EPA-600/2-76-073
March 1976
Environmental Protection Technology Series
WIND TUNNEL EVALUATION OF
PARTICLE SIZING INSTRUMENTS
Industrial Environmental Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and application of
environmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the ENVIRONMENTAL PROTECTION
TECHNOLOGY series. This series describes research performed to develop and
demonstrate instrumentation, equipment, and methodology to repair or prevent
environmental degradation from point and non-point sources of pollution. This
work provides the new or improved technology required for the control and
treatment of pollution sources to meet environmental quality standards.
EPA REVIEW NOTICE
This report has been reviewed by the U.S. Environmental
Protection Agency, and approved for publication. Approval
does not signify that the contents necessarily reflect the
views and policy of the Agency, nor does mention of trade
names or commercial products constitute endorsement or
recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-76-073
March 1976
WIND TUNNEL
EVALUATION OF
PARTICLE SIZING INSTRUMENTS
by
Charles H. Gooding
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, NC 27709
Contract No. 68-02-1398, Task 23
ROAP No. 21ADL-18B
Program Element No. 1AB012
EPA Project Officer: W. B. Kuykendal
Industrial Environmental Research Laboratory
Office of Energy, Minerals, and Industry
Research Triangle Park, NC 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
-------
CONTENTS
Number Page
Figures iv
Tables vi
Acknowledgments vii
Conclusions 1
Recommendations 3
Sections
1.0 INTRODUCTION 5
2.0 CHARACTERISTICS OF THE WIND TUNNEL AND TEST
CONDITIONS 6
3.0 OPERATING PROCEDURES FOR THE PARTICLE SIZING
INSTRUMENTS 12
3.1 Brink Impactor 12
3.2 Andersen Impactor 15
3.3 Southern Series Cyclones 17
3.4 Environmental Systems Corporation PILLS IV 17
3.5 GCA In-Stack Beta Impactor 18
3.6 Celesco Piezoelectric Microbalance Impactor 19
3.7 Thermo-Systems 3030 Electrical Aerosol Size
Analyzer 21
4.0 DATA REDUCTION 22
4.1 Differential Particle Size Distributions-
050 Method 22
4.2 Brink Impactor 24
4.3 Andersen Impactor 24
4.4 Southern Series Cyclones 26
4.5 Celesco Piezoelectric Microbalance Impactor 27
4.6 GCA In-Stack Beta Impactor 27
4.7 Environmental Systems Coporation PILLS IV 28
5.0 COMPARISON OF INSTRUMENTS BASED ON PRIMARY DATA 32
5.1 Procedure for Successive Runs 32
5.2 Comparison of Instruments in Successive Runs 33
5.3 Comparison of the GCA In-Stack Beta Impactor
with the Brink Impactor 44
6.0 RESULTS OF SIMULTANEOUS RUNS 50
Appendix - Calculation of Differential Size Distribution 61
References 64
in
-------
FIGURES
Number Page
1 Isometric sketch of the particulate aerodynamic test
facility. 7
2 Dust concentration profiles with and without an
upstream obstruction (velocity =9.1 m/s). 9
3 Dust concentration profiles with and without an
upstream obstruction (velocity = 27.3 m/s). 10
4 Electronic and flow schematics of Celesco Impactor. 20
5 Determination of mean concentration for stages
of the GCA In-Stack Beta Impactor. 29
6 Comparison and curve fit of Brink and Andersen
Impactor data (concentration = 0.089 g/Nm3). 37
7 Comparison and curve fit of Brink and Andersen
Impactor data (concentration^ 0.955 g/Nm3). 38
8 Comparison of Southern Series Cyclone data with
impactor curve (concentration = 0.089 g/Nm3). 40
9 Comparison of Southern Series Cyclone data with
impactor curve (concentration = 0.955 g/Nm3). 41
10 Comparison of Celesco Piezoelectric Microbalance
Impactor with impactor curve (concentration = 0.089 g/Nm3). 42
11 Comparison of Environmental Systems Corporation PILLS IV
data with impactor curve (concentration = 0.089 g/Nm3). 45
12 Comparison of Environmental Systems Corporation PILLS IV
data with impactor curve (concentration = 0.955 g/Nm3). 46
13 Comparison of GCA In-Stack Beta Impactor data with
corresponding Brink curve (concentration = 0.267 g/Nm3). 47
14 Comparison of GCA In-Stack Beta Impactor data with ~
corresponding Brink curve (concentration = 0.955 g/Nm ). 48
15 Comparison of simultaneous Brink data with primary
Brink data curve fit (concentration = 0.089 g/Nm3). 51
16 Comparison of simultaneous Brink data with primary
Brink data curve fit (concentration = 0.955 g/Nm3). 52
17 Comparison of simultaneous Andersen data with primary
Andersen data curve fit (concentration = 0.089 g/Nm3). 53
IV
-------
Figures (cont'd)
Number Page
18 Comparison of simultaneous Andersen data with primary
Andersen data curve fit (concentration = 0.955 g/Nm3). 54
19 Comparison of simultaneous Southern Series Cyclone data
with primary Cyclone data curve fit (concentration =
0.089 g/Nm3). 55
20 Comparison of simultaneous Southern Series Cyclone
data with primary Cyclone Data curve fit (concentration =
0.955 g/Nm3). 56
21 Comparison of simultaneous Celesco data with primary
Celesco data curve fit (concentration = 0.089 g/Nm3). 57
22 Comparison of simultaneous PILLS IV data with primary
PILLS IV data curve fit (concentration = 0.089 g/Nm3). 58
23 Comparison of simultaneous PILLS IV data with primary
PILLS IV data curve fit (concentration = 0.955 g/Nm3). 59
-------
TABLES
Number Page
1 Wind tunnel operating conditions 8
2 ' Particle sizing instruments tested 13
3 Instrument specifications 14
4 D5o's of Brink Stages and Cyclone 25
5 D5o's °f Andersen Stages and Cyclone 25
6 Size increments of PILLS IV 31
7 Position of instruments in the wind tunnel and
. measured total mass concentration 34
8 Normalized variance for low concentration runs-
primary data 35
9 Normalized variance for high concentration runs-
primary data 36
A-l Reduced data table for Brink run No.905-2 63
-------
ACKNOWLEDGMENTS
The cooperation and assistance of the following individuals in
various phases of this project are acknowledged with sincere thanks.
Messrs. A.D. Brooks and D.W. VanOsdell and Miss M.B. Hardy of RTI
made major contributions in the experimental work and data analysis.
Mr. Billy Bowles of Monsanto Research Corporation and several
members of his staff operated the wind tunnel facilities.
Dr. Don Wallace of IBC/Celesco (Irvine, Calif.), Mr. Jim Congdon of
GCA/Technology Division (Bedford, Mass.), and Dr. Gerhard Kreikebaum of
Environmental Systems Corporation (Knoxville, Tenn.) provided direction
and assistance in the operation of their respective instruments.
Mr. Joe McCain of Southern Research Institute (Birmingham, Ala.)
provided information on the operating aspects of several of the instruments
and on the data reduction technique.
Mr. Neal Hill of the Hill Environmental Group (Chapel Hill, N.C.)
provided information on the Andersen cyclone precollector.
Mr. W. B. Kuykendal of the EPA Industrial Environmental Research
Laboratory (Research Triangle Park, N. C.) provided overall direction
for this work in his capacity as Project Officer.
-------
CONCLUSIONS
The conclusions derived from this experimental work and subsequent
data analysis are as follows:
1) Although the particulate aerodynamic test facility is
equipped with 8 sampling ports, simultaneous sampling with more
than one instrument in the tunnel did not yield reliable data.
The concentration of particulate across the wind tunnel was
not uniform, and instruments placed in upstream ports created a
disturbance which was detected over the length of the tunnel.
For instrument comparison, more consistent results were obtained
by operating the wind tunnel continuously with steady state
conditions and sampling with the instruments one at a time in the
same port.
2) The Brink and Andersen Impactors and the Southern Series
Cyclones yielded comparable results when measuring the particle
size distribution of fly ash.
3) The Celesco Piezoelectric Microbalance Impactor yielded
results consistent with those obtained using the Brink and
Andersen Impactors and Series Cyclones. However, the extractive
testing procedure resulted in an absence of data from the first
three of the ten stages of the impactor due to sample line losses.
Data from the last three of the ten stages were also lost
because a measurable mass of particulate could not be collected
on these stages without overloading the middle stages.
4) The GCA In-Stack Beta Impactor appeared to classify particles,
but the distribution calculated from the data did not agree well
with the Brink and Andersen results. More study is needed to
clarify and perhaps improve on the conversion from beta
attenuation to collected mass. The zero fluctuation problem
must be eliminated before the instrument can be used with
confidence. The amplifer gains should also be optimized to
provide a detectable signal from the lower stages and to avoid
pegging of the chart readout on the upper stages.
5) The differential particle size distributions obtained with
the PILLS IV instrument did not agree at all with those obtained
from the inertial classification devices. This result was not
unexpected since the PILLS IV is an optical single-particle
1
-------
counter. For the two types of instruments to be comparable,
all of the particles should be spherical and of the same
density. Neither of these conditions exists with fly ash.
6) No conclusions could be reached concerning the operation
of the Thermo-Systems Electrical Aerosol Size Analyzer because
erratic results were obtained. The problem was postulated to be
inadequate static charge neutralization of the particles prior to
entry into the instrument.
-------
RECOMMENDATIONS
This project was a first effort to evaluate and compare the results
of several particle sizing instruments which operate on various theoretical
principles. Expanded studies that include other instruments as well as
other industrial dusts are obviously needed to extend the value of
existing experimental data and to better define the options available to
the researcher faced with instrument selection for particle sizing
studies.
Specific recommendations related to the equipment and instruments
utilized in this study are enumerated below.
1) The EPA particulate aerodynamic test facility has been
equipped with a new dust feed system since the completion of
this study. Before particulate studies similar to this one are
undertaken, the modified system should be thoroughly tested to
characterize the stratification and upstream disturbance phenomena
identified in this work. If these phenomena still exist, both
total particulate mass and particle sizing measurements should
employ duct traversing techniques or at least identical-point-
sampling by all instruments to be compared.
2) The Brink and Andersen Impactors and the Southern Series
Cyclones may be assumed to give equivalent particle size distribution
data when operated under conditions similar to this study. Use
of the Andersen cyclone precollector at a sample rate more than a
few percent greater or less than 21.2 &/min is not recommended
until more extensive calibration tests are performed at other
flow rates to determine the DrQ and the shape of the collection
efficiency curve in relation to the efficiency curves of the
upper cascade stages.
3) The "brush" technique now employed to remove particles
deposited in the cyclones and on impactor walls can lead to
significant errors due to particles sticking to the brush and
brush hairs contaminating the sample. A reasonable alternative
might be washing of the surfaces with small volumes of a
volatile solvent, followed by evaporation, dessication, and
weighing. This and other possible alternative techniques should
be investigated.
-------
4) The PILLS IV instrument should not be used in a comparative
situation with inertial separation devices until a method is
established to correlate the results. Such a correlation would
likely involve, as a minimum, the expression of particle shape
and density as a function of size.
5) If the development of the GCA In-Stack Beta Impactor is
to continue, the source of the zero fluctuation must be
identified and eliminated. The correlation between beta
attenuation and mass of particles on the substrates should be
verified, and the gain of the individual stages should be
optimized to allow data retrieval from the lower stages and to
avoid pegging of the chart readout on the upper stages.
6) Consideration should be given to the development of an
in-stack cascade impactor using piezoelectric crystals as stage
substrates. Such an instrument could offer the advantages of
near real-time operation and reasonably small physical size.
The primary obstacle to overcome will be the low mass capacity
of the crystals. An integrated, in-stack dilution system could
possibly be developed to alleviate this limitation.
-------
1.0 INTRODUCTION
In the last few years, particulate control technology has been
developed to the extent that several methods are now available to
collect large particles with an efficiency of greater than 99 percent.
Emphasis on particulate control has now shifted toward the fine particle
size range, particularly to particles which have diameters between 0.3
and 3 micrometers. These particles are of particular interest because
they tend to remain in the atmosphere for long periods of time, thus
contributing to atmospheric haze. They also tend to deposit in the
human respiratory system when inhaled.
Many instruments have been developed to classify airborne particulate
according to size. This research project was undertaken to evaluate and
compare several particle sizing devices in a wind tunnel. Particulate
concentrations more typical of stack conditions rather than atmospheric
conditions were used. The scope of the project was originally intended
to encompass three different types of particulate with different pro-
perties; however, because of unexpected complications with the dust
dispersion system, the project was limited to a comparison of the
instruments with fly ash from a coal-fired power plant at two concen-
tration levels.
Two experimental approaches were utilized. In the first, three or
more instruments were tested simultaneously in different ports of the
wind tunnel. Replicate runs were made, alternating ports and instruments
to average out any port bias. In the second experimental approach,
which yielded more consistent data, particular care was taken to operate
the wind tunnel at steady state conditions and the instruments were
tested successively, one at a time, in the same port. Analysis of the
data obtained by both approaches resulted in an evaluation of wind
tunnel performance characteristics as well as a comparison of the particle
sizing instruments.
-------
2.0 CHARACTERISTICS OF THE WIND TUNNEL AND TEST CONDITIONS
The experimental portion of this project was conducted at the
particulate aerodynamic test facility located at EPA, Environmental
Research Center, Research Triangle Park, N. C. This facility, is
basically a low speed wind tunnel. The lower leg of the tunnel is a
modular 61 cm diameter duct, 12.2 m in length. Each of four test
sections located in this leg is equipped with two opposing 15 cm test
ports. Dust is injected just upstream of the first test section and is
removed in a baghouse dust collector located at the end of the return
air loop. The tunnel is thus closed-loop with respect to gas flow but
open-loop with respect to particulate. Figure 1 is an isometric sketch
of the wind tunnel showing the port designations utilized in this report.
Table 1 summarizes the operating conditions for the test program.
The low concentration is a reasonable approximation of clean stack
emissions and the high concentration is approximately an order of
magnitude higher. The velocities are higher than those normally en-
countered in a stack test, but were necessary in order to minimize
settling of large particles and to achieve the desired concentrations.
A moderate temperature was chosen to avoid the complications of preheating
and condensation devices in the sample trains.
Initially the experimental program was begun with the assumptions
that the wind tunnel provided a uniform cross-sectional particulate
concentration at all ports as indicated in the design manual (ref.l),
and that upstream particulate sizing instruments would present a negligible
disturbance to downstream positions. After several series of simultaneous
tests were completed and the data were analyzed, it was apparent that
these assumptions were invalid. A special test was designed to evaluate
the constancy of total mass particulate concentrations across a chosen
cross-section of the wind tunnel and to determine the effect of an
upstream obstruction on the particulate distribution across the cross-
section. An IKOR model 206 portable air quality monitor, which is a
real-time instrument, was utilized for this test.
-------
•CONTROL RfXNEL
Figure 1. Isometric sketch of the particulate aerodynamic test facility.
-------
Table 1. WIND TUNNEL OPERATING CONDITIONS
High Concentration Low Concentration
Velocity (m/s)
Temperature (°C)
Relative Humidity
Gas Composition
Fly Ash Feed (kg/hr)
0
Concentration (g/Nm )
9.1
27
50-60%
Air
8.3
0.955
27.4
27
50-60%
Air
2.3
0.089
In the IKOR model 206 a sample is continuously drawn through an
electronic sensing head where the particulate generates an electric
current by charge transfer. The current is proportional to total mass
concentration. The IKOR was placed in port H to run horizontal traverses
across the tunnel. The bottom portion of Figures 2 and 3 show the
results of traverses with no upstream obstructions at velocities of 9.1
m/s and 27.4 m/s, respectively. These data indicate that the profile is
definitely not uniform, and is possibly assymmetric with respect to the
horizontal centerline of the wind tunnel.
To evaluate the effect of upstream obstructions, a section of 3.3 cm
O.D. pipe was placed in the center of the flow stream, extending from
top to bottom of the tunnel. The diameter of this pipe was approximately
the same as that of a Brink Impactor. Horizontal traverses were then
run at each velocity with the IKOR in port H and the pipe 3, 6, and 9 m
upstream, corresponding to the upstream port locations. The results are
plotted in Figures 2 and 3. The general tendency at both velocities is
for the obstruction to depress the particulate concentration in the
center and create peaks on each size between the vertical centerline and
the wall. This effect is noticeable even when the obstruction is 9 m
upstream.
8
-------
OBSTRUCTION 9m UPSTREAM
J I
J 1 I I I I I
OBSTRUCTION 6m UPSTREAM
S '
UJ
I I I
I I I I
OBSTRUCTION 3m UPSTREAM
*" NO OBSTRUCTION
3
2
I
-12 -10 -8 -6 -4 -2
8
POSITION OF IKOR PROBE IN TUNNEL
(HORIZONTAL INCHES FROM CENTERLINE)
10 12
Figure 2. Dust concentration profiles with and without an
upstream obstruction (velocity = 9.1 m/s).
-------
<
UJ
o:
OL
o
1.0
0.5
1.0-
O5
1.0-
^ 05-
0.5
j i
I I
j I
-12 -10 -8
OBSTRUCTION 9m UPSTREAM
I I i I I
OBSTRUCTION 6m UPSTREAM
j i
OBSTRUCTION 3m UPSTREAM
j I
j i
NO OBSTRUCTION
J I i I
j i
I i
j I
-6-4-2024 68
POSITION OF IKDR PROBE IN TUNNEL
(HORIZONTAL INCHES FROM CENTERLINE)
10 12
Figure 3. Dust concentration profiles with and without an
upstream obstruction (velocity = 27.3 m/s).
10
-------
Since the scope of the project did not include a thorough evaluation
of the wind tunnel's characteristics, no additional tests were conducted
to determine the severity of interference created by instruments in
opposing ports or to compare the profiles in different test sections
with no obstructions or to run vertical traverses. The results discussed
above and shown in Figures 2 and 3 were sufficient to cast serious doubt
on the validity of the data obtained with more than one instrument in
the tunnel at the same time. The results also suggested that comparative
instrument runs should be made with all of the instruments sampling at
identically the same point inside the wind tunnel.
In addition to the fly ash evaluation, comparison of the instruments
with dusts from a basic oxygen furnace and from a cement kiln was originally
planned. Neither of these dusts could be utilized, however, because of
severe plugging which occurred in the dust feeder and in the tubes that
transport the redispersed dust to the wind tunnel. Several minor
modifications to the dust feed system were attempted with no success.
The system provided a satisfactory feed stream of fly ash, which generally
has better flow properties than either the BOF or cement kiln dust.
Since the completion of this project, the dust feeder has been replaced
by a system of more sophisticated design.
11
-------
3.0 OPERATING PROCEDURES FOR THE PARTICLE SIZING INSTRUMENTS
The seven particle sizing instruments listed in Table 2 were tested
during this study. Specifications of these instruments are summarized
in Table 3. The operating and sample analysis procedures utilized for
each instrument are described in the following paragraphs.
3.1 BRINK IMPACTOR
The modified Brink impactor used by EPA is a low flow rate impactor
consisting of seven stages numbered 0 through 6. In this study, a
cyclone precollector designed and fabricated by Southern Research
Institute was also utilized.
An aluminum foil substrate, coated with a 20 percent solution of
Dow Corning High Vacuum Grease in benzene, was used on each of the seven
stages of the cascade. The aluminum foil substrates were initially
fitted to the shape of the collection plates. An eye dropper was then
used to place two or three drops of the benzene-grease solution on each
foil. The foils were baked overnight at 110°C and desiccated for at
least 12 hours prior to weighing. Tare weights were also determined for
an aluminum foil cup prepared for the cyclone catch and a backup filter
used to collect the material that passed through the last impaction
stage. All of the foils and filters were handled with tweezers to avoid
contamination.
Before each run, the Brink was assembled from the bottom up; first
placing a teflon washer and a support filter disc behind the actual
filter, then following with the 6 to 0 stages in series, and finally
placing the cyclone and sample nozzle in position. Because of the high
gas velocities, it was necessary to use nozzles of 1.2 and 1.7 millimeter
diameter at the low and high concentrations, respectively, to achieve
isokinetic sampling. These nozzle diameters are smaller than the
recommended 2 millimeter minimum, but no experimental problems attributed
to the nozzle diameter were identified during the data analysis.
To take a sample, the Brink was placed horizontally in the chosen
port, taking special care not to bump the Brink against the side of the
wind tunnel and thus dislodge the aluminum foil substrates. A vacuum
12
-------
Table 2. PARTICLE SIZING INSTRUMENTS TESTED
Instrument
Company
Status
co
Brink Impactor
Andersen Impactor
Series Cyclones
PILLS IV
In-Stack Beta Impactor
Piezoelectric Microbalance
Impactor
Electrical Aerosol
Size Analyzer
Monsanto Envirochem, Inc.
St. Louis, Mo.
Andersen 2000, Inc.
Atlanta, Ga.
Southern Research Institute
Birmingham, Ala.
Environmental Systems Corp.
Knoxville, Tenn.
GCA Corp.
Bedford, Mass.
Celesco, Inc.
Irvine, Ca.
Thermo-Systems, Inc.
St.Paul, Minn.
Commercially Available
Commercially Available
Research and Development
Commercially Available
Research and Development
Commercially Available
Commercially Available
-------
Table 3. INSTRUMENT SPECIFICATIONS
Instrument
Operating Principle
Mode of Operation
Size Range (um)
(Unit density)
Number cf Size
Intervals
Mass Concentration
Range (g/m3)
Nominal Instrument
Flow Rate U/m!n)
Data Rate
Approximate Cost
Brink Andersen
Impactor Impactor
Manual Impactor Manual Impactor
with Cyclone with Cyclone
In Stack In Stack
<0.4 to >14 <0.4 to >8.3
9 10
0.1 to 6 0.02 to 3
1.4 20
Manual Manual
$2.5K $3K
Southern
Series
Cyclone
Manual
Cyclone
In Stack
<0.7 to>3.5
4
Q.I to 25
20
Manual
R&D
ESC
PILLS IV
Single Particle
Dual angle
Light
Scattering
In Stack
0.3 to 3.0
10
103 to 106*
NA
Batch
{5-60 min)
$35K
GCA
Beta
Impactor
Impactor with
Beta Detection
of Mass
In Stack
0.3 to >6.5
7
0.3 to 20
9
Real time
(5 min lag)
R&O
Celesco
Piezoelectric
Impactor
Impactor with
Piezoelectric
Detection of
Mass
Extractive
0.09 to >35
10
50 x 10"6
to 0.08
0.2
Batch
(10-15 min)
$15K
Thermo-Sys terns
Model 3030 Electrical
Aerosol Size Analyzer
Electrostatic Charging
and Mobility Analysis
Extractive
0.0032 to 1.0
10
10"6 to 10"3
4
Near real time
('2 min scan)
$12K
*Particles/cc
-------
pump was used to draw the sample and a calibrated orifice was used to
determine the sample flow rate. Sample rates of 1.25 and 1.86 x,/min
were utilized at the high and low concentrations, respectively. During
each run, the sample rate was held constant by manual adjustment of a
needle valve as necessary. Sample times were 90 minutes for the low
concentration runs and 30 minutes for the high concentration runs.
After each test run the Brink was carefully removed from the duct
and returned to the laboratory. Disassembly proceeded from the cyclone
down. All materials from the cyclone and nozzle were carefully brushed
into the tared aluminum foil cup. The aluminum foil substrates and
backup filter were then removed sequentially and placed in a desiccator
along with the cyclone cup. The samples were desiccated overnight and
weighed to a precision of 1 microgram on a Cahn Model G2 electrobalance.
The cyclone generally contained 80 to 90 percent of the total mass.
Stage weights ranged from approximately 10 micrograms to 10 milligrams.
Occasionally negative weights were observed on the number 5 and 6 stages
and the filter. These anomalous results were attributed to handling
problems and weighing inaccuracies and were deleted prior to data reduction.
In several of the early runs, unusually high and inconsistent
cyclone catches were noted. This problem was attributed to leakage
detected around the nozzle flange of the cyclone. Apparently large
particles impacted on the surface of the cyclone were being sucked in
through the leak. A gasket was fitted between the flange and cyclone
(and one also between the two lower parts of the cyclone as a precaution),
and the problem was corrected. Data from the previous runs were discarded.
3.2 ANDERSEN IMPACTOR
The Andersen is a high flow rate impactor containing 8 stages in
series with an optional cyclone precollector and backup filter. The
Andersen uses preformed glass fiber filter substrates purchased from the
manufacturer. Handling of the substrates was simplified by cutting
aluminum foil squares slightly larger than the substrates and inscribing
identifying numbers. Tweezers and surgical gloves were used to prevent
contamination of the foils or substrates.
Each substrate was desiccated overnight and weighed as a unit with
its foil square. An aluminum foil cup was also cut, desiccated, and
weighed to receive the cyclone catch. The Andersen was carefully assembled
from the filter up according to the manufacturer's instructions. Nozzles
15
-------
of 3.0 to 7.0 millimeters were employed with the cyclone precollector to
allow isokinetic sampling.
For each run, the Andersen assembly was horizontally positioned in
the appropriate port. The sample was drawn with a vacuum pump and a
calibrated orifice and needle valve were utilized to maintain an iso-
kinetic sample rate. Sample rates of 3.9 and 21.1 £/min with a sample
time of 30 minutes were used for high concentration runs, and sample
rates of 15.8 and 20.7 &/min with a sample time of 90 minutes were used
for low concentration runs. As discussed further in sections 5.0 and
6.0, the higher sample rates yielded more reproducible results in each
case.
When each test run was completed, the Andersen was returned to the
laboratory and disassembled from the cyclone down. The material in the
nozzle and cyclone was carefully brushed into the tared aluminum foil
cup. The substrates were then removed and placed on the appropriate
foil square using tweezers and gloves to minimize loss or contamination
of the samples. Difficulty was consistently experienced with the filters
sticking to the stainless steel 0-rings and plates. Occasionally
particulate was also found on the steel plates rather than the substrates.
In all of these cases, the material was carefully brushed or scraped
onto the appropriate foil square. The squares were then loosely folded
over the substrates to prevent spillage, and the samples were desiccated
overnight.
Weighing was accomplished on a Mettler balance to a precision of
0.1 milligram. The cyclone generally contained about 90 percent of the
total mass. Stage weights ranged from less than 1 milligram to about 40
milligrams. The low and occasionally negative weights observed on the
last three stages and filter resulted in some loss of consistency at the
smaller particle sizes. As noted above, particulate scouring and spillage
on the steel plates were occasionally observed with the highest stage
weights. Obviously, in any sampling situation a compromise must be
reached between overloading of the upper stages and a reduction of
weighing accuracy on the lower stages. Use of the Cahn electrobalance
can partially solve this problem, but it is extremely difficult because
of the large size of the Andersen substrates. A balance of comparable
precision to the Cahn but larger weighing pans would remedy some of
these difficulties.
16
-------
3.3 SOUTHERN SERIES CYCLONES
The series cyclone system tested during this study was developed by
Southern Research Institute to satisfy the specific objectives of
allowing longer sample times in high dust load situations and of allowing
larger quantities of size-fractionated particulate to be collected for
chemical analysis. The system consists of three cyclones in series,
each with a different cut point, and a Gelman backup filter. No internal
substrates are required. To prepare this system, the cyclones are
simply assembled with a tared filter in the filter holder and a nozzle
fitted to the inlet of the largest cyclone. The cyclones were designed
for a nominal flow rate of 1 actual cubic foot per minute. During these
tests they were utilized with the same sampling system as the Andersen
Impactor. For high concentration runs, 30 minute samples were drawn at
a flow rate of 21.1 &/min (0.746 acfm) using a 7 mm nozzle. Low con-
centration samples were obtained in a 90 minute sample period using a
rate of 20.7 &/min (0.730 acfm) and a 4.0 mm nozzle. The cyclones had
a vertical orientation while they were in the wind tunnel.
When each test run was completed, the cyclones were returned to the
laboratory and disassembled. The material in the nozzle and first
cyclone was carefully brushed into a tared aluminum foil cup. Similar
cups were used to receive the catch from the second and third cyclones
and the filter. The samples were then desiccated overnight. A Mettler
balance was used to weigh the particulate with a precision of 0.1
milligram. The first cyclone generally contained about 98 percent of
the total particulate catch, leaving only a few milligrams for the
second and third cyclones and filter.
3.4 ENVIRONMENTAL SYSTEMS CORPORATION PILLS IV
The PILLS IV is an optical, single particle counter. Particles are
counted and sized as they pass through a small viewing volume through
which a laser is focused. Light is scattered by each particle into two
cones at near-forward angles. The light is collected and the size of
particles greater than 0.6 micrometer is determined from a ratio of the
collected light intensities. Small angle scatter is used directly to
determine the size of particles smaller than 0.6 micrometers. For these
tests, the PILLS IV was set up to yield particle counts in 10 increments
from 0.3 to 3.0 micrometers.
17
-------
The PILLS IV operates with an in-stack probe, approximately 10 cm
in diameter and 2 m long, which contains the viewing volume. The optics
are protected by a purge air system. At the beginning of each run, the
limits were set and checked and the probe was inserted. The "count
enable" switch was then manually thrown. The length of time over which
particles were counted was also manually controlled. At the end of the
count, the number of particles counted in each size increment was printed
out on tape. The number of laser pulses was also printed to give a
precise time reference. During the low concentration runs a minimum
count time of one hour was required to insure that at least 5 to 10
particles were counted in each size increment.
3.5 GCA IN-STACK BETA IMPACTOR
The GCA In-Stack Beta Impactor consists of seven cascade stages and
operates on the same classification principle as the Brink and Andersen
impactors. Unlike the Brink and Andersen, the GCA is a real-time
instrument, and manual weighing of the collected particulate is not
necessary. Instead the collection substrate of each stage is a Mylar
film coated with petroleum jelly. In operation the film rolls at a
predetermined rate between two cassettes, passing under the stage jet.
On each side of the stage jet the film passes between a beta source and
a Geiger-Mueller tube. The increased beta attenuation observed on the
dirty side of the stage jet is, in principle, proportional to the mass
of particulate collected. The difference in upstream and downstream
beta attenuation for each stage is compared internally and displayed on
a strip chart. Periodically, by manual control, a reverse flow of air
is directed through the cascade to prohibit entry of particulate and
allow for zero check and adjustment.
Since the Beta Impactor is a recently developed instrument and had
not been tested previously, time was devoted to establishing an operating
procedure, studying the zero fluctuations apparently caused by in-
consistencies in the film or petroleum jelly coating, and adjusting the
sensitivity of the output of each stage to give a reasonable response.
It was not possible to operate the Beta Impactor at the low concentration
used for the other instruments because of the sensitivity limitation or
to accomplish port switching since an oversized port was required.
18
-------
Comparative runs were made simultaneously with the Brink Impactor at the
nominal high concentration and at a lower concentration three times that
of the nominal low concentration runs. (This particular concentration
was used for convenience since it resulted from a simple combination of
the low dust feed rate and the low velocity used in other runs).
3.6 CELESCO PIEZOELECTRIC MICROBALANCE IMPACTOR
The Celesco Impactor is based on the same particle classification
principle as the Brink and Andersen. In this instrument, however, the
collection substrate of each stage is a piezoelectric quartz crystal.
The mass collected on each stage is determined by the vibration frequency
shift observed in the crystal (See electronic schematic in Figure 4).
The Celesco Impactor was developed for low concentration particle
sizing. Without dilution it has a nominal operating range of 50 to
3
6,000 yg/m total mass concentration. At higher concentrations, the
crystals will overload quickly invalidating the linear relationship
between frequency shift and mass. A dilution system could be used to
control mass concentration, but one was not available for this test
program. Instead, for this first attempt to utilize the Celesco at a
3
"clean stack concentration" of 0.089 g/Nm , the sample time was reduced
to 5 seconds. No attempt was made to sample at the higher concentration.
Extractive samples were drawn through a 1/4-inch O.D. tube approximately
40 cm long. A 1 mm nozzle was used since it was the smallest available.
(For the Celesco's standard 200 m£/min sample flow, a 0.4 mm nozzle
would have been required for isokinetic sampling.)
At the beginning of each test, a filtered air flow of 200 nu/min
was started through the Celesco using its integral pump and a detachable
rotameter (See flow schematic in Figure 4). At the end of a one-hour
warmup, the flow was adjusted back to 200 nU/min, and repetitive crystal
scans were run to determine the baseline frequency of each crystal.
When all crystals had stabilized (frequency drift less than 10 Hz in a
10 minute interval), the sample was drawn. A manually operated, quick-
opening, three-way valve was used to switch from the filter purge to the
sample mode. At the end of seven seconds, the valve was switched back
to purge. To allow for the dead space in the sample line, the sample
time was treated in the calculations as if it were 5 seconds rather than
7 seconds.
19
-------
SENSING MODULE
CONTROL MODULE
ro
o
SENSING
CRYSTAL
REFERENCE
CRYSTAL
I
1
SENSING
OSCILLATOR
REFERENCE
OSCILLATOR
Id 1
MIXER 1* FREQ' T° RATE
MixtK p> VOLTAGE
1 1
OUTPUTS 66 6
THREE-WAY
VALVE
SAMPLE
FILTER
IMPACTION
JET
AIR
FREQ. SATURATION
ELECTRONIC SCHEMATIC
CONCENTRATION
EXHAUST
CONTROL
VALVE
REFERENCE XTAL.
SENSING XTAL.
FLOW SCHEMATIC
BLOWER
Figure 4. Electronic and flow schematics of Celesco impactor.
-------
After the sample was drawn, a repetitive scan of the crystal
frequency was initiated and continued until the drift criterion was again
achieved. The frequency shift in each crystal was then recorded, and a
check was made to ensure that no crystal had exceeded the manufacturer's
recommended allowable frequency shift. When any crystal approached the
allowable shift after several runs, it was replaced with a fresh crystal.
3.7 THERMO-SYSTEMS MODEL 3030 ELECTRICAL AEROSOL SIZE ANALYZER
The Thermo-Systems Electrical Aerosol Size Analyzer operates on the
principle of electrostatic charging and mobility analysis to determine
particle size. Particles may be classified in ten size increments
between 0.0032 and 1.0 micrometers. The instrument is primarily for
laboratory and ambient use with a nominal concentration range of 1 to
3
1,000 yg/m when used with typical atmospheric aerosols. Since the
redispersed fly ash utilized in this study was expected to have a
smaller fraction of small particles than a normal atmospheric aerosol,
operation of the EASA without dilution was attempted.
A sample tube was affixed to the aerosol inlet for extractive
sampling, ignoring isokinetic conditions since no particles greater than
1 micrometer could be detected by the instrument. During several days of
testing the results were quite erratic and the digital readout was
always of negative sign. These problems were described to a manufacturer's
representative by telephone and he suggested that the particles were
probably highly negatively charged. A 10 millicurie Krypton-85 source
was incorporated in the sample line, but no appreciable change in the
results could be determined. The instrument was removed from the wind
tunnel to a laboratory where ambient air samples were drawn and normal
results were obtained. Since the erroneous results thus seemed to be
related to the condition of the particles rather than a malfunction of
the instrument and no other particle conditioning equipment was avail-
able, operation of the EASA was discontinued.
-------
4.0 DATA REDUCTION
Several techniques can be used to analyze and present particle size
distribution data obtained with cascade impactors and other sizing
instruments. The method utilized in this project is frequently used by
EPA and involves the presentation of data as differential particle size
distributions using the D™ concept. The standard calculations and all
exceptions applied in this project are described in the following para-
graphs. Further information can be obtained in Reference 2.
4.1 DIFFERENTIAL PARTICLE SIZE DISTRIBUTIONS^ METHOD
The DCQ method is presently used for the majority of cascade impactor
data reduction. It is also applicable to cyclones. The method is
fairly straightforward and can be hand-calculated, but it is recognized
as a considerably simplified picture of the real distribution and can
cause some loss of information (ref.2).
The D5Q of a stage in a cascade impactor or of a cyclone is the
particle diameter at which the device achieves 50 percent efficiency;
one half of the particles of that diameter are captured and one half are
not. Larger particles are captured with greater than 50 percent efficiency,
and smaller particles with less than 50 percent efficiency. The DSQ at
a particular set of conditions is determined by experimental tests or
from theoretical and semi-empirical equations based on previous experimental
work. The latter approach was used in this work.
The DCQ reported can be either aerodynamic diameter (that is,
diameter based on the behavior of unit density particles) or approximate
physical diameter, which is based on an estimate of the true particle
density. In either case, the particles are assumed to be spherical. In
this project all calculations were based on an assumed fly ash particle
3
density of 2.5 g/cm .
The Dr« analysis method simplifies calculations by assuming that a
given stage captures all of the particles with a diameter equal to or
greater than the D™ of that stage and less than the D™ of the preceding
stage. For the first stage (or cyclone), it is assumed that all of the
particles caught have diameters greater than, or equal to, the DrQ for
that stage (or cyclone), but less than the maximum particle size. If
the maximum particle size is not known, some arbitrary large value is
used. In these tests a maximum of 100 ym was assumed.
22
-------
With these simplifications, the mass collected on a given stage is
assigned to a particular diameter; usually the geometric mean of the
stage DrQ and the preceding stage DrQ is used.
If the true particle-size distribution constituted a continuum, the
amount of material having diameters between D and D+dD could be represented
by dM. Then the integral
would yield the total mass having diameters between D-j and $2- In tnis
integral the term dM/dD is referred to as the differential particle size
distribution.
Because the intervals between the stage DCQ'S are in most instruments
logarithmically related, the differential particle size distribution is
normally calculated as dM/d(log D) rather than dM/dD. This modification
also minimizes scaling problems when the data are plotted in the usual
manner on log-log or semi-log paper with dM/d(log D) as the ordinate and
log D as the abscissa.
To calculate the differential distribution, the mass on stage "n"
is designated by AMn and is, in approximation, the mass of particulate
with diameters between (Dcg)n and (D5Q)n+-,. Tne Mlog °) associated
with AM is log (D50)n+-, - log (DcnL. Using these approximations, the
derivative term associated with stage "n" is
"n II
AM Mass on Stage "n
«/d(l.,D)V a(logD")| ' 1og(D > . log(n )
50n
The diameter corresponding to stage "n" is
T I1/2
Dgeo = [n ' ^B
23
-------
Plotting this approximation of dM/d (log D) versus log D
results in a histogram. If an impactor with an infinite number of
stages were available, the histogram would approach a continous curve of
differential distribution, dM/d(log D), as a function of particle
diameter. Such an impactor does not exist, but the histogram is usually
plotted as a smooth curve by connecting the available data points. This
curve is then a continuous function approximating the actual particle
size distribution. The accuracy of the approximation is limited by the
number of points, and by the basic inaccuracy of neglecting the non-
ideal behavior of the impactors, especially overlapping collection
efficiencies for adjacent stages or cyclones.
Despite its limitations, the differential distribution method of
data analysis offers several advantages over cumulative distribution
representations. Experimental errors associated with one stage of the
sizing device are not propagated to other points of the distribution.
Also since the method does not involve the use of total mass concen-
tration or total size distribution from diameters of zero to infinity,
it is especially useful in comparing instruments with overlapping but
different size fractionation ranges and different stage cut points. To
normalize the differences in mass of sample collected by various in-
struments, the mass on each stage is usually divided by the standard
volume of the sample, yielding concentration units, i.e., dC/d(log D) in
g/Nm3.
A sample calculation using Brink Impactor data is presented in the
Appendix of this report.
4.2 BRINK IMPACTOR
To reduce the Brink data, D5Q's of the impactor stages and cyclone
were obtained for the operating conditions described previously from a
computer program which EPA has on file. The calculated Dr^'s for the
various sample rates are shown in Table 4. No exceptions to the general
calculational scheme explained in Section 4.1 and the Appendix were
required.
4.3 ANDERSEN IMPACTOR
To reduce the Andersen data, D5Q's of the impactor stages were
obtained from the EPA computer program. To determine the D5Q of the
cyclone, RTI contacted the manufacturer and was referred to a consultant
24
-------
Table 4. D5Q'S OF BRINK STAGES AMD CYCLONE
Stage
Cyclone
0
1
2
3
4
5
6
D50(ym)
@1.25 £/min
9.39
5.73
3.25
1.93
1.32
0.71
0.46
0.28
D50(um)
01.86 £/min
7.69
4.68
2.65
1.56
1.07
0.56
0.36
0.21
Table 5. D^'S OF ANDERSEN STAGES AND CYCLONE
Stage
Cyclone
1
2
3
4
5
6
7
8
D5Q(um)
@21.1 a/min
5.30
7.27
4.50
3.04
2.06
1.32
0.64
0.39
0.24
D5Q(ym)
020.7 A/min
5.35
7.31
4.54
3.07
2.08
1.33
0.65
0.39
0.25
D50(ym)
015.8 j,/nrin
6.08
8.36
5.20
3.52
2.39
1.53
0.75
0.46
0.29
D50(ym)
03.9 £/min
12.3
17.0
10.6
7.20
4.91
3.17
1.60
0.99
0.68
25
-------
(ref.3). The consultant stated that the cyclone has been calibrated
only at a flow rate of 21.2 £/min (0.75 cfm). For other sample rates,
it was assumed that the D5Q was inversely proportional to the square
root of the flow rate. The calculated D50's of the cyclone and stages
for the various sample rates are shown in Table 5.
The design of the Andersen cyclone presents an anomaly in that the
Drg at the calibrated flow rate is less than the D^Q of the first stage
at the same flow. To handle this irregularity, the mass of the cyclone
and first stage were added in the calculations and assigned to the
smaller D™. The consultant concurred with this approach. No other
deviations from the standard calculations shown in the Appendix were
encountered.
4.4 SOUTHERN SERIES CYCLONES
The Drg's of the Series Cyclones were determined from a set of
equations derived by Southern Research Institute from calibration data.
Cyclone 1 D^m^ = 225>1
Cyclone 2 D5Q(ym) = 88.25-^y/pF
Cyclone 3 D5o^ym) = 43.29^ y/pF
where y = gas viscosity, poise
3
p = particle density, g/cm
3
F = sample flow rate, ft /min
No deviations from the standard calculations shown in the Appendix
were necessary.
26
-------
4.5 CELESCO PIEZOELECTRIC MICROBALANCE IMPACTOR
The DCQ'S of the Celesco stages were provided by the manufacturer
for a particle density of 2.0 g/cm . To correct these values to a
3
particle density of 2.5 g/cm , a simple ratio was used, assuming that
DJ-Q is inversely proportional to the square root of particle density.
The DCQ'S of the stages on which data were obtained were as follows:
Stage 4 D5Q = 2.87 ym
Stage 5 D5Q = 1.42 ym
Stage 6 D5Q = 0.71 ym
Stage 7 D = 0.35 ym
To convert from frequency shift on each stage to incremental mass
concentration with the standard sample rate and a five-second sample,
the following equation was used:
C = 1.02 x lo"4 x AF
where C = mass concentration, g/m
AF = frequency shift, Hz
The remainder of the data reduction was similar to the example
shown in the Appendix.
4.6 GCA IN-STACK BETA IMPACTOR
The D5Q' s of the Beta Impactor were provided by GCA for a particle
density of 1.0 g/cm and were corrected to a density of 2.5 using the
same relationship employed with the Celesco. The results were as follows:
Stage 1 D5Q = 4.1 ym
Stage 2 D5Q = 2.5 ym
27
-------
Stage 3 D5Q = 1.5 ym
Stage 4 D5Q = 0.89 ym
Stage 5 D5Q = 0.54 ym
Stage 6 D5Q = 0.33 ym
Stage 7 D5Q = 0.19 ym
Raw data were obtained from the Beta Impactor in the form of seven
pen traces on a strip chart, corresponding to the seven stages. Scale
3
factors were provided by GCA to convert from pen displacement to mg/m .
Since the continuous output of each stage showed considerable fluctuation,
a graphical integration was performed over the sample period to determine
the time-averaged particle concentration in each size increment. Figure
5 is a reproduction of the typical response of two adjacent stages,
illustrating the selection of zero and the graphical integration procedure.
Once the average incremental concentration was determined for each
stage, the remainder of the calculational procedure was similar to that
shown in the Appendix.
4.7 ENVIRONMENTAL SYSTEMS CORPORATION PILLS IV
Since the PILLS IV counts individual particles in a given size
increment rather than collecting an agglomerate of particles for mass
determination, a different technique is required to obtain the differential
particle size distribution in mass concentration terms. The data printout
from the PILLS IV gives the total number of times that the laser fired
and the number of particles counted in each of the ten size increments.
(The sample time can be accurately determined from the number of laser
pulses since the pulse rate Is 1000 times per second.) The incremental
number concentration is then determined by
Ani ni
vs - K-
28
-------
IV)
MEAN CONCENTRATION =
TIME OF INTEGRATION
Figure 5. Determination of mean concentration for stages of the GCA In-Stack Beta Impactor.
-------
An. 3
where —y— = incremental number concentration, particles/m
n. = number of particles in increment i
K = number of laser pulses
V = viewing volume = 2xlo~ m
In each set of calculations the viewing volume was corrected to standard
conditions as shown in the Appendix.
Assuming that all particles were spherical and of equal density,
the incremental mass concentration was then determined by
An.
» 3
where AC. = incremental mass concentration, g/m
3
p = assumed particle density, g/cm
d* = geometric mean of the upper and lower boundaries of
the increment, ym.
A differential particle size distribution equivalent to those
calculated by the procedure described in the Appendix was then calculated
by
AC.
where
3
dC/d(log D) = differential particle size distribution, g/m
Alog(d.) = difference of the common logarithms of the upper
and lower boundaries of the increment.
30
-------
The size increments of the PILLS IV used in this project are presented
in Table 6.
Table 6. SIZE INCREMENTS OF PILLS IV
Channel
5
6
7
8
9
10
11
12
13
14
Size Range, ym
0.35 - 0.38
0.38 - 0.48
0.48 - 0.60
0.60 - 0.75
0.75 - 0.95
0.95 - 1.19
1.19 - 1.50
1.50 - 1.89
1.89 - 2.38
v 2.38 - 3.00
Mean Diameter, ym
0.365
0.42
0.53
0.67
0.85
1.06
1.34
1.69
2.12
2.67
Alogfdj)
0.036
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
31
-------
5.0 COMPARISON OF INSTRUMENTS BASED ON PRIMARY DATA
5.1 PROCEDURE FOR SUCCESSIVE RUNS
As discussed previously in this report, it was originally assumed
that the most valid approach to instrument comparison was to operate the
instruments simultaneously in different ports of the wind tunnel. This
approach required that the paniculate concentration in the wind tunnel
be uniform and that upstream instruments had a negligible effect on the
particle concentration and distribution arriving at downstream sample
points. When the problems of analyzing the data obtained in this manner
became apparent, additional runs were made operating the instruments one
at a time in succession in the same port. Although the remaining
authorized time for project completion allowed only two runs with each
instrument at each concentration, the limited data proved to be of
primary use in comparing the particle sizing instruments.
During the successive runs, frequent checks were made of wind
tunnel operating parameters to ensure steady state operations. The IKOR
was also used in port H as an additional check of steady state conditions.
In general, the IKOR indicated a "steady periodic" dust concentration in
the tunnel during any run, oscillating within about 10 percent of the
mean value with a frequency on the order of one cycle per minute. All
of the instruments with the exception of the GCA, which is discussed in
Section 5.3, were operated in port D with their respective sample
nozzles positioned as closely as possible to the same point. Table 7
shows the variation of sample points among the instruments and the total
mass concentration measured by each.
At both the high and low concentrations the Brink, Andersen and
Series Cyclones all indicated a significantly higher total mass loading
than the homogenous mass loading calculated from the dust feeder calibration.
Wide disparity between dust loadings obtained with mass trains and with
cascade impactors is not uncommon (ref. 4), and is frequently attributed
to stratification of dust in the duct. Stratification is the likely
cause in this case also. At the higher concentration stratification of
the dust streams from the individual feed nozzles could be observed with
a flashlight shining into the sample ports.
32
-------
The total mass concentration indicated by the Brink measurements
differs markedly from the Andersen and Series Cyclone measurements.
Part of this difference might also be attributed to dust stratification
in the wind tunnel since the sample points did not coincide identically.
Another credible explanation is the presence of a relative error in the
flow rate measurements of the Brink sample train and the separate sample
train used for the Andersen and Series Cyclones. These two sample
trains were calibrated with different flow standards since the flow
rates differed by approximately an order of magnitude.
Duplicates were run with all instruments at both concentrations.
From these duplicates a normalized variance of the differential distribution
at each stage was calculated for both the high and low concentrations
according to the following equation.
xi -x
"-' 1-1
where v = normalized variance at a specified stage
n = number of runs
x.j = differential distribution of run i
x = mean differential distribution of all runs, specified stage
In general,a higher variance indicates a lower level of repeatibility.
The variance in selected size intervals is tabulated for each of the
instruments in Tables 8 and 9. These tabulations are useful as an
indication of the relative variance of the instruments, but must be
viewed with some caution since they were calculated from only two
duplicate runs. Part of the variance might also be attributed to wind
tunnel variations and not to instrument performance. The conclusions
from Tables 8 and 9 are included in the following discussion of instrument
comparison.
5.2 COMPARISON OF INSTRUMENTS IN SUCCESSIVE RUNS
Figures 6 and 7 are plots of differential distribution versus
particle size as calculated from the Brink and Andersen data at low and
high concentrations, respectively. The curves on each plot represent a
least square fit to the experimental data of both instruments in the
33
-------
Table 7. POSITION OF INSTRUMENTS IN THE WIND TUNNEL AND MEASURED TOTAL MASS CONCENTRATION
Instrument
Brink
Andersen
Cyclones
PILLS IV
Celesco
Dust feeder
Calibration
Horizontal Position
23 cm into port D
23 cm into port D
23 cm into port D
23 cm into port D
IS cm into port D
Vertical Position
1 .3 cm below
centerline
2.5 cm above
centerline
5.1 cm below
centerline
On centerline
On centerline
"
3
Indicated Total Mass Concentration, g/Nm
Low Runs
0.577 and 0.660
0.376 and 0.386
0.376 and 0.377
--
--
O.OG9
High Runs
1.841 and 1.867
2.200 and 2.384
2.373 and 2.765
--
--
0.955
-------
Table 8. NORMALIZED VARIANCE FOR LOW CONCENTRATION RUNS-PRIMARY DATA
Particle Size
Range, ym
>10
5-10
2-5
1-2
0.5-1
0.2-0.5
Brink
0.004
0.001
0.03
—
0.02
Andersen
0.0004
--
0.002
0.009
0.5
Cyclones
9 x 10"7
--
—
0.0005
<0.005
PILLS IV
—
—
0.04
0.09
0.3
0.4
Celesco
—
--
0.05
0.02
0.1
GCA*
0.02
--
3
1
--
CO
en
*GCA variance calculated from three special comparative runs with Brink at a concentration of 0.267 g/Nnf
-------
Table 9. NORMALIZED VARIANCE FOR HIGH CONCENTRATION RUNS-PRIMARY DATA
Particle Size
Range, ym
>10
5-10
2-5
1-2
0.5-1
0.2-0.5
Brink
0.005
0.5
0.06
0.03
0.02
0.7
Andersen
0.004
--
0.008
0.0003
0.1
3
Cyclones
0.01
--
--
0.02
0.02
—
PILLS IV
__
--
0.0003
0.03
0.02
0.01
Celesco
—
—
--
--
--
--
GCA*
0.004
--
0.03
0.03
0.06
8
GO
*GCA variance calculated from two special comparative runs with Brink at same concentration as
successive runs.
-------
IOV
Id'
10,
'£
o>
u
TJ
Q
.63
.65
O BRINK DATA
A ANDERSEN DATA
O.I
1.0 K)
PARTICLE DIAMETER , Dgeo ,Mm
100
Figure 6.
Comparison and curve fit of Brink and Andersen Impactor
data (concentration = 0.089 g/Nnr).
37
-------
10
ro
E
I
CP
o
u
•o
5
<
F
o
16*
O BRINK DATA
A ANDERSEN DATA
0.1
1.0 10
PARTICLE DIAMETER , Dgeo ,/xm
100
Figure 7.
Comparison and curve fit of Brink and Andersen Impactor
data (concentration = 0.955 g/Nm3).
38
-------
form
log[dC/d(log D)] = aQ + a^log DgeQ) + a2 (log DgeQ)2
where aQ, a-| and a2 are constants determined by the least square procedure.
The Brink and Andersen data were treated together because these
instruments have been shown by EPA to yield comparable particle size
distributions in past experiments. The agreement was good in these
tests also. The figure "(2)" beside a data point indicates that two
individual runs yielded data points so close together as to be indistinguish-
able on the plot.
The Brink Impactor showed a large variation in the results obtained
with the zero stage at high concentration. This discrepancy was noticed
throughout the test (see also Section 6). The effect apparently results
from a difference in the shape of the particle collection efficiency
curves of the cyclone precollector and the zero stage. The cyclone
efficiency curve overlaps with the zero stage curve (ref. 2), which
leads to a particle collection phenomenon not well handled by the DSQ
data reduction technique. The effect is apparently more pronounced at
the lower sample rate used with the high concentration results. The
Andersen runs shown in these figures were made with sample flow rates
within 3 percent of the calibration flow rate of the cyclone precollector.
The variance of the Andersen (Tables 8 and 9) appears to be slightly
less than that of the Brink.
In Figures 8 through 12 the experimental data obtained with the
Series Cyclones, Celesco and PILLS IV are compared with the appropriate
least square curve of the Brink and Andersen data. Figures 8 and 9 show
this comparison for the Series Cyclones at the low and high concentrations
respectively. The Series Cyclone data are in good agreement with the
Brink and Andersen curves. At both the low and high concentrations the
cyclone data are actually within the scatter band exhibited by the Brink
and Andersen data used to calculate the least square fit. At the high
concentration, the cyclone seems to have a slight tendency toward
higher readings relative to the Brink and Andersen. This effect would
be in agreement with the usual observation that cyclone collection
efficiency increases with increasing dust load, but a definite con-
clusion would be presumptious due to the limited amount of data in this
39
-------
ic
Id1
id2
o
•o
Q
I io3
10
Id5
O (2)
O SERIES CYCLONE DATA
0.1
1.0
K>
100
Figure 8.
f'ARTlCLE DIAMETER , Dgeo ,
Comparison of Southern Series Cyclone data with
impactor curve (concentration = 0.089 g/Nm3).
40
-------
10'
K>v
e
0»
O
8*
^ id1
u.
O
.63
O SERIES CYCUDNE DATA
0.1
1.0
10
100
Fi gure 9..
RARTICLE DIAMETER , Dgeo ,
Comparison of Southern Series Cyclone data^with
impactor curve (concentration = 0.955
41
-------
iow
10
8
6
Irf
33
i
O CELESCO DATA
10
0.1
1.0
10
100
Figure TO-
PARTICLE DIAMETER , Dgeo
Comparison of Celesco Piezoelectric Microbalance Impactor
with impactor curve (concentration = 0.089 g/Nm3).
42
-------
case. The variance of the Series Cyclone data is comparable to that of
the Brink at the high concentration and is considerably lower than the
variance of the Brink or Andersen at the lower concentrations.
Figure 10 compares the Celesco data with the Brink/Andersen curve
fit at low concentration. Data were obtained only from the fourth,
fifth, sixth and seventh stages of the ten stage cascade in the Celesco
instrument. The loss of data at the top and bottom of the cascade was
not unexpected with the given conditions and sampling apparatus. The
first three stages correspond to D5 's of 22.3, 11.5, and 5.6 microns,
respectively. With a lag time of 2 seconds in the 1/4-inch extractive
sample line, a simple settling velocity calculation shows that no particles
of 5.6 microns or greater would ever reach the cascade. The same cal-
culation partially explains the low tendency of the fourth stage data
(DCQ =2.9 micrometers). Data were lost from the last three stages of
the cascade because of the short sample time. The sample time was limited
to 5 seconds to prevent overloading of the middle stages of the cascade.
With this short sample time and the given particle size distribution,
one would expect a frequency shift on the eight stage of less than 3 Hz.
Since the drift criterion was 10 Hz or less in ten minutes, a 3 Hz shift
could obviously not be detected.
The data points corresponding to the fifth, sixth and seventh
stages of the cascade are significantly.higher than the Brink/Andersen
curve but show a similar slope. At least three experimental factors may
have bearing on this effect.
1) Anisokinetic sampling was necessary because a sample nozzle
small enough to allow isokinetic sampling was not available. The
slow sampling would tend to oversample larger particles and shift
the distributions upward. However, this effect is probably not too
significant since particles of approximately 1 micron diameter and
smaller are usually sampled correctly regardless of the anisokinetic
conditions (ref. 5).
2) Sample time and rate were determined with a stop watch and
rotameter supplied by the instrument manufacturer. Errors in
either or both of these measurements could bias the differential
distribution curve, but the combined error would have to be on the
order of 100 percent to fully account for the discrepancy between
the Celesco data and the Brink, Andersen curve.
43
-------
3) As shown in Table 7, the Celesco sample point was several
cm away from the Brink and Andersen sample points, and thus dust
stratification in the tunnel could have contributed to the error.
The variance of the data that were obtained with the Celesco is
slightly higher than the variance of the Brink Impactor data.
Figures 11 and 12 compare data obtained with the PILLS IV to the
Brink/Andersen curves. The pronounced disagreement was not unexpected
since the instruments operate on completely different theoretical
principles. Coincidence of the PILLS IV data and the Brink/Andersen
curves would only be expected if all of the particles were spherical and
of the same density. There is no known explanation for the indicated
increase in the differential distribution as the particle diameter
decreases from about 0.5 microns. The variance of the PILLS IV data
appears to be slightly less than that of the Brink data.
5.3 COMPARISON OF THE GCA IN-STACK BETA IMPACTOR WITH THE BRINK IMPACTOR
Special tests were arranged for the GCA In-Stack Beta Impactor so
that it could be operated and adjusted by a GCA engineer. To obtain a
comparison with the Brink Impactor, the two instruments were operated
simultaneously with the Brink in port D and the GCA in port F. The GCA
instrument did not show sufficient sensitivity to use the low concentration
used with the other instruments. Therefore runs were made at a medium
concentration three times the normal low concentration and at the normal
high concentration. The results of these two comparisons are shown in
Figures 13 and 14. In both figures a least-square curve of the same
form used previously has been calculated from the Brink data and the GCA
data are plotted for comparison.
At both concentrations, the data points corresponding to the first
stage of the GCA Impactor are more than an order of magnitude lower than
the Brink curve and appear to be inconsistent with the remainder of the
GCA data. This effect was probably caused in part by the heavy dust
accumulations observed on the inside walls of the upper stages and the
inlet nozzle and cone of the GCA probe. Accumulations of dust were also
noted on the interior walls of the lower stages, but were not nearly as
pronounced as at the inlet. Furthermore, the graphical integration for
first stage was biased by the occasional pegging of the chart pen (see
Figure 5).
44
-------
Id1
10
6
IT
•* -2
O
T3
1
i
0>
a
-------
10'
10
"fe
z
O(2)
C 10
V.
V
»
1
i
E
5
_j
I
u.
o
8
I06
8
.6*
o
o
O PILLS IV DATA
8
'0(2)
8
10
0.1
Figure 12.
1.0 10
PARTICLE DIAMETER, Dgeo ,
100
Comparison of Environmental Systems Corporation PILLS IV
data with impactor curve (concentration = 0.955 g/Nm3).
46
-------
10'
10°
ro
o»
o"
5»-'
O
•o
1
£
< -2
i 10
g
,63
o
o
O 6CA DATA
0.1
1.0
10
100
PARTICLE DIAMETER , Dgeo
Figure 13.
Comparison of GCA In-Stack Beta Impactor data with
corresponding Brink curve (concentration = 0.267 g/Nnr)
47
-------
10'
.0°
1:
z
o
•o
UJ
fc
o
,63
IQ.
O 6CA DATA
O.f
1.0 10
PARTICLE DIAMETER , Dgco.^m
100
Figure 14.
Comparison of GCA In-Stack Beta Impactor data with 3
corresponding Brink curve (concentration = 0.955 g/Nm ),
48
-------
At the high concentration the remainder of the GCA data falls in a
line roughly parallel to the Brink curve. The differential distribution
indicated by the GCA is roughly 30 to 50 percent lower than the Brink
curve at corresponding particle diameters. Part of this discrepancy can
certainly be attributed to the obstruction effect and to the dust
stratification in the wind tunnel. The Brink sample nozzle was inserted
23 cm into port D (corresponding to the "-3" position in Figure 2),
while the GCA sample nozzle was at the center of the port and approximately
2 m downstream from the Brink. Figure 2 indicates that a 50 percent
variation in mass concentration, which would lead to a comparable variation
in the differential distribution curve, could possibly be attributed to
the stratification and obstruction effects alone. Additional tests will
have to be run under more ideal conditions to determine if the scale
factors used to convert beta attenuation to particle mass also need
correction.
At the high concentration the variance of the GCA data is comparable
to that of the Brink data for particle diameters greater than 1 micron,
but increases sharply on the lower stages. At the medium concentration,
the variance is consistently greater than any of the other instruments.
The fluctuation of the GCA response with time has already been
illustrated in Figure 5. Since no other instrument except the IKOR
operated in real time, it could not be determined whether or not these
fluctuations reflected actual time variations in the total mass concen-
tration of particle size distribution. The IKOR did show oscillations
of smaller magnitude than the GCA, but the GCA also showed significant
fluctuations in the zero, which would tend to indicate an inconsistency
in the grease coating of the substrate or in the substrate itself.
49
-------
6.0 RESULTS OF SIMULTANEOUS RUNS
In the early part of the experimental program, tests runs were made
with several instruments operating simultaneously in different ports of
the wind tunnel. At each operating condition, runs were repeated until
each instrument was tested twice in each port in which it could be
operated. The PILLS IV instrument was operated only in ports D and F
because of the physical size of the probe. All of the data which were
recoverable from these simultaneous runs are presented in Figures 15
through 23, with appropriate port identifications. In each figure the
curve drawn is a second-order, least-square fit, not of the data shown,
but of the data obtained under equivalent conditions with the same
instrument during the successive runs; i.e., the data presented in
Figures 6 through 12.
Analysis of these figures reveals four important effects which
support and extend the conclusions drawn from the primary data and
profile studies.
1) With each instrument, excluding the Andersen Impactor
which is discussed below, the data are in reasonable agreement
with the least-square fit of the successive test run results.
This agreement tends to support the conclusions drawn from the
limited number of primary data runs.
2) The spread of the experimental data at individual particle
diameter is much greater than that obtained in the successive
runs, indicating port variations and the detrimental effect of
instruments interfering with other instruments downstream.
3) Although it does not appear in each of the figures, there
is a definite overall trend of decreasing particle concentration
as one moves downstream in the wind tunnel; i.e., the differential
distribution measured at port B > ports C and D > ports E and F.
This trend is another indication of the interference and dust
stratification created by upstream instruments in relation to
downstream sampling ports. It might also indicate particle
50
-------
10
10°
IO
Z
X.
o>
2
5
«
1
25
<
16'
O PORT B
D PORT C
A PORT D
-------
10'
8*
o
•o
.6'
to
S
i
.63
10
O PORT B
D TORT C
A PORT D
tf PORT E
x PORT F
a
O.I
Figure 16.
1.0 10
PARTICLE DIAMETER, Dgeo,/*m
100
Comparison of simultaneous Brink data with^primary Brink
data curve fit (concentration = 0.955 g/Nm3).
52
-------
10
.0°
io
E
o»
o
2
v r
5
3
$
10
a
O PORT B
D PORT C
& PORT D
-------
10
10
"e
*>.
a>
••
s
$
2
16'
i
id*
lO
10
O PORT B
O PORT C
PORT D
PORT E
x PORT F
O.I
1.0
10
100
Figure 18.
PARTICLE DIAMETER, Dgeo,Mm
Comparison of simultaneous Andersen data with primary
Andersen data curve fit (concentration = 0.955 g/Nm3).
54
-------
10
to
z
o
I
o
Q
m
1
Q
I
Ul
10"-
10
.62
10
-4
10
Ql
1.0
PARTICLE DIAMETER, Dgeo ,
10
100
Figure 19. Comparison of simultaneous Southern Series Cyclone data
with primary Cyclone data curve fit (concentration = 0.089 g/Nrn3)
55
-------
10
10*
"l
O
•o
O
I
cc
10
io2
O PORT B
D PORT C
& PORT D
tf PORT E
x PORT F
UJ
U.
U.
5
.63
-4
10
0.1
Figure 20.
1.0 10
PARTICLE DIAMETER, Dgeo ,
100
Comparison of simultaneous Southern Series Cyclone data with
primary Cyclone data curve fit (concentration = 0.955 g/Nm3).
56
-------
10
S
-------
10"
10
£
io
03
I
<
I-
i
u.
u.
6
A PORT D
x PORT F
10*
x
A x
A
A*A
io8
0.1
1.0 10
PARTICLE DIAMETER , Dgeo ,
Figure 22.
100
Comparison of simultaneous PILLS IV data with primary
PILLS IV data curve fit (concentration = 0.089 g/Nm3).
58
-------
10°
-I
10
E
z
•v
o>
~ 10*
•o
o
TJ
O
i
or
5
a
< id3
H
u.
o
•of
A PORT 0
X PORT F
0.1 1.0 10 100
PARTICLE DIAMETER, Dgeo,
Figure 23. Comparison of simultaneous PILLS IV data with primary
PILLS IV data curve fit (concentration = 0.955 g/Nm3).
59
-------
settling. Terminal velocity calculations indicate that settling
could have been significant for particles of greater than 50 microns
diameter, particularly in the high concentration—low velocity
runs. No experimental evaluation of vertical stratification was
attempted.
4) Figure 18 for the Andersen Impactor at high concentration
exhibits a flattening of the differential distribution curve.
This effect is believed to be related to the sample flow rate
of 3.9 £/min used in these early runs. This sample rate is much
lower than the calibration flow rate of the cyclone precollector,
and it is likely that the cyclone efficiency curve at this flow
rate is distorted to such an extent that it interferes with the
behavior of several stages in the cascade. The data shown in
Figure 17 at low concentration were obtained at a flow rate of
15.8 £/min and the effect is not as pronounced although considerable
scatter is evident. There was no explanation for the extremely
low results obtained in the port B test runs in this particular
case.
60
-------
APPENDIX
CALCULATION OF DIFFERENTIAL SIZE DISTRIBUTION
The procedure for reducing the data from stage weights to a
differential distribution is illustrated in the following sequence of
equations. Table A-l is a table of reduced data taken from a run with
the Brink Impactor. These data are plotted in Figure 7.
STEP 1: Calculate the standard volume of the sample.
VN
\i
" Vm
[" Pa-APT-
760
APm
273
Tm
1
1
+
y
3
where V = standard volume, Nm
3
V = measured volume, m
P, = atmospheric pressure, mm Hg
a
AP-, = pressure drop from atmosphere to the tunnel, mm Hg
AP = pressure drop from tunnel to the flow measuring device
m mm Hg
T = temperature of the flow measuring device, °K
y = volume fraction of water at the flow measuring device
STEP 2: Calculate the incremental concentration on the stage.
Mi -3
AC. = —~- x 10 J
1 VN
3
AC- = incremental concentration on stage i, g/Nm
M.J = mass collected on stage i, mg
61
-------
STEP 3: Calculate the difference in logarithms of the D5Q's of two
adjacent stages.
(Alog D5Q)1 = (log D50).+1 - (log D5Q)i
where D™ is expressed in pm.
STEP 4: Calculate the geometric mean of two adjacent stages.
1/2
where (DQeo)i is the particle diameter associated with the mass
collected on stage i, expressed in ym.
STEP 5: Calculate the point on the differential distribution corresponding
to
-------
Table A-l. REDUCED DATA TABLE FOR BRINK RUN NO. 905-2
Stage
Arbitrary
maximum
Cyclone
0
1
2
3
4
5
6
Filter
D50' ym
100
9.39
5.73
3.25
1.93
1.32
0.71
0.46
0.28
— —
logD50
2.0
0.973
0.758
0.512
0.286
0.121
-0.149
-0.337
-0.553
— —
AlogD50
—
1.027
0.215
0.246
0.226
0.165
0.270
0.188
0.216
— —
Dgeo' ym
--
30.6
7.34
4.32
2.50
1.60
0.97
0.57
0.36
~ —
M, mg
--
47.840
6.336
3.848
1.992
0.576
0.296
0.056
0.008
0.280
ACr g/Nm3
--
1.454
0.193
0.117
0.0605
0.0175
0.0090
0.0017
0.00024
~ •"
dC/d(logD)
--
1.416
0.896
0.475
0.268
0.106
0.0333
0.0090
0.0011
~ ~
OJ
Additional Data
Pa = 765.y mm Hg
APT = 1.0 mm Hg
AP_ = 17.8 mm Hg
= 300°K
m
y = 0.019 (54% relative humidity)
Vm = 0.0374 m3
... VN = 0.0329 Mm3
m
-------
REFERENCES
1. Blann, D.D., K.A. Green, and L.W. Andersen (Aerotherm/Acurex
Corporation). Design, Fabrication, and Installation of a Participate
Aerodynamic Test Facility. Environmental Protection Agency,
Research Triangle Park, N.C. Publication No. EPA-650/2-74-103.
October 1974. 66 p.
2. Smith, W.B., K.M. Cushing, and J.D. McCain (Southern Research
Institute). Particulate Sizing Techniques for Control Device
Evaluation. Environmental Protection Agency, Research Triangle
Park, N.C. Publication No. EPA-650/2-74-102. October 1974.
120 p.
3. Private communication with Mr. Neal Hill, Hill Environmental Group,
Chapel Hill, N. C. August 5, 1975.
4. Smith, W.B., K.M. Cushing, G.E. Lacey, and J.D. McCain (Southern
Research Institute). Particulate Sizing Techniques for Control
Device Evaluation. Environmental Protection Agency, Research
Triangle Park, N. C. Publication No. EPA-650/2-74-102-a.
August 1975. 124 p.
5. Rouillard, E.E.A. Experimental Errors in Sampling Dust Laden Gas
Streams. Chemical Engineering Group, South African Council for
Scientific and Industrial Research. CSIR Special Report 66/51/4510/2.
December 1971. 21 p.
64
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/2-76-073
2.
3. RECIPIENT'S ACCESSION NO.
. TITLE AND SUBTITLE
Wind Tunnel Evaluation of Particle Sizing Instruments
5. REPORT DATE
March 1976
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Charles H. Gooding
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, NC 27709
10. PROGRAM ELEMENT NO.
1AB012; ROAP 21ADL-18B
11. CONTRACT/GRANT NO.
68-02-1398, Task 23
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND,PERIOD COVERED
Task Final; 6-10/75
14. SPONSORING AGENCY CODE
EPA-ORD
IB. SUPPLEMENTARY NOTES Project Officer for this report is W. B. Kuykendal, Mail Drop 62,
Ext 2557.
16. ABSTRACT
The report gives results of an experimental study, undertaken to evaluate
and compare several particle sizing instruments. Fly ash from a coal-fired power
plant was redispersed and fed into a wind tunnel at concentrations corresponding
roughly to clean and dirty stack conditions. Data were obtained with two standard cas-
cade impactors (using gravimetric mass determination), a set of series cyclones, a
cascade impactor with piezoelectric crystal sensors, a cascade impactor using beta
attenuation to determine collected mass, and an optical single-particle counter using
a. laser light source. The standard impactors and the series cyclones yielded com-
aarable results. Data from the piezoelectric cyrstal cascade were in reasonable
agreement with the standard impactors but were limited because of the required
extractive sampling mode and the mass capacity limitation of the crystals. The beta
impactor showed general agreement but needs further development in the areas of zero
stability, sensitivity, signal-to-noise ratio, and scale conversion from beta attenuation
o collected mass. No simple correlation could be established between the results of
he optical instrument and those of the inertial classification devices, due to the
probable nonuniformity of particle shape and density.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPCN ENDED TERMS
c. COSATI Held/Group
\ir Pollution
Size Determination
Particles
nstruments
Evaluation
Vind Tunnels
Fly Ash
Impactors
Cyclone Separators
Piezoelectricity
Beta Particles
Lasers
Air Pollution Control
Stationary Sources
Cascade Impactors
13B
14 B
14A
21B
131
07A
20C
20H
20E
8. DISTRIBUTION STATEMENT
Unlimited
19. SECURI1 Y CLASS (Tliis Report)
Unclassified
21. NO. Of PAGES
72
20. SECURITY CLASS (Tin's page)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
65
------- |