1
10
26.15
1.28
Run 2
4
21.685
1.21
Run3
1
14.245
0.995
(ACFM for 760 °R and 22%
HO by volume)
^ 3
Grain loading (grain/ft 0.0464 0.13 22 0.3 25 8
at stack conditions, dry)
Despite these difficulties, it is possible from the data collected to infer that
particle re-entrainment did occur in the Andersen sampler. The difference in the
shape of the distribution below 1.9 /xm between the 4 min and 10 min runs indicates
that re-entrainment caused an increase in the weight of the samples collected on
the lower stages during the 10 min run. By comparing the impactor distributions
with the microscopically determined distributions, it is obvious that the impactor
data significantly overestimates the concentration of small particles. Therefore,
the only plausible explanation is particle re-entrainment.
-22-
-------
Four other difficulties were encountered in using the Andersen impactor:
(1) It was difficult to weigh the sample catch on plates having 25 g
tare weight;
(2) it was very difficult to clean the collected particulate matter from the
plates;
(3) visual inspection of the plates indicates that a significant amount
of particulate matter was collected in and around the jet holes
(see Figures 7 and 8); and
(4) from 4-8% of the total sample catch was impacted on the zero
stage by the entrance nozzle to the impactor.
6
The Environmental Research Corporation report on the evaluation of the
Andersen impactor was reviewed to determine if any of their results confirmed
those reported above. However, the main purpose of their evaluation was to
determine the reproducibility of the results obtained from the Andersen sampler
for several samples taken under identical conditions. The results with a poly-
dispersed aerosol having a mass medium diameter of 0. 64 /im indicate that the
Andersen impactor produces reproducible results. Tests with mono-dispersed
aerosols of 1-to-l mixtures of uranine and methylene blue may indicate that
particle re-entrainment occurs and that the calibration constants for the impactor
are not accurate. Unfortunately, aerosols of this type have densities which vary
from 15-20%, depending upon the concentration of solvent in the aerosol generator
and variations in water content of the bulk materials. Although the aerosol den-
7
sity was not reported by ERC, the work of others indicates that this aerosol has
a density varying from 1.2 to 1.43. This variation is large enough to account for
some of the particle re-entrainment results, but it does not account for others.
For example, results shown in Table 10 indicate that significant re-entrainment of
2-^im particles from stage 5 to stage 6 occurred at 0.5 cfm flow rate. This same
result could have been produced by a change in aerosol density of the magnitude
indicated above. However, this does not account for the significant amount of mass
collected by the filter after the impactor during the same runs. Clearly, for this
type of test to be conclusive, it must be performed using aerosols of constant den-
sity and their size must be monitored during the test.
-23-
-------
FIGURE 7
Photo shows particle deposits on stage 6 of the Andersen
Impactor. Deposition around jet holes is clearly visible.
FIGURE 8
Photo shows deposition in and around jets on stage 4.
-24-
-------
Table 10
ERG Andersen Stack Sampler Tests—2/nm Mono-disperse
Aerosol (fluorometric analysis of collected dye)
Percent
of Total
Collected Mass
Stage
Filter
8
7
6
5
4
3
2
1
0
(1.0
5-minute
sample
3.7
1.0
1.0
4.8
57.8
30.2
0.3
0.3
0.7
cfm)
15-minute
sample
14.7
4.6
1.2
1.0
47.0
29.4
0.3
0.3
1.4
Percent
of Total
Collected Mass
(0.5
10-minute
sample
0.7
4.9
92.0
0.9
0.1
1.2
cfm)
30-minute
sample
3.3
1.0
45.6
48.4
1.0
0.7
The failure of the Andersen impactor to perform adequately during the field
test led to the consideration of other types of samplers. We concluded that an
arrangement of small sampling cyclones might be feasible since cyclones are
capable of collecting a large amount of sample. Our goal during the rest of this
porject was then two-fold: first we had to eliminate the major sources of sampling
errors in the complete sampling — for example, deposition in the transport tube,
sample bias due to non-isokinetic sampling, arid poor filter performance; and,
secondly we had to design and evaluate a new type of cyclone sampler.
-25-
-------
3. Design of a Parallel Cyclone Size Selective Sampler
3.1. Conceptual design of the sampler and train
A conceptual design for a parallel cyclone sampling train is shown in Figures 9
and 10; it consists of three units: (1) the sample box, (2) a control unit which con-
tains pumps and gas metering equipment, and (3) a coolant supply system which sup-
plies coolant for the water vapor traps for all sampling trains being used. The
components used in the sampling train are listed in Table llA.
The coolant system uses a small refrigeration unit to cool a water/glycol
mixture which is being pumped through the water vapor traps and back to the re-
frigeration unit. This eliminates the inconvenience of using ice to cool the traps.
Particulate samples will be removed from the stack by a 3-ft x 1/2-in
diameter pitobe containing a thermocouple and s-type pitot tube for measuring
stack velocity and temperature. The nozzle of the probe will be designed so that
a minimum of dust deposition occurs in the bend. This will be achieved in a half-
inch probe by using a 4-in radius of curvature. Our dust deposition experiments
indicate that a minimum deposition will occur in a 1/2-in probe when used at the
flow rate planned.
When sampling at the scrubber inlet, a cyclone precollector is used to pre-
vent large particles from entering the small cyclones, since in the small cyclones
high gas velocities are obtained and could lead to the loss of the large particles
by particle interaction and/or bounce. The outlet of the precollector functions as
a gas manifold to divide the gas flow into seven branches. Since the gas flow
in the precollector outlet is a vortex, the best aerodynamic method of dividing the
flow is seven tangential outlets, as shown in Figure 13A. An inverted cone in the
middle of the outlet manifold is used to maintain the vortex motion. The seven outlet
flow rates are arranged so that the total flow into any quadrant of the manifold is
equivalent to that of any other quadrant. For sampling at the scrubber outlet, the
precollector body is replaced by an adapter which connects the probe to the outlet
manifold.
-26-
-------
Flow
Duct
Probe
Pitote
o-
Thermo-
couple
Coolant supply connectors
Flow control valves
IT
1.963
Cyclone pressure
drop indicator
Sample box
-Water trap
Umbilical connector
Bypass valve
Dry gas meter
Manometer
Control box
FIGURE 9
Layout of size selective gravimetric sampler
-------
Probe
to
00
ByF
>
(
V
I
CONTROL UNIT
Bypass valve
Pump
Dry
gas
meter
Umbilical connector
nometer
Thermocou
COOLANT SUPPLY UNIT
FIGURE 10 Components of size selective sample showing one of the
seven flow streams
-------
Table 11A Particulatc sampling equipment and cost summary
Sampling Train Components
Sampling probe (pitobe)
Probe nozzle
Seven stage cyclone with precollector
Gas flow valves
Gas manifold and tubing
Temperature controller and probes
Heater and blower
1/2 inch quick disconnect
Magnehilic pressure guage
Silicone o-ring
Insulated sample box
Condenser
Control Unit
Umbilical connector
By -pass valves
Control pump (no. 0322)
Gas pump (no. 0822)
Orifice meter
Dual column manometer
Dry gas meter
Electrical switches and accessories
Toggle valve
Pipe fittings
Box
Gas meter thermometer
Pyrometer
Coolant Supply Unit
Refrigerator, pump and heat exchanger
(can be used simultaneously with all
seven samplers)
Quantity
1
1
1
7
1
2
1
4
6
1
1
1
1
4
1
1
1
1
1
1
1
1
1
2
1
1
TOTAL
Cost
$ 125
25
3500
56
100
90
35
64
164
40
150
150
150
32
89
160
32
187
100
50
5
50
150
20
45
1000
$6569
-29-
-------
Outlet manifold of T-3A
Filter
Stage O
Stage
Stage 4
FIGURE 11 Top view of cyclone set showing tangential
outline on cyclone T-3A
- 30 -
-------
The seven parallel stages consist of six cyclones followed by filters and a
filter stage. The pressure drop across each cyclone is indicated by magnehelic
differential pressure gauges. Each cascaded cyclone-filter is made of three parts:
a collection pot where collected particles are trapped, the cyclone body, and a
cover plate-filter holder combination which contains the outlet pipe of the cyclone
and filter. All three parts are connected by flanges with o-ring seals. This sys-
tem functioned perfectly during the TVA field test. Flow control in the seven
stages is maintained by valves on the exhaust side of each stage. The cyclones are
maintained at 275 °F by a thermostatically controlled heater/blower in the sample
box.
The control unit contains the pumps with by-pass valves, dry gas meter,
pyrometer, manometers and an orifice meter. The design is similar to that used
in the EPA mass train. The pump size is based on the pressure drop requirements
of the filters. One pump supplies the vacuum for the six cyclones while a second smalle:
pump is used on the filter. Both pumps exhaust into the dry gas meter. Instan-
taneous flow rate measurements are made with the orifice meter/manometer. A
second manometer is used to measure the velocity pressure produced by the pitot.
3.2. Operation of the sampler
Two flow conditions must be satisfied while sampling: isokinetic conditions
must be maintained, and a constant cut point for the small cyclones must be main-
tained. The first condition can be maintained by varying the flow rate through the
filter stage of the sampler from 0 to 2 cfm. This will vary the total sampler flow
rate from 4 to 6 cubic feet per minute without affecting the flows through the six
cyclones. Isokinetic conditions are maintained by the same method used in the
EPA particle train. We have determined that the most economic and precise way
of maintaining constant cut points in the cyclone is to maintain constant pressure
drop across them. By adjusting the flow rate with the control valve, the operator
can maintain a constant pressure drop throughout the sampling period. It is clear
then that for isokinetic sampling two people will be required to perform sampling:
one to maintain isokinetic conditions and one to maintain constant cut points in the
size-selective sampler.
-31-
-------
After a thirty minute sampling period has been completed, the sampler will
be shut off and the parallel water vapor trap will be removed and transported to the
laboratory. There the filter will be removed, dried in a desiccator, weighed and
the amount of water collected in the water vapor trap measured. The filters can
be used as the substrate for x-ray fluorescence analysis of the chemical constituents
of the particulate matter collected to determine the size distribution of calcium
moieties. If the precollector is being used, the particles collected in its collection
pot will be weighed. To determine the total mass collected, the filter sample from
stage 0 and the precollector catch, if there is one, can be used. To determine
g
the size distribution, a mathemetical procedure can be used to convert the filter
weights to a size distribution.
3.3. Uses of data from the size selective mass sampler
As an example of how this data can be used, two hypothetical distributions are
shown in Figure 12. A differential and cumulative distribution for the scrubber inlet
and scrubber outlet are shown. The distribution for the inlet is an extrapolation
of data obtained from a Bahco analysis of TVA flyash limestone mixtures. If we
assume a scrubber mass collection efficiency of 98% and the outlet particle size
distribution as shown, it is possible to calculate the scrubber size efficiency
curve as also shown. The calculation is :
where
A = the fraction of particulate mass in the scrubber inlet of size X;
B = the fraction of particulate mass in the scrubber outlet of size X;
K = the fractional mass removal efficiency of the scrubber; and
E = the scrubber collection efficiency of particle size X.
-32-
-------
5. 5
5 5
5-4.5
,§.3.5
3
2.5
2
1.5
1
0.5
0
-Differential distribution at inlet
-Cumulative distribution at inlet
•Differential distribution
at outlet
Cumulative distribution at outlet
Scrubber size
efficicncv curve
Percent mass'
0.05-0.1-0.2-0.5-1
FIGURE 12
5 10 20 30 40 50 60 70 80 90 95 98 99
Hypothetical particle size distribution and collection effiency curve
for a wet scrubber
99.8 99.9 99.99
-------
There is actually another way of handling the particle size data which is
simpler and more accurate. All particle collectors—whether scrubbers, cyclones-
or lungs—select particles based on the particle settling velocity and not
its diameter. It is possible to calculate particle diameters from settling velocities
if the particle density p , gas density p , and gas viscosity n are known. For par-
o
tides of one micrometer or smaller, an additional term — the particle slip correction
factor—must be calculated which allows for particle slip. The effect of these com-
plications is readily apparent by comparing the collection efficiency equations for
cyclones when stated for settling velocity W* and particle diameter d :
i)U
V
w* =
Uf/r.g;
where
V = the radial velocity;
U. = the peripheral velocity; and
g = the acceleration of gravity.
where
5 = the mean free path in the stack gas.
Similar equations can be written for the collection of a wet scrubber. Since the gas
stream conditions and particle density will be different for the inlet and outlet streams,
it will be necessary to measure these parameters while sampling if particle size
data is used for scrubber efficiency determinations. It would be difficult and expen-
sive to obtain experimental values for the slip correction factor although estimates
can be calculated. Since cyclones are settling velocity selectors, the particle size
data obtained with them would contain fewer uncertainties if it were used in the
form of particle settling velocities instead of particle diameters.
-34-
-------
The isokinetic experiments to improve the accuracy of our theoretical model
will be performed during Phase A of our work plan. We will use the light scattering
particle counter to sample, size and count particles in wind tunnel tests. Such a
test can be performed in approximately 160 man hours.
Manpower requirements
According to Bechtel's data, a typical test with the scrubbers will require
8 hours. Sampling may begin as soon as steady state conditions are reached.
Preparations for sampling include cleaning and assembling equipment and loading
preweighed filters into the sampler. This will require about one man hour per
sample and the task can be performed by either sampling personnel or a lab tech-
nician. Transporting the sampling unit to the sampling point and allowing the cyclones
to reach operating temperature requires about 15-30 minutes. After a sample has
been collected, the reverse of the above steps is required with about the same time
requirements. The only difference is that cleaning the equipment is unnecessary.
A total of about 3 man hours per sample will be required before and after sampling
and the lab analyses will require one man hour per sampling.
Summary of Manpower Requirements
Operation No. of Men Total Man Hours
Sampling
1. Clean and assemble equipment 2 1
2. Adjust equipment 2 1/4
3. Sampling 2 1
4. Filter removal and transport to lab 1 1/2
2-3/4
Laboratory Analysis
1. Preweigh filters 1 1/4
2. Dry and reweigh filters 1 1/2
3. Calculate and report data 1 1/4
-35-
-------
3.4. Introduction to cyclone design
The major advantage of using a cyclone as a size selective sampler is its
ability to collect a large sample without re-entrainment. This property is inherent
in the way of cyclone functions. Gas enters the cyclone through the side by tangential
inlet pipe and leaves the cyclone through an axial exit pipe. This arrangement pro-
duces a vortex in the body of the cyclone which exerts considerable centrifugal force
on particles suspended in the air stream. For a particle to be collected by the
cyclone, the centrifugal force acting on it must be greater than the opposing viscos
forces of the gas. When this occurs, the particles accumulate at the walls where
a downward secondary flow of the gas causes them to be collected at the bottom of
the cyclone in a collection pot and all that is required to collect a given mass of sample
is a large enough collection pot.
The prediction of cyclone performance is the greatest difficulty in the use of
cyclones in a size selective sampler. The major difficulty is the inability to cal-
culate flow conditions in a cyclone. A few methods of making these calculations after
considerable simplification of the flow pattern have been developed. As a result,
3
no sound theoretical methods are available for determination of the effects of change
in cyclone geometry on particle efficiency.
In one simplified theoretical approach, the centrifugal force acting on a
particle in a simple vortex is calculated and set equal to the momentum of a particle
falling in still air with a terminal velocity of w. Under these circumstances where
the two forces acting on the particle are exactly equal, we would expect that a particle
would have exactly equal probability of escaping or being collected by the cyclone.
In other words, particles with this settling velocity are collected with 50% efficiency.
The equation thus derived in equation 17, where v is the radial velocity, u. is the
peripheral velocity at i, r is the radius of the exit pipe, and g is the gravitational
£t
constant (see Figure 13). By substituting Stoke's law for the settling velocity of the
particle, equation 7 can be derived for particle diameter d (see equation 18), where
D U
M is the gas viscosity, p is the particle density, p is the gas density and the last
r O
term in parenthesis in the denominator is a simplified form of the slip correction
factor. As might be expected, the greatest difficulty in using these equations is
the uncertainty in knowing the two velocities in the cyclone. Using the empherical
-36-
-------
FIGURE 13 Schematic drawing of a cyclone showing relevant dimensions and velocities
-37-
-------
data to be described later, it may be possible to use these equations with reasonable
accuracy.
In an ideal cyclone all particles smaller than d would escape the cyclone while
ou
all those greater would be collected. This would lead to a vertical collection efficiency
curve. Actual cyclones on the other hand have S-shaped collection efficiency curves
having a straight central portion whose slope varies with cyclone geometry in an un-
known way (see Figure 14).
The first task in a systematic design of a size selective sampler was to de-
termine its operating constraints. The calculated gas viscosity, density, temperature,
humidity and particle density determined in the process variable study were used for
all calculations in the design of the cyclones.
The major factors affecting the number of stages used in the size selective
sampler are the range of the size distribution to be sampled and the line width or
size resolution of each stage. The line width of the cyclone is the experimental
uncertainty in knowing exactly the collection efficiency curve for the cyclone.
Cyclone Design Equations
V
W* = r Eq. (17)
18UV r
r-2 Eq. (18)
. (19)
a* 2 2
d50
K= 0.4043+ 0.28G3Q Eq. (21)
-38-
-------
N C
w .2
<« "o
0 g
•g -
0) 0)
<+H 03
-w M
a «
0 o
o o
100
75
50
25
0
/I
r! I
D,25% -1 I1- D,75%
D,50%
Particle Diameter, D (micron)
FIGURE 14 A comparison of the ideal and actual
collection efficiency
-39-
-------
At best the cut point for each stage can be one line width apart over the
entire range. A more reasonable spacing, however, would be three line widths.
In the case of a cyclone, two factors affect its resolution: the error in controlling
the collection parameters and the error in the calibration method. The error in
controlling the collection parameters of the cyclone can be estimated using the method
g
of sensitivity analysis described elsewhere.
If the performance of a system can be predicted mathematically, it is possible
using this technique to determine the variability in the operation of the system using
equation 19, where a is the standard deviation of the function of F(x,y) and standard
r
deviations a and a are encountered in the parameters concerning the operation of
the system. By applying this equation to the mathematical model for the prediction of
d , it is possible to estimate the standard deviation of the cut-off point. By as-
o U
suming that cyclone dimensions do not vary, and ignoring the variations in particle
density and slip correction factor, it is possible to derive equation 20 for the standard
deviation in d ... where the terms on the right are the fractional errors in knowing
o U
the absolute gas viscosity and flow rate in the cyclone. By assuming the flow rate
can be controlled to within 2-3% and the viscosity is known within 1-10% of the actual
values, the variation in d will be 2.25-10.5%.
oO
For d equal to 1 Mm, the variation in collection parameters produce a
o u
standard deviation of 0. 02-0.1 Mm, which clearly indicates that scanning electron
microscopy will be required for calibration of the collectors. By using the epidiaescope
described elsewhere in conjunction with the SEM, it is possible to size particles with
an accuracy equal to the resolution limit of the SEM, or ~0.025 Mm. However, a
large number of particles will have to be counted to obtain this accuracy and a more
reasonable calibration error would be 0.1 Mm. The sum of these errors, then,
reduces the size of the resolution of the cyclone with d equal to 1 Mm to 0.1-0.14 Mm.
OU
A larger d , the size resolution will be limited by the errors in flow and viscosity
o u
measurements. For d equal to 4.5 Mm, the size resolution will be about 0.14-0.49 Mm.
5 U
A more rigorous development of the sensitivity of the cyclone cut-off point to vari-
ations in the operating parameters indicates that the variations in wall friction may
be as significant as viscosity and flow rate changes. Wall friction variations are
due to the deposition of particles on the cyclone walls causing increase drag on the
-40-
-------
rotating gas. The pressure drop across the cyclone will indicate the variations in
these parameters, therefore, the most logical way of maintaining constant d is to
oO
maintain a constant pressure drop across the cyclone.
By spacing a number of cut points along the particle size distribution obtained
during the Kansas Power and Light field test, we determined that a size selective
sampler having seven stages and the d 's as indicated in the optimization section,
ol)
would be optimum. Each stage is placed three line widths apart and the calculated
percent mass and surface area of the particles collected by each stage is indicated
in the table.
Again using the size distribution obtained from the KPL test and assuming a
particle concentration of 0.1 grain/cfm, it was possible to calculate the necessary
flow rate for each stage to collect a minimum of 10 mg during a 30-minute sampling
period. The more efficient stages of the sampler must have a flow rate of at least
0.75 cfm to collect the necessary sample.
Since there will be a large population of particles greater than 5 Mm in
diameter at the scrubber inlet, it will be necessary to use a precollector to remove
these particles. The selection of the cut point for the precollector must be made with
care since a compromise must be achieved between two conflicting requirements.
The large particles must be prevented from enetering the small cyclones for large
particles. On the other hand, d_0 for the precollector should not be so small that
50
there is considerable overlap between efficiency curves of the precollector and those
in some of the stages in the sampler.
Using published collection efficiency curves , it was possible to arrive at
a reasonable compromise between the two contradictory requirements for the
precollector. By setting the D for the precollector equal to D for stage 1, the
50 90
overlap between the collection efficiency curve for stage 1 and the precollector will
be small and the quantity of particles greater than 10 Mm entering the small cyclones
will be minimized. From the published data we estimated that D would be 1.5 times
*J U
greater than D for stage 1 which leads to a D for the precollector of 6. 75 Mm.
O U 1} U
D for the precollector will be slightly larger than 10 Mm.
-41-
-------
This produced some overlap with stage 1, but the small error caused can
be corrected by using the particle size information obtained by weighing the pre-
collcctor catch.
Now that the design criteria for the sampler have been developed, it is
necessary to determine the geometry and size of cyclone which meet these requirements.
In order to develop as accurate a method as possible for a design of cyclones, a
considerable literature search was performed. Our early attempts at cyclone design
were based on the prediction equation developed by Rosin and Rammler. This includes
the number of turns the gas makes in the cyclone to estimate the cut-off point of the
cyclone . A cyclone was designed and constructed using the geometric configuration
12
given by the U. S. Public Health Service (see Table 11B). Although no direct method
exists for measuring the number of turns of the gas in a cyclone, an attempt was
made to estimate it based on the pressure drop across the cyclone.
The results of the pressure drop tests on the small cyclone proved un-
satisfactory as no sound relationship between the pressure drop in the cyclone and
the number of turns could be developed. As a result, equations using the number of
gas turns in the cyclone for predicting performance were abandoned. We also de-
13
termined that the prediction model for pressure drop developed by First could not be
used for small cyclones since it was only possible to fit First's equation to our
data for pressure drop by allowing a constant of proportionality, K, in the equation
to vary with flow rate.
By fitting a least squares polynomial to the pressure drop data, we found that
K varied with Q according to equation 21.
, r i ^ u 13,14,15,16 , .
After reviewing the work of several German researchers , a design
technique was developed which allowed the adjustment of all cyclone parameters and
the optimization of cyclones for the collection of specific size particles.
-42-
-------
Dimensions and geometry of a small cyclone based on U.S.P.H.S.
design (for definition of variables, see Figure 15)
r =0.142"
h = 6r =0.850"
o 1
r =4r = 0.568"
h = 16r = 2.272"
h = 16r = 2. 272"
i* JL
r = 2r =0.284"
ft J.
r, = r, = 0.142"
4 1
The optimization process consists of adjusting the various dimensions of a
cyclone so that the energy losses in a cyclone are minimized while maintaining a
constant cut-point. There are three advantages to be gained by using such an op-
timization process: first of all, no more work than necassary is expended by the
cyclone air stream to remove particles of a given size. A minor result of this
is smaller pump size requirements. More importantly, greater energy losses
could yield increased turbulance in the cyclone and large deviations from predicted
performance. We would also expect that a cyclone with minimum turbulence would
have a very sharp collection efficiency curve. Lastly, the optimization design
method combines all the dimensions of the cyclone together. For example, adjust-
ments in one dimension allow more convenient construction, as appropriate adjust-
ment of other dimensions can be made to maintain a constant cut-point.
This theoretical approach is similar to an empherical method developed
17
by Stairmand for determining the optimum geometry for a cyclone. Stairmand
injected a dye into an operating, transparent cyclone and observed its vortex.
By adjusting various cyclone dimensions he was able to obtain a stable vortex.
The instabilities in the vortex indicate an increase in the energy loss for a given
collection efficiency in the cyclone.
- 43 -
-------
0.2 0.4 0.6 0.8
FIGURE 15 The relationship between a , A, and b/r from
the measurements of Muschelknautz and Brumer.
-44-
-------
3.5. Theory of cyclone optimization
The following discussion on the theory of cyclone optimization will be limited
to a rudimentary discussion of the major features. A critical review of the flow
in a cyclone leading to the determination of the collection efficiency has been given
by First. Extensive measurements of flows in cyclones made with a directional
pitot are also reported by First.
To use the optimization process, one needs to know the basic relationships
concerning flow and pressure drop in cyclones. Experimental investigations have
shown that flow into the separating chamber through a tangentially attached pipe
causes the jet to be constricted and therefore results in an increase in gas velocity
practically without any energy loss. Mathematically, the constriction is described
by
V
Ua=^f Eq. (22)
Earth derived a correction factor, a, from an analysis of impulse and friction
moments which relates the geometry of the inlet to the constriction,
V • r
*=-^7~ Eq. (23a)
a' 3
or
r3
a1 =a— Eq. (23b)
e
Measurements by Muschelknautz and Brumer have shown that the ratio of inlet to
outlet cross sectional areas, A, effects the value of a.
A = — Eq. (24)
a2
The values of a used in our designs are taken from the data in reference 10 for slot-
shaped inlets (see Figure 15). Muschelknautz suggested that a for a circular inlet
is determined using the equivalent slot width given by the length of the side of a
square having the same area as the circle. However, by replacing the height and
- 45 -
-------
width dimensions of a slot by the diameter of a circle, one degree of freedom is removed
and a can no longer be determined independent of A. Thus, a trial and error method
must be used.
To use equation 17 to determine the critical settling velocity W*, it is necessary
to calculate the radial velocity, v , and the peripheral velocity at i, u.. The mean
value of v is easily calculated from
V = 0 Q Eq. (25)
r 27rr2h2
Measurements have shown that v departs considerably from its mean value
near the bottom of the outlet pipe. This is one likely cause of the "S" shape of the
collection efficiency curve.
The value of u. can be estimated from an analysis of the moments of flow in
the cyclone. The moment of momentum, M., required to produce a given u. is the
momentum supplied by the inlet gas, M , less the momentum losses, M , due to
c r
wall friction. By modeling a cyclone so that the friction surface is unrolled and
approached by the gas as if it were a flat plate, the wall friction moment is obtained
as
M =X-u.u • Trhr.r (y/g). Eq. (26)
r i a i a
This serves to define the coefficient of friction, X . From this analysis, an equation
for u. is obtained
i
77"
Eq. (27>
v. a, a + hr TT\
i e
Introducing the dimensionless parameters
u.
U. =—- Eq. (28a)
i
ro
R = -=• Eq. (28b)
A=— Eq. (28c)
- 46 -
-------
-T Eq. (28d)
H. =— Eq. (28e)
3
and substituting equation 23b into equation 27, a dimensionless equation is obtained
U = — - - Eq. (29)
This clearly illustrates the effect of wall friction and the inlet correction factor on
cyclone velocities. Until measurements by Muschelknutz proved otherwise, X. was
assumed to be a constant. Frictional losses are dependent on the dust loading of
the gas and the roughness and surface area of the walls. For small cyclones the
frictional losses caused by the suspended dust is an order of magnitude smaller
than wall friction based on experimentally verified calculations for typical dust loadings.
Muschelknautz has attempted to use a Nikurodse diagram to represent the
relationship between wall friction coefficients and the Reynolds number, Re, of the
flow in a cyclone. The Reynolds number at the cyclone walls, Re , is calculated
R
by ratioing the mass inertia and viscosity forces causing the wall friction of the flat
plate representing the cyclone walls. This leads to the equation
2r v /Y ,
Be = - e a - - Eq. (30)
'R 2h(rR[lMva/Um)2j
which can be transposed using the dimensionless cyclone parameters and the
following relations
Eq. (31a)
u =\]u u Eq. (31b)
m Y a i
v = v./R(R-l) Eq. (31c)
3. I
- 47 -
-------
to „
Re.
1 Eq. (31d)
2H(R-1)[ 1 + o'A/UR (R-l)2)]
where Re. is the Reynolds number of the outlet pipe calculated.
Based on measurements of the velocity of compressed air, oil, and water in
small cyclones, the relationship between Re and X is shown in Figure 16. Ac-
K
cording to these measurements, wall roughness has no affect on X for Re < 100
R
and the following equation can be fitted to the data,
x =Q(1.1787 -0.627 in ReR) ^ (JJ2)
During field trials with a small cyclone at TVA, we found that the pressure drop
across the cyclone dropped steadily although the flow rate remained constant. A
build-up of particle deposits on the walls was also observed, indicating that wall
roughness may have an effect on X even for this range of Re0.
K
The total pressure loss for a cyclone consists of the losses in the separating
chamber, Ap , and the losses in the cylinder volume below the outlet tube, Ap.:
6 1
P2 2 P2 2
Ape = ^a + T V " ^i+ TUi * Eq' (33a)
Ap. = (p. + ~uf) - (p + -^vf) Eq. (33b)
1 L Lt \ 111 £t I
Ap = Ap + Ap. Eq. (33c)
6 1
The pressure loss in the separation chamber was calculated by Earth as the difference
of the speed in the intake and discharge of the assumed friction surface:
Eq. (34)
- 48 -
-------
10
-1
a
o
o
o
4->
a
£ 10
a>
o
10
-3
A
n
4
III
= 6x1)
2.6 x 10~5 to (
,-3
x H
-4
Re0, Reynolds Number
ri
FIGURE 16 Dependence of the wall friction coefficient \
on the Reynolds number and the relative wall
roughness ks/ro. The relative wall roughness
is the ratio of particle diameter deposited on
the wall to the cyclone radius.
Range I: Measurements with silicone oil M20
Range II: Measurements with water
Range HI: Measurements with compressed air of 2 to 13 atm.
- 49 -
-------
By defining a dimensional pressure drop coefficient as
= 0 Eq" (35)
Y-vV2g
and substituting equations 28, 29 and 34, we get
The much larger loss of the outlet pipe pressure, Ap., and the corresponding loss
coefficient t,. are determined from the measurements of Muschelknutz. The de-
pendence of the loss coefficient on Re. and U is shown in Figure 17. To allow
computer assistance in the calculative procedure, a polynomial was fit to the
curve for Re. = 2000.
Api 2
t,. =— = 2.571 + 2.542u + 1. 724u Eq. (37)
1 vf/2g
which fits the data to within ±6%. The values of Re. for the cyclones designed for
the sampler were all close enough to 2000 to allow the use of equation 37.
Specifying the three parameters H, R, and U completely defined a cyclone
since all the other variables can be derived from the above relationship if Re. or,
more properly, the flow rate, is given. Several types of cyclones having different
values of the three free parameters, different size and pressure loss can be de-
signed which will solve a particular dust removal requirement. The choice of cyclone
type then must be based on size and pressure loss. To determine which cyclone
from a group having equal volume has the lowest pressure loss, two parameters are
introduced which serve as evaluation criteria for designing an optimum cyclone:
Eq. (38)
*v*
U*V
B* =- Eq. (39)
-50-
-------
100
E
v\
to
Cjiculated zee. to
71 W.BJilh/7/ p
FIGURE 17 Pressure loss coefficients of sharp-edged
outlet pipes as a function of the relative
inner peripheral velocities and the Re num-
ber according to measurements.
-51-
-------
The representative velocity v* is calculated from
v*=—&j- Eq. (40)
where
r* = (r h) Eq. (41)
This velocity is related to the frontal face area of a cyclone of diameter r* and
is proportional to the actual velocities in the cyclone. The parameter £* indicates
the effect of an increased pressure loss due to the result of an increased removal
efficiency, as reflected by larger v*'s.
Comparing two cyclones of the same radius, (e.g., equal face area and equal
gas throughput and therefore equal v*; also identical separation performances as
expressed by the settling velocity, W*( of the particles collected with 50% efficiency)
the respective separation index, B*, will also bo identical, therefore, the unit
having a lower pressure loss or £* has to be considered as better. By proper substitution
of the above equations, the performance can be obtained in the form:
Eq. (42)
B*= r-TT Eq. (43)
4H.HR U
i
The optimization process consists of first estimating the wall friction coefficient,
X, and then selecting values of B*, H, R, H. to solve equation 43 for U. Once U has
been determined, t, and A T can be determined. By successive iterations based on
values of H and R and H., values of £* can be scanned to find the minimum.
After the dimensionless cyclone has been designed, it is necessary to calculate
the required cyclone radius to collect particles of a given settling velocity for a given
flow rate, Q. The following relationship between these variables has been derived:
1/3
r (QW*R/277gB*) Eq. (44)
O
-52-
-------
All cyclone dimensions can now be calculated. The value of Re calculated for the
cyclone is then used to estimate a new value of X to evaluate the original estimate.
If X is not accurate, the optimization process is repeated using the new value of X .
Values of B* are chosen arbitraily. From inspection of equation 23 it is
clear that small values of B* correspond to small values of W*. In other words,
to collect small particles, a cyclone must be designed which has a small B*. Com-
plete freedom of choice does not exist however if R is too small or x too large for
a given B*, as the cyclone inlet area approaches zero. To obtain high spinning
velocity in such a cyclone the design procedure requires the inlet velocity to approach
infinity.
An example of the optimization procedure is given below by the computer output
(Table 12) for the design of cyclone T-2A. In this case the minimum value of £* occurs
for R = 3 and H = 23. However, by accepting slightly less (6%) than optimum perfor-
mance, a shorter and therefore more easily constructed cyclone is obtained by using
R = 3.5 and H = 13.
Table 13 gives the values of important parameters for the three cyclone types
designed by this method.
3.6. Experimental evaluation of a sampling cyclone
The TVA field trial was the first experimental test of the optimization process
as applied to small cyclones. A high efficiency cyclone designated T-1A was designed
and constructed for the test. The important dimensions for this cyclone are listed
in Table 14. Since X is dependent on the Reynolds Number in the cyclone, both X
and U vary with flow. Cyclone T-1A was designed during the early development
stages of the optimization procedure and as a result of some errors the cyclone is
not optimum. The value of a' was assumed to be 1 based on past experience with
cyclones having R = 3; however, this is not accurate for R = 8. Actually a' is less than
1 and thus the indicated inlet radius is far too small.
-53-
-------
Table 12. The computer printout for the optimization calculations for cyclone
T-2A. The numbers inputed after the question are the values of R,
B*, \ , and (I-I-H.). The five columns of output are the values of H
scanned and the corresponding values of the parameter £*, the di-
mensionless pressure drop C> inlet ration A a' and peripheral velo-
city U.
What are R, B, Lambda, and H-H. (must be
i
H XI*xl04 Zeta
5 1.9 ******
21 1.7 25.327
23 1.5 19.858
25 1.4 16.249
27 1.3 13.729
29 1.2 11.893
31 1.2 10.503
33 1.1 9.435
35 1.1 8.584
37 1.1 7.897
39 1.1 7.333
What are E, B, Lambda, and H-H. (must be
H XI*xl04 Zeta
5 3.3 ******
11 1.9 40.734
13 1.4 23.665
15 1.1 16.485
17 1.0 12.679
19 0.98 10.375
21 0.96 8.854
23 0.95 7.787
What are R, B, Lambda, and H-H. (must be
H XI* x 104 Zeta
5 8.0 ******
7 4.5 ******
9 1.7 32.064
11 1.2 17.941
13 1.1 12.563
15 1.0 9.814
17 1.0 8.177
What are R, B, Lambda, and H-H. (must be
H XI*xl04 Zeta
5 15.0 ******
7 3.2 59.925
9 1.6 21.321
11 1.3 13.152
13 1.2 9.741
15 1.2 7.909
< 5) ?
FA1
-.195
.367
.428
.489
.550
.611
.671
.732
.792
.853
.913
< 5) ?
FA'
-.168
.325
.465
.602
.738
.873
1.007
1.141
< 5) ?
FA'
-.117
.192
.447
.687
.921
1.152
1.381
< 5) ?
FA'
-.104
.318
.671
1.004
1.329
1.650
2.5, 1E-4, .038, 4
U
8.944
1.059
.957
.873
.803
.743
.691
.647
.607
.572
.541
3, 1E-4, .038, 4
U
7.454
1.899
1.541
1.297
1.121
.987
.882
.797
3.5, 1E-4, .038, 4
U
6.389
3.117
2.130
1.628
1.321
1.118
.961
4, 1E-4, .041, 4
U
5.590
2.728
1.863
1.425
1.156
.973
-54-
-------
Table 13
Cyclone Dimensions
Cyclone Type
Parameter
B*
«*
R
H
A ,
a
u
C
hl
h2
h3
h4
rl
r2
r3
r4
b
e
Table 14
T-1B
2.5 x 10~5
6.2xl04 1
7
17
2.167
0.961
12.56
0.306
0.996 .
0.230
0.080
0.077
0.536
0.092
24°
Parameters for cyclone
H = 15
R = 8
Hi = 11
U = 1. 133
B = 2.5 x 10~5
X = 0.040
A = 2. 24
T-2A
1C'4
. 1 x 104
3.5
13
0.921
1.321
15.45
0.826
1.859
0.620
0.198
0.207
0.723
0.250
1415'
T-1A with
r3 =
r2 =
rl =
4 =
hl =
e =
T-3A
7.5 x 10~4
1.3 x 103
2
7
0.822
1.992
2.708
2.031
2.031
1.720
0.677
1.500
0.677
0.750
22°
Q = 1 cf m
0.559"
0.070"
0.177"
0.209"
0.768"
0.279"
31°
-55-
-------
The results and procedures for the TVA trials are fully reported elsewhere.
The theoretically predicted cut-off point was 0.43 Mm and 1 cfm with a pressure drop
of 33. G" of HO. The actual values achieved were 0. 74 Mm and 26" of HO. The
M ^
possible causes for this descrepancy are two-fold. According to theory and the
measurements of Muschelknautz, the cut-off point of the cyclone varies with (1/r ) ' .
tt
These reported measurements were made for cyclones having values of R = 2 to 4.
In data obtained by First for cyclones with R = 1.5 to 3, the total mass efficiency
0.2
for a cyclone varies with (1/r ) . However, these two results are not directly
£1
comparable because of differences in the way cyclone performance was measured.
It is possible however that beyond some point an increase in R (smaller r ) will
Lt
not actually produce a proportionate increase in efficiency as predicted by theory.
Measurements of cyclone T-2B with R = 3 will reinforce this point.
Another difficulty is the undersized inlet pipe. Since a' is smaller than assumed,
A and therefore r, must be larger to obtain the optimum value of A ,. The smaller
1 a
r causes the inlet velocity to be much higher than required. Ideally, the inlet
velocity should equal the peripheral velocity, u., of the vortex which is set primarily
by r and Q. The flow in the cyclone will be undisturbed if these conditions are set.
o
It may be possible that the increased inlet velocity produces enough turbulence to reduce
the efficiency of the cyclone. The lower than predicted pressure drop in the cyclone
may indicate that this is not the case since increased turbulence should increase the
pressure drop.
A theoretical evaluation of the effect of increasing the wall friction, X , indicates
that this factor alone cannot account for the reduced pressure drop. However, as
pointed out earlier, X was found to be a function of wall roughness in these small
cyclones. The measurements of Muschelknautz show that X is only a function of
Re_. when Re > 100. Even at a flow rate of 1 cfm, Re = 28.5 for cyclone T-1A.
K 1\ i\
A measure of the steepness of the collection efficiency curve for a device
is given by the geometric standard deviation of the collection efficiency S where
Particle diameter at 50% efficiency
Particle diameter at 84.13% efficiency
We arc pleased to reveal that the optimization procedure for the cyclone design produced
a very steep efficiency curve having a geometric standard deviation of 0.94. This
-5G-
-------
17 18 19
surpasses the performance reported for several cyclones ' ' as well as for
20
inertial impactors
4. Review of Laboratory Equipment
4.1. Wind tunnel and laboratory preparation
Several modifications were made to the wind tunnel after it was moved from
Gainesville, Florida and reassembled in our facility at 2600 South Michigan Avenue,
Chicago, Illinois, to allow better performance (see Figure 18). The sections of wind
tunnel are supported by small screw jacks and sealed together using silicon rubber
o-rings. Two difficulties in tunnel operation were experienced in Florida:
1) to obtain low velocities in the wind tunnel, the opposed vane damper
had to be shut completely which caused high particle deposition on
the damper, and
2) short-term variations in air velocity in the tunnel interfered with
turbulence measurements.
An old auto transmission was therefore modified and connected to the fan drive to
allow a reduction of air velocities without the use of the damper (see Figure 19).
In addition, to determine the source of the velocity variations., a tachometer was
installed on the prop to measure prop speed. The two highest gears in the car's
transmission produced velocities of 4372 and 3012 feet per minute at room temperature.
Using the tachometer, we determined the prop speed in high gear varied from 1880
to 1876 rpm in a period of from 5-8 seconds. This corresponded to a velocity
variation of 53. 9 to 54. 6 miles per hour at room temperature and accounted for
approximately half of the velocity variations encountered in Florida. The other
half was due to temperature variations caused by the large dead band in the temperature
controller used on the heat exchanger. A new solid-state temperature controller
was installed which reduced the temperature dead band from 10T to 1°F. The old
controller has been installed as a redundant safety device and senses the temperature
inside the heat exchanger by turning off the gas supply and setting off an alarm when-
ever the heat exchanger rises above its safety limit, 550T. This prevents burn
out of heat exchanger even if the gas supply is accidentally left on after the wind tunnel
has been turned off. A second solid-state tempcratui-e controller has been installed
-57-
-------
FIGURE 18 Overall view of the wind-tunnel test facility
- 58 -
-------
FIGURE 19 Modified auto transmission installed
on the wind tunnel
- 59 -
-------
as a dew point sensor to control humidity in the wind tunnel. The dew point sensor is
an ordinary thermal couple surrounded by glass wool which is soaked in water from
a small reservoir. Air from the wind tunnel is withdrawn by a small probe and blown
over the wet bulb. The dew point in the wind tunnel is controlled by regulating the
amount of water fed into the tunnel from a small reservoir with a solenoid valve con-
trolled by the dew point sensor. The amount of water which must be added to the
wind tunnel to obtain the conditions that will exist at the scrubber of the TVA has been
calculated and is shown in Table 15.
Table 15
Amount of water required to maintain a given humidity in
the wind tunnel
Position
Inlet
Outlet.
before reheating
after reheating
Dry
Temperature (°F)
300
128
250
Wet
Temperature (°F)
112
124
137
Relative
Humidity (%)
1.76
90.5
7.73
Amount
of Water
Required (ml)
257
570
612
The participate concentrations which occur at the scrubber site will be simu-
lated in the wind tunnel by adding glass beads. The amount of glass beads required
to obtain the expected dust loadings are detailed in Table 16.
Since the size-selective sampler will be calibrated using microscopic techniques
counting single particles, it is important that the glass beads enter the sampler as
single particles and not as agglomerates. Agglomeration causes errors in the collection
efficiency calibrations and occurs for particles smaller than 5 ^m due to electro-
21
static charging. A bi-polar ion generator described by Whitby has been added to the
wind tunnel to neutralize electrostatic charges on the particulate matter. This de-
+11 3
vice uses a Corona discharge to create an air jet containing 10 bi-polar ions/cm
which are injected into the wind tunnel.
-GO-
-------
Table 16
Amount of flyash and limestone required in the wind tunnel to
maintain a given particle concentration
Inlet
Outlet
Condition
Particulate Weight of flyash Participate Weight of flyash
Concentration and limestone Concentration and limestone
(lb/ft3) needed (g) (lb/ft3) needed (g)
0% excess limestone
injection
0% excess dolomite
injection
150% excess limestone
injection
150% excess dolomite
injection
8.2x 10
-4
7.8x 10
-4
1.3 x 10
-3
1.2x 10
-3
93
88
147.5
133.6
8.35x 10
-6
7.23 x 10
-6
1.22x 10
-5
l.llx 10
-5
0.95
0.82
1.39
1.26
-61-
-------
Two difficulties in obtaining a satisfactory sample of glass beads have been
encountered and solved. The beads obtained from a manufacturer must be separated
to obtain uniform density by using the density gradient tube. The density gradient
separation technique developed during a previous contract had to be modified because
absorbed vapor on the glass beads altered the density of the tube and a gradient
could not be maintained. Oven drying the beads removed the absorbants.
We also found that these beads contained no particles smaller than 0.9 Mm in
diameter. A sample of glass beads was purchased from Particle Information Service
which contained a large number of submicrometer particles, the smallest of which
is 0.3 Mm in diameter. A spinning riffler was designed and constructed to produce
a uniform mixture of glass beads. The riffler was constructed from a discarded ro-
tating display table and seven pie-shaped plastic dishes. A vibrating hopper filled
with glass beads feeds the material into the pie-shaped dishes which are rotating
beneath the hopper.
Since the size resolution of most size-selective sampler is quite high, it was
necessary to improve the Climet light scattering particle counter for sampler testing
(see Figure 20). The resolution was improved in two ways: the old tungsten filament
light source has been replaced by a quartz iodine lamp with about twice the illumination
and a new photo multiplier high voltage power supply has been installed with a far
better voltage stability than the old supply. These modifications reduce the signal
to noise ratio and therefore improve the size resolution by a factor of 3. The orifice
in the sampling section of the Climet has been increased from 0.42 to 0.72 inches in
diameter to allow use of the Climet at flow rates of 0. 8 cfm.
Since we found that optical microscopy could not be successfully used with
the automatic image analyzer for particle size distribution analysis in the submicro-
meter range, it was necessary to construct an epidiascope attachment for the analyzer
to view photomicrographs taken with the scanning electron microscope and determine
the particle size distribution of the particles in the micrograph. The epidiascope
uses a 50 mm camera lens to image 35 mm negatives on the face of the TV camera
tube. The 35 mm negatives are held in a special negative carrier and illuminated
by transmitted light. A size standard is included on each film roll taken with the
-G2-
-------
p * » *
FIGURE 20 Climet light-scattering particle counter
with the new photomultiplier power supply
and a digital volt meter for setting pre-
cise discriminator levels
- 63 -
-------
scanning electron microscope and allows calibration of the HMC and eliminates the
source of inaccuracy. Any size standard can be used, but for these analyses we
used a 5 Mm diameter electron microscope aperture. A prime calibration of the SEM
is made using a replica diffraction grading of known accuracy. In this way size
distributions of particles greater than 0.05 Mm can be obtained with an accuracy of ±5%.
4.2. EMC evaluation
As was reported in our March progress report, the FIMC automatic image
analyzing computer was ordered and installed in March.
The computer was connected to a Leitz research polarizing microscope equipped
with a special set of plan apochromat objectives. This scope allowed the computer
to resolve details in the sample much better than would be possible with ordinary achro-
mats or fluoride objectives.
So that the instrument could be evaluated correctly, we prepared several
different type samples in several different mediums. Each sample was counted
several times and finally the same areas were sized and compared. Different types
of illumination—bright field, dark field and monochromatic light—were also used.
Through experimentation we found that the best medium is a viscous liquid
with a high refractive index (n); but yet relatively inert and noncorrosive.
The best medium was Aroclor®5442, a chlorinated polyphenol resin manu-
factured by Monsanto Laboratories. The refractive index is 1.662 which is actually
not high enough for this type sample. We have found that the small particles, less
than 0.5 Mm, cannot be detected by the computer because the light rays are not
diffracted enough to provide good contrast between the particle and background
light.
Statistically the I1MC unit has an overall accuracy of about 7% and this is
only achieved under good sample preparation techniques.
Because the system is connected to a light microscope, there is automatically
a 1-3% increase in the particle size due to the diffraction rings surrounding the
particle. However, these diffraction rings prevent a perfect image at the edge of
the particle. There is no way to negate these rings without losing the image quality
required to activate the automatic image analyzer. With good plan aprochromatic
-04-
-------
objectives and low magnification, these rings are reduced in size with a 0.5-1.0%
increase in size of the particle and with good plan aprochromatic objectives and
high magnification the size increase is 1-2%, and with an achromat objective the
increase is 3-4%.
Another factor which must be considered is the sample preparation. The
samples for analysis must be taken from different areas of the filter so that the
cumulative sample analyzed is as accurately representative as possible. The com-
puter contains a built-in particle size discrimination factor. Line voltage variation,
internal binary noise, discriminator settings and operator error add together an-
other 3% to the total of 7%.
One of these problems we have been able to eliminate. By using a scanning
electron microscope (SEM) for the image requirements, the associated error re-
duces to less than 0. 25% for the image requirements.
5. Field Test: Kansas Power and Light Company
5.1 Background
The Kansas Power and Light Company (KPL), Lawrence, Kansas, uses a
pulverized coal boiler with injected limestone and wet scrubber to control SO
emissions. This unit, designed by Combustion Engineering, Inc., was under
test by APCO's division of compliance, Durham, North Carolina. Emissions
testing for compliance was performed by York Research on 22-26 March 1971.
Waldon Research also had permission to test SO monitors the same week we
£i
performed our experiments.
5. 2. Experimental plan
Our experiments were planned to obtain:
1) The amount of particle deposition in sampling tubes as a function
of their size and flow rate,
2) changes in particle mass and size distribution due to an isokinetic
sampling,
3) the accuracy of the Andersen Stack Sampler as a size selective
impactor, and
4) the size range efficiency of two cyclone samplers.
-05-
-------
5.3. Description of KPL tests
5.3.1. Time priorities
R. Battles, H. Humecki and T. Smith of McCrone Associates met with
Mr. J. Rohm of APCO, coordinator for all testing, and Mr. M. Funston, plant
manager for KPL, 22 March 1971 at KPL. Mr. Rohm informed us that York
Research would test at both scrubber outlet stacks, occupying all the sampling
ports all day Tuesday and until noon on Wednesday and Thursday. Walden Research
would be sampling the inlet during this time and the outlet on Wednesday, Thursday
and Friday afternoons, simultaneously with us, and at noon on Friday the unit would
be switched from coal to gas.
This allowed us a day and a half at the outlet and two days at the inlet. Our
original sampling time plan required three days at the outlet with a day and a
half at the inlet. This time discrepancy plus several equipment breakdowns cut
heavily into completing all the planned tests.
5.3.2. Completed tests
The sequence of tests were arranged to conform to the time available. Our
tests and operating experiences are described below in chronological order.
(a) Tuesday—cyclone samplers at inlet
Our sampling train consisted of two external cyclones in series, followed
by a filter. We intended sampling the scrubber inlet for one hour with this train,
using a one-half inch diameter nozzle and probe at a 5.5 cfm sampling rate. The
nozzle diameter and sampling rate were chosen—based on our preliminary dust de-
position data—to minimize particle deposition in the probe. At this sampling rate
the large cyclone would have a 50% collection efficiency of 5 microns and the second
cyclone would collect particles down in the submicron size range. Both these
cyclones and the filter holder were placed inside the sampling box on loan from APCO.
For a water trap, we connected a heat exchanger, three bubblers and a silica gel
trap in series. A 5 cfm carbon vein pump was followed by a dry gas meter and a
calibrated orifice meter.
The ducts to the sampler scrubber inlet were horizontal and, since there
-66-
-------
was no room to insert the probe into the side of the duct, dust collection with the
cyclones could not fall into the collection pots; but instead remain inside the cyclones.
This consequently caused a re-entrainment of the particles collected. Since our
large cyclone was designed for laboratory use, it is actually inconvenient for field
use because of the method for fastening the cover. If it is necessary for a cyclone
to be taken apart for cleaning, an easily removable clamp should be used to hold
the cyclone together, We obtained a 49-minute duration cyclone and back up filter
sample from the inlet.
Several other minor problems were encountered; these resulted from the
design of these particular cyclones. The pressure drop across the sampling train
was so great that it was impossible to sample at a rate greater than 2.5 cfm. Another
problem was that our freon cooled heat exchanger plugged with ice due to high humidity
of the inlet gas. Also, the swivel adapter connecting the 0.50 inch probe to the 1.0
inch inside diameter cyclone inlet caused significant deposition at their junction. A
considerable quantity of dust deposited immediately before the adapter. Also, the
transition from the very large outlet diameter of the first cyclone to the small inlet
diameter of the second cyclone caused some deposition. The practice of using a
probe nozzle whose diameter is smaller than the transport tube diameter causes in-
creased dust deposition in the tube, and should not be used for accurate sampling.
This dust deposition is mainly due to the decreased velocity through the carrier tube.
(b) Tuesday—Andersen Sampler at the inlet
After we had completed sampling with the series cyclones, we connected the
large cyclone in front of the Andersen sampler. A 0. 25 inch inside diameter probe was
used with a 0.75 cfm flow rate to minimize deposition in the sampling tube. We
attempted sampling with this train at the scrubber inlet, but the large cyclone came
apart. As a result, further attempts to sample with the large cyclone at the scrubber
inlet were abandoned.
-67-
-------
(c) Wednesday—Andersen sampling at outlet
Wednesday afternoon we were able to move our equipment to the scrubber
outlet and begin sampling with the Andersen Stack Sampler. The sampling train
consisted of a 0. 25 inch probe and nozzle connected to the inlet of the Andersen
Sampler followed by a filter and two heat exchanger cyclones to remove water
vapor. The heat exchanger cyclones were cooled with a bottle of freon. The sam-
pling port and position we used was the same as port number 3 sampled by York
Research, and was 23.6 inches inside the stack.
The Andersen Stack Sampler was used for three sampling times: 1, 4 and
10 minutes at a flow rate of 0.75 cfm. Several difficulties were encountered in
sampling with this train. The Andersen Sampler and back-up filter were heated
with temperature controlled heating tapes and since the sample box had no thermal
baffling, it was necessary to surround the impactor and filter holder with glass
fiber insulation to prevent overheating the other components in the sampling box.
Since the heating tapes were secured with metal clamps, it was cumbersome re-
moving the tapes from the then hot components so the impactor plates could be removed.
Also the thermal couple for the temperature controller was located at the outlet
of the filter holder. As a result, unless gas is coming through the filter holder,
the thermocouple is isolated from the Andersen impactor and filter holder. Thus,
the temperature of the impactor and filter holder will be considerably higher than
the set point of the temperature regulator. The 0. 25 inch diameter probe used for
sampling with the impactor was cleaned between each of the three runs and little
deposition was observed.
The small cyclone heat exchanger froze and plugged during the four minute
sampling run, necessitating a rerun. Although the gas temperature stayed between
SOT and 40 T, the small inlet diameter caused the plugging, since the freon flowed
through the small (second) cyclone cooling jacket and then through the larger cyclone
heat exchanger. The two cyclones were connected in series, both for freon coolant
and for sampling gas, however, it was impossible to maintain a temperature in the
first cyclone lower than 100°F, even when the second cyclone indicated a temperature
of SOT. The cooling jackets should have been connected in parallel. Alternately, a
-68-
-------
higher boiling freon could be used to prevent clogging the heat exchangers. The life
of the 12 pound freon bottle was roughly one sampling hour.
Upon changing the impactor plates, we observed that the particles formed small
mounds just below each jet. Each plate was placed in a petri dish along with its
spacer ring for transport for lab analysis. This method of packing the particle
loaded plates is not completely satisfactory since some of the particles could be
dislodged during transportation. If time permitted, a transporter container could
have been made to hold the entire set of plates and spacers similar in a way as they are
held in the impactor.
(d) Thursday—Particle deposition in probes at the outlet
On Thursday morning the sampling trains were assembled so that two deposition
studies could be performed. Two large filter holders were placed side by side in
the APCO sampling box and the probes were connected to the filter holders with
swagelock fittings. One sampling train consisted of the APCO pump and dry gas
r
meter, while the other was a standard RAC train. Two bubblers, connected in series,
were used in each train to remove water vapor. By Thursday afternoon York Research
had completed their sampling and we had completed assembly of the two sampling
trains. We attempted to calibrate the orifice meter in the APCO train, but the dry
gas meter failed. After some searching we were able to find a small wet test meter
that Combustion Engineering had used, and we then finally calibrated the orifice meter
at 0.75 cfm. As a result of the dry gas meter failure, we were not able to accomplish
any simultaneous iso- and aniso-kinetic sampling Thursday.
The week before our field sampling we tested the dry gas meter in the APCO
train and determined that it was in error by 2.5%. We also discovered two leaks
in the APCO train which were corrected. It was necessary to operate the RAC
unit some 30-60 minutes in the field to free the dry gas meter in that train. This
problem has never been observed in the lab, but only occurs when temperatures
are low, but still above freezing.
-69-
-------
(c) Friday—Particle deposition in probes at the outlet
On Friday morning we started early and were able to perform a number of probe
deposition studies. One tube was used at a flow rate of 0.75 cfm at position number
3 in the stack for a sampling time of 30 minutes. One filter sample was also collected
using a 0.375 inch diameter probe at the flow rate of 3 cfm and a sampling time of
30 minutes. Since this filter sample was fairly representative of the particles in
the stack, it was used to determine the particle size distribution.
5.4. Scrjes cyclone efficiency
As we have already discussed, only one 49 minute test run was completed with
the series cyclones at the inlet on Tuesday since we found that the system had several
problems—one of which is design. The cyclones were designed to be used in an
upright position during the sampling. However, since the inlet port was on top of the
duct, it was necessary to lay the sampling train on its side. As a result we found
a great amount of particle deposition inside of each cyclone. The transport tubes and
connections also aided in particle deposition. This allowed particle re-entrainment
and reduced the cyclone's efficiency. Nearly all of the small particles were allowed
to pass through as well as some of the large particles.
We have reached these conclusions by direct observation of the deposition
and a comparison of the laboratory tested efficiency of 85% and the less than 30%
efficiency found in the KPL test.
From these conclusions we feel that the field test efficiency could have
been increased to a realistic efficiency given more time to better prepare ourselves
about the actual source.
5.5. Probe deposition
On Thursday and Friday we performed tests with four probes of different
diameters to learn about probe deposition. As we had learned from our laboratory
tests, particle deposition is decreased as the probe transport tube diameter decreases.
-70-
-------
This was found to be true in the field test at KPL when the tubes are horizontal.
We also found that there is an increased particle deposition when the same tube is
held vertically. The highest efficiency obtained was with a small transport tube used
with the proper nozzle diameter and held in a verticle position during sampling. The
efficiency was determined to be approximately 98% for a 0.25 in tube held horizontally
and 87% for a 0.50 in tube which held vertically.
6. Field Test: Shawnee Power Plant
A field test of the small cyclone was carried out to prove the viability of
cyclones as size selective samplers. The test was conducted on the Zurn pilot
plant scrubber connected to unit 10 of Tennessee Valley Authority's Shawnee Power
Plant. The pilot plant is a cyclone and scrubber connected in series to the precipitator
inlet. During the tests limestone was injected into the boiler at a stoichiometric
ratio of 1 to 3.
All our sampling was done using a standard RAG sampling train at the outlet
of the scrubber (see Figure 21). A special 3/8 in I.D. x 2 ft long stainless steel,
heated probe was used to heat the gas to 275 T before it entered cyclone T-1A. The
dimensions of this brass cyclone are given elsewhere in the report. The particles
escaping the cyclone were collected on a 90 mm, 0.25 Mm, Solvinert filter. These
filters were selected for evaluation because of the high tare weight and their ability
to withstand high temperatures. The cyclone and filter holder were housed in the
RAC sample box and heated to 275°F during sampling to prevent water condensation.
Small pressure taps were connected to the inlet and outlet of the cyclone so that
the pressure drop could be monitored while sampling.
Several samples were collected, but were later discarded because of equipment
malfunctions. Another problem encountered was the short running time of the scrubber
during most of the experiments. One sample was successfully taken which allowed
an evaluation of the cyclone's performance. During this sample the stack gas was 127°F,
contained 9. 70% water vapor by volume and the flow rate through the cyclone was
l.Olcfm.
-71-
-------
FIGURE 21 The pilot scrubber. The sampling equipment
used for testing the sampling cyclone is shown
at top center and bottom left.
- 72 -
-------
The sample filter was dried and weighed and the particles in the cyclone were
removed and weighed. The cyclone collected 17.4% of the particles. Both samples
were dispersed on aluminum stubs for analysis with the Cambridge Stereoscan
scanning electron microscope. A high contrast 35 mm film was used for the photo-
micrographs. These photomicrographs were in turn analyzed using the epidiascope
attachment to the I1MC. A calibration standard was photographed on each roll of
film. The particle counts were made in groups; two hundred particles each from
five portions of the sample. The five number distributions were converted to weight
fractions using the geometric mean particle size of each interval. The five sets
of data were averaged and the standard deviation determined. The standard de-
viation estimates all the errors encountered in the counting process, including ran-
dom variations, operator error, and instrument variations. The results for the
filter sample and cyclone catch are shown in Tables 17 and 18 and Figure 22.
- 73 -
-------
Table 17
Filter catch size distribution*
Lower
Size
0.030
0.100
0.200
0.300
0.400
0.500
0.600
0.700
0.800
Upper
Size
0.100
0.200
0.300
0.400
0.500
0.600
0.700
0.800
0.900
Geometric
Mean Size
0.055
0.141
0.245
0.346
0.447
0.548
0.648
0.748
0.849
* The size distribution data and
filter
Lower
Size
0.058
0.383
0.766
1.150
1.535
1.923
2.305
2.690
3.070
and cyclone
Table 18
Upper
Size
0.383
0.766
1.150
1.535
1.923
2.305
2.690
3.070
3.450
catch.
Number
Fraction
0.122
0.337
0.282
0.151
0.079
0.020
0.011
0.001
****
Standard
Deviation
0.020
0.038
0.013
0.032
0.024
0.004
0.005
0.001
****
error statistics for the 1000
Weight
Fraction
0.001
0.040
0.171
0.254
0.280
0.132
0.114
0.007
****
particles
Standard
Deviation
0.000
0.010
0.029
0.047
0.062
0.012
0.050
0.017
****
counted from the
Cyclone catch size distribution*
Geometric
Mean Size
0.149
0.542
0.939
1.329
1.718
2.105
2.490
2.874
3.254
Number
Fraction
0.398
0.264
0.140
0.097
0.046
0.025
0.016
0.011
0.004
Standard
Deviation
0.021
0.021
0.026
0.018
0.006
0.010
0.003
0.001
0.003
Weight
Fraction
0.001
0.029
0.081
0.152
0.156
0.153
0.166
0.173
0.090
Standard
Deviation
0.000
0.007
0.027
0.005
0.012
0.040
0.035
0.010
0.044
The size distribution data and error statistics for the 1000 particles counted from
the filter ami cyclone catch.
- 74 -
-------
Cyclone efficiency curve
o
N
CO
s
3
o
w
UI
rt
30
20
10
5
2
1
0.5
0.2
0.1
0.05
0.01
Distribution of filter catch
Distribution of cyclone catch
•3E
I
.3 .4 .5
FIGURE 22
. (i .7 .8 .9 1.0 1.1 1.2
Size distribution of TVA samples collected
with small cyclone
-75-
1.3 1.4 1.5 l.« 1.7
-------
Calculating the size efficiency curve from the two weight fraction distributions is
straightforward. The mass fractional collection efficiency is defined as
M. - M M
K = --2- =
i c o
Where M., M , and M are the masses entering, collected by, and leaving the cyclones.
The fractional collection efficiency for particles of size X is defined similarly
X
Where X and X are the mass of particles of size X that are collected by and escape
C O
the cyclone.
X = C M Eq. (29a)
c x c
X = O M Eq. (29b)
oxo
C and O are the mass fractions of particles of size X found in the cyclone and
X X
filter catch as given by the two mass distributions. Therefore
EX = - ~T*~ Eq- (30)
x c +o Vs
x x k
7. Deposition of Particles in a Horizontal Sampling Tube
Sampling errors also result from particle deposition in the sampling line
down stream of the sampling probe. Therefore another problem in designing a
size selective sampler is how to design a sampling line which will minimize particle
deposition.
Three mechanisms contribute to the total particle deposition in a horizontal
tube: diffusional deposition, gravitational settling and turbulent deposition. Each
of these mechanisms is a function of several variables such as the length and size
of sampling tube, sampling time, particle size and characteristics, Reynolds number
21
of the gas, and flow rate
-7G-
-------
Since many variables are involved in particle deposition, the total transport
tube problem is very difficult to analyze theoretically. Therefore, empirical data
will be of most value for a practical solution to the deposition problem.
The purpose of the preliminary experiment was to obtain information to serve
as a guide to design a detailed field test. It is not designed to test the significance of
relative effect of each variable on deposition but to observe the tendency of deposition
due to variations of several important variables.
Three types of polydisperse aerosols were tested in this experiment: the
mixture of flyash and limestone particles, fine glass beads and coarse glass beads.
The weight average sizes of these aerosols were about 11, 24 and 85 micrometers
respectively. Figure 23 shows the schematic arrangement of equipment for the par-
ticle deposition study. When each experiment was completed, the nozzle, bend
and the tube were thoroughly cleaned with a specially designed cleaning tube as il-
lustrated in Figure 24. A special tip with six 1/32 inch nozzles around its periphery
was attached to one end of a 3/8 in diameter x 10 ft long tube. This end of the tube
was stuck into the sampling tube and the other end was conected to the exhaust end
of the pump with rubber tubing. By circulating the air through the system at the
rate of about 5.5 cfm and moving the cleaning tube back and forth several times in
the sampling tube, the air jet created by the nozzle was able to knock out the par-
ticles deposited in the tube. The particles trapped in the nozzle, bend and tube and
collected by the cyclone and filter are respectively collected and weighed and the
percentage of particles that pass through the transport tube is computed. The ex-
periment for each aerosol was repeated several times by using various flow rates
ranging from 0. 7 to 5.1 cfm and two different sizes of nozzle and transport tube:
1/2 and 1 in.
In summary, the experiments were performed with the following conditions:
a. A mixture of flyash and limestone particles passing through a
1 in tube with a 90°-bend of small curvature (radius of curvature=
diameter of tube).
b. A mixture of flyash and limestone particles passing through a
1 in tube with a 90° bend of large curvature (radius of curvatures
8X diameter of tube).
c. A mixture of flyash and limestone particles passing through a
1/2 in nozzle and 1/2 in tube.
-77-
-------
FILTER
10'SAMPLING
TUBE
90° BEND
CO
I
AIR
IN
•i>
CYCIDNE
ACOUSTIC
DUST FEEDER
By-pass valve
ORIFICE
METER
DRY GAS METER
-MANOMETER
BY-
PASS
VALVE
FIGURE 23
Schematic arrangement of equipment for particle
deposition study
-------
CLEANING TUBE
6-1/32" NOZZLES
FILTER
FIGURE 24 Arrangement of equipment to clean particles deposited in
the sampling tube
-------
d. A mixture of flyash and limestone particles passing through a
1/2 in nozzle, a cyclone and a 1/2 in tube.
e. Fine glass beads passing through a 1/2 in nozzle and a 1/2 in
tube.
f. Coarse glass beads passing through a 1/2 in nozzle and a 1/2 in
tube.
The results of the experiments under various experimental conditions are shown in
Figure 25.
From these experiments we observed that for certain types of particle samples
with a given sampling line the deposition increases progressively as the flow rate
22
decreases. This is just contradictory to what Sehmel observed. In his study of
methylene blue-uranine particle deposition in a 10 ft vertical tube, Sehmel reports
that the percent deposition within the tube increases progressively as the flow rate
increases. Such differences are probably due to the differences in the adhesive
characteristics, density of the particles and flow rates. Methylene blue-uranine
is water soluble and has a density of 1.5 gm/cc while the glass beads and flyash
are insoluble in water and have a density of about 2.5 gm/cc. The flow rates
used ranged from 0. 7 to 5.1 cfm and are much lower than the 0.5 to 50 cfm
used by Sehmel. We also concluded from our preliminary experiment that when a
1/2-in sampling tube with a 1/2-in nozzle is used to sample the particles at the
scrubber inlet, a flow rate of about 2.5 cfm is required to minimize deposition.
Whereas, at the outlet about 5.5 cfm is needed. If a flow rate of 0. 75 cfm is used with the
Andersen Stack Sampler, as recommended by its manufacturer, a transport tube with
an inside diameter of > 0.5 in should be used for minimum deposition.
The findings obtained from this preliminary dust deposition experiment were
applied in our field test at KPL. A one-half inch diameter nozzle and probe were
used for sampling at the wet-scrubber inlet. At a flow rate of 2.5 cfm only 4.6%
(by mass) of the particles deposited in the probe with the most depositable particles
in the range from 1 to 6 micrometers.
One point that deserves special mention is the segregation of particles in
the sampling nozzle and the transport tube. Visual inspection of the samples col-
lected in the deposition experiment indicates that particles trapped in the nozzle
were mainly limestone and those deposited in the transparent tube were mostly
flynsh.
-80-
-------
100
I
oo
Flyash + Limestone
1/2 in. tube with 1/2 in.
nozzle.
Flyash + Limestone
1 in. tube
90° bend with large
curvature, 8D
Large glass beads
(Flex lite 140 mesh)
1/2 in. tube with
V2 in. nozzle
Fine glass beads
(Potter Bros. no. 5000)
1/2 in. tube with 1/2 in
nozzle.
Flyash + Limestone
After cyclone
1/2 in. tube with
in.nozzle.
Flyash + Limes-
tone
1 in. tube
90° bend with
small curvature,
i n
40 _
30
20 _
10 _
FIGURE 25
Deposition of particles in horizontal sampling tube as a
function of particle size and sampling conditions
-------
Although our preliminary experiments on dust deposition provide valuable
information on the design of a sampling line, we recommend that further analyses
be performed to obtain a sampling line with minimum deposition. Since porous
23
wall tubes have been recommended by previous investigations we recommend that
such a tube be considered and tested. Such a tube has been obtained and a study
can begin once the research funds are available.
8. Prediction of Sample Bias Due to Non-isokinetic Aspiration
For particle sampling in a stack and/or duct it is generally considered
necessary to have isokinetic sampling conditions in order to obtain a representative
sample. Studies on sampling bias due to anisokinetic conditions can be found in
24, 25, 26
various publications . A general conclusion from these studies is that
the magnitude of sampling bias is related to particle size, diameter of sampling
probe and degree of variation from isokinetic condition.
In the Bechtel report on sampling and analytical determinations for an alkali
scrubbing test facility, they mention that particles larger than five micrometers
do not get through the scrubber, and consequently that errors due to size classi-
fication are not introduced by anisokinetic procedures. On the basis of this in-
formation, the report concludes that isokinetic sampling is not needed at a wet
scrubber outlet. In order to justify this statement and determine if isokinetic sam-
pling is really not necessary at the scrubber outlet in this project, we performed a
theoretical analysis of the predictive errors of particle sampling at the wet
scrubber outlet. The mathematical model describing the sampling bias presented
21
in our previous report was used to predict the bias. The ranges of the process
variables mentioned in the previous section of this report were used to compute
the values of those variables involved in the model. The prediction was made for
1, 5, and 10-^m particles sampled with 1/8 and 1 in nozzles, the reasonable upper
and lower limits of the proposed probe size. The results,as shown in Figures 26,
27 and 28, show the variation of nozzle bias with velocity ratio.
-82-
-------
10 urn
8 -
i I l I I I
I I
1 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
• Velocity ratio
0.4 0.6
1-in. nozzle diameter
1/8-in. nozzle diameter
FIGURE 26 Comparison of predicted nozzle bias with the same sampling time
and nozzle cross section
-83-
-------
8 L
I
CO
+ 1.645
Standard
deviation
Mean
0.4 0.6
FIGURE 27
Predicted range of nozzle bias for 5-^m particles sampled with
1/8 in. nozzle at various given velocity ratios at 90% confidence
level
- 1.645
Standard
deviation
3.0
-------
I
oo
OT
I
0.4 0.6
1.6 1.8 2.0 2.2
Velocity ratio
- 1.645
Standard
deviation
Mean
- 1.645
Standard
deviation
2.4 2.6 2.8 3.0
FIGURE 28 Predicted range of nozzle bias for 5-^m particles sampled
with 1-in. nozzle at various given velocity ratios at 90%
confidence level
-------
Numerically, the analysis indicates that errors due to anisokinetic sampling
are not very significant. For example, at the 90% confidence level the nozzle bias
for 5 Mm particles would be within -2 and +8.3% if one uses a 1-in nozzle with sam-
pling velocity ranging from 0.6,and 3 times the free stream velocity. With a
1/8-in probe the bias would be much less.
The problem of whether isokinetic sampling is required at the wet scrubber
outlet depends on how much nozzle bias is considered acceptable. Considering the
errors introduced by other factors in sampling particles, (e.g., the readings of
manometer and flowmeter scales) the error due to anisokLnetic sampling may be
insignificant. We concluded, therefore, that isokinetic sampling may not be needed
at the wet scrubber outlet if it is proved that particles larger than 5 micrometers
do not get through the scrubber and that other errors are more significant. We
recommend, however, that a field test be made to test the validity of this theoretical
determination.
9. Evaluation of Filters for Parallel Cyclone Samplers
Since a filter will be the particle collection substrate for the size selective
sampler, filter selection is an important factor. There are several desirable
qualities that a filter for this application should have. It must be able to operate
at temperatures of 250-300°F, it must have a very high particle collection ef-
ficiency for submicron particles, it must withstand the attack of corrosive gases,
and it must allow accurate weighing of the particles collected on the filter. In
addition, a low pressure drop is desirable since this allows the use of smaller
diameter filters and smaller pumps.
A preliminary evaluation of three types of filters was performed to determine
if any of these materials met all of the above requirements. The filters tested
were: a 90-mm diameter Solvinert filter made by Millipore Corporation with a 0. 25
pore size, a 110-mm diameter high efficiency glass fiber filter made by Gelman,
and a 47-mm Nuclcpore filter made by General Electric Corporation with a pore size
of 0.4 Mm in diameter. All three of these filters meet the corrosive gas and tem-
perature stated above. The weight stability of the Solvinert and glass fiber filters
-86-
-------
were tested to determine the amount of sample which must be collected to obtain
3cr accuracy in the sample weight. Fifteen of the 90-mm Solvincrt filters were
dried in an oven for two hours at 135°C and then each weighed twice. The arithmetic
mean of the weight deviations and the standard deviation in the mean was calculated.
For a confidence level of 99%, the weight deviation was 0.201% or approximately
0.6 mg, since the filters has a tare weight of approximately 300 mg. Therefore,
to obtain a 3cr accuracy of 1% using these filters, a sample of 60 mg must be col-
lected. Glass fiber filters are very fragile and can lose considerable weight
during sampling and subsequent filter handling. Three of the Gelman filters
were loaded into a filter holder ten times each and weighed to determine the amount
of weight loss due to the assembly procedure. The weight loss at a 99% confidence
level was 0.126% or about 1 mg for a 800 mg tare weight filter. Using these fil-
ters, approximately 100 mg of sample must be collected to obtain the 1% accuracy.
Greater weight loss can be expected during the actual use of these filters due to
blow-off of glass fibers during sampling and water moisture variations.
The pressure drop characteristic of the Solvinert and the Nuclepore filters
were tested in holders having stainless steel screen supports and Teflon back-up
filters. The Teflon filters decreased the pressure drop by reducing the area re-
striction of the stainless steel screen. The 0. 25- Mm Solvinert filter was tested in
2 2
a filter holder having an area of 0.06 in , while a 1.5 in filter holder was used to
test the 0.4 Mm Nuclepore filters. The pressure drop curves determined for these
two filters are shown in Figure 29. By using the manufacturer's information, it
was possible to extrapolate this data to filters of different sizes and pore diameters.
Besides having a lower pressure drop, the Nuclepore filters have two
other advantages over the Solvinert filter: low tare weight and greater strength.
The Nuclepore 47-mm filter weighed only 20 mg. The Solvinert filters are very
brittle and were difficult to handle during the field test at TVA. We do not expect
there will be a weight stability problem with the Nuclepore filters because of their
low tare weight. The Pallflex Products Corporation's Ultipore 0.35 filter may also
be usable. The manufacturer's data shows that the pressure drop-flow rate curve for
a 47-mm filter is the same as for the 0.5 Mm-90 mm Solvinert filter. The weight
stability for this filter and the Nuclepore filter need to be evaluated.
-87-
-------
3
2.5 .
2 .
1.5
2.5 3 45
Pressure drop
6 7 8 9 10
—>(in. of Hg)
1.5
FIGURE 29 Pressure drop vs. flow rate for several filters
in typical filter holders with teflon backup fil-
ter to reduce the pressure drop.
-88-
-------
m. Summary of Phase II
An automatic particle monitor is attractive mainly because manual
collection and analysis of four particulate samples in eight hours from six gas stream
is difficult and costly as it introduces high manpower requirements and a great
possibility of missed or improperly treated samples. An automatic monitor would
provide, or at least approach, continuous monitoring, more timely data, and
additional information about the particles such as their iron, sulfur and magnesium
content.
A conceptual design for an automatic particle monitor system which is
25
technically and economically feasible has been developed and reported previously.
The monitor will sample six points simultaneously using a cascaded cyclone and slot
impactor to size-selectively collect the sample; a /?-ray mass gauge and an x-ray
fluorescence spectrometer to analyze the mass and composition of the particle-size
fractions collected; and a computer to control the monitors as well as collect and
analyze the data. Six particle-collection and -analysis subsystems will be controlled
by two computers. At the inlet of each of the three scrubbers, one particle size
fraction will be collected and analyzed each minute while only one analysis every two
to three minutes will be feasible at the outlet of the scrubber. The /3-ray mass gauges
will require 10-30 sec counting times for each size fraction analyzed. After mass
analysis, the particles will be moved to the x-ray fluorescence analyzer where an
analysis of the elemental composition will be made after a 10-sec count. One of
the major functions in data reduction will be the elimination of interferences in the
x-ray fluorescence spectrometer due to particle-size and matrix-composition variations;
this will be accomplished by physical and/or mathematical methods.
The major advantage of the design is the use of proven techniques. )9-ray
absorption and x-ray fluorescence are well understood phenomena and have been used
in many process control applications. Particle sampling and collection using cyclones
and impactors are being thoroughly investigated under the same contract for which the
automatic monitor is being developed. The integration of these techniques with
computer control is certainly feasible.
-89-
-------
The maximum cost for constructing and maintaining the automatic
monitor is estimated to be $358, 710. The estimated cost for manual sampling
includes the development program for the manual methods and the equipment and
manpower required for sampling during 30 weeks of scrubber tests but excludes
the cost of analyzing the collected samples and training the sampling personnel.
This cost alone is $425,228. If the demonstration program lasts longer than 30
weeks and if the two cost items excluded are added, the total cost for using the
manual method will be much greater. Based on the figures described, the automatic
monitor is economically feasible (see Table 19).
From the cost data for the automatic monitor, several conclusions about
cost reductions have been made. The large number of duplicate components make
searching the market for the least expensive components important. In addition,
a design which eliminates a few components will be considerably cheaper; as is
shown, a more efficient computer system which would reduce the number of compo-
nents for that system is feasible. In this way, price reductions of $20, 000 to $30, 000
can be achieved without any sacrifice in instrument performance. By reducing
instrument performance or the number of sampling points to be measured simul-
taneously, even larger reductions in hardware and development costs can be achieved.
Table 19 Cost Data Summary
Cost Comparison
Item Manual method Automatic method
Hardware, (including replacement parts)
Labor:
Development Program
Sampling manpower
(assuming 30 weeks
demonstration program)
Cost of chemical analysis
of collected samples
Total
$156,728 $173,768
104,000 184,942
(includes repairs at TVA
during demonstration
program)
164,500 automatic
unknown automatic
$425,228 $358,710
-90-
-------
References
1. Bechtel Corporation, Alkali scrubbing test facility, Phase I: preliminary
engineering, report to NAPCA, NAPCA contract PH 22-68-67 (May, 1969).
2. Bechtel Corporation, Alkali scrubbing test facility, Phase II: Design engineer-
ing, design criteria, report to NAPCA, NAPCA contract PH 22-68-67 (May,
1970).
3. S. M. Blacker, Evaluation of the Andersen Stack Sampler, A report of field
test, NAPCA, 1970.
4. Environmental Research Corporation: Report of results of Andersen Stack
Sampler evaluation tests, Submitted to HEW, 1970.
5. Walter C. McCrone Associates, Inc., Particulate measurements for fossil-
fuel combustion sources. A report to NAPCA on the state of the art, sec-
tion 6, NAPCA contract CPA 22-69-130 (July, 1970).
6. Environmental Research Corporation, "Report of Results of Andersen Stack
Sampler Evaluation Tests," December 2, :970.
7. Schemel, G. A. ,• "The Density of Uranine Particles Produced by a Spinning
Disc Aerosol Generator. "J. of Amer. Indust. Hygiene Assoc., September,
October 491-492.
8. Wiland, Geisel, "Calculating the Particle Size Distribution of a Duct by
Means of Fractional Separation Efficiency Curves and Total Efficiency Curves,"
Staub-Reinhart, Luft. 28, 25-28 (1968).
9. Hilbert Schenck, Theories of Engineering Experimentation, 50-53, McGraw-
Hill Book Co., (1968).
10. Muschelknautz, E., and K. Brunner, "Experiments with Cyclones," Chem.-
Ing. -Techn., 39, pp 531-538 (1967).
11. Rosin, P., and E. Rammler, Journal of Inst. of Fuel, T_, p. 29 (1933).
12. "Air Pollution Engineering Manual," PHS publication, 99-AD-40, U. S. Dept.,
HEW (1967).
13. N. W. First, "Fundamental Factors in the Design of Cyclone Dust Collectors,"
Sc. D. discertation, Harv. Univ. (1950).
14. W. Barth, "Calculation and Design of Cyclone Separators on Basis of Recent
Investigations," etc.
15. Muschelknautz, E., and W. Krambrock, The aerodynamic coefficients of the
cyclone separator as based on recent, improved measurement, Chem. Ing.
Tech. 42, 247-255 (1970).
16. Muschelknautz, E., Design of cyclone separators in the engineering practice,
Staub-Rcinholdt Luft 3J), 1-12 (1970).
-91-
-------
17. Stairmand, C. J., The design and performance of cyclone separators,
Trans. Inst. Chcm. Engrs. 29, 356-383 (1951).
18. Lipman, M. and A. Kydonicus, A multistage aerosol sampler for extended
sampling intervals, Am. Ind. Hyg. Assoc. J., 730-7 (1970).
19. Freudcnthal, P., High collection efficiency of the Aerotec-3 cyclone for sub-
micron particles, Atmos. Environ. 5, 151-4 (1971).
20. Cochman, J. C., and II. M. Moseley, Simplified method for determining
cascade impactor stage efficiencies, Am. Ind. Hyg. Assoc. J., 62-67 (1967).
21. Walter C. McCrone Associates, Inc., Participate measurements for fossil-
fuel combustion sources. A report to NAPCA on the state of the art, sections
3, 4 and 5 NAPCA contract CAP 22-69-130 (July, 1970).
22. G. A. Sehmel, Particle sampling bias introduced by Anisokinetic Sampling
and deposition within the sampling line, American Industrial Hygiene Asso-
ciation Journal, Nov-Dec., 1970.
23. Personal communication with Meryl Jackson, Freeman Laboratory, Chicago.
24. W. Strauss, Industrial Gas Cleaning, Pergamon Press Ltd., First Edition,
1966.
25. N. A. Fuchs, The Mechanics of Aerosols, Pergamon Press, Ltd., 1964.
26. S. Badzioch, Correction for Anisokinetic Sampling of Gas-borne Dust Par-
ticles, Journal of the Institute of Fuel, March, 1960.
-92-
-------
BIBLIOGRAPHIC DATA
SHEET
1. Report No.
EPA-650/2-73-024
2.
3. Recipient's Accession No.
4. Title and Subtitle
Measurement and Characterization of Particles in Wet
Scrubbing Process f or SO.. Control
A
5* Report Date
July 1972
6.
7. Author(s)
8. Performing Organization Kept.
No' MA Proi 2001
9.'"Performing Organization Name and Address
WaUer C. McCrone Associates, Inc.
2820 South Michigan Avenue
Chicago, Illinois 60616
10. Project/Task/Work Unit No.
11. Contract/Grant No.
EHSD 71-25
12. Sponsoring Organization Name and Address
EPA, Office of Research and Development
NERC-RTP, Control Systems Laboratory
Research Triangle Park, North Carolina 27711
.epprt & I
Covered Final
8/15/70-7/7/72
14.
15. Supplementary Notes
16. Abstracts The report gives results of the development of a technique for size-
selective, high-capacity particulate sampling to be used in measuring and
characterizing the particles in the wet-scrubbing process for SOx control. It
provides information on both wet-scrubbing system process variables, and the
efficiency of scrubbers to be used at the TVA test station. The sampler, existing as
a manually operated bench-scale prototype, is capable not only of fractionating the
particles in the desired range, but also of tolerating the process stream environment.
A preliminary evaluation indicated not only that existing hardware did not meet all
requirements, but that automatic sampling was more cost effective (despite higher
initial costs) than manual sampling, by at least 16 percent. However, program
economics resulted in the decision to design a manually operated model.
17. Key Words and Document Analysis.
Air Pollution
Particle Size
Measurement
Sampling
Sulfur Oxides
Cost Effectiveness
Washing
Scrubbers
Wind Tunnels
17b. Identificrs/Opcn-Endcd Terms
Air Pollution Control
Particulates
Characterization
Particle Collection
Cascade Impactors
17e. COSAT1 Field/Group
13JJ
17a. IVscriplors
Field Tests
Dust Filters
Dust
Dust Collectors
Fly Ash
Cyclone Separators
Elutriators
Flue Gases
Wet Scrubbing
Andersen Stack Sampler
14B'
18. Availability Statement
Unlimited
19..Security Class (This
Report) '
UNCLASSIFIED
20. Security Class (This
Page
UNCLASSIFIED
21- No. of Pages
100
22. Price
FORM NTIS-3S (REV. 3-72)
-93-
USCOMM-DC M9S2-P72
-------
INSTRUCTIONS FOR COMPLETING FORM NTIS-35 (10-70) (Bibliographic Data Sheet based on COSATI
Guidelines to Format Standards for Scientific and Technical Reports Prepared by or for' the Federal Government,
PB-180600).
1. Report Number. Each individually bound report shall carry a unique alphanumeric designation selected by the performing
organization or provided by the sponsoring organization. Use uppercase letters and Arabic numerals only. Examples
FASIiB-NS-87 and FAA-RD-68-09-
2. Leave blank.
3. Recipient's Accession Number. . Reserved for use by each report recipient.
4. Title and Subtitle. Title should indicate clearly and briefly ilic subject coverage of the report, and be displayed promi-
nently. Set subtitle, if used, in smaller type or otherwise subordinate it to main title. When a report is prepared in more
than one volume, repeat the primary title, add volume number and include subtitle for ihe specific volume.
5. Report Date, l-.uch report shall carry a date indicating .11 least month and year. Indicate the basis on which it was selected
(e.g., date of issue, date of approval, date of preparation.
6. Performing Organization Code. Leave blank.
•7. Aothor(s). Give namc(s) in conventional order (e.g., John K. Doe, or J.Robert Doe). List author's affiliation if it differs
from the performing organization.
8. Performing Organization Report Number. Insert if performing organization wishes to assign this number.
9. Performing Organization Name and Address. Ciive name, street, city, state, and zip code. List no more than two levels of
an organizational hierarchy. Display the name of the organization exactly as it should appear in Government indexes such
as USGRDR-I.
10. Project/Task/Work Unit Number. Use the project, task and .work unit numbers under which the report was prepared.
11. Controct/Gronf Number. Insert contract or grant number under which report was prepared.
12. Sponsoring Agency Nome and Address. Include zip code.
13. Type of Report and Period Covered. Indicate interim, final, etc., and, if applicable, dates covered.
14. Sponsoring Agency Code. Leave blank.
IS. Supplementary Notes. Kntcr information not included elsewhere but useful, such as: Prepared in cooperation with . . .
Translation of ... Presented at i tinfert-ncc of ... To be published in-. . . Supersedes . . . Supplements
16. Abstract. Include a brief (200 words or less) factual summary of the most significant information contained in the report.
If the report contains a significant bibliography or literature survey, mention it here.
17. Key Words and Document Analysis, (a). Descriptors. Select from the Thesaurus of Kngincering and Scientific Terms the
proper authorized terms that identify the major concept of the research and are sufficiently specific and precise CO be used
as index entries for cataloging.
(b). Identifiers and Open-Ended Terms. Use identifiers for project names, code names, equipment designators, etc. Use
open-ended terms written in descriptor form for those subjects for which no descriptor exists.
(c). COSATI Field/Group. Field and Group assignments are to be taken from the 1963 COSATI Subject Category List.
Since the majority of documents arc muliidisciplinary in nature, the primary Field/Group assignment(s) will be the specific
discipline, area of human endeavor, or type of physical object. The application(s) will be cross-referenced with secondary
Field/Group assignments that will follow the primary posting(s).
18. Distribution Statement. Denote relcasability to the public or limitation for reasons other than security for example "Re-
lease unlimited". Cite any av.iilability to the public, with address and price.
19 & 20. Security Classification. Do not submit classified reports to the National Technical
21. Number of Pages. Insert the total number of pages, including this one and unnumbered pages, but excluding distribution
list, if any.
22. Price. Insert the price set by the National Technical Information Service or the Government Printing Office, if known.
FORM NTIS-3S (REV. 3-721 -94- USCOMM-DC 14MZ-P72
------- |