EPA-600/2-77-201
September 1977
Environmental Protection Technology Series
GENERATION AND SIMULATION
OF METALLIC PARTICULATE AIR POLLUTANTS
BY ELECTRIC ARC SPRAYING
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 Protec-
tion Agency, have been grouped into nine series. These nine broad categories were
established to facilitate further development and application of environmental tech-
nology. Elimination of traditional grouping was consciously planned to foster technology
transfer and a maximum interface in related fields. The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL PROTECTION TECHNOLOGY
series. This series describes research performed to develop and demonstrate instrumen-
tation, equipment, and methodology to repair or prevent environmental degradation from
point and non-point sources of pollution. This work provides the new or improved tech-
nology required for the control and treatment of pollution sources to meet environmental
quality standards.
REVIEW NOTICE
This report has been reviewed by the participating Federal Agencies, and approved for
publication. Approval does not signify that the contents necessarily reflect the views and
policies of the Government, nor does mention of trade names or commercial products
constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Information
Service, Springfield, Virginia 22161.
-------
EPA-600/2-77-201
September 1977
GENERATION AND SIMULATION
OF METALLIC PARTICULATE AIR POLLUTANTS
BY ELECTRIC ARC SPRAYING
by
B. Linsky, R. Hedden, M. Naylor, and F. Dimmick
University of West Virginia
Morgantown, West Virginia 26506
Grant No. R801858
ROAP No. 21ADM-025
Program Element No. 1AB012
EPA Project Officer: Dennis C. Drehmel
Industrial Environmental Research Laboratory
Office of Energy, Minerals, and Industry
Research Triangle Park, North Carolina 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, D.C. 20460
-------
PREFACE
The growing concern about fine particles was barely detectable among
air pollution specialists in the late 1950's. It surfaced in the emission
inventory of the San Francisco Bay Area Air Pollution Control^District
Technical Report of 1958. By the early 1970's recognition of the need to «
control fine particles separately had been established. It became U.S.
Environmental Protection Agency policy to deal with these microscopic and
sub-micrometer particulates differently than larger ones from a technological
control viewpoint, from an undesirable effects viewpoint, from a regulatory
viewpoint, and from a measurement viewpoint.
The fine particles that were technically available for research and
engineering development were either surrogates for realistic fine partic-
ulates, such as DOP, or were stale dusts that were redispersed in a gas
stream. Some of the surrogates came from aqeous solutions or suspensions
and were high in water vapor content as they entered the air or other gas
streams. Caught in a dust collector weeks earlier and redispersed, the
metal or metal oxide fine particulates were not fresh even though it is
generally recognized that fresh metal fumes are often more reactive than
stale ones.
There seemed to be a growing demand for a fresh dry fine particle gen-
erator of realistic metallic oxides such as a fine particle generating
ii
-------
system with a fresh, dry particle output somewhat similar to that from an
electric arc furnace. It was hoped that the mass per unit time (mass
emission rate) and mass per unit volume of gases (mass volumetric concen-
tration) would be sufficient to allow a realistically high dust-in-exhaust-
gas stream for testing bench scale or small pilot scale dust collector to
"improve the breed" of such collectors. The U.S. Environmental Protection
Agency having an immediate need for such a fine particle generator because
of work they were doing and extramural work they were having done for them
on charged droplet fine particle scrubbers and other fine particle control
equipment research and development purchased identical equipment to the
metallizer generator when the ongoing findings of this research grant indi-
cated a resonable possibility of success in mass volumetric concentration
and mass emission rate of fine particles.
This report contains, as appendices, the theses and problem reports
of the graduate students who did most of the research work on this project.
As will be noted, most of the electron micrography and all of the elemental
and chemical compound analytical work were done by a specialist contractor,
Walter C. McCrone Associates. Much of the fine particle sizing was also done
by the McCrone organization.
ill
-------
ABSTRACT
This research program was a pragmatic experimental activity that was
conceived to provide a long-needed research tool for those concerned with
very fine particles and their technological control. In order to try to
provide a generated output with an appropriate mass and concentration of
fresh dry fine metal oxide particles for bench or pilot scale collector
research and development work, electric arc generators of fresh fine par-
ticle aerosols were modified and tested with zinc metallizing wire and with
steel welding and metallizing wire at West Virginia University.
Work with two electric arc aerosol generators is reported; one generator
employed a single consumable electrode of welding wire and the other generator
employed two consumable wire electrodes of a commercially available electric
arc metallizer. The generated aerosols were exhausted into a duct system
and sampled using membrane filters.
The single electrode generator produced 0.67 grams per cubic meter of
0.1 micrometer diameter fresh iron oxide agglomerated particles as sampled
by an Andersen Stack Sampler. The mass emission rate with an average of 1.95
grams per minute varied within a plus or minus 12 percent range. The basic
particles that formed the 0.1 micrometer agglomerated particles were charac-
terized to be log normally distributed with a mass median diameter of 0.0167
micrometer as observed and measured by electron microscopy.
iv
-------
The two consumable electrode aerosol generator produced agglomerated
submicron particles (as measured by scanning electron microscopy) formed
from 2-20 nanometer diameter basic particles (as measured by transmission
electrode microscopy). Mass volumetric concentrations ranged from 0.7 -
2.0 grams per cubic meter for zinc oxide aerosols. The mass emission rate
averaged 8.6 grains per minute for the zinc oxide aerosols and 7.7 grams per
minute for the iron oxide aerosols. Selected area electron diffraction
analyses found the zinc oxide to be ZnO and the iron oxide to be FejD..
The two electrode aerosol generator was further tested and validated for
reproducibilities of total mass volumetric concentration and basic particle
diameter distributions. Variables of operation were investigated to deter-
mine their effect on the mass volumetric concentration of the fresh dry fine
particle aerosol produced by the generator.
This report was submitted in fulfillment of Grant No. R-801858-01-1
by West Virginia University under the sponsorship of the U. S. Environmental
Protection Agency. This report covers the period July 20, 1973 to January 15,
1975, and work completed outside of the grant period through April 29, 1977
was included.
-------
CONTENTS
Page
PREFACE ii
ABSTRACT iv
FIGURES vii
TABLES viii
CONVERSIONS FROM METRIC TO ENGLISH ix
ACKNOWLEDGMENTS x
A. INTRODUCTION 1
B. CONCLUSIONS AND RECOMMENDATIONS 6
C. EXPERIMENTAL PROCEDURES 11
Apparatus 12
Sampling 17
Analytic Techniques 19
Experimental Designs 19
D. RESULTS AND DISCUSSION 21
APPENDICES
1. THE HEDDEN REPORT 1-i
2. THE NAYLOR REPORT 2-i
3. THE DIMMICK REPORT 3-i
vi
-------
FIGURES
Number Page
1 Simplified diagram of aerosol generator
apparatus (Hedden) 13
2 Metallizer equipment used by Naylor 15
3 Shrouded head metallizer equipment used
by Dimmick 16
vii
-------
TABLES
Number Page
1 Summary of mass volumetric concentration and emission
rate with approximate range of agglomerated particle
size indicated (Redden) . '. 22
2 Relative basic particle size frequency distribution
in percent (Hedden) , 24
3 • Zinc oxide mass concentration mean and variance
summary (Naylor) 25
4 Means and variances of zinc oxide basic particle size
5
6
7
8
Analysis of variance for Latin Square experimental
design mass volumetric concentration data (Dimmick) .
Statistics for agglomerated samples (Dimmick)
Statistics for very fine basic particles (Dimmick) .
. 28
. 29
. 30
. 31
viii
-------
CONVERSIONS FROM METRIC TO ENGLISH
1 meter - 1010 A = 3.28 feet
1 meter/sec = 3.28 feet/sec
3 3
1 cubic meter (m ) = 35.315 cubic feet (ft )
1 m3/min = 35.315 ft3/minute
_3
1 gram = 15.43 grains = 2.2x10 pounds
3 33
1 gram/m = 0.44 grain/ft = 0.0624 pounds/1000 ft
1 gram/min. = 0.138 pounds/hr.
°C = (°F-32)/1.8
ix
-------
ACKNOWLEDGMENTS
The cooperation of several present and former materials identification
specialists at West Virginia University is recognized and appreciated. They
are named in the component reports and theses. Assistance beyond the ordi-
nary level of equipment and service supplier's advice and guidance are also
appreciated and recognized from Wall Colmonoy, Flamespray Industries, and
McCrone's firms.
There were also inumerable discussants by phone and at various air pol-
lution specialists' gatherings, ranging from a Gordon Research Conference
through an Air Pollution Control Association session to several EPA supported
conferences on fine particulates. It would not be possible to name all of the
contributors of ideas, experiences, cautions, etc.
Considerable financial support, other than that which was provided in
the research grant itself, was in the form of U.S. Environmental Protection
Agency traineeships for graduate study from the Office of Air Pollution
Manpower Development, as is shown in the component reports and theses.
-------
Section A
INTRODUCTION
The need for a generator of fresh fine particle aerosols and the lack
of a suitable generator led to the proposal of research and development of a
fine particle aerosol generator for use in fine particle air pollution con-
trol research and development. The general purpose of this research was to
prepare and characterize high mass volumetric concentration, fresh fine
metallic oxide particle aerosols. In particular, the electric arc process
of particle generation was chosen as most likely to provide the desired
product. This required characterizing the fine particle aerosols and the
mass concentrations produced by an electric arc aerosol generator.
Two electric arc aerosol generators were investigated and developed
by graduate students at West Virginia University under the guidance of
Professor Benjamin Linsky. Robert Redden, having developed an initial single
consumable electrode aerosol generator, reported results on a modified ver-
sion of his generator in his Masters report. Michael Naylor and Fred Dimmick,
working with a commercially available metallizer employing two consumable
electrodes, reported results in their Masters reports. These three Masters
of Science reports are the basis of this summary report; each of the Masters
reports is attached as an appendix.
In order to place into proper perspective the information from these
-------
Masters reports, a short developmental history from Hedden's first consumable
electrode electric arc aerosol generator to the current form of the metalllzer
is presented. Again, differences in equipment configurations, differences in
aerosol sampling methods and differences in analytic techniques are presented
tp provide insight into the background of the information.
During his graduate study, Robert Hedden developed a number of very fine
particle generators which used an electric arc welder system. His systems
changed based on his experience; one of his systems used two consumable wires,
another used consumable wire and a non-consumable fixed electrode and another
used one consumable wire and a non-consumable rotating electrode. He
examined the last system in detail; it consisted of a consumable wire feed
as the anode and a relatively non-consumable tungsten electrode as the cathode.
Carbon steel welding wire was used for the anode. Runs were limited to 30
minutes due to the tungsten cathode being consumed after that length of time.
Modifications of Hedden's system were developed in 1973-74 by Professor
Linsky, Robert Hedden, Dr. William Fischer, Arland Johansen, Michael McCawley,
Rasool Nekooi and Dr. Richard Sears. This system consisted of using a Hobart
continuous electric arc welder with two consumable wires from 2 separate heads
each with its own wire feed motor and motor speed regulator. This system was
unstable because of inadequate control over several factors including the
guidance of the wires; for example, sometimes gaps between the wires were
greater than 0.25 inch. Data collected using this system was insufficient
to establish conclusive results.
In December, 1974, a commercial metallizer was found on the industrial
consumer market that seemed to have suitable properties. The Electrospray
-------
metallizer unit (Wall Coltnonoy Corp., Model VT-500) was selected after exam-
ining alternatives such as Model 10E metallizing gun of METCO, Inc.
The Electrospray metallizer was put into operation by Professor Linsky,
Michael McCawley, Joe McCauley, Dr. Sears and Anthony Angotti. After estab-
lishing the generation system consisting of the metallizer, a collection
barrel and an exhaust system for the aerosol, Robert Redden, John Garbak,
Don Stone and Mike Naylor proceeded to calibrate the equipment used in the
process and tried to establish stability on generation runs. Even though
reliable stability on runs was not established, periodic stable runs allowed
some sampling.
Early work with the metallizer, which initially consisted of generating
aerosols with metallizer steel wire (coated with copper), was followed by a
change to 15 gauge zinc metallizing wire. Continued difficulty in estab-
lishing reliably stable generation runs occurred using the zinc wire.
In December, 1975, Naylor, Stone, and Dimmick made modifications and
ran several successful runs with the zinc metallizing wire on the aerosol
generator. Naylor collected samples in January, 1976 for his report.
During the summer of 1976, Dimmick, Stone and Craig Repp worked on aerosol
generation with steel welding wire and copper coated steel metallizing wire.
After a few modifications and additions, Dimmick took samples with welding wire
and metallizing wire.
Differences in generator configurations, in sampling methods, and in
analytical techniques are evident when comparing the three Masters reports.
The obvious differences between the Hedden design and the commercial
metallizer design may overshadow the subtle differences in the generator
3
-------
configurations of Naylor and Dimmick. For example, Naylor positioned the
separation and collection barrel vertically while Dimmick positioned the
barrel horizontally; Dimmick employed an electrode tip shroud and Naylor
did not. Redden's power source was a constant voltage source; Naylor and
Dimmick's was not. Also, Hedden's arc chamber was above atmospheric pressure
and Naylor's and Dimmick's was below atmospheric pressure.
Differences in sampling methods center around two points: (1) methods
used and (2) proficiency of each experimenter. Redden used an Andersen
Stack Sampler followed by a membrane filter. Naylor and Dimmick used mem-
brane filters to collect particle diameter count and mass volumetric concen-
tration samples. While Naylor's mass volumetric concentration samples were
collected generally under isokinetic conditions, Dimmick's were not and
Redden did not report an isokinetic estimate.
Differences in analytical procedures may have led the research astray.
In particular, Redden employed typical electron microscopy techniques of
particle measurement including dispersion of particle aggregates. Although
Hedden explained that chained and agglomerated particles should be expected
from electric arc production of aerosols, Naylor and Dimmick, through the
McCrone organization, measured particles in the 2-20 nanometer range while
an awareness of an apparent agglomerated-particle aerosol grew when properly
loaded membrane filters were evaluated from scanning electron micrographs.
Also, the inherent differences in sampling methods led to other differences
in analytical measurements.
Despite the differences in equipment configuration, sampling techniques
and analytical methods, the three Masters reports have fulfilled the objectives
of aerosol characterization and apparatus investigation. Certainly some
-------
questions have been answered, some have been partially answered and many more
have been raised. The report that follows should be taken in the light of
the differences and similarities between each Masters report.
-------
Section B
CONCLUSIONS AND RECOMMENDATIONS'
Hedden concluded that based on the results of limited experimental
tests the feasibility of employing an electric arc to produce aerosols of
very fine iron oxide particles from a feedstock of consumable wire has been
demonstrated. More than 30 grains per minute (1.95 g/min) of discrete,
spherical particles with a mass median diameter of 0.0167 micrometers were
repeatedly produced within a range of plus or minus 12 percent. Also, a
conservative examination of the mass distribution data obtained from the
Andersen Stack Sampler particle sizing technique indicated that approx-
imately 75 percent of the particles have an effective diameter less than
0.1 micrometers.
Hedden recommended:
(1) With modification, the unit might be adapted for the production
of aerosols of non-conductive refractory materials. This might
be accomplished by replacing the wire torch with another non-
consumable electrode holder, establishing an arc between them,
and guiding the non-conductive feedstock into the high-intensity
arc.
(2) Elimination of the tungsten electrode as one side of the arc could
be accomplished by utilizing two wire torches and establishing an
arc between the two consumable wire "electrodes."
-------
(3) If agglomerating effects prove undesirable for a particular study,
an Ion generator or polonium grid could be employed to neutralize
the charged particles.
(4) Optimization of main operating parameters - arc voltage, wire feed
rate, and wire diameter - could significantly improve the feed-
stock-to^particulate conversion ratio of 4.75 percent.
Naylor concluded that for at least one operating condition, the generator
can produce very fine agglomerated zinc oxide particles that are reproduc-
ible with respect to both the distribution of the basic particle diameters
and mass volumetric concentration. Therefore, this metallizer is considered
worthy of continued attention by researchers as a device for generating very
fine particles.
3
The generation of mass concentration ranging from 0.67 - 2.05 g/sm
(1.5 - 4.7 grain/scf) was reproducibile at three of four operating conditions.
The mass volumetric concentration was sensitive to changes in the operating
conditions. The mass emission rate was 4.7 - 14.5 gram/minute.
Basic particle sizes ranged from 6.0 - 10.8 nm (0.006 - 0.0108 jim). At
one operating condition, the basic particle diameters' mean and standard
deviation were reproducibile, while at another the basic particle's mean,
standard deviation and cumulative frequency distribution diameter was repro-
ducibile. Comparatively, the mean particle diameters from these two operating
conditions were significantly different.
\
Virtually all of the basic particles were agglomerated into chains or
clusters of various sorts. The agglomeration initiated immediately after
basic particle generation at the electric arc. Its fast rate undoubtedly
-------
resulted from the high initial number of basic particles per volume, their
very fine particle size, sonic agglomeration from the noisy arcing process,
and thermal turbulence of the gas stream near the arc.
At the one condition for which mass concentration was not reproducible
the mean basic particle diameter was and is not considered to be reproducible.
In summary, Naylor recommended:
(1) Improve the research tool.
a) Further development of equipment to improve operational
quality.
b) Investigation of other metals and expansion of generating
and sampling program.
(2) Further study of the agglomeration of very fine particles.
a) Define appropriate dimensions for describing the agglomerates.
b) Based on these dimensions, determine whether or not the metallizer
generates predictably reproducible distributions of sizes.
c) If predictable reproducibility is established determine if there
is a significant correlation between agglomerate size and such
variables as chamber-duct retention time, number of very fine
basic particles per unit volume, and diameters of the very
fine particles.
d) Related to "c" above, determine if there is a significant
correlation between very fine basic particle diameter, agglom-
erate size and mass volumetric concentration.
(3) Use the research tool; for example, modify various experimental
collection equipment designs to achieve and improve the effective-
ness of small particle control.
8
-------
Dimmick concluded that the electric arc generator produced an aerosol
of agglomerated iron oxide fine particles in mass volumetric concentrations
3
averaging 1.4 gm/m and mass emission rates averaging 7.8 grams per minute.
The generator showed some controllability of mass volumetric concentration
and variability of very fine basic particle populations with respect to some
operating variables. The analysis of thei production of the mass volumetric
concentration of the aerosol showed the mass volumetric concentration to be
dependent on the wire feed rate and the open voltage across the electrodes.
The iron oxide fine basic particles,which composed the agglomerated
particles having an estimated size of 0.4 um, have been characterized to
have means ranging from 5.4 nm to 9.1 nm (0.0054 pm to 0.0091 ym). Differences
in size means and size distributions between sets of operating conditions
indicated that changing the operation variables changed the size character-
istics of the pppulation of fine particles.
A simple comparative test for discerning the effect, if any, of changing
the exhausting rate of the arc chamber on the aerosol's particle size was
inconclusive. The agglomerated particle size was found not to change with an
increase in exhausting rate. However, the very fine basic particle mean size
was found to be greatly affected by the increase in exhausting rate. This
change in generating properties indicated a need for further experimentation.
Dimmick recommended:
(1) Further investigation of the evidence found in the simple compara'-
tive test.
(2) The compressed air-jet that atomizes and quenches the melted and
9
-------
vaproized metal should be studied to ascertain its effect on
the aerosol's particle size mean and distribution. The open vol-
tage would require very precise control if its effect on an aerosol's
particle mean size or size distribution were to be investigated; a
more appropriate power supply is needed.
(3) In the application of this aerosol generator, an on-line con-
tinuous mean size and size distribution analyzer, such as a Whitby-
Liu mobility analyzer, would provide the necessary "before control
equipment" and "after control equipment" data in an engineering
evaluation of a piece of air pollution control equipment. Better
control of the separation of the larger particles within the barrel
could also help approximate real life circumstances.
The overall conclusion is that electric arc generation of fine particle
aerosols provides an adequate emission rate at high enough mass volumetric
concentrations to test fine particle control equipment. The fine particles
are formed by agglomeration from many 2-20 nanometer sized particles with the
metallizer generator.
An overall recommendation is to provide continual monitoring of the
aerosol's particle size mean and distribution while further investigating
any effects of the many independent variables present in the system. Although
development is still needed, the electric arc generation of fresh fine metallic
oxide particle aerosols provides a generating means for those interested in
research and development of fine particle air pollution control technology.
10
-------
Section C
Experimental Procedures
The two aerosol generators that were developed by Redden and by Naylor
and Dinmick use the energy from a direct current electric arc to heat materials -
steel and zinc - in the production of fine particle aerosols. The heat (up
to 7000°K) was provided from an arc between an electrode and the material to
be melted (direct arc heating) or by an arc between two electrodes (indirect
arc heating). In both direct and indirect heating, charged electrodes initiate
an intense electric arc when brought into close proximity to each other. The
result of this heat generating arc is the melting and vaporizing of the
electrodes - particularly the anode.
In both generators, a nozzled gas (compressed air) jet converged about
the zone of melted and vaporized materials. This jet had the effect of atom-
izing the melted materials into molten droplets as well as facilitating the
condensation of the vaporized materials into very fine particles. These very
fine particles, being newly formed, were considered to be fresh.
Simplistically, two products are produced by the electric arc aerosol
generators : the solidified molten droplets from the atomizing spray and the
solidified condensate of metallic vapors. Both generating apparatus incorp-
orated settJing chambers to separate the much larger particles from the product.
Presumably, the fresh metallic fine particles were not affected by the gravi-
metric and centrifugal separation even though it became evident that the high
11
-------
particle number concentrations and the presence of abundant electric charges
agglomerated the very fine basic particles into larger agglomerated particles -
thereby increasing the effective size of the particles. This assumption does
not exclude the possibility that discrete, very fine particles remain in the
gas stream.
Apparatus
Hedden's final apparatus (Figure 1) consisted of an adjustable, constant
voltage direct current power supply, a consumable electrode wire feed assembly
with control panel, an arc chamber, a non-consumable tungsten electrode with
cooled holder and shielding gas supply system, and a system to connect the
arc chamber to an air mover upstream and a settling chamber and exhaust system
with sampling port downstream.
The power supply utilized in Hedden's apparatus, conventionally used to
provide power for semiautomatic welding processes such as gas shielded open
air and submerged arc, was a direct current, self contained motor driven
generator unit rated for a 100% duty cycle. This rating allowed continuous,
uninterrupted production of power under full load. Once set, the output of
a constant voltage power supply has essentially the same voltage no matter
what the welding current may be. The voltage could be varied over a range
of approximately 10 - 50 volts.
The automatic welding head assembly consisted of two main components:
(1) the wire feed assembly with torch, and (2) remote control panel. The ,
wire feed assembly combined an insulated wire supply reel, guides, drive motor
with gear reduction box, and feed rolls, that constantly feed a consumable
electrode wire through the air-cooled torch to the arc. The remote control
12
-------
u>
AIR
FLOW
EXHAUST
VACCUM
POWER SUPPLY
WIRE REEL
£
WIRE TORCH
ARC CHAMBER
TUNGSTEN
ELECTRODE
HOLDER
SHIELDING
GAS
SUPPLY
AIR
MOVER
SHIELDING
GAS NOZZLE
COOLING
WATER
TANK
Figure 1. Simplified diagram of aerosol generator
apparatus (Hedden)
-------
panel was electrically connected to both the welding machine and the non-
consumable electrode holder cooling water supply through solenoid valves.
The arc chamber was fabircated from a 33 centimeter length of 12.7
centimeter (5 inches) diameter steel pipe. The attachments between the arc
chamber and the wire feed assembly were designed to allow adjustment in all
axes to obtain perfect alignment between the feed wire electrode and the
non-consumable electrode. A shielding gas nozzle and supply system provided
for displacing the air surrounding the non-consumable electrode thus preventing
rapid consumption of the tungsten by oxidation .
Air flow through the system was provided by a "Gelman Hurricane Air
Sampler" set on low speed. The ductwork leading from the arc chamber incor-
porated a settling chamber and elutriation column to separate the molten
globules of steel from the gas stream. The duct leaving the elutriation
column made a 180 degree turn before passing the sampling port and nozzle.
An adequate length of straight duct was established to allow optimum position-
ing of the sample probe - a minimum of 10 duct diameters downstream and 5
duct diameters upstream from any disturbance to air flow.
The Naylor and Dimmick aerosol generating apparatus consisted of an
electric arc two consumable electrode metallizer that incorporated an elec-
tric arc and a compressed air jet to spray melted and vaporized consumable
wire electrodes, a 55 gallon oil drum that was used as the arc chamber and
was connected to a supplemental air supply cleaned by a HEPA (High Efficiency
Mr Particulate) filter, a draft fan with exhausting ductwork from the oil
drum, and a cloth filter particulate collector. (For Naylor's see Figure 2,
for Dimmick's, Figure 3) The .metallizer system supplied the necessary power
and wire feed delivery. The ductwork and fan provided for exhausting and
14
-------
Phenolic spray head Cross-section
I E
lid of
barrel i
A) Power source of metallizer
B) Control console
C) Dual wire spools
D) Wire straightening, drive and conduit feed mechanism
E) Flexible conduit for wire
F) Flexible compressed air line - main air
G) High amperage electrical cable
H) 55 gallon barrel
I) Phenolic spray head
J) Dilution air from Hepa filter
K) Flexible compressed air line - secondary air
L) Exhaust duct - 6 inch diameter
M) Sampling port
N) Venturi
O) Stack sampler
Figure 2. Metallizer equipment used
by Naylor
15
-------
Wire
dispenser
Secondary
Arc gun air
Secondary air
connection
Electrode tip
gas jet
•Electrode wire
Lid
Figure 3. Sfiroulded Read metallizer equipment
used by Dlmmick
16
-------
sampling the generated aerosol. The oversized baghouse cleaned the particle
laden gas stream of most of the particulates before returning the process air
to the atmosphere.
The metallizer of the Naylor and Dimmick apparatus is used commercially
to provide coatings on metal or to fill in or build up worn areas on shafts,
etc. It consisted of a non-constant voltage power source which converts 3
phase 220 volt alternating current to a direct current that is variable from
0 to 40 volts, up to about 400 amperes. A console was located on top of the
power source from which air pressure to spray gas jet, consumable wire delivery
rate, and arc open-circuit voltage are adjusted. Flexible conduits conveyed
the wires from reels passing through the console to a phenolic spray head
that guided the wires into fixed position electrode tips.
The spray head was fastened to the lid of a 55 gallon barrel with the
electrodes extending through a slot into the barrel. The arcing takes
place in the barrel. The larger particles of the metallized wire are directed
by the high pressure air jet to the other end of the barrel where they are
deposited. The dispersed air jet stream and a low pressure air stream enter-
ing through a HEPA filter conveyed the smaller particles through an exhaust
duct past the sampling port to a cloth filter collector and exhaust blower.
Sampling
Hedden's system was allowed to operate for approximately one minute to
establish steady-state conditions for sampling with steel welding wire. During
this period the temperature .of the exhaust gas stream stabilized producing
a constant velocity past the sampling nozzle. A ten minute sampling period
then commenced.
17
-------
The sampling train consisted of a one-half inch (12.7 mm) stainless steel
hook nozzle attached to an Andersen Stack Sampler, in series with an aluminum
Millipore filter holder supporting a 47 mm diameter Gelman type GA-4 membrane
filter to separate the fume from the sample gas stream.
A calibrated orifice was used to indicate instantaneous sample flow rate,
with total flow rate recorded on a dry gas meter.
The fine particle aerosol mass concentration was determined gravimetri-
cally. Also, each sample of collected fume; was resuspettded in 2 ml of a 1%
nitrocellulose solution and dispersed by placing the test tube in an ultra-?
sonic cleaner; the fume collected on the membrane filter was carefully
scraped from the membrane and placed directly in 2 ml of the nitrocellulose
solution and dispersed. A disposable pipette was used to place one drpp of
i
the solution on an uncoated electron microscope grid.
The metallizer generator was operated for at least two minutes before
samples were taken either for mass conversion analysis or particle size dis-
tribution analysis. Randomly ordered samples were taken with a shut-off
between each sampling run. Naylor took samples using zinc metallizing wire
and Dimmick took samples using steel metallizing and welding wire.
Sampling for mass concentration data consisted of pre-trial determinations
of isokinetic sampling conditions and then the actual sampling. After adjust-
ing the stack sampler for isokinetic sampling, samples were taken such that
at least 10 mg of mass was deposited on the tared filter.
Sampling the aerosol for subsequent particle size distribution and
analysis consisted of estimating isokinetic conditions and then collection
on appropriate membrane filters with sampling times of three, five and seven
seconds. Then, the filters were enclosed in a plastic filter holder and
18
-------
were shipped by United Postal Service to McCrone Associates, Inc. in Chicago
for analysis by scanning electron microscopy (SEM), transmission electron
microscopy (TEM), and selected area electron diffraction SAED).
Analytic Techniques
Hedden used gravimetric and flow measurement techniques to determine a
aerosol mass concentration. Hedden had electron micrograph from an RCA
electron microscope at West Virginia University to measure particle diameter
counts with the use of a semi-circle graticule. '
Naylor and Dimmick used gravimetric and flow measurement with an RAC
Stak sampler to determine aerosol mass concentration. Naylor and Dimmick
both recieved fine basic particle measurements made by McCrone Associates
using TEM and SEM. McCrone Associates also performed chemical identification
using SAED. Naylor characterized the agglomerated particles from McCrone
and Dimmick measured the agglomerated particles using the enlarged SEM
photographs.
Experimental Designs
Hedden investigated the performance characterisitcs of the aerosol
generator under semi-continuous operation. A series of three controlled
test runs were made and a portion of the resultant fume was collected. All
tests were conducted under identical conditions of arc voltage, wire feed
rate, shielding gas flow rate, and dilution air flow rate. The test program
was structured to determine the mass rate of production of iron oxide fume,
the particle size (not cecognized at first) distribution of that fume, and
the degree of reproducibility of production rate and size distribution.
19
-------
Naylor investigated the reproducibility of the metallizer aerosol
generator using zinc metallizing wire. He used four settings with the
generator to obtain data for particle size measurement and mass volumetric
concentration determination. Naylor varied the consumable wire feed rate, open
circuit voltage and the gas jet pressure.
Dimmick characterized the aerosols' particle size under four different
operating conditions using steel welding wire. He also performed a simple
comparative experiment investigating the effect of increasing the exhaust flow
within the ductwork. Dimmick designed an experiment to investigate the
effect of generator variables - wire feed rate, open circuit voltage, and
gas jet pressure - on the aerosols' mass volumetric concentration.
20
-------
Section D
Results and Discussion
Mass Volumetric Concentration and Emission Rate (Hedden)
With the mass of particulate in the sample gas stream determined gravi-
metrically, the average mass volumetric concentration of iron oxide fume in
the sample gas stream was 0.292 grains/scf (668.1 mg/m ). Assuming that the
mass volumetric concentration of particulate matter in the total gas stream
is the same as for the sample gas stream, the average total particulate pro-
duction rate was calculated to be 30.08 grains/min (117.0 grams/hr). Total
particulate emission rate, given in common units of measurement, is listed
in Table 1.
Another performance factor directly related to the rate of production
is the conversion ratio of consumable wire feed stock to very fine particles.
The wire consumption rate was 5.43 Ib/hr (2463 gram/hr). The average partic-
ulate production rate was 0.258 Ib/hr (117.0 g/hr). Therefore, 5.43 pounds
(2463 g) of wire was required to produce 0.258 pounds (117.0 g) of very fine
particles for a conversion ratio of 4.75 percent.
Particle Sizing (Hedden)
Two methods were employed, with differing results, to determine particle
size distribution: (1) an electron microscope and (2) an Andersen Stack
Sampler.
Because of the high magnification, and consequently small field, of the
electron microscope, photographic plates having a final magnification of
87,600 times were made of a minimum of three fields of each grid. Individual
particles of iron oxide were spherical; and individual spherical particles
21
-------
TABLE 1. Summary of mass volumetric concentration and emission rate with approximate range
of agglomerated particle size indicated(Redden)
Approximate
range of
particle
STAGE size (urn)
0 > 30
1 9.2-30
2 5.5-9.2
3 3.3-5.5
4 2.0-3.3
5 1.0-2.0
6 0.3-1.0
7 0.1-0.3
8 < 0.1
Filter < 0.1
(1) TOTALS
RUN
Collected
weight
(mg)
2.5
0.4
0.4
0.8
1.3
2.7
8.1
16.4
9.0
91.6
1
Cumulative
percent
total
100.10
98.13
97.83
97.53
96.93
95.95
93.92
87.84
75.53
68.77
133.2
RUN
Collected
weight
(mg)
1.5
0.3
0.2
0.2
0.7
3.1
7.7
11.2
12.8
86.1
2
Cumulative
percent
total
100.00
98.79
98.55
98.39
98.23
97.66
95.16
88.94
79.89
69.55
123.8
(2) Sample volume
(scf)
(3) Mass volumetric
concentr. (mg/scf)
(4) Volumetric flow
rate (scfm)
(5) Emission rate
( grams /min)
7.23
18.42
103
1.90
7.21
17.17
103
1.77
RUN
Collected
weight
(mg)
2.4
0.5
0.4
0.5
1.5
3.3
8.4
17.4
13.4
103.4
3
Cumulative
percent
total
100.00
98.42
98.09
97.83
97.50
96.51
94.34
88.82
77.38
68.57
152.1
7.17
21.21
103
2.18
OVERALL
Mean
weight
(mg)
2.13
0.4
0.33
0.5
1.17
3.03
8.07
15.0
11.73
93.7
136.0
Standard
deviation
(mg)
0.50
0.08
0.09
0.25
0.34
0.25
0.29
2.72
1.95
7.22
14.42
7.20
18.89
103
1.95
-
2.1
-
0.17
NS
NJ
-------
composing hirger agglomerates were easily distinguished. The agglomerates
wcro not. MM.slgned a composite or equivalent particle size but the basic
pnrtlcU'8 composing an agglomerate were classified individually.
With a total of 12,355 basic particles sized the raw data was grouped into
10 classes with a basic particle size interval of 0.01 micrometers each to
simplify calculation of various statistical values. Three basic particles
exceeded 0.2 micrometers at 0.22, 0.25 and 0.24 micrometers. A simplified
method for calculating the arithmetic mean and standard deviation of the
basic particle size distribution was employed and relative particle size
frequency was presented in Table 2.
The arithmetic mean size of the basic particles was found to be 0.022
micrometer with a standard deviation of 0.0163 micrometers. Using log-
normal probability paper revealed an indication that the basic particle size
distribution was log normal. The median, sometime known as the mass median
diameter (HMD) divides the frequency distribution (by mass) in half and was
found to 0.0167 micrometer.
The second method involved the relatively new Andersen Stack Sampler
used in conjunction with a back-up membrane filter. Combining the mass distri-
bution data with calibration information supplied by the manufacturer,
cumulative particle (agglomerated) size mass distributions can be obtained
and were presented in Table 1. Although this mass distribution data indicates
approximately 69 percent of the sampled particulate passed through the
Andersen Stack Sampler without collection, the previous data presented con-
cerning the size of the discrete basic particles suggests that essentially
all the particulate matter should have passed through the sampler without
impacting on the collection plates.
23
-------
Table 2. Relative basic particle size frequency distribution in percent(Hedden)
STAGE
0
1
2
3
4
5
6
7
8
Filter
Mean
Standard
Deviation
Range
Particle Size, micrometers
0.005-
0.015
46.3
43.0
54.3
50.5
43.0
46.3
47.4
44.3
48.0
42.8
46.6
3.73
11.3
0.015-
0.025
22.0
22.4
19.1
19.2
23.3
22.3
20.5
21.2
21.6
22.1
21.4
1.38
4.2
0.025-
0.035
13.6
17.7
12.3
14.4
17.1
16.4
17.3
15.1
14.9
15.7
15.5
1.72
5.4
0.035-
0.045
7.7
6.9
7.0
6.6
8.7
5.3
6.6
8.8
6.3
8.3
7.2
1.15
3.5
0.045-
0.055
4.5
4.5
3.5
4.2
3.1
3.7
3.4
5.1
6.2
4.5
4.3
0.92
3.1
0.055-
0.065
2.1
2.6
2.1
2.2
2.2
2.6
2.7
2.1
1.3
3.6
2.4
0.59
2.3
0.065-
0.075
1.5
1.5
0.9
1.1
0.7
2.3
0.7
1.1
0.6
1.4
1.2
0.52
1.7
0.075-
0.085
1.3
0.4
0.2
0.8
0.6
0.3
0.3
1.0
0.4
0.8
0.6
0.36
1.1
0.085-
0.095
0.1
0.2
0.1
0.6
0.3
0.2
0.4
0.2
0.3
0.4
0.3
0.16
0.5
0.095
and>
0.8
0.9
0.4
0.1
0.9
0.6
0.7
0.9
0.4
0.4
0.6
0.27
0.8
99.9
100.1
99.9
99.7
99.9
100.0
100.0
99.8
100.0
100.0
100.0
-------
Two methods were used to estimate the reproducibility of the relative
particle size distributions both using the mass distribuiton data obtained
from the Andersen Stack Sampler and the back up membrane filter. The results
of the three runs are presented in Table 1. The arithmetic mean and standard
deviation for the weights found for each of the three runs on each stage of
the sampler and membrane filter are shown. The second estimate of agglomerated
size distribution reproducibility was obtained by calculating the arithmetic
mean and standard deviation of the cumulative relative mass distribution for
each of the various size intervals as shown in Table 2.
Mass Volumetric Concentration and Electron Diffraction (Naylor)
Naylor stated that the mass concentration results were more meaningful
when the statistical parameters of mean, standard deviation, and coefficient
of variation were calculated. The coefficient of variation is one indicator
of reproducibility or precision of mass volumetric concentration generation.
These parameters are summarized in Table 3.
Table 3. Zinc oxide mass concentration
mean and variation summary (Naylor)
Condition
(1) Mean (g/sm3) 1.03 0.68 2.05 1.12
(2) Stand. Dev. (g/sm3) 0.20 0.05 0.14 0.10
(3) Coefficient 100x(2)/(l) 19% 8.4% 6.9% 8.7%
of variation
25
-------
Naylor tested the null hypothesis that the mean mass concentrations
were equal using standard statisitcal methods. These hypothesis tests
support the conclusion at the 5% level of significance that the mean of
Condition 3 was greater than the mean of Condition 1, and that the mean of
Condition 1 was greater than the mean of Condition 2.
Selected area electron diffraction results, as received from McCrone
Associates, were compared to standard electron diffraction information (ASTM)
and were found to indicate that the particles generated were ZnO.
Particle Sizing (Naylor)
The electron microscopy photographs of the zinc oxide samples were all
similar in terms of particle appearance. Particle size means and variances
were presented in Table 4. SEM photographs show agglomerates of different
shapes and sizes. The agglomerates range in size from less than 0.1 vim to
over 2 pm and appear to be interconnected. The shapes vary from spherical
to chainlike. These agglomerates are composed of very fine particles as
shown in the TEM photos of sample 101.
Table 4. Means and variances of zinc oxide basic
particle size data (Naylor)
Sample #
Condition II
5 (nm)
S2 (nm2) 14.6
S (nm)
Sample Size
101
1
6.0
4.6
3.8
297
125
1
10.2
29.2
5.4
280
133
2
9.1
35.6
6.0
350
134
2
10.0
41.1
6.4
280
104
3
10.8
28.1
5.3
376
128
3
10.4
22.0
4.7
281
26
-------
The basic particles mean diameters for samples 101 (D,0,) and 125 (D,25)
of condition 1 were named significantly different by inspection. The
values of the pairs D and D for Condition 2, and D , and D . for con-
dition 3 were analyzed by a statistical method for significant difference.
This analysis indicated that the two replicates for Conditon 2 have diameter
means and variances that were not significantly different. The analysis also
indicated that the means and variances for Condition 3 were not significantly
different.
An important question is whether the means and variances for basic
particle diameter populations generated under different conditions are
different. Since the mean of Condition 1 cannot be estimated within a rela-
tively tight range, only the relationship of Condition 2 to Condition 3 is
examined. Hypothesis testing demonstrated that the mean basic particle
diameters for Conditions 2 and 3 are significantly different. Thus changing
the operating conditions apparently changed the mean basic particle diameters.
Mass Volumetric Concentration and Electron Diffraction (Dimmick)
The Latin Square analysis (Table. 5) found the wire feed rate to be sig-
nificant at the 0.01 significance level. This implied the wire feed rate
had an effect on the resultant mass volumetric concentration. The variables
of open voltage and jet pressure were not found to have any effect on the mass
volumetric concentration of the aerosol even at the 0.1 significance level.
The analysis of variance statistics for replication was also calculated and
found to indicate a significant difference between trials at the 0.05 signi-
ficance level but not the 0.01 significance level.
27
-------
Table 5. Analysis of variance for Latin Square experimental design
mass volumetric concentration data (Dimmlck)
Variable
Wire feed
rate
Open Voltage
Main air- jet
pressure
Replication
Experimental
error
Total
df
3
3
3
1
21
31
Sum of
Squares
1.473
0.319
0.106
0.335
0.952
3.185
Mean F , F rt ... F n ni
cal a=0.05 a=0.01
Square
0.4908 10.86 3.07 4.87
0.1065 2.35 3.07 4.87
0.0352 0.77 3.07 4.87
0.3349 7.39 4.32 8.02
0.0453
The SAS GLM analysis of the combination of the Latin Square and another
set of data found the results to significantly describe the variability of
all the data. With this total set of 40 observations, a statistical procedure
calculated F statistics for determining the effect or noneffect of the indepen-
dent variables in the experimental model. The procedure found the variables
of wire feed rate, open circuit voltage, and sample replication to have a
significant effect on the outcome: mass volumetric concentration of the gas
stream.
The SAED analysis showed two possible resultant materials. Two different
phases of iron oxide were possible: magnetite and maghemite.
Particle Sizing (Dimmick)
The size means of the agglomerates (Table 6) were found to be 0.41 ym
(S=0.53 jjm) for sample 715 and 0.40 ym (S=0.54 ym) for sample 714. These
28
-------
means were also found to be equal; this implied that they were from the same
population. The coefficient of variances were rather large, thus the agglom-
erated particle distributions representative of the aerosols were not "mono-
disperse."
6. Statistics for agglomerated samples (Dimmick)
pie No.
714
715
Mean
(ym)
0.40
0.41
Standard
Deviation
(ym)
0.54
0.53
Standard Error
of the Mean (ym)
0.03
0.03
Coefficie:
Variance
135
129
The simple comparative analysis of increasing the duct exhausting
velocity from 1300 fpm to 1850 fpm showed only a decrease in the agglomerated
particle size from 0.41 ym to 0.40 ym. This difference was not significant
at a 0.1 level of significnace.
Fine basic particle statistics were presented in Table 7. Statistical
tests indicated a variety of results. The within group t-test analysis
showed that the sample means in all the sets were equivalent within groups.
The between group t-test analysis showed a variety of equal and unequal
means. The within group very fine basic particle size distributions for sets
2, 3, and 4 were found to be equivalent. Size distributions between groups
were found to be all different at an alpha level of 0.05. This implied that
at three of four different operating conditions three different fine basic
particle size distributions were generated. The t-test analysis between the
two comparative samples 712 and 716 showed sample 712 to have a significantly
29
-------
Table 7. Statistics for very fine basic par tides (Dimmick)
Set Number
1
2
3
4
5
6
Sample No .
600
601
602
603
604
605
607
611
609
612
716
712
Number of
Particles
386
446
434
369
368
356
296
274
298
297
449
497
Mean
(nm)
8.00
7.83
7.53
8.45
8.32
8.31
9.05
8.57
7.86
8.13
8.05
6.42
Standard
Deviation
(nm)
5.35
4.56
5.23
5.37
5.51
5.41
5.49
5.40
5.40
5.37
5.21
4.79
Standard of Error
of the Mean (nm)
0.27
0.22
0.25
0.28
0.29
0.29
0.32
0.33
0.31
0.31
0.25
0.21
Coefficient of
Variance (%)
66.9
58.2
69.5
63.5
66.3
65.1
60.5
62.9
68.8
66.0
64.7
74.6
-------
smaller very fine basic particle size than sample 716.
An overall characterization of the fine particle generators is presented
in Table 8. This table list basic particle size statistics for the basic
Table 8. Hedden-Naylor-Dimmick summary of results
Limit of microscope resolution (nm)
Width of diameter count interval (nm)
Mean basic particle diameter (nm)
Range of basic diameters (nm)
Standard deviation (nm)
Cumulative Frequency Distribution
3
Mass volumetric concentration (gm/m )
3
Mass volumetric concentration (gm/m )
Mass emission rate (gm/min)
Percent conversion (%)
Redden Naylor Dimmick
(iron oxide) (zinc oxide) (iron oxide)
5
10
22
5->100
16.3
log normal
0.67
.61-0.75
1.95
4.8
1
1
6.0-10.8
1-50
3.8-6.4
normal
1.21
.67-2.05
8.57
1
1
6.4-9.1
1-720
4.7-5.5
1.38
0.81-2.16
7.73
8-15
particles - not the agglomerated particles. Redden did not directly measure
what may be presumed to be the aerosol; that is, he measured agglomerated
particles by inference from the Andersen Stack Sampler and not directly
from micrographs. Naylor indicated a range of sizes for the agglomerated
particles from less than 0.1 ym to greater than 2 ym. Dimmick measured the
agglomerated particles and found an average size of 0.4 ym. The table also
shows various production characteristics related to the efficiency of mass
31
-------
conversion from wire feed to particle aerosol.
A question that should be opened for discussion and research is centered
around the physical explanation of the electric arc generation of fine particle
aerosols. What are the relations and factors involved in the dichotomous
formation of solidified molten droplets and agglomerated solid fine particles?
Also, what effects on the chemical processes involved in the aerosol production
does the dichotomous formation have?
32
-------
ELECTRIC ARC GENERATION
OF POLYDISPERSED IRON OXIDE
AEROSOL IN AN AIR STREAM
PROBLEM
Submitted to the Graduate School
of
West Virginia University
In Partial Fullfillment of the Requirements for
the Degree of Master of Science in Engineering
by
Robert E. Hedden, B.S.
Morgantown
West Virginia
1972
-------
ABSTRACT
Freshly formed metal and metal oxide particles in the sub-micron
size range are known to have different behavior characteristics
than those of collected, classified, aged, and sedispersed participates.
In the past, significant quantities of freshly formed ultrafine particles
have not been available for research work in air pollution control
techniques, instrument calibration, emission simulation, animal
inhalation or particle behavior studies. This has caused considerable
difficulty in translating laboratory research to field application.
An experimental aerosol generator that supplies reproducible
amounts of spherical, solid, fresh particles of metals and metal
oxides directly in a hot or cooled gas stream is described in this
report. The generator employs a direct current power source to create
an arc between a relatively non-consumable tungsten cathode and a
consumable feed-stock of wire. The feed-stock is vaporized in the high
current density of the electric arc and particles are formed on cooling
by a gas stream moving past the arc area.
•Rate of continuous production of 0.1 micrometer diameter fresh iron
oxide particles exceeds 100 grams/hour. The moving gas stream was
sampled using a high temperature Andersen Stack Sampler probe followed
by a membrane filter to obtain mass concentration. The size, shape
and particle size distribution of the particles were determined using
an electron microscope.
Other operating parameters, performance characteristics, and
development background are discussed.
-------
ACKNOWLEDGEMENTS
The author IB grateful to all those who have contributed to the
successful preparation of this report. Special appreciation is
extended to his graduate advisor, Professor Benjamin Linsky, for his
sincere interest in this study; and to Frank Noonan, then Director of
Air Pollution Control Engineering Laboratories at West Virginia
University, for his laboratory assistance and constructive criticism.
Gratitude is also expressed to Dr. Howard W. Butler, Chairman of
Mechanical Engineering, and Hasin T. Gencsoy, Professor of Mechanical
Engineering, for making special equipment and facilities of their
department available for this study. The cooperativeness of
Don Garletts and Harold Martin of the Mechaincal Engineering
Laboratory, and Fran Culler of the Civil Engineering Laboratory, during
the mechanical design and fabrication of the apparatus was sincerely
appreciated.
Special thanks is extended to Dr. M. R. Friedman of the
Anatomy Department, West Virginia University School of Medicine,
under whose guidance electromicroscopist Betsy Walker obtained
numerous electronmicrographs of the material produced in this
experiment. Further thanks is expressed to J. Reginald Dietz,
Director of Research and Development for National Steel Corporation,
Weirton, West Virginia, under whose direction R. E. Brien, Senior
Research Metallurgist, prepared electronmicrographs of material
collected from preliminary experiments.
-------
The author also appreciated the encouragement and assistance
given by Benjamin Euseblo and Bruce Harris, of what was then the
National Air Pollution Control Administration, Cincinnati, Ohio.
Several training grants, 5 T01 APOOOO 9-07 and 5 T01 APOOOO 9-08,
from what is now the Air and Water Programs Division of the
U. S. Environmental Protection Agency made this study possible. The
author is further indebted to the U. S. Bureau of Mines, Morgantown,
West Virginia, for supplying the helium used in these experiments
and to Two Thousand, Inc. for the use of the Andersen Stack Sampler.
1-iv
-------
TABLE OF CONTENTS
Page
ABSTRACT l_ii
ACKNOWLEDGEMENTS 1-iii
LIST OF TABLES 1-vi
LIST OF FIGURES 1-vii
I. INTRODUCTION 1-1
II. LITERATURE REVIEW 1-5
III. APPARATUS 1-10
IV. EXPERIMENTAL 1-25
SAMPLING PROCEDURE 1-25
RESULTS 1-29
Mass Concentration and Rate of Production 1-29
Particle Size Distribution 1-34
Reproducibility of Production Rate and Relative
Particle Size Distribution 1-49
V. CONCLUSIONS 1-52
VI. LIST OF REFERENCES 1-55
APPENDICES 1-58
APPENDIX A. Selected Bibliography of Readings Concerning
Electric Arc Discharges 1-A-l
APPENDIX B. Developmental Background of the Tungsten
Electrode Holder 1-B-l
APPENDIX C. Operating Procedure for Arc Aerosol
Generator 1-C-l
APPENDIX D. Electron Microscope Particle Count Raw Data . . . 1-D-l
APPENDIX E. Statistical Calculations 1-E-l
1-v
-------
LIST OF TABLES
Page
Table 1. Operating and sampling test conditions 1-27
Table 2. Weight (mass) distribution of particulate matter
retained by the Andersen Stack Sampler and
backup membrane filter 1-30
Table 3. Mass concentration and total production rate of
iron oxide particles 1-31
Table 4. Electron microscope particle count data 1-36
Table 5. Relative particle size frequency distribution in
percent from the data of Table 4 1-37
Table 6. Cumulative relative particle size distribution
from data of Table 5 1-43
Table 7. Arithmetic mean and standard deviation of the
cumulative relative mass distribution
from data of Table 2 1-46
Table 8. Arithmetic mean and standard deviation of the weight
(mass) distribution from data of Table 2 1-51
Table D-l. Electron microscope particle count raw data 1-D-l
Table E-l. Transformation for calculating the arithmetic mean
and standard deviation from data of Table 4 1-E-l
Table E-2. Tabulation for calculating median and geometric
mean from data of Table 4 l-E-2
Table E-3. Calculation of arithmetic mean and standard
deviation of the relative frequency within each
particle size interval from the data of Table 5 l-E-3
Table E-4. Calculation of arithmetic mean and standard
deviation of the cumulative relative mass
distribution from data of Table 7 l-E-7
Table E-5. Calculation of arithmetic mean and standard
deviation of weight (mass) distribution from
data of Table 8 l-E-8
1-vi
-------
LIST OF FIGURES
Page
Figure 1. Sketch of the components of a direct current arc .... 1-3
Figure 2. Graphical representation of thermal plasma in
the arc region 1-3
Figure 3. Simplified diagram of aerosol generator apparatus . . . 1-11
Figure 4. Photographs showing assembled and exploded view
of the non-consumable electrode holder 1-15
Figure 5. Photographs showing general arrangement of arc
chamber apparatus 1-19
Figure 6. Simplified pictorial sketch of the arc chamber
and ductwork system 1-23
Figure 7. Histogram representing particle size data
from Table 5 1-38
Figure 8. Distribution curve representing data from Table 5 ... 1-38
Figure 9. Ogive representing particle size data from Table 6 ... 1-39
Figure 10. Cumulative distribution data from Table 6 plotted
using normal probability scales 1-39
Figure 11. Cumulative particle size distribution data from
Table 6 plotted using log-normal
probability scales V. . . 1-40
Figure 12. Cumulative particle size distribution data from
Table 7 plotted using log-normal
probability scales 1-47
Figure B-l. Sketches showing developmental configurations of the
tungsten electrode-arc chamber apparatus l-B-3
Figure C-l. Simplified cross-sectional, plan view sketch showing
location of components inside arc chamber l-C-3
Figure C-2. Calibration curve of wire speed dial versus wire speed
of 0.045 inch diameter wire l-C-5
1-vii
-------
I. INTRODUCTION
The generation of significant quantities of solid aerosols with
reproducibility as to mass, shape and particle size distribution is
an essential sine gua non for research in air pollution control
techniques, instrument calibration, emission simulation, animal inhalation
and particle behavior studies, and in related fields of research (1,2,3,4).
Therefore, the gas-solid phase system of a test aerosol is usually made
up of components to simulate an actual natural or industrial situation.
Redispersion of collected, aged, and classified powders, the largest
present source of significant quantities of solid aerosols, requires
large energy inputs to overcome adhesive and agglomerating forces inherent
In bulk powders (5). Even then, experience has shown that laboratory
redispersions of particles exhibit different behavior characteristics
than freshly formed particles (6,7). Significant quantities of freshly
formed particles, synthesized in the same manner that they are produced
industrially, have not been available for research work. These factors
have caused specific difficulties in translating laboratory control
engineering research and medical research to field applications and
diagnosis.
One industrial process that is adaptable to the primary formation of
solid aerosols in the laboratory is the heating of materials by the
energy from a direct current electric arc. The heat is provided by an
arc between an electrode and the material to be melted (direct arc
heating) or by an arc between two electrodes (indirect arc heating).
1-1
-------
Because the source of most of the heat is nonchemical, electric arc
discharge heating is especially desirable in controlling contamination
of the heated material.
The arc discharge occurs when two electrodes are brought into close
proximity to one another such that a high electrical resistance is
developed at their boundary and the tips begin to glow. This glow
indicates the emission of electrons from the cathode, which because of
high emission temperatures, up to 7000 K (8,9,10), produces ionization
of the air. The bombardment of electrons received by the anode causes
it to become white hot and to erode either by melting or vaporization.
(Refer to Figure 1.)
In the actual direct current arc, the molecules of air dissociate.
They lose some of their orbiting electrons and form a mixture of posi-
tive ions and electrons. This mixture of electrically charged atoms,
and electrons, with their negative charge, is called thermal plasma
(Figure 2). As a result, the ionized air becomes conductive to elec-
tricity and current flows without a mechanical connection between
electrodes. Although a more detailed discussion of what happens in an
electric arc is beyond the scope of this paper, a bibliography has been
included in Appendix A listing specific readings which examine arc dis-
charges .
In industry, the intense heat produced by electron activity in the
electric arc is utilized for melting and refining of ferrous and nonfer-
rous metals, and for the production of refractories. As the electric arc
furnace is one of the most difficult to control sources of particulate
emissions in industry, it is an appropriate and adaptable method for the
simulation and primary formation of solid aerosols in the laboratory.
1-2
-------
Cathode (-)
Anode (+)
Direct Current
Electromotive Force
Figure 1. Sketch of the Components of a Direct Current Arc
Free Electrons
Q.o
o
o
Figure 2. Graphical Representation of Thermal Plasma
in the Arc Region
1-3
-------
Using the technique of direct arc heating, an electric arc aerosol
generator has been designed and built and has undergone limited tests
to evaluate and maximize its performance characteristics and the
properties of the aerosol produced. The generator appears suitable for
production of particles of metals, oxides of metals, and other refractory
materials, utilizing a constant potential, direct current power supply.
An arc is sustained between a relatively nonconsumable tungsten cathode
and a consumable feedstock of wire. Commercially available steel
welding wire is the consumable feedstock used in the present investigation.
The wire feed is maintained at a constant rate by a modified automatic
welding head mechanism with a variable speed controller. A specially
designed, water-cooled holder allows simultaneous manual advancement and
rotation of the tungsten cathode to compensate for and maintain uniform
erosion. The continuous arc discharge is contained in a 13 inch (33 cm)
length of 5 inch (12.7 cm) standard steel pipe through which passes the
entraining gas stream.
1-4
-------
II. LITERATURE REVIEW
Many methods have been utilized to produce solid aerosols in
significant and reproducible quantities for gas cleaning, animal
inhalation and other aerosol research. In the past, aerosols of
ultrafine solid particles made from materials to reproduce actual
environments have not been available; this has caused difficulty
in translating laboratory research to practical applications.
There are two basic types of aerosols: (1) aerosols which are
homogeneous with respect to size, shape, and specific gravity, and
(2) heterogeneous aerosols which simulate particles generated in
nature or by an industrial process. Homogeneous aerosols are used
for fundamental research and product testing and have uniform
properties in order to minimize the number of variables which must
be considered. This facilitates characterization of factors involved
in a study and structuring standardized tests for determining accept-
ability of control and test equipment and working atmospheres.
Although aerosols produced by homogeneous aerosol generators have
many inherent advantages, the output concentrations from these gener-
ators is in the range of 0.1 to 10 milligrams per liter with flow
rates of 1 to 4 liters per minute; further, dilution ratios of 10 to
100 are required to prevent agglomeration (11, 12). Consequently,
these units are not adequate for the evaluation of gas cleaning equip-
ment, where large volumes with high cdncentrations of particulate
matter are required, as the concentration would be too low for realistic
results.
1-5
-------
For most applied and engineering research, heterogeneous aerosols
are used. Silverman and Billings (13) have reviewed many of the
methods and problems associated with generating solid aerosols in the
laboratory for a variety of materials. A more complete treatise by
Silverman on the generation of heterogeneous aerosols can be found in
section 12.4.3 of the Air Pollution Handbook (14).
Of the many methods discussed by Silverman and Billings, only the
redispersion method is capable of producing large quantities of solid
aerosol for large scale laboratory testing. Large quantities — grains
to grams to tons — are available with a wide variety of physical,
electrical, and chemical characteristics for resuspension in a gas
/
stream.
The major difficulty in dry dispersion of particles is limiting
agglomeration. This limitation can be minimized by employing feeders
utilizing techniques discussed by Olive (15) and by considering
dispersion principles as discussed by Fuchs (16) to overcome bulk
powder adhesive and agglomerating forces. However, the lower limit
of particle diameter is 0.1 micrometers (17) and almost all aerosols
produced from dry powders seem to be heterogeneous with respect to
particle size.
The method which most closely approximates an actual industrial
situation is that of direct attrition. "Fresh" aerosols can be
produced by various mechanical and electro-mechanical means such as
friction, evaporation, impact, explosion and condensation and burning.
Small hammer and ball mills have been employed to produce fine
materials on the order of 1 micrometer diameter. And a number of
1-6
-------
eorapauJes, most notably manufacturers of air filtration equipment,
produce size reduction equipment usable as aerosol generators (18).
There are many other primary methods reported in the literature but
none have been capable of achieving wide acceptance because constant
and reproducible production rates are very difficult to achieve in
the submicron size range. At best, it is usually hoped a constant
volumetric particle feed rate can be achieved but the resulting
concentration and particle size distribution must still be determined
by measurement.
The problems of reproducibility appear to be directly attributable
to two major factors: (1) the control of the composition of the
feedstock, and (2) the control of input energy, whether it be mechanical,
chemical, electrical, and so forth.
In the aerosol generator under present consideration the feedstock
is "common" welding wire available with various metal bases (aluminum,
iron, chromium, copper, and nickel) in a wide range of trace metal
compositions. Feed rate is controllable by one of a number of
commercially available, constant speed, wire-feed motors. The energy
input is electrical and is supplied by a constant voltage (CV) direct
current power supply. An electrical potential is created between a
consumable wire feedstock (the anode) and a relatively non-consumable
tungsten electrode (the cathode). When the anode and cathode are
brought in close proximity to each other, an intense electric arc is
initiated which melts and evaporates a portion of the feedstock
material. After leaving the vicinity of the electric arc, the
vaporized metal rapidly condenses producing a desirable ultra-fine
1-7
-------
metal fume which can be conveyed by an air stream to sampling equip-
ment, control devices, inhalation subjects, and so forth.
Although the use of an electric arc has been previously adapted for
production of submicron particulates, vaporization of materials by this
method has been primarily to obtain quantities of submicron particles
as an end'product in themselves. For example, Holmgren, et. aJ. (19)
used various electric arc systems to generate particulates to be col-
lected for later study of the particles using chemical analysis, x-ray
diffraction, chemical reactivity tests, sinterability tests and electron
microscopy. This and similar work by Harris, et. al. (20) has been
performed to determine commercial applications and production methods
for particulate materials. No consideration was given to maintaining
the gas-solid phase system as a viable tool in itself.
The TAFA Division of the Humphreys Corporation (21) has developed a
commercial carbon arc processing system capable of producing spherical
particles. However, this system requires preliminary size reduction of
the feed material as the feedstock must be in powder form. The end
product in this system, once again the particles themselves, is produced
by melting the powder (as compared to vaporization and subsequent con-
densation) whereupon each discrete particle assumes a spherical config-
uration. It is apparent that the size and mass of the spheroidal
particles produced is directly limited by the size characteristic of
the powder fed. Also, as opposed to a workable aerosol generator, the
design emphasis of this unit is on the generation and immediate col-
lection of the particles themselves. The inclusion of air or other gases
is usually undesirable for their purposes.
1-8
-------
In addition to carbon arcs, high voltage condenser discharge through
thin foil and wire (22), and high voltage sparks (23) have been used to
produce metallic aerosols for study in chambers. Truitt et. al. (24,25),
under several studies for the United States Atomic Energy Commission,
have used two "spark generators" which differ only in the power source.
A Tesla coil was used as a power source for one and the high-frequency
starter attachment on a welding machine supplied the power for the other.
The electric "spark was set up across the axis of a 2-in. diameter glass
pipe" between various metal electrodes. However, the design of the spark
chamber limited aerosol generation to about one minute because of the
mechanical limitations of electrode consumption. To avoid the problems
of non-uniform and non-sustained generation, tanks, plastic bags, and
stainless steel drums were used as holding chambers from which relatively
small volumes, at rates up to 2 liters per minute, were withdrawn. Not
only are mass concentrations and flow rates low and suitable only for
small scale laboratory studies, but the problems of maintaining a con-
stant aerosol particle size distribution by achieving a steady-state
between agglomeration and preferential settling of these larger particles
must be considered. Even then, the mass concentration of the aerosol
will decrease due to settling.
In the present apparatus, the undesirable, as well as the desirable,
characteristics of other uses of electric arcs have been integrated
into the prototype of a continuous, steady state, and reproducible source
of relatively large quantities of fresh metallic aerosol.
1-9
-------
Ill. APPARATUS
The apparatus, diagrammatically sketched in Figure 3, consists of
an adjustable, constant voltage, direct current power supply, a wire
feed assembly with control panel, a non-consumable (tungsten) electrode
holder with cooling system, an arc chamber, a shielding gas supply
system, and a ductwork system connecting the chamber to an air mover
upstream and a settling chamber and exhaust system, with sampling
port, downstream.
The conventional use of the power supply utilized in the present
apparatus is to provide power for semiautomatic welding processes such
as, gas shielded, open arc, and submerged arc. This "constant voltage
(CV)" type of power source is a direct current, self-contained motor
driven generator unit rated for a 100% duty cycle. This rating allows
continuous, uninterrupted production of power under full load. Various
instruments and controls are mounted on the front of the sheet metal
enclosure of the motor-generator unit. In addition to an ammeter and
voltmeter, a 110 volt receptacle, fuse holder, remote control receptacle,
and electrode and ground terminals are located on the welding machine.
Controls include start-stop pushbuttons and a field rheostat to regulate
the output voltage. The voltmeter located on the machine reads open
circuit voltage (while not welding) and arc voltage (while welding).
Once set, the output of a CV power supply has essentially the same
voltage no matter what the welding current may be. The voltage can be
iflobart Motor Driven Constant Voltage Welder for Automatic Welding,
Model MC-900, Hobart Brothers Company, Troy, Ohio.
1-10
-------
WIRE REEL
CONTROL
PANEL
SHIELDING
GAS
SUPPLY
WIRE TORCH
ARC CHAMBER
SETTLING
CHAMBER
SHIELDING
GAS NOZZLE
TUNGSTEN
ELECTRODE
HOLDER
POWER SUPPLY
COOLING
WATER
TANK
Figure 3. Simplified diagram of aerosol generator apparatus
-------
set over a range of approximately 10-50 volts; there is no current
control on a CV type machine as the welding current flow is determined
by the wire feeder. That is, as the wire feed speed is increased, the
machine automatically supplies additional current, at the same constant
voltage, to maintain the arc (26, 27). Thus, the wire feed speed
control knob "sets" the welding current. The welding current is read
on either the ammeter on the welding machine or the ammeter on the
remote control panel for the "automatic welding head."
The automatic welding head assembly^ consists of two main components:
(1) wire feed assembly with torch, and (2) remote control panel. The
wire feed assembly combines an insulated wire supply reel, guides, drive
motor with gear reduction box, and feed rolls, to constantly feed a
consumable electrode wire^ through the air-cooled torch to the arc. The
consumable electrode wire used in the present apparatus is solid steel
(because an iron oxide fume was desired) and bare, except for a very
thin coating of copper to prevent surface oxidation. The consumable
wire is maintained at a positive (+) potential with respect to ground.
The remote control panel is electrically connected to both the welding
^Hobart Automatic Welding Head, Model AI-22, Hobart Brothers Company,
Troy, Ohio.
3The steel wire used is Hobart, Type 18 (E 705-G). The spool was
further designated "6719-67, MIG 18B, .045." A typical wire trace
element composition by per cent is given (27)
C .12
Mn 1.90
Si .80
P .020
S .020
Mo .50
1-12
-------
machine and the non-consumable electrode holder cooling water supply
through solenoid valves.
The front of the control panel contains the following controls
and instruments (28).
•'•• JLuse ~ This 4 ampere fuse fuses the input power to the control
panel from the welding machine.
2. Ammeter - This ammeter indicates the amount of current flow
at the arc.
3. Wire Speed Control - This controls the speed of the wire through
the torch. (Figure C-2 is the calibration curve for this
control).
4. Voltage Control - This controls the amount of voltage across
the arc.
5. Voltmeter - This voltmeter indicates the amount of voltage
across the arc.
6. Pilot Light - This light indicates when power is available at
the control panel.
7. Wire Feed Control - This switch controls the direction of travel
of the wire through the torch.
8. Inch Button - By depressing this button, the wire inches through
the torch, forward or reverse, depending on the position of the
"wire feed control."
9. Start and Stop Button - These start and stop the wire feed motor
and therefore start and stop the arc.
10. Purge Button - This button purges the gas through the shielding
gas nozzle.
1-13
-------
11• On-Off Toggle Switch - This switch controls power to the control
panel.
12. Water Valve - This valve controls the on and off cycle of water
to the non-consumable electrode holder.
13. Gas Valve - This valve controls the on and off cycle of gas to
the shielding gas nozzle.
The non-consumable electrode holder, shown in an exploded view in
Figure 4, was designed and fabricated by the author. Materials of con-
struction include stainless steel, steel and brass as indicated in the
diagram. The non-consumable electrode is a 1/4 inch (6.35 mm) diameter
tungsten arc welding electrode with a ground finish and 2 per cent thoria
content. The ground finish is smoother and therefore makes better
mechanical and electrical contact than the swaged type. Thoria is added
to tungsten electrodes to improve arc stability, to make initiation of
an arc easier, and to increase the amount of current carried per electrode
diameter (29).
Several preliminary configurations of the tungsten electrode holder-
arc chamber combination were constructed to investigate the feasibility
of developing a workable semi-continuous model capable of achieving the
desired result. Many of the problems encountered in the construction and
operation of the models described in Appendix B have thus been eliminated
in the construction of the present holder. Final design of the tungsten
electrode holder involved three prime considerations: (1) compensation
for electrode erosion, (2) electrical connection between power supply and
electrode, and (3) dissipation of heat build-up.
1-14
-------
Figure 4. Photographs showing assembled and exploded view
of the non-consumable electrode holder
Figure 4a. Assembled view
Figure 4b. Exploded view
PARTIAL PARTS LIST
1 Electrode guide (S)
2 Sliding seal components
3 Tungsten electrode
4 Feed screw (B)
5 Electrode holder body (M)
6 Cooling water fittings (B)
7 Rubber "0" ring
8 Packing retainer (A)
9 Silicone rubber packing
10 Feed screw guide (S)
11 Rubber bellows
12 Jam nut and lock washer (M)
13 Compression spring
14 Electrical connector
15 Rotating contact disks (B)
16 Plexiglas insulator
17 Feed screw knob
Letter in parenthesis indicates material of construction:
aluminum (A), stainless steel (S), brass (B), and mild steel (M)
1-15
-------
Among other reasons, tungsten was chosen as a "non-consumable"
electrode for its inherent property of having the highest melting
point of all metals, 6170 F (3410 C) (30). But even with its
refractory nature, the intense heat of the arc, the availability of
air for oxidation, and the physical forces in the arc cause erosion
of the electrode. The major portion of the erosion is due to melting
of the electrode and subsequent removal of the molten material from
the arc region by gravitational force, arc blow, and the moving gas
stream. It is theorized a much smaller portion of the tungsten is
lost by evaporization as its boiling point is 10,700 F (5927 C) (30).
No trace of discrete tungsten particles could be detected on examin-
ation of the electronmicrographs of the collected fume as condensed
tungsten has a morphology similar to that of iron oxide.
To advance the electrode into the arc, a manually operated screw
feed with a pitch of 20 threads per inch (7.87 threads per cm) was
employed. To compensate for the increased erosion rate on the down-
stream side of the electrode due to arc blow, the electrode was fixed
to the feed screw causing the electrode and screw to rotate together.
This arrangement allowed a relatively rounded point to be maintained
in the arc throughout the tests.
Transfer of power from the stationary power supply to the rotating
feed screw was accomplished through a sliding contact between two
4 inch (41.7 cm) diameter spring-loaded, brass disks (see Figure 4).
A thin coating of graphite paste was placed between the disks for
lubricating purposes and to improve conductivity. During the test the
1-16
-------
feed screw was turned approximately four revolutions per minute; upon
disassembly no trace of arcing was observed between the contact
surfaces of the disks.
To take advantage of electron flow, the tungsten electrode is
grounded (-) making it the cathode of the direct current circuit.
The characteristics of electron flow, negative to positive potential,
cause the consumable wire (anode) to be bombarded with electrons from
the relatively non-consumable cathode thereby concentrating the heat
in the wire to be vaporized. At a current flow through the arc region
of 150 amperes, the ratio of current density at the consumable wire
versus that for the non-consumable tungsten electrode is approximately
30 to 1. Current density is defined as the current flow in amperes
divided by the cross-sectional area of the conductor. The current
density for the wire is,
current _ 150 amperes x 4
area IT (0.045 inches)2
2
current density (wire) = 94,300 amperes/inch
(= 14,620 amperes/cm2).
The current density in the tungsten electrode is,
current _ 150 amperes x 4
area ir (0.25 inches)^
current density (tungsten) = 3056 amperes/inch2
(= 474 amperes/cm2)
As this is a direct current arc, the power (P) consumed in the arc
is the product of electro-motive force (E), in volts, and current (I),
in amperes.
1-17
-------
Therefore,
P = El
P = 23 volts x 150 amperes
P = 3450 watts or 11,775 Btu/hr
(= 49.44 kg-calories/min).
This relatively high rate of heat input to the system must be
dissipated by one of the various methods of heat transfer: convection,
radiation, or conduction. Air flow through the arc chamber past the
arc region at the rate of 103 scfm (2.92 m3/min) supplies significant
cooling for the overall system. However, the major source of concern
for heat build-up to adverse levels is the tungsten electrode holder
itself. An easily replaceable 1/8 inch (3.175 mm) thick steel disk
(see Figure C-l) is used to protect the exposed end of the electrode
holder from arc splatter and to reflect the radiant energy of the arc.
To dissipate the heat build-up in the tungsten electrode, a portion of
which is conducted to the electrode holder through intimate metallic
contact, the entire electrode holder is water cooled. Cooling water
is supplied by a self-contained tank and pump assembly , through a
solenoid valve located on the control panel assembly to the inlet port
on the electrode holder. The return water flows unrestricted back to
the holding tank which acts as a heat sink.
No attempt was made to cool the arc chamber itself as no problems
of dimensional stability due to heat build-up were encountered. The
arc chamber was fabricated from a 13 inch (33 cm) length of 5 inch (12.7 cm)
Hobart Circoolator, S-3568B, Hobart Brothers Company, Troy, Ohio.
1-18
-------
Figure 5a.
Electrode holder being inserted
into mounting block of arc chamber
Figure 5.
Figure 5b. Arc chamber mounting
arrangement with electrode
holder in place
Photographs showing general
arrangement of arc chamber apparatus.
-------
diameter, Schedule 40, steel pipe. The pipe axis was supported in
a horizontal plane by steel supports (see Figure 5) attached to the
chamber by stainless steel clamps; the steel supports were then
bolted to a 20 x 24 inch (50.8 x 61.0 cm), 1/2 inch (12.7 mm) thick
aluminum plate. (The wire feed assembly with torch are also bolted
to this base plate through insulators. This arrangement assured
positive orientation between wire feed torch, tungsten electrode
holder and arc chamber.) The attachments between the chamber and
the base plate, and between the wire feed assembly and base plate,
are designed to allow adjustment in all axes to obtain perfect
alignment between the feed wire and the non-consumable electrode.
The feed wire torch passes through the chamber wall 90 degrees
to both the center line axis and the vertical tangent of the circular
chamber. A concentric phenolic collar insulates the positive potential
of the wire torch from the ground (-) potential of the chamber, base
plate, and electrode holder. The tungsten electrode holder passes
through an aluminum mounting block, milled to the contour of the arc
chamber and bolted in place, on the same axis as the wire torch. A
sliding fit exists between the mounting block and electrode holder,
the latter being held in place by set screws. Various adjustments
allow the arc to be maintained on the center line of the chamber.
Refer to Appendix C, Figure C-l for a detail of the positioning of the
various components inside the arc chamber.
A 2 x 4 1/4 inch (5.1 x 10.8 cm) view window is located on the top
of the chamber to allow visual evaluation of tungsten electrode erosion
1-20
-------
and observation of arc shape. The window is a standard lift-front,
lens holder from a weldor's helmet modified to fit the chamber. A
Shade Number 12 welding lens is used in the holder as protection from
the injurious rays of the arc. The lift-front feature of the lens
holder permits raising the window for a clear view of the interior
of the chamber and the opening allows insertion of tools for making
adjustments within the chamber.
Another salient feature of the arc chamber is the shielding gas
nozzle and supply system. The shielding gas is supplied by a standard
steel cylinder containing approximately 244 standard cubic feet
(6.91 m3) of gas under a pressure of 2200 psig (154.7 kg/cm2). To
reduce cylinder pressure to a usable level and assure constant flow,
a single stage regulator and flowmeter is used and is mounted on top
of the cylinder. A valve is used in the line to control flow through
the flowmeter which is graduated in standard cubic feet of helium per
hour (cfh). The gas flows to a solenoid on the control panel which
shuts off the gas supply after the arc is extinguished, thus conserving
the inert gas. From the solenoid, the gas flows to a ceramic nozzle
located in the arc chamber which directs the gas at the tip of the
tungsten electrode. Effective shielding of the tungsten electrode dis-
places the air surrounding the electrode; preventing rapid consumption
of the tungsten by oxidation. However, effective shielding is a dif-
ficult problem. Turbulence in the inert gas stream itself tends to
pull in air; this problem is compounded by the relatively high velocity
air stream being forced past the arc region. In an attempt to compensate
for gas stream turbulence and maintain effective coverage of the electrode
1-21
-------
tip, an unusually high gas flow is used - 100 cubic feet per hour
(47 1/min). Because of the relatively high rate of electrode consump-
tion experienced it is still doubtful whether adequate coverage was
obtained.
The usual choice of inert gas for shielding during tungsten inert
gas. (TIG) arc welding is between argon and helium. Each gas has its
own particular advantages for different types of welding processes and
materials. In the present apparatus, helium was the choice because of
its availability. However, the choice of helium over argon appears
advantageous for several reasons: (1) helium allows a 40 percent
greater heat input per unit length of arc because of its higher ioniza-
tion potential, and (2) it gives a more stable arc when using direct
current power.
Figure 6 is a pictorial diagram of the arc chamber and its associated
ductwork. Air flow through the system is provided by a "Gelman Hurricane
Air Sampler" set on low speed. The air discharge of the blower passes
through a 1 1/8 inch (28.6 mm) orifice, which limits flow to a nominal
100 cubic feet per minute (47.2 liters/second), through a detachable
muffler into a 4 inch (10.2 cm) diameter horizontal duct. All ductwork
throughout the system is fabricated from 22 gauge (0.794 mm) galvanized
steel sheet. A transition piece connects the 4 inch (10.2 cm) diameter
duct to a 2 foot (0.61 m) length of 5 inch (12.7 cm) duct leading directly
to the arc chamber. From the chamber, a 2 foot (0.61 m) length of 5 inch
round duct diverges into a 10 inch (25.4 cm) diameter section over a
distance of 2 feet (0.61 m) and intersects a vertical duct at a 90 degree
angle forming a tee. The bottom of the tee is closed by a removable
-------
t
KJ
U)
ELUTRIATION
COLUMN
10"O.D.
VIEW
WINDOW
EXHAUST
SETTLING CHAMBER
AIR
MOVER
TO POWER SUPPLY
Figure 6. Simplified pictorial sketch of the arc chamber and duct work system
-------
cap; the top of the tee converges over a 2 foot (0.61 in) length to
a 3 inch (7.62 cm) diameter duct.
To separate the larger molten globules of steel from the gas
stream, the ductwork leading from the arc chamber incorporates a
settling chamber and elutriation column section in its design. The
actual velocity in the 10 inch (25.4 cm) diameter vertical duct is
220 feet per minute (67.1 m/min) or 3.7 feet per second (1.12 m/sec).
At this velocity steel particles larger than approximately 1/8 inch
(3.175 mm) diameter are effectively eliminated from the gas stream by
gravitational force. Further, inertial separation takes place in
this section as the gas stream is caused to make a 90 degree change
in direction.
A view window, again adapted from an arc welder's helmet, is located
on the vertical portion of the tee in such a manner as to allow observa-
tion of the arc by sighting through the inside of the 5 inch (12.7 cm)
diameter duct.
The 3 inch (7.6 cm) duct leaving the elutriation column makes a
180 degree turn, and passes a 1/2 inch (12.7 mm) sampling nozzle. An
adequate length of straight duct was established to allow optimum
positioning of the sample probe — a minimum of 10 diameters downstream
and 5 diameters upstream from any disturbance to air flow. The 3 inch
(7.62 cm) duct then diverges into a 4 inch (10.2 cm) duct, makes a
180 degree turn and exhausts the contaminated gas stream to the atmos-
phere through a roof ventilator.
Temperature sensors (bimetallic helix, dial thermometers) were placed
in the duct at locations where significant temperature changes were
1-24
-------
expected to occur. A thermometer was also placed in close proximity
to the sampling port. Location of thermometers is indicated on
Figure 6.
* * *
A detailed procedure for the assembly and operation of the electric
arc aerosol generator apparatus is found in Appendix C.
1-25
-------
IV. EXPERIMENTAL
To Investigate the performance characteristics of the aerosol
generator under semi-continuous operation, a series of three con-
trolled test runs were made and a portion of the resultant fume was
collected. All tests were conducted under identical conditions of
arc voltage, wire feed rate, shielding gas flow rate, and dilution
air flow rate. Refer to Table 1 for a complete listing of test
conditions. The test settings for voltage and wire feed were estab-
lished on the bases of limited preliminary studies and are not claimed
to be optimized values. The test program was structured to determine
the mass rate of production of iron oxide fume, the particle size
distribution of that fume, and the degree of reproducibility of pro-
duction rate and size distribution.
SAMPLING PROCEDURE
After the arc was initiated according to the procedure outlined in
Appendix C, the system was allowed to operate for approximately one
minute to establish steady-state conditions for sampling. During this
period the temperature of the exhaust gas stream stabilized producing
a constant velocity past the sampling nozzle. A ten minute sampling
period then commenced.
The sampling train consisted of a one-half inch (12.7 mm) stainless
steel hook nozzle attached to an Andersen Stack Sampler-5, in series
5Andersen Stack Sampler manufactured by 2000 Inc., 5899 South State
Street, Salt Lake City, Utah 84107.
1-26
-------
Table 1. Operating and sampling test conditions
OPERATING PARAMETERS English Units Metric Units
Wire Feed Rate 200 inches/min 5.1 meters/min
Wire Diameter 0.045 inches 1.14 mm
Cathode Feed Ratea 0.2 inches/min 5.08 mm/min
Cathode Diameter 0.25 inches 6.35 mm
Arc Chamber Pressure +0.72 inches W.C. +1.34 mm Hg
Shielding Gas Flow Rate 100 cubic feet/hr 47.2 liters/min
Air Flow Rate 103 scfm 2.9 cubic meters/min
A power supply voltage control setting of 23 volts produced a current
flow through the arc which fluctuated between 150 and 165 amperes.
SAMPLING PARAMETERS
Sample Time 10 minutes per run
Sample Rate 0.72 cfm @ 70F 20.41 1pm @ 21C
Exhaust Gas Temperature 167F 75C
at Sample Port
Sample Nozzle Diameter 0.25 inches 6.35 mm
Ambient Temperature 70F 21C
aCathode (tungsten electrode) was manually advanced approximately
four turns per minute to compensate for erosion.
1-27
-------
with an aluminum Millipore filter holder supporting a 47 mm diameter
Gelman Type GA-4 membrane filter to separate the fume from the sample
gas stream.
The Andersen Stack Sampler is approximately 2.75 inches (6.99 cm)
in diameter and 5.0 inches (12.7 cm) in length, and weighs 3 pounds
(1.36 kg). Construction is of stainless steel to withstand high
temperatures and resist corrosion. The sample contains nine stages
which are numbered 0 through 8, the number 8 stage having the same
characteristics the number 2 stage has. The stages are 2.5 inches
(6.35 cm) diameter gold plated brass plates with asbestos gaskets
between the plates for spacing and sealing. Each stage has 300 air
jets (holes) arranged in concentric circles, the holes decreasing in
size moving from 0 through 7 (31). As the air flow through each stage
is the same, the velocity of air through the air jets increases as hole
size decreases. When the velocity imparted inertia of a particle is
sufficient to overcome the aerodynamic drag, the particle will leave
the gas stream and impact on the collection surface. Therefore,
selective impaction of a particle on a particular plate depends on its
aerodynamic dimensions - size, shape and density - and their relation-
ship to gas velocity (32).
A calibrated orifice was used to indicate instantaneous sample flow
rate, with total flow rate recorded on a dry gas meter.
Fume concentration was determined gravimetrically. After the sample
was collected on the pre-weighed collection plates of the Andersen
sampler and the back-up membrane filter, the plates and filter were
1-28
-------
dried at 248 F (120 C) for two hours. The plates and filter were then
removed to a. dessicator containing indicating type silica-gel and
allowed to come to room temperature. When the plates and membrane fil-
ter were removed for weighing they were immediately placed on the en-
closed balance pan and weighed. A small container of indicating-type
silica-gel was placed in the balance enclosure. After the gravimetric
determinations were made for Run 1, the collection plates were cleaned
using minimal amounts of distilled water and a rubber policeman. The
wash water from each plate was collected in separate test tubes, and
the particulate allowed to settle for one week; the bulk of the water
was then decanted and the remaining water evaporated in a drying oven.
Each sample of collected fume was then resuspended in 2 ml of a 1%
nitrocellulose solution and dispersed by placing the test tube in an
ultrasonic cleaner. The fume collected on the membrane filter was
carefully scraped from the membrane and placed directly in 2 ml of the
nitrocellulose solution and dispersed.
A disposable pipette was used to place one drop of the solution on
an uncoated electron microscope grid. A separate grid was prepared
using the solution containing particles from each collection plate
and the membrane filter from Run 1. The 200 mesh grids (specimen
screens) were bombarded by electrons and the projection recorded on
photographic film for later particle size determination. An RCA
Type EMU-4 electron microscope was used and produced a primary
magnification of 29,200 times.
6No. 1120 Nitrocullulose Solution, 1%, 30 ml bottles. Ernest F.
Fullan, Inc., P.O. Box 444, Schenectady, New York.
1-29
-------
Table 2. Weight (mass) distribution of particulate matter retained
by Andersen Stack Sampler and back-up membrane filter
STAGE
0
1
2
3
4
5
6
7
8
Filter
TOTAL
RUN 1
Weight
(grams)
.0025
.0004
.0004
.0008
.0013
.0027
.0081
.0164
.0090
.0916
0.1332
Percent
Total
1.88
0.30
0.30
0.60
0.98
2.03
6.08
12.31
6.76
68.77
100.01
Cumulative
Relative
Frequency
(Percent)
100.01
98.13
97.83
97.53
96.93
95.95
93.92
87.84
75.53
68.77
RUN 2
Weight
(grams)
.0015
.0003
.0002
.0002
.0007
.0031
.0077
.0112
.0128
.0861
0.1238
Percent
Total
1.21
0.24
0.16
0.16
0.57
2.50
6.22
9.05
10.34
69.55
100.00
Cumulative
Relative
Frequency
(Percent)
100 . 00
98.79
98.55
98.39
98.23
97.66
95.16
88.94
79.89
69.55
RUN 3
Weight
(grams)
.0024
.0005
.0004
.0005
.0015
.0033
.0084
.0174
.0134
.1034
0.1521
Percent
Total
1.58
0.33
0.26
0.33
0.99
2.17
5.52
11.44
8.81
68.57
100.00
Cumulative
Relative
Frequency
(Percent)
100.00
98.42
98.09
97.83
97.50
96.51
94.34
88.82
77.38
68.57
Co
o
-------
Table 3. Mass concentration and total production rate of iron oxide particles
CD (2) (3) (?) (5) (6) (5) .(8) (9)
RUN
1
2
3
AVE
Table 1
Sample
Time
(mins)
10
10
10
10
Total
Sample
Volume
(scf)
7.23
7.21
7.17
7.20
Table 1
Sample
Weight
(grams)
0.1332
0.1238
0.1521
0.1364
15'43%
22880
Mass Concentration
(gr/scf)
0.284
0,265
0.327
0.292
(mg/m )
649.8
606.3
748.2
668.1
Table 1
Total
Gas
Flow
(scfm)
103
103
103
103
©X{6)
.008570
Total Particulate
(gr/min)
29.25
27.30
33.68
30.08
(Ibs/hr)
0,251
0.234
0.289
0.258
.06480
3.890
Production Rate
(g/min)
1.90
1.77
2,18
1.95
(g/hr)
113.8
1Q6. 2
131.0
117.0
-------
RESULTS
In all test runs to determine performance characteristics,
identical apparatus and operating and sampling conditions were
maintained as listed in Table 1.
Mass Concentration and Rate of Production
The mass of particulate in the sample gas stream was deter-
mined from the gravimetric differential between the tare weight of
the clean Andersen collection plates and membrane filter and the
weight of the plates and filter after sampling. The weight
differential for each of the tests runs is shown in Table 2.
Combining the information from Table 2 with knowledge of the total
gas volumei'sampled and the sample time, the mass concentration of
particulate matter in the sample gas stream was calculated. Average
mass concentration of iron oxide fume in the sample gas stream was
Q
0.292 grains/scf (668.1 mg/m ). Assuming the mass concentration of
particulate matter in the total gas stream is the same as for the
sample gas stream, the average total particulate production rate was
calculated to be 30.08 grains/min (117.0 grams/hr). Additional
information concerning mass flow rate and mass concentration cal-
culated for each run for both the sample and total gas streams is
presented in Table 3. Total particulate production rate, given in
various common units of measurement, is also listed in Table 3.
Another performance factor directly related to the rate of pro-
duction is the conversion ratio of consumable wire feed stock to
ultrafine particles. For the feed rate of 200 inches (5.1 m) per
1-32
-------
minute of 0.045 inch (1.14 mm) diameter mild steel wire, 12 thousand
inches (305 m) are consumed in one hour. (Refer to Table 1). As
there are 2210 inches per pound (25.46 m/kg) of wire (33), the wire
consumption rate was
12.000 inches/hr ,
2,210 inches/lb = 5'43 lb/hr <2'46 k§/hr>-
From Table 3, the average particulate production rate was
0.258 lb/hr (117.0 g/hr). Therefore, 5.43 pounds (2463 g) of wire was
required to produce 0.258 pounds (117.0 g) of ultrafine particles for
a conversion ratio of 4.75 percent. Although no complete mass balance
was performed, the remaining portion of consumed metal can be accounted
for by the following observations. A relatively large portion of the
feed stock was not vaporized as some material melted and formed a
puddle in the bottom of the arc chamber. A significant number of
relatively larger particles, approximately 1000 micrometers, settled
out in the settling chamber and elutriation column section of the
ductwork. The loss of smaller particles is due to the electrophoretic
and thermophoretic forces acting on the particles suspended in the gas
stream. These forces cause the smaller particles to migrate to the
walls of the ducting and attach themselves through various adhesive
and electrostatic forces. A thin reddish-brown coating was observed
on the inside of the ductwork after completion of the test program.
It is also interesting to note the number of discrete spherical
particles generated at the above production rate. Considering a mass
mean diameter of 0.022 micrometers and specific gravity of approximately
1-33
-------
5.2 (34), the numerical generation rate is calculated in the
following manner.
Volume of single sphere = i^d3
6
" !ir(0.022)3 = 5.58 x l(f6 pm3
6
Specific gravity of Fe203 =5.2
rt O
Weight of single sphere = 5.58 x 10"6 pm3 x ICT12 cm ^m x 5.2 g/cm3
—18
= 29.0 x 10~ grams
Generation rate = _ 117.0 g/hr _ ,, . .18 . . „
- TH - ~ 4 x 10 particles/hr
29.0 x 10 g/particle
This is more than four quintillion particles per hour.
Particle Size Distribution
Two methods were employed, with differing results, to determine
particle size distribution: (1) an electron microscope and
(2) an Andersen Stack Sampler.
Because of the high magnification, and consequently small field,
of the electron microscope, photographic plates were made of a mini-
mum of three fields of each grid. The photographic plates were en-
^ *
larged and printed on 8 x 10 inch (20.3 x 25.4 cm) photographic paper
giving a final magnification of 87,600 times. The electron micrographs
obtained in this manner have a two-dimensional appearance and discrete
particles are readily discernible. Individual particles of iron oxide
are spherical* and individual spherical particles composing larger
agglomerates were easily distinguished. The particle size
1-34
-------
distribution data presented in Tables 4 and 5 and graphically
represented in Figures 7, 8, 9, 10, and 11 are based on direct
counts of discrete, spherical particles. That is, agglomerates are
not assigned a composite or equivalent particle size but the
particles composing an agglomerate are classified individually.
A specially constructed semi-circle graticule was used to facilitate
particle sizing.
Examination of the raw particle count data found in Appendix D
lists thirty fields for which particle size determinations were
made. The number of discrete particles observed on each 2.275 x
2.875 micrometer field ranged from 177 to 742 with an arithmetic
mean of 412. A total of 12,355 particles were individually sized.
In order to simplify calculation of various statistical values,
the raw data was grouped into 10 classes with a particle size
interval of 0.01 micrometers each. Although the following mathe-
matical descriptions of particle size were calculated on the assump-
tion that all particles were classified in one of ten equal-interval
classes, this was not exactly the case. As a small number of particles
exceeded the upper boundary of the tenth class (0.105 micrometers),
the tenth class was actually open-ended and contained all particles
0.095 micrometers in diameter and larger. This assumption introduces
a slight error in the following calculations, but this error is
relatively insignificant as only 30 particles of the 12,355 particles
counted exceeded 0.105 micrometers. Three of these particles exceeded
0.2 micrometers at 0.22, 0.25 and 0.24 micrometers and were found on
stages 5, 6, and 7, respectively.
1-35
-------
Table 4. Electron microscope particle count data'
STAGE
0
1
2
3
4
5
6
7
8
Filter
TOTALS
RANGE
Particle Size, micrometers
0.005-
0.015
522
351
709
431
773
592
345
436
749
818
5726
493
0.015-
0.025
248
183
249
164
419
285
149
209
336
421
2663
270
0.025-
0.035
153
145
161
123
308
210
126
149
233
300
1908
185
0.035-
0.045
87
56
91
56
157
68
48
87
98
159
907
111
0.045-
0.055
51
37
46
36
56
47
25
50
97
86
531
72
0.055-
0.065
24
21
27
19
39
33
20
21
20
68
292
49
0.065-
0.075
17
12
12
9
13
29
5
11
9
27
144
20
0.075-
0.085
15
3
3
7
10
4
2
10
7
15
76
13
0.085-
0.095
1
2
2
5
5
3
3
2
4
7
34
5
0.095
and>
9
7
5
1
16
8
5
9
6
8
74
15
TOTALS
1127
817
1305
851
1796
1279
728
984
1559
1909
12355
1181
1
LJ
Refer to APPENDIX D for raw particle count data.
-------
Table 5. Relative Particle Size Frequency Distribution in Percent from data of Table 4
STAGE
0
1
2
3
4
5
6
7
8
Filter
Mean
Standard
Deviation
Range
Particle Size, micrometers
0.005-
0.015
46.3
43.0
54.3
50.5
43.0
46.3
47.4
44.3
48.0
42.8
46.6
3.73
11.3
0.015-
0.025
22.0
22.4
19.1
19.2
23.3
22.3
20.5
21.2
21.6
22.1
21.4
1.38
4.2
0.025-
0.035
13.6
17.7
12.3
14.4
17.1
16.4
17.3
15.1
14.9
15.7
15.5
1.72
5.4
0.035-
0.045
7.7
6.9
7.0
6.6
8.7
5.3
6.6
8.8
6.3
8.3
7.2
1.15
3.5
0.045-
0.055
4.5
4.5
3.5
4.2
3.1
3.7
3.4
5.1
6.2
4.5
4.3
0.92
3.1
0.055-
0.065
2.1
2.6
2.1
2.2
2.2
2.6
2.7
2.1
1.3
3.6
2.4
0.59
2.3
0.065-
0.075
1.5
1.5
0.9
1.1
0.7
2.3
0.7
1.1
0.6
1.4
1.2
0.52
1.7
0.075-
0.085
1.3
0.4
0.2
0.8
0.6
0.3
0.3
1.0
0.4
0.8
0.6
0.36
1.1
0.085-
0.095
0.1
0.2
0.1
0.6
0.3
0.2
0.4
0.2
0.3
0.4
0.3
0.16
0.5
0.095
and>
0.8
0.9
0.4
0.1
0.9
0.6
0.7
0.9
0.4
0.4
0.6
0.27
0.8
99.9
100.1
99.9
99.7
99.9
100.0
100.0
99.8
100.0
100.0
100.0
Refer to APPENDIX E, Table E-3 for calculations.
-------
Figure 7. Histogram representing particle size data from Table 5
60*
, percent
.e- m
0 0
i i
1 30-
-------
l''lKur<» (). Ogive representing cumulative distribution data from Table 6
c 100""
eg
£ 90"
en
8 80-
g
CJ
Vi
0)
PL.
70-
60-
I 40
0)
°" 30 —
a) -ju—i
(U
>
CO
20-
10-
Median
CO
n
a)
4J
§
o
Vf
u
Q)
I
M
T3
H
O
.01
Figure
.10-
.09-
.08
.07-1
.06
.05-
.04-
.03-
.02 -
.Ol-
I
.02
' I i^ I IT
.03 .04 .05 .06 .07 .08
Particle diameter, micrometers
.09 .10
lO. Cumulative distribution data from Table 6
plotted using normal probability scales
0
Mediani
I ' I ' I ' § ' I T
|50 |70 I 90 95 98J99.:
40 60 80 99
99.99
1 2 5 10 20 30
Cumulative relative frequency, percent less than
1-39
-------
Figure 11. Cumulative particle size distribution data from
Table 6 plotted using log-normal probability scales
CO
n
-------
The raw particle count data of Appendix D has been somewhat
reduced arid presented in a more usable form in Table 4. A simpli-
fied method for calculating the arithmetic mean and standard devia-
tion of the particle size distribution as described by Cadle (35)
is employed. Calculation of the arithmetic mean, d, is based on
the following equations,
_ n n
d = c 1 E u f +d I If
n £-^ i i on ^-^ i
d~ = cu + dQ
where c is the class interval, n is the total number of particles
counted, u. is the class mark of the new scale (-1,0,1,2. . .8) ,
d is the class mark corresponding to u 0, and f , is the frequency
o i x
for each class. Using the transformations found in Appendix E,
Table E-l, the arithmetic mean is,
d - .01 / 2763 \ + 0.02
/ 2763 \ + 0.
VL2.355/
df = 0.022 micrometers
The equation for standard deviation (s) employs the same notation
and is calculated using the following equation.
-2*
a - c / i y u. f, -i
, £-cl _ JL -I-
Using the values of Appendix E, Table E-]
s - 0.0:
^12355/ \12355
s = 0.0163 micrometers
1-41
-------
In order to present the particle size distribution data in
the several graphical forms of Figures 9, 10 and 11, it was
necessary to transform the numerical distribution into a relative
frequency distribution expressed in percent. The results of this
conversion are presented on Table 6. The histogram and distribu-
tion curve of Figures 7 and 8, respectively, indicate a highly
positive skewed distribution based on the particle size range
examined. As the lower limit of resolution of the electron micro-
scope was approximately 0.005 micrometers (50 A), no information
can be presented on particles that may have formed with diameters
less than 0.005 micrometers.
The range of relative frequency found on each of the specimen
screens for each size interval is also presented on Figure 8 and
indicates the size distribution of discrete particles collected on
each plate of the Andersen Stack Sampler and the membrane filter is
approximately the same. Standard deviation of the relative frequency
within each size interval is presented in Appendix G. The range of
standard deviations also seems to indicate similar particle size
s
distributions.
The cumulative frequency distribution tabulated in Table 6 is
presented as an Ogive in Figure 9 and, more conventionally, using
normal probability and log-normal probability scales in Figures 10
and 11 respectively. The log-normal scales of the latter reveal a
relatively straight line indicating a log-normal particle size
distribution.
1-42
-------
Table (>. Cumulative distribution of data from Table 5
Class
boundaries
L
U
Percent
in
Class
Percent less than
upper class limit
0.005
0.015
0.025
0.035
0.045
0.055
0.065
0.075
0.085
0.095
0.015
0.025
0.035
0.045
0.055
0.065
0.075
0.085
0.095
0.1053
46.3
21.6
15.4
7.3
4.3
2.4
1.2
0.6
0.3
0.6
46.3
67.9
83.3
90.3
94.9
97.3
98.5
99.1
99.4
100.
Refer to page 34 for explanation of upper limit for
this class.
1-43
-------
The median and geometric mean are calculated using the procedures
presented by Freund and Williams (36). By definition, the median
divides the frequency distribution in half, e.g. there will be the
same number of particles above the median size as there are below.
For the present frequency distribution containing 12,355 ranked
observations, the median corresponds to the calculated particle size
of the 6178th particle. For the grouped data of Table 4 the median (M)
is calculated using the following equation (37),
where L is the lower class boundary of the interval containing the
median value, c is the class interval, J is the difference between
the median observation and the cumulative frequency of all classes
preceeding the class containing the median, and fM is the frequency of
the class containing the median. Using the tabulated values of
Appendix F, the median is
M - 015 + O.Q1 (6170-5726)
2666
M = 0.0167 micrometers.
The median, sometimes known as the mass median diameter (HMD), has
been graphically inserted in Figures 7 and 9. Inspection of Figure 9
confirms the calculated value of the median as the ogive curve repre-
senting the cumulative frequency distribution passes through coordinates
of 50 percent cumulative relative frequency and 0.0167 micrometer
particle diameter. -11
-------
Geometric mean (G) is calculated based on the following
relationship (38),
log G = I log x- (f)
where x is the class mark of an interval, f is the frequency of
that class interval, and n is the total number of observations
(£f). Again using the tabulated values found in Appendix F, the
geometric mean is calculated as follows,
Log G - "27,715.1 = 2". 24322
12,355
antilogarithm "2.24322 = 0.01751
G = 0.0175 micrometers
The geometric mean has been plotted on Figures 10 and 11 and at
the "50 percent less than" point intersects the frequency dis-
tribution curves as expected.
By inspection of Table 4, the mode for this particle size dis-
tribution occurs in the 0.005-0.015 micrometer diameter interval.
As stated previously, two methods were employed to determine
particle size distribution. The second method involved the relatively
new Andersen Stack Sampler used in conjunction with a back-up membrane
filter. Combining the mass distribution data of Table 2 with cali-
bration information supplied by the manufacturer, cumulative particle
size mass distribution can be obtained and is presented on Table 7.
Although this mass distribution data indicates approximately
69 percent of the sampled particulate passed through the Andersen
1-45
-------
Table 7. Arithmetic mean and standard deviation of cumulative relative mass distribution
from data of Table 2
STAGE
0
1
2
3
4
5
6
7
8
Filter
Approximate
range of
particle
size3
(micrometers)
30 and larger
9.2-30
5.5-9.2
3.3-5.5
2.0-3.3
1.0-2.0
0.3-1.0
0.1-0.3
less than 0.1
Upper
Class
Limit
(micrometers
unknown
30
9.2
5.5
3.3
2.0
1.0
0.3
0.1
0.1
RUN 1
RUN 2
RUN 3
Cumulative relative mass
(percent less than
upper class limit)
100.10
98.13
97.83
97.53
96.93
95.95
93.92
87.84
75.53
68.77
100.00
98.79
98.55
98.39
98.23
97.66
95.16
88.94
79.89
69.55
100.00
98.42
98.09
97.83
97.50
96.51
94.34
88.82
77.38
68.57
Mean
cumulative
relative
massc
(percent less
than)
100.03
98.45
98.26
97.92
97.55
96.71
94.47
88.53
77.60
68.96
Standard
deviation0
(percent)
0.330
0."386
0.437
0.652
0.872
0.631
0.603
2.188
0.518
c
s
X
.0033
.0039
.0044
.0066
.0090
.0066
.0068
.0281
.0075
-p-
3 Farrell Yeates, it. at., "Calibration and Evaluation of the Andersen Stack Sampler", Technical
Report No. FY70 MCR-1, submitted to Two Thousand, Inc., Salt Lake City, Utah (April, 1970), p. 18.
b From Table 2.
c Refer to APPENDIX E, Table E-4 for calculations.
-------
Figure 12. Cumulative particle size distribution data from
Table 7 plotted using log-normal probability scales
-------
Stack Sampler without collection, the previous data presented concerning
the size of the discrete particles suggests essentially all the
particulate matter should have passed through the sampler without
impacting on the collection plates. As the particles were produced in
a direct current arc, it was expected that chains and agglomerates
would form, thereby increasing the effective aerodynamic size of the
particles. In fact, a limited number of these formations were observed
on the electron micrographs. However, since the electron microscope
sample preparation technique was designed to disperse any particle
aggregates to facilitate counting, the effective size could not be
confirmed by direct observation.
The Andersen Stack Sampler utilizes aerodynamic sizing to equate
the size of all particles sampled irrespective of shape or density
to the aerodynamic equivalence of unit density spheres. Vomela and
Whitby (39) have made measurements on aggregates consisting of from
10 to 300 particles having a mass mean diameter of 0.05 micrometers
and concluded the fluid drag on agglomerates of particles is nearly
the same as that of a sphere having a volume equivalent to that of
the aggregates. For example, a mass consisting of one hundred,
0.022 micrometer diameter particles is aerodynamically equivalent to
one particle 0.1 micrometers in diameter. At present, the only
calibration data available from the manufacturer is for unit density
spheres and is listed on Table 7.
The mass distribution developed in table 7 and presented graph-
ically in Figure 12 is plotted as if the particles were spherical
and of unit density. In reality this is known not to be the case as
1-48
-------
the specific gravity of iron oxide (Fe203) is in the range of
5.12 - 5.24 (34), thereby increasing the relative inertial force
of spherical iron oxide particles when compared to the same size
unit density spheres. That is, inside the sampler an iron oxide
sphere will overcome the aerodynamic drag force and leave the turning
gas stream (and impinge on the collection plate) at a much lower
velocity than would the same size particle of unit density.
Reproducibility of Production Rate and Relative Particle Size
Distribution
The production rate of particulate matter for each of the three
runs was calculated and presented on Table 3 with the arithmetic
mean calculated at 30.08 grains/min (1.95 g/min). As an indication
of variability about the mean, the standard deviation (s) was
calculated using the following equation,
n
- x)'
n
where x represents an individual observation, x is the arithmetic
mean, and n is the total number of observations.
29.25
27.30
33.68
-0.83
-2.78
3.60
.6889
7.7284
12.9600
m.3773
1--49
-------
Therefore,
\i
21.3773 - 2.68 grains/min
Two methods were used to estimate the reproducibility of the
relative particle size distribution both using the mass distri-
bution data obtained from the Andersen Stack Sampler and membrane
filter. The results of the three runs are presented in Table 2.
Table 8 presents the arithmetic mean and standard deviation
for the weights found for each of the three runs on each stage of
the sampler and the membrane filter. As a further, and possibly
more meaningful, comparison of the standard deviations of the various
stages with each other, the last column on Table 8 presents the
standard deviation as a percent of the arithmetic mean of the weights.
The mean and standard deviation of the total particulate mass captured
during each run is also given and is another indicator of the repro-
ducibility of production rate.
The second estimate of size distribution reproducibility is obtained
by calculating the arithmetic mean and standard deviation of the cumu-
lative relative mass distribution for each of the various size inter-
vals. The results are presented on Table 7. Again the last column
represents the standard deviation as a percentage of the mean, in this
case the mean cumulative relative mass.
1-50
-------
Table 8. Arithmetic mean and standard deviation of the
weight (mass) distribution from data of Table 2
STAGE
0
1
2
3
4
5
6
7
8
Filter
TOTALS
RUN 1
RUN 2
RUN 3
Collected weight
(milligrams)
2.5
0.4
0.4
0.8
1.3
2.7
8.1
16.4
9.0
91.6
133.2
1.5
0.3
0.2
0.2
0.7
3.1
7.7
11.2
12.8
86.1
123.8
2.4
0.5
0.4
0.5
1.5
3.3
8.4
17.4
13.4
103.4
152.1
X
Mean
Q
weight
(mg)
2.13
0.40
0.33
0.50
1.17
3.03
8.07
15.00
11.73
93.70
136.00
Mean
s
Standard
•a
deviation
(rag)
0.4496
0.0812
0.0943
0.2449
0.3398
0.2493
0.2867
2.7178
1.9482
7.2171
14.4202
1.3629
_s_a
X
0.2110
0.2030
0.2857
0.4898
0.2904
0.0822
0.0355
0.1561
0.1660
0.0770
0.1060
0.1997
a Refer to APPENDIX E, Table E-5 for calculations,
1-51
-------
V. CONCLUSIONS
On the basis of the results of the limited experimental tests
reported herein, the feasibility of employing an electric arc to
produce aerosols of ultra-fine iron oxide particles from a feed-
stock of consumable wire has been demonstrated. Although little
attempt was made to optimize the operating parameters, more than
30 grains per minute (1.95 g/min) of discrete, spherical particles
with a mass median diameter of 0.0167 micrometers was repeatedly
produced within a range of plus or minus 12 percent. Due to
agglomeration, a limited number of aggregates and chains of particles
were formed thereby increasing the mean effective aerodynamic size
of the particle distribution. Even so, a conservative examination
of the mass distribution data obtained from the Andersen Stack
Sampler particle sizing technique indicates approximately 75 percent
of the particles have an effective diameter less than 0.1 micrometers.
If the agglomerating effects prove undesirable for a particular study,
an ion generator or polonium grid could be employed to neutralize the
charged particles. The rate of particle production is believed to be
essentially uniform as the consumable wire was fed at a constant rate
and the power supply was regulated at a constant voltage. It is felt
the feedstock-to-particulate conversion ratio of 4.75 percent could
be significantly improved through optimization of the main parameters:
arc voltage, wire feed rate and wire diameter. The interior dimensions
of the electrode holder coupled with an erosion rate of 0.2 inches
(5.08 mm) per minute limited length of continuous production to
1-52
-------
approximately 30 minutes. Optimization of the above parameters
complemented by improved shielding of the tungsten electrode should
serve to reduce the erosion rate of the relatively non-consumable
tungsten, thereby increasing the time interval between start-up and
replacement of the tungsten electrode. A redesign of the electrode
holder through addition of an automatic advancing mechanism capable
of accepting longer tungsten electrodes could greatly improve the
endurance of the generator.
The morphology of the iron oxide generated in this study closely
approximates the characteristics of emissions from various heavy
industrial processes: open hearth and basic oxygen steel making,
scarfing, welding, plasma and oxygen-acetylene slab cutting, and of
course, the electric arc furnace. As the behavior of collected and
re-dispersed iron oxide particulate matter varies significantly from
that of "fresh" iron oxide fume, the primary aerosol formation tech-
nique demonstrated here may prove invaluable for control equipment
research. The practical implications of a device such as this are
much more far-reachiag. With no modification to the basic generator
apparatus, essentially all metals and metal alloys could be vapor-
ized in the electric arc producing oxides of most metals. Among
hazardous metals of current interest because of their physiological
effects are lead, chromium, beryllium, vanadian, tin and copper.
To further increase the research applications, special cored wires
can be produced to introduce additional contaminants into a system
at a controlled rate along with a primary metal fume. The basic
system is also amenable to introduction of gaseous constituents for
1-53
-------
study of possible or known synergistic relationships.
With modification, the unit could also be adapted for the production
of aerosols of non-conductive refractory materials. This could be
accomplished by replacing the wire torch with another non-consumable
electrode holder, establishing an arc between them, and guiding the
non-conductive feedstock into the high-intensity arc. Conversely,
elimination of the tungsten electrode as one side of the arc could be
accomplished by utilizing two wire torches and establishing an arc
between the consumable wire "electrodes". This modification will
inherently eliminate the problems of tungsten electrode erosion
(requiring periodic replacement) and potential contamination of the
solid phase of the aerosol. Ability of the generator to operate
continuously would also be greatly enhanced.
1-54
-------
LIST OF REFERENCES
(1) Lodge, James P., "Production of Controlled Test Atmospheres",
Air Pollution. Arthur C. Stern, ed., Vol. 2, Academic Press,
New York (1968) pp. 465-483.
(2) Amdur, Mary 0., "The Physiological Response of Guinea Pigs to
Atmospheric Pollutants", International Journal of Air Pollution,
Vol. 1 (1959) pp. 170-183";
(3) Fuchs, N. A., The Mechanics of Aerosols. The Macmillan Company,
New York (1964) p. viii-x.
(4) Silverman, Leslie, "Experimental Test Methods", Air Pollution
Handbook. Paul L. Magill, Francis R. Holden and Charles Ackley
eds., McGraw-Hill, New York (1956) pp. 12-1-12-2.
(5) Fuchs, op. cit.. pp. 367-377.
(6) Ibid, p. 374.
(7) Silverman, op. cit.. p. 12-20.
(8) Udin, Harry, Funk, Edward R., and Wulff, John, Welding for
Engineers, John Wiley & Sons, Inc., New York (1954) p. 141.
(9) Holmgren, J. D., Gibson, J. 0., and Sheer, C., "Some Character-
istics of Arc Vaporized Submicron Particulates", Ultra Fine
Particles. W. Kuhn, ed., John Wiley, New York (1963) p. 130.
(10) Amick, James and Turkevich, John, "Electron Microscopic Examina-
tion of Aerosols Formed in a Direct Current Arc", Ultra Fine
Particles. W. Kuhn, ed., John Wiley, New York (1963) p. 147.
(11) Whitby, K. T., Lundgren, D. A., and Petersen, C. M., "Homogeneous
Aerosol Generators", International Journal of Air and Water
Pollution, Vol. 9 (1965) p. 265, 276.
(12) Silverman, Leslie and Billings, Charles E., "Methods of Generating
Solid Aerosols", Journal of Air Pollution Control Association,
Vol. 76, No. 6, (1956) p. 82.
(13) Ibid, pp. 77 ff.
(14) Silverman, "Experimental Test Methods", op. cit., pp. 12-19-12-38.
(15) Olive, T., "Solids Feeders", Chemical Engineering. Vol. 59,
No. 11 (1952) pp. 163-178.
1-55
-------
(16) Fuchs, op. cit. chapter VIII.
(17) Silverman, "Methods of Generating Solid Aerosols", op. cit.
p. 77.
(18) Silverman, "Experimental Test Methods", op. cit., pp. 12-20,
12-22-12-23.
(19) Holmgren et. al., op. cit.. p. 129.
(20) Harris, V., Holmgren, J. D., Korman, S., and Sheer, C.,
"Arc Decomposition of Rhodonite", Electrochemical Society Journal.
Vol. 106, No. 10 (Oct. 1959) pp. 874-876.
(21) Tafa Division, Humphreys Corporation, 180 North Main Street,
Concord, New Hampshire 03301.
(22) Karioris, Frank G., and Fish, Birney R., "An Exploding Wire Aerosol
Generator", Journal of Colloid Science. Vol. 17, (1962) pp. 155-161.
(23) Silverman, "Methods of Generating Solid Aerosols", op. cit.. p. 81.
(24) Davis, R. J., Truitt, J., and Gill, J. S., "Aerosol Studies in the
Reactor Safety Program at Oak Ridge National Laboratory", Journal
of the Air Pollution Control Association, Vol. 18, No. 10
(Oct. 1968), pp. 675-676.
(25) Truitt, J. and Davis, R. J., "The Function of Condensing Steam
in Aerosol Scrubbers", Reactor Chemistry Division, Oak Ridge
National Laboratory, Oak Ridge, Tennessee, unpublished report
of research sponsored by the National Air Pollution Control
Administration (1970).
(26) "Operation Instructions for Electric Motor Driver Constant Voltage
Welder for Automatic Welding, Model MC-900, Specs 4136, 4210".
Hobart Brothers Company, Troy, Ohio.
(27) "Hobart Welding Wire and Flux", Bulletin EW-382 (NWSA-630).
Hobart Brothers Company, Troy, Ohio.
(28) "Installation-Operation-Maintenance Manual for Hobart Automatic
Welding Head, Model AI-22, Specs 3974A." Hobart Brothers Company,
Troy, Ohio.
(29) Welding Data Book. Robert N. Williams, ed., Industrial Publishing
Company, Cleveland, Ohio (1967) pp. A/72-A/73.
(30) Rose, Arthur and Rose, Elizabeth, eds. The Condensed Chemical
Dictionary. Van Nostrand Reinhold Company (1967) p. 982.
1-56
-------
(31) Yeates, Parrel, et. al.. "Calibration and Evaluation of the
Andersen Stack Sampler", Technical Report No. FY70 MCR-1,
submitted to Two Thousand, Inc., Salt Lake City, Utah
(April 1970) p. 1.
(32) Andersen, A. A., "A Sampler for Respiratory Health Hazard
Assessment", AIHA Journal. Vol. 27 (March-April 1966) pp. 160-165.
(33) Welding Data Book, op. cit.. p. A/79.
(34) Rose, op. cit.. p. 412.
(35) Cadle, Richard D., Particle Size. Reinhold Publishing Corporation,
New York (1965) pp. 20-22.
(36) Freund, John E. and Williams, Frank J., Modern Business Statistics,
Prentice-Hall, Englewood Cliffs, New Jersey (1958) pp. 60 ff.
(37) Ibid., p. 62.
(38) Ibid., pp. 66-68.
(39) Vomela, R. A. and Whitby, K. T., "The Charging and Mobility of
Chain Aggregate Smoke Particles", Journal of Colloid and Interface
Science. Vol. 25, No. 4 (December 1967) pp. 568-576.
1-57
-------
APPENDICES
APPENDIX A Selected Bibliography of Readings Concerning
Electric Arc Discharges
APPENDIX B Developmental Background of the Tungsten
Electrode Holder
APPENDIX C Operating Procedure for Arc Aerosol Generator
APPENDIX D Electron Microscope Particle Count Raw Data
APPENDIX E Statistical Calculations
1-58
-------
APPENDIX A
Selected Bibliography of Readings Concerning Electric Arc Discharges
Aidala, Joseph B., "Electric Heating", in Chemical Engineers' Handbook,
Robert H. Perry at. at., eds., McGraw-Hill Book Co., New York
(1963) pp. 25-41—25-42.
Fleming, John Ambrose, "Electric Lighting", in Encyclopaedia Britan-
nica, llth ed., vol. 16, Encyclopaedia Britannica Inc., New
York (1910) pp. 659-666.
Jackson, Clarence E., "Welding", in Encyclopedia Americana, vol. 28,
Americana Corp., New York (1969) pp. 599-603.
Rose, David John, "Conduction [of Electricity] in Gases", in
Encyclopedia Britannica, vol. 8, Encyclopedia Britannica Inc.,
Chicago (1970) pp. 206-208.
Thomson, Joseph John, "Electric Conduction", in Encyclopaedia
Britannica, llth ed., vol. 6, Encyclopaedia Britannica Inc.,
New York (1910) pp. 884-887.
Udin, Harry, Funk, Edward R. and Wulff, John, Welding for Engineers,
John Wiley and Sons Inc., New York (1954) pp. 114-115 and
136-169.
1-A-l
-------
APPENDIX B
Developmental Configurations of the
Arc Chamber and Electrode Holder Apparatus
Thoughts of utilizing an electric arc to create large, reproducible
quantities of ultra-fine particles was precipitated by the author's
experience with various welding and arc-cutting techniques, and
knowledge of the characteristics of the resulting process emissions.
Tungsten inert gas (TIG) welding techniques were first developed in
the 1930's and have been refined for primary use in the aircraft *
industry for welding magnesium and aluminum. A refractory,
"non-consumable", tungsten electrode is used to initiate an arc
between the base metal and itself, and filler wire is manually fed
into the molten puddle. Helium and/or argon is directed at the weld
area to reduce oxidation of the tungsten electrode and molten metal
in the weld. Tungsten arcs are also used as a heat source for cutting
of metals. Another relatively recent development in the welding
industry is the metal inert gas (MIG) process. It was not until 1948
that the first consumable electrode welding equipment was patented.
This technique has now seen wide use in both automatic and semiauto-
matic applications.
The electric arc aerosol generator apparatus discussed in this report
is a combination of equipment from these two welding processes.
An automatic wire feed and torch assembly is used to feed consumable
wire into an arc established between itself and a tungsten
electrode using the constant potential MIG welder to supply the
electromotive force. In order to establish the feasibility of
1-B-l
-------
proceeding with the design of a semi-continuous model, several
preliminary arrangements were tried.
Figure B-l, pictorially illustrates, in a simplified manner, the
evolutionary configurations investigated which lead to the design
and construction of the arc chamber and electrode holder used to
produce the results reported in this paper. Configuration I utilized
a standard, rack mounted consumable electrode welding head and feed
motor assembly (the same units used in the final configuration) to
direct the wire at the tip of a one-eighth inch diameter tungsten
electrode. The tungsten electrode was held in place be a compression-
type, one-eighth inch brass tubing fitting (Swagelok Part No. 200-1-2)
bored through to allow the electrode to slide in and out of the
chamber. By tightening and loosening the compression nut, the fitting
acts as a chuck to lock the electrode in position. The pipe thread
end of the fitting was place in a tapped hole in the arc chamber on
the same centerline axis as the wire torch. The electrical hookup
was greatly simplified as the tungsten electrode was placed at
negative potential and the whole arc chamber could be safely and
easily grounded. On the other side, passage of the wire torch through
the chamber wall had to be insulated.
This configuration met with only limited success as the rate
of tungsten melting and erosion was excessive, and continuous
operation was limited to a maximum of thirty seconds. Two changes
"Swagelok" fittings are manufactured by Crawford Fitting
Company, Solon, Ohio 44139.
l-B-2
-------
Figure B-l.
Simplified sketches showing developmental configurations
of the tungsten electrode-arc chamber apparatus
CONFIGURATION I
CONFIGURATION II
CONFIGURATION III
l-B-3
-------
were deemed appropriate to retard erosion and thereby increase
operation time: (1) increase electrode diameter, and (2) supply
a water-cooled electrode holder to dissipate heat.
Configuration II used the same wire torch arrangement, but the
non-consumable electrode was increased in size to 0.25 inches in
diameter and placed in a specifically constructed water-cooled holder.
The holder was fabricated from an eight inch length of one inch stan-
dard pipe with a steel plug welded in one end. The plug was threaded
and held a compression-type, stainless steel tubing fitting (Swangelok
Part No. 400-1-316), again bored through to act as a water tight chuck
to hold the tungsten electrode. The other end of the holder contained
inlet and outlet fittings for the cooling water supply. An aluminum
electrode holder mounting block was machined to fit the curvature of
the chamber and bolted in place in such a manner as to locate the
electrode centerline on the same axis as the consumable wire center-
line. The holder was held in place by a set screw.
This configuration operated satisfactorily for up to two minutes
before the build-up of molten wire and tungsten on the compression
fitting forced a shut down. After melting one of the stainless steel
fittings beyond use, a replaceable steel disc was placed over the end
of the protruding electrode, as shown in Figure B-l, Configuration II,
to protect the fitting and holder body from arc splatter and molten
build-up. This addition did little to increase length of operation,
as the same problem of build-up of molten metal recurred, this time
on the disc protecting the holder.
l-B-4
-------
To eliminate the problem of molten metal build-up it became
obvious the arc chamber and its attachments could be rotated 90 degrees
allowing the molten metal droplets to fall harmlessly to the bottom
of the chamber. This change in orientation required substantial
changes in the chamber mounts and fabrication of a special insulated
mounting system for the feed motor and wire spool. The arrangement
has been denoted Configuration III in Figure B-l.
With this change, length of operation was doubled, but this
only meant continuous operation for a maximum of four minutes. It
became apparent the limiting factor was excessive tungsten melting
and erosion. Therefore, to increase endurance, a holder had to be
designed to allow advancement of the electrode into the arc to
compensate for melting and maintain uniform erosion. A water-cooled
holder was finally designed and fabricated which satisfactorily incorporated
these features. This unit, shown in Figure 4, was used successfully to
obtain the results of this report.
l-B-5
-------
APPENDIX C
Operating procedure for Arc Aerosol Generator
1, Assemble apparatus as shown in Figures 3, 5 and 6, omitting
inlet duct.
2. Insert assembled cathode holder into arc chamber mounting block
to blue line on holder and tighten set screws.
3. Place heat shield over end of cathode holder and tighten set
screw by inserting alien wrench through view window opening.
4. Install shielding gas nozzle through view window opening.
5. Advance cathode by turning knob until tip is centered in gas
nozzle outlet. Refer to Figure C-l.
6. Make electrical, shielding gas and water supply connections.
7. Turn on power to Control Panel only.
8. Set WIRE SPEED control dial to desired setting. (Refer to
Figure C-2 for calibration curve for 0.045 inch diameter wire).
9. Clip off approximately 1/2 inch from protuding wire to remove
any oxidized metal.
10. Place WIRE FEED switch in DOWN position.
11. Press INCH button until wire (anode) protrudes approximately
1/2 inch (12.7 mm) from torch tip. (See Figure C-l).
12. Adjust arc chamber on its slide mounts until gap between anode
and cathode is approximately 1/8 inch (3.175 mm) and tighten
hold down screws. (See Figure C-l).
13. Spray interior of arc chamber with welding splatter anti-stick
spray.
14. Install inlet duct.
15. Close view window and secure with tape.
16. Turn on exhaust fan switch.
17. Turn on main power switch at buss and press START button on
power supply.
1-C-l
-------
18. Adjust VOLTAGE control dial on control panel to desired open
circuit voltage. Voltage is read on power supply voltmeter.
19. Open valve on shielding gas tank.
20. Depress PURGE button on control panel and adjust gas flow to
desired rate. Release button.
21. Turn on power to air blower (low speed) and water pump.
22. Depress PURGE button approximately one second and immediately
press START button on control panel to initiate arc.
23. For continuous operation while conducting tests, manually turn
feed screw on cathode holder to maintain tungsten tip* centered
in the axis of the shielding gas nozzle outlet.
24. To cease operating, place WIRE FEE3 switch in OFF position.
25. Press STOP button on power supply.
26. Allow shielding gas and cooling water to flow approximately
5 seconds for each 10 amperes of current flow, e.g. for a
current flow of 150 amperes, flows should continue for 70
seconds after the arc is extinguished. Alternately, allow
cooling water and air to circulate until air temperature
immediately downstream from arc chamber cools to 100°F (38eC).
27. Press STOP button on Control Panel.
28. Close valve on shielding gas tank to conserve gas.
29. Turn off all power.
l-C-2
-------
Figure C-l. Simplified t-ross-sectionnl, p.lan view sketch
showing location of components Inside arc chamber
O GAS SO
Dashed lines denote
view window opening
1-03
-------
Figure C-2. Calibration curve of wire speed dial versus
wire speed of 0.045 inch diameter wire
10 _
9 .
8 .
7 .
g
03 r-
a5
4 -
3 -
2 -
0
I I I
100 200 300
Wire speed, inches/minute
I
400
500
l I 1 I i i I
5 10
Wire speed, meters/minute
1-0-4
-------
Table D-l. Electron microscope particle count raw data
W Q
. O i-J
3
4
2
3
2
2
2
3
0
0
1
0
5
7
4
3
5
0
TOTAL
372
493
262
354
267
196
315
668
322
235
200
416
539
620
637
726
237
316
M
X
-------
Table D-l (continued). Electron microscope particle count raw data
W Q
O i-J
2
2
1
1
3
5
5
0
1
2
5
1
74
2
7
7
TOTAL
304
247
177
205
368
411
723
366
470
651
516
742
12355
565
1236
412
NJ
-------
APPENDIX E
Table E-l. Transformation for calculating the arithmetic mean and
standard deviation from data of Table 4
Class mark
di
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.10
frequency
f
5726
2663
1908
907
531
292
144
76
34
74
u
-1
0
1
2
3
4
5
6
7
8
uf
-5726
0
1908
1814
1593
1168
720
456
238
592
u2f
5726
0
1908
3628
4779
4672
3600
2736
1666
4336
n=12,355
22763 233,451
1-E-l
-------
APPENDIX E
Table E-2. Tabulation for calculating median an geometric
mean from data of Table 4
Class
boundaries
L
.005
.015
.025
.035
.045
.055
.065
.075
.085
.095
U
.015
.025
.035
.045
.055
.065
.075
.085
.095
.105
Class mark
Xi
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.10
log x±
"2.00000
"2 . 30103
2". 47712
1.60206
1.69897
1.77815
I. 84510
1.90309
1.95424
I. 00000
frequency
f
5726
2663
1908
907
531
292
144
76
34
74
Cumulative
frequency
5726
8389
10297
11204
11735
12027
12171
12247
12281
12355
-log x±(f)
11452.0
6127.6
4726.3
2360.1
1433.2
811.1
409.7
220.6
100.4
74.0
n=12,355
227,715.1
l-E-2
-------
APPENDIX E
Table E-3. Calculation of arithmetic mean and standard
deviation of the relative frequency within each
particle size interval from the data of Table 5
X
i
46.3
43.0
54.3
50.5
43.0
46.3
47.4
44.3
48.0
42.8
465.9
22.0
22.4
19.1
19.2
23.3
22.3
20.5
21.2
21.6
22.1
213.7
13.6
17.7
12.3
14.4
17.1
16.4
17.3
15.1
14.9
15.7
154.5
X
46.6
21.4
15.5
x.-x
i
-0.3
-3.6
7.7
3.9
-3.6
-0.3
0.8
-2.3
2.0
-3.8
0.6
1.0
-2.3
-2.2
1.9
0.9
-0.9
-0.2
0.2
0.7
-1.9
2.1
-3.2
-1.1
1.6
0.9
1.8
-0.4
-0.6 0.
0.2
_
(x -x)
i
0.09
12.96
59.29
15.21
12.96
0.09
0.64
5.29
4.00
14.44
124.97
0.36
1.00
5.29
4.84
3.61
0.81
0.81
0.04
0.04
0.49
17.29
3.61
4.41
10.24
1.21
2.56
0.81
3.24
0.16
0.36
0.04
26.64
s
3.726
1.386
1.720
s
~
X
0.0799
0.0647
0.1109
-------
Table E-3 (continued).
Calculation of arithmetic mean and standard
deviation of the relative frequency within
each particle size interval from the data
of Table 5
x
i
7.7
6.9
7.0
6.6
8.7
5.3
6.6
8.8
6.3
8.3
72.2
4.5
4.5
3.5
4.2
3.1
3.7
3.4
5.1
6.2
4.5
42.7
2.1
2.6
2.1
2.2
2.2
2.6
2.7
2.1
1.3
3.6
23.5
X
7.2
4.3
2.4
x -x
i
0.5
-0.3
0.2
0.6
1.5
-1.9
-0.8
1.6
-0.9
1.1
0.2
0.2
-0.8
-0.1
-1.2
-0.6
-0.9
0.8
1.9
0.2
-0.3
0.2
-0.3
-0.2
-0.2
0.2
0.3
-0.3
-1.1
1.2
_
(x -x)
i
0.25
0.09
0-04
0.36
2.25
3.61
0.64
2.56
0.81
1.21
11.81
0.04
0.04
0.64
0.01
1.44
0.36
0.81
0.64
3.61
0.04
7.63
0.09
0.04
0.09
0.04
0.04
0.04
0.09
0.09
1.21
1.44
3.17
s
1.146
0.921
0.593
s
"=
X
0.1591
0.214
0.247
l-E-4
-------
Table E-3 (continued).
Calculation of arithmetic mean and standard
deviation of the relative frequency within
each particle size interval from the data
of Table 5
X.
i
1-5
1.5
0.9
1.1
0.7
2.3
0.7
1.1
0.6
1.4
11.8
1.3
0.4
0.2
0.8
0.6
0.3
0.3
1.0
0.4
0.8
6.1
0.1
o.2
0.1
0.6
0.3
0.2
0.4
0.2
0.3
0.4
2.8
X
1.2
0.6
0.3
x —x
1
0.3
0.3
-0.3
-0.1
-0.5
1.1
-0.5
-0.1
-0.6
0.2
0.7
-0.2
-0.4
0.2
0.0
-0.3
-0.3
0.4
-0.2
0.2
-0.2
-0.1
-0.2
0.3
0.0
-0.1
0.1
-0.1
0.0
0.1
<
/„ _x\ ^
i
0.09
0.09
0.09
0.01
0.25
1.21
0.25
0.01
0.36
0.04
2.40
0.49
0.04
0.16
0.04
0.00
0.09
0.09
0.16
0.04
0.04
1.15
0.04
0.01
0.04
0.09
0.00
0.01
0.01
0.01
0.00
0.01
0.22
0.516
0.357
0.156
s
""^
X
0.430
0.596
0.521
l-E-5
-------
Table E-3 (continued)
Calculation of arithmetic mean and standard
deviation of the relative frequency within
each particle size interval from the data
of Table 5
xi
0.8
0.9
0.4
0.1
0.9
0.6
0.7
0.9
0.4
0.4
6.1
X
0.6
x.-x
0.2
0.3
-0.2
-0.5
0.3
0.0
0.1
0.2
-0.2
-0.2
(x..-x)2
0.04
0.09
0.04
0.25
0.09
0.00
0.01
0.04
0.04
0.04
0.64
s
0.267
s
X
0.444
210.788 2.867
Mean 1.079 0.287
l-E-6
-------
APPENDIX E
Table E-4. Calculation of arithmetic mean and standard deviation
of the cumulative relative mass distribution data
of Table 7
STAGE
1
2
3
4
5
6
7
Xi
98.13
98.79
98.42
295.34
97.83
98.55
98.09
294.47
97.53
98.39
97.83
293.75
96.93
98.23
97.50
292.66
95.95
97.66
96.51
290.12
93.92
95.16
94.34
282.42
87.84
88.94
88.82
265.60
x
98.45
98.26
97.92
97.55
96.71
94.47
88.53
x -x
-0.32
0.34
-0.03
-0.43
0.29
-0.17
-0.39
0.47
-0.09
-0.62
0.68
-0.05
-0.76
0.95
-0.20
-0.55
0.69
-0.13
-0.69
0.41
0.29
— 2
.1024
.1156
.0009
0.2180
.1849
.0841
.0289
0.2979
.1521
.2209
-0081
0.3811
.3844
.4624
.0025
0.8493
.5776
.9025
.0400
1.5201
.3025
.4761
.0169
0.7955
.4761
.1681
.0841
0.7283
s
0.330
0.386
0.437
0.652
0.872
0.631
0.603
s
X
0.0033
0.0039
0.0044
0.0066
0.0090
0.0066
0.0068
l-E-7
-------
Table E-4 (continued).
Calculation of arithmetic mean and
standard deviation of the relative mass
distribution data of Table 7
w
a
CO
8
F
Xi
75.53
79.89
77.38
232.80
68.77
69.55
68.57
206.89
X
77.60
68.96
x -x
-2.07
2.29
-0.22
-0.19
0.59
-0.39
(x^x)2
4.2849
5.2441
.0484
9.5774
.0361
.3481
.1521
0.5363
s
2.188
0.518
s
X
0.0281
0.0075
l-E-8
-------
APPENDIX E
Table E-5. Calculation of arithmetic mean and standard deviation
of weight (mass) distribution from data of Table 8
w
o
3 g
Cfl Pd
0
1
2
3
1
1
2
3
2
1
2
3
3
1
2
3
4
1
2
3
5
1
2
3
6
1
2
3
Xi
2.5
1.5
2.4
6.4
0.4
0.3
0.5
1.2
0.4
0.2
0.4
1.2
0.8
0.2
0.5
1.5
1.3
0.7
1.5
3.5
2.7
3.1
3.3
9.1
8.1
7.7
8.4
24.2
X
2.13
x. -x
0.37
-0.63
0.27
0.40
0.00
-0.10
0.10
0.33
0.07
-0.13
0.07
0.50
0.30
-0.30
0.00
1.17
0.13
-0.47
0.33
3.03
-0.33
0.07
0.27
8.07
0.03
-0.37
0.33
(Xj-x)
0.1369
0.3969
0.0729
0.6067
0.0000
0.0100
0.0100
0.0200
0.0049
0.0169
0.0049
0.0267
0.0900
0.0900
0.0000
0.1800
0.0169
0.2209
0.1089
0.3467
0.1089
0.0049
0.0729
0.1867
0.0009
0.1369
0.1089
0.2467
s
0.4496
s
X
0.2110
0.0812
0.2030
0.0943
0.2857
0.2449
0.4898
0.3398
0.2904
0.2493
0.0822
0.2867
0.0355
l-E-9
-------
Table E-5 (continued). Calculation of arithmetic mean and standard
deviation of weight (mass) distribution from
data of Table 8
w
o
-------
VERY FINE PARTICLE
GENERATION BY ELECTRIC ARC:
SAMPLING AND ANALYSIS
PROBLEM
Submitted to the Graduate School
of
West Virginia University
In Partial Fulfillment of the Requirements for
the Degree of Master of Science in Engineering
by
Mike Naylor, B. S.
Morgantown,
West Virginia
1976
2-i
-------
ABSTRACT
Established methods of generating useful reproducible con-
centrations of very fine particles for use in research possess
various limitations. The development of an improved electric arc
metal generating method used a new apparatus: a commercial electric
arc metallizer. It is capable of generating very fine particles of
the metals available in the form of metallizing or welding wire.
The generation, sampling and analysis of zinc oxide particles are
examined.
In addition to studying the effects of changing the arc
metallizer's operating variables of air spray pressure and rate of
wire consumption, the reproducibilities of total mass concentration
and distribution of particle diameters were tested.
Particles were examined on Nuclepore 0.2 (j,m membrane filters
by electron microscopy and electron diffraction. Particles were
counted by diameter. Two replicate samples for each of three arc
metallizer operating conditions were analyzed. One condition is
not reproducible with respect to the mean diameter of sampled parti-
cles. The other two conditions are reproducible with respect to
diameter means and variance. One of these conditions also has
particle diameter distributions that are shown not significantly
different. The mean diameters of the conditions are significantly
different, indicating that by changing the air spray pressure and
wirefeed rate, the particle diameters are changed.
The diameters of the particles generated at specific conditions
2-ii
-------
are distributed approximately normally. Mean diameters range from
6 to 11 nm (0.006 to 0.011 |j,m).
Virtually all of the very fine particles agglomerated,
forming interconnecting chains. This phenomena apparently resulted
from combined effects of the very fine particle diameters and the
large initial concentration of the particles.
Mass concentrations were measured with Millipore 0.45 |j,m
membrane filters. Four different operating conditions were defined,
with three of them being the same as the ones used for particle
sizing. These yield mean mass concentrations ranging from 0.65
3
to 2.0 gram/m.
For three of these conditions, the replicate mass concentra-
tion coefficient of variation (C.V-) is less than 9%. The fourth
has a C.V- of 19%. This is the same condition for which the particle
diameter means are not reproducible. The mass concentrations gen-
erated at different conditions are found significantly different by
testing. The mass concentration is increased by increasing the air
spray pressure or the rate of wirefeed or both.
Since the influence of the operating conditions has been
demonstrated as well as varying levels of reproducibility for each
condition, this metallizer is considered worthy of continued develop-
ment and potential utilization by researchers as a device for
generating very fine particles.
2-iii
-------
ACKNOWLEDGEMENTS
The investigator is grateful to his graduate advisor, Pro-
fessor Benjamin Linsky, who introduced him to the topic of fine
particle generation. He offered suggestions and encouragement in
preparing for and writing this report.
For their research assistance, appreciation is extended to
Dr. D. R. Sears, then Director of the Air Pollution Laboratories at
WVU, and to graduate students, John Garbak, Don Stone and Fred Dimmick.
The investigator thanks Fran Culler of the Civil Engineering Labora-
tory and Don Garletts of the Mechanical Engineering Laboratory for
fabricating equipment parts.
The investigator acknowledges the suggestions from Charles
Ghastin and Ed Morantz, sales representative for and developer of,
respectively,the Wall Colmonoy electrospray metallizer.
The investigator expresses thanks to Richard Shimps of McCrone
Associates for his advice on obtaining appropriate samples for electron
microscopy.
The investigator appreciates the critical review of the sta-
tistical analysis by Dr. E. Z. Damewood of the Industrial Engineering
Department at W.V.U.
Several types of financial assistance made this research
possible. The metallizer and electron microscopy services were
funded by Environmental Protection Agency Research grant R-801858-01-l>
which was administered by Bennis Drehmel. The investigator received
a study traineeship from EPA guant T-90Q487-10-0,and tuition compensa-
tion for one semester from the Civil Engineering Department at WVU.
2-iv
-------
TABLE OFCONTENTS
ABSTRACT 2-ii
ACKNOWLEDGEMENTS 2~iv
LIST OF FIGURES 2-viii
LIST OF TABLES 2-ix
CONVERSION FACTORS, EQUIVALENTS, AND ABBREVIATIONS 2-x
I. INTRODUCTION 2-1
II. REVIEW OF LITERATURE 2-3
III. EXPERIMENTAL 2-6
A, GENERATION 2-6
B. SAMPLING 2-7
1) General Considerations 2-7
2) Description of Sampling Components 2-11
3) Filters 2-12
4) Calculation of Nucleopore Efficiency 2-15
5) Sampling Procedure for Mass Concentration .... 2-16
6) Sampling Procedure for Electron Microscopy .... 2-21
7) Sensitivity of Results to Deviations from
Isovelocity Sampling . 2-22
C. ANALYTICAL TECHNIQUES 2-24
1) Mass Concentration Analytical Techniques 2-24
2) Particle Sizing and Identification Techniques . . 2-25
D. SELECTION OF EQUIPMENT OPERATING CONDITIONS FOR
SAMPLING 2-26
IV. RESULTS 2-28
A. BRIEF EVALUATION OF EQUIPMENT OPERATION 2-28
B. SAMPLE NUMBER IDENTIFICATION 2-29
2-v
-------
Page
1) Mass Concentration , 29
2) Electron Microscopy and Electron Diffraction ... 29
C. CHEMICAL COMPOSITION 29
1) Electron Diffraction Data 29
2) Assay of Wire 30
D. MASS CONCENTRATION 31
1) Means Testing 32
2) Analysis of Variance 32
E. PARTICLE SIZING BASED ON ELECTRON MICROSCOPY 33
1) Analysis 36
2) Computation of Means and Variances 36
3) Reproducibility Among Replicates 37
4) Differences Between Conditions . 37
5) Frequency Distributions 38
i) Computation of Normal Distribution 47
ii) Goodness of Fit 47
F. AGGLOMERATION 48
G. COMPARISON WITH HEDDEN'S RESULTS 50
V. A. CONCLUSIONS 52
RECOMMENDATIONS 53
VI. LIST OF REFERENCES 56
VII. APPENDICES
Appendix A A-l
Appendix B B-l
Appendix C C-l
2-vi
-------
Page
Appendix D , 2-D-l
Appendix E 2-E-l
Appendix F 2-F-l
Appendix G 2-G-l
Appendix H 2-E-l
2-vii
-------
LIST OF FIGURES
Figure
1
2
3
4
5
6
7
8
9
10
11
12
13
14
H-l
H-2
H-3
Schematic of EPA-RAC Sampling Train . , .
Actual Stack Sampling Flow System
Schematic of Sampling Point and Venturi Meter ....
Approximate Relationships of Sampling Nozzle ....
Typical Form for Mass Concentration Data
Plot of Line of Equal Bias as Function of R and K . .
Comparison of Cumulative Frequencies for Condition 2
Comparison of Cumulative Frequencies for Condition 3
Observed Frequency of Sample 134 and Normal
Frequency ....... .
Observed Frequency of Sample 133 and Normal
Frequency
Arithmetic Probability Plot of Sample 128 Diameters .
Log Probability Plot of Sample 128 Diameters ....
SEM Photograph of Sample 133 (5000 X)
SEM Photograph of Sample 133 (10,000 X)
TEM Photograph of Sample 101 (300.000 X)
Page
2-8
2-9
2-10
2-13
2-14
2-18
2-19
2-23
2-41
2-42
2-43
2-44
2-45
2-46
2 -H-2
2 -H-2
2-H-3
2-viii
-------
LIST OF TABLES
Table
Page
1 Summary of Nuclepore Impaction, Interception,
Diffusion and Total Collection Efficiencies ...... 2-16
2 Summary of Possible Sampling Bias ........... 2-23
3 Zinc Wire Sampling Conditions ............. 2-27
4 Identification of Mass Concentration Samples ..... 2-29
5 Identification of McCrone Samples ........... 2-29
6 Electron Diffraction Data-Sample 104 ......... 2-30
7 Diffraction Data — Sample 133 ............ 2-30
8 Standardized Zinc Oxide Mass Concentrations for
All Samples ...................... 2-31
9A Zinc Oxide Mass Concentration Mean and Variation Summary 2-31
9B Mean Mass Concentration Matrix ............ 2-32
10A Zinc Oxide Particle Size Data ............. 2-34
10B Zinc Oxide Particle Size Data ............. 2-35
11 Means and Variances of Zinc Oxide Particle Size Data . 2-37
12 Cumulative Observed and Expected Frequencies for
Condition 2 ...................... 2-39
13 Cumulative Observed and Expected Frequencies for
Condition 3 ......................
14 Hedden-Naylor Data .................. 2~50
2-ix
-------
CONVERSION FACTORS. EQUIVALENTS AND ABBREVIATIONS
acfm = actual cubic feet per minute
acmm = actual cubic meters per minute
°C = (°F - 32)/1.8
1 cm = 0.03281 ft.
cm/sec =» 1.97 fpm
1 cubic meter (m3) = 35.315 cubic feet (ft3)
g/acf - gram/actual cubic feet
3
g/sm = gram /standard cubic meter
1 grain = 0.0648 gram
1 grain = 1/7000 pound
grain/acf = grains/actual cubic foot
3
2.29 grains/scf = 1.00 g/sm
HEPA = high efficiency particulate air
mg/acf = milligrams/actual cubic foot
-3 -7
1 nm = 1 x 10 |j,m - 1 x 10 cm
SAED = selected area electron diffraction
SEM - scanning electron microscopy
scfm = standard cubic feet/minute
3
sm = standard cubic meters
TEM - transmission electron microscopy
-4 +3
|j,m = 1 x 10 cm = 10 nm
2-x
-------
INTRODUCTION
This report examines a method of producing in the laboratory
reproducible concentrations of freshly made very small, metal oxide
particles. The method represents a continuation of a long range pro-
ject of the Air Pollution Engineering Laboratory, the goal of which
is to develop a valid research tool to generate very fine freshly
made particles for extended periods of time and to characterize these
particles. The results of this project are applicable to the fields
of respiratory, control equipment, instrument design, vegetation and
cloud physics research.
The particle generator simulates general industrial electric
arc processes. The concentrations of particles generated may be
typical of those to which workmen are exposed.
A coonercially available electric arc metallizer has been
four.i. that is potentially capable of delivering reproducible very
sr^i.11 particles evolved from at least a dozen different metals.
Metals in wire form are utilized by the machine as consumable elec-
rr^ces. "The wire passes through the high heat zone of the arc,
beccning molten. Air pressure then sprays the molten particles onto
tie surface being coated."*1 '
The process evolves metal vapor at high temperatures, which cools
in an air stream and condenses into very fine particles commonly
called metallurgical smoke or fresh metal fume. The nucleation of
vapor is facilitated by the intense concentration of ions produced
around the arc. Most of the original particles are believed to be
(2)
formed by condensation on ions rather than by self nucleation.
2-1
-------
The unit pMrtlcJ.es are electrically charged. The complexes consist
of long chains, frequently of hundreds of fine particles, most of
which are too small to be resolved by the optical microscope so
that electron microscopy must be used to observe their structure
and form; electron diffraction shows the composition and nature of
(2)
their surface regions more sensitively than Xray diffraction. '
The particles generated were analyzed for mass concentration
and characterization, and distribution of sizes.
A problem report describing the historical development of
the equipment in this project is presently being written by Don
Stone.
2-2
-------
REVIEW OF LITERATURE
The entire subject area of very fine particles is growing
rapidly. Those working in the field know that the published litera-
ture has fallen far behind the real developments. This was clearly
depicted during the Aerosol Measurement Workshop at Gainesville,
Florida, in March 1976.
This literature review covers a few topics relating to very
fine particle generation, sampling and analysis.
For years, Professor Linsky has pushed for recognition of
small particles in air pollution separate from total emitted par-
ticulates. In 1958, he presented a paper explaining the need for
/3\
such a listing by particles sizes in an emissions inventory. He
has also pointed out the lack of acceptable methods for producing
fine fresh metal particles as distinguished from "stale particles."
(4)
Robert Redden developed a series of modifications of gene-
rators of very fine particles that utilized an electric arc welder
with a consumable wire feedstock as an anode and a relatively non-
consumable tungsten cathode.
Redden"s literature review, which was not comprehensive,dis-
cussed the following limitations in particle generation to date (1972)
1) The homogeneous aerosol generators yield con-
centrations that are too dilute for air pollu-
tion control device research and development.
2) Higher concentrations can be achieved by re-
dispersing large quantities of previously pro-
duced and collected powder but these particles
lack the quality being "fresh" and are agglom-
erated .
2-3
-------
3) Direct attrition methods — friction, evapora-
tion, impact, explosion and combustion approach
the situations in the industrial environments
but do not generally achieve particles smaller
than one micrometer. Submicrometer ranges avail-
able are not reproducible due to control of feed-
stock and input energy.
4) Electric arc generators have been used pre-
viously but particle characterization was given
little consideration
5) Finally, a. high voltage continuous discharge
method lacks the features of high concentration
and runs lasting an hour or more.' '
Hedden's equipment provided data which established reprodu-
cibility of mass concentration. For a particular set of operating
3
conditions he indicated an average concentration of 0.65 gram/sm
(1.49 grain/scf) at standard conditions of 25°C and 760 mm. He also
studied particles collected on one filter by electron microscopy.
Due to limitations of time and money, he was not able to investigate
the reproducibility of particle diameter sizes by count nor the in-
fluence of operating parameters on mass concentration and particle
size. He used a carbon steel welding wire. Runs were limited to
30 minutes as the tungsten electrode would be consumed by that time.
Due to limited facilities, Hedden prepared the samples col-
lected on membrane filters for electron microscopy analysis by dis-
persing the particles on a suitable grid. The dispersion probably
altered agglomeration characteristics.
A major revision of Hedden's modifications was developed in
1973-74 by Hedden, Professor Linsky, Dr. William H. Fischer, Arland
Johansen, Michael McCawley, Rasool Nekooi, and Dr. Sears. ' The
revised apparatus was an electric welder which used two consumable
2-4
-------
wires Instead of one. This machine suffered stability problems due
to inadequate control of the wires meeting and forming an arc. Gaps
of greater than fc inch resulted in unstable runs. ' Mass concen-
trations were measured for varying voltage and rates of wire feed
but, due to equipment limitations, the data was insufficient to
establish conclusive results.
A method was cited by Brain^ ' in 1974 for producing ferric
oxide by combustion of iron pentacarbonyl, He observed concentra-
Q
tions in an animal exposure chamber ranging from 1.00 to 4.00 g/m
and noticed agglomerates of particles about O.G2(j,m in diameter averaging
0.4 (J.m in average count diameter. He said the diameters of the
agglomerates were distributed log normally. While he argued that
the iron oxides provide excellent test particles for animal exposure,
his technique was obviously limited to iron oxide.
Hedden's size analysis of iron oxide particles also indicated
a log normal distribution of diameters by count. However, neither
Hedden nor Brain used statistical techniques of goodness to fit
tests to check whether their observed size distributions were rea-
sonably close to the generally claimed log normal distributions.
This report examines a method of producing metal oxide
particles to determine its acceptability with respect to continuous
generation, influence of operating conditions and reproducibility.
2-5
-------
EXPERIMENTAL METHODS
GENERATION
The metallizes, manufactured by Flame Spray Industries, has
many commercial applications. It is widely used to provide coatings
on metal or to fill in and build up worn areas on shafts, etc. One
use for zinc wire in a continuous production line is to replace the
rust proof coating along the machined or rolled threaded ends of elec-
trical conduit that had been galvanized before the thread cutting.
The metallizer operates in a manner similar to an electric
arc welder except that the product is blown onto another piece of
material that is not connected electrically to the electrodes or
power supply. The consumable wires are energized by the high
amperage low voltage power source forming a plasma arc through
which current will flow. The heating volatilizes more wire while
the compressed air upstream dissipates and directs the condensing
particles. A steady state of an arc generating fine fresh metal-
oxide particles is established.
The specific metallizer used in this research (Fig. 1) con-
sists of a power source which converts 3 phase 220 volt alternating
current to direct current that is variable from 0 to 40 volts and
up to about 400 amps; a console on top of the power source from
which air pressure, wire feed rate, and open circuit voltage are
adjusted; flexible conduits which convey wire, from reels secured
to the console, to a phenolic spray head that guides the wires into
fixed position electrode tips. Electrical cables transfer the direct
current from the power source to the wires at the spray head.
2-6
-------
The spray head was fastened to the lid of a 55 gallon barrel,
with the electrodes extending through a slot into the barrel. The
arcing takes place in the barrel. The larger particles of the
metallized wire are directed by the high pressure air stream to the
other end of the barrel where they are deposited.
A low pressure air stream entering through a HEPA (high effi-
ciency particulate air filter) and the dispersed compressed air convey
the much smaller particles through an exhaust duct to a cloth filter
collector and exhaust blower (baghouse system). The entire layout
of generation, exhaust and sampling is shown in Figure 1.
The size of the deposited particles in the barrel not drawn
into the exhaust stream can be approximately calculated by treating
the barrel as a cyclone. As estimated by the formulas and computa-
tions in Appendix A, the barrel bottom would collect 50% by weight
of the particles 7 jj,m in diameter and 95% by weight of those 30 M-m
in diameter.
SAMPLING
General Cons iderat ions
To assure the best representation of the very fine particles,
sampling was done under iso-velocity conditions. The procedure
followed recommendations of the EPA Method Five^ ' where possible.
The recommended sampling train, as interpreted by Research Appli-
ance Company, ' is indicated in Figure 2. This was modified as
shown in Figure 3.
The membrane filter holder was seated on top of the sampling
2-7
-------
Phenolic spray head Cross-sect ion
I E G F G E I
|
A) Power source of metallizer
8) Control console
C) Dual wire spools
D) Wire straightening, drive and conduit feed mechanism
E) Flexible conduit for wire
F) Flexible compressed air line - main air
G) High amperage electrical cable
H) 55 gallon barrel
I) Phenolic spray head
J) Dilution air from Hepa filter
K) Flexible compressed air line - secondary air
L) Exhaust duct - 6 inch diameter
M) Sampling port
N) Venturi
O) Stack sampler
Figure 1. Layout of equipment
2-8
-------
N>
I
vo
3)
4)
5)
6)
3
9)
1GJ
11)
12)
13)
14)
15}
16)
17)
19
19)
20)
21)
Probe
Cyclone "\ ,_.. . .
Flask J Eliminators
Particulate filter
Impingers (Greenburg-Smith)
Thermometer
Check valve
Umbilical cord
Vacuum gage
Course flow adjust valve
Fine flow adjust valve
Oiler
Vacuum pump
Filter
Dry gas meter
Orifice tube
Incline manometer
Solenoid valves
Pitot
Thermocouple
Pyrometer
21
Figure 2. Schematic of EPA-RAC sampling train
-------
T)16
I
M
O
13
1) Probe
2) Filter holder
3) Umbilical cord
4) Impingers
5) Vacuum gage
6) Course flow adjust valve
7) Fine flow adjust valve
8) Oiler
9) Vacuum pump
10) Filter
11} Thermometer
12) Dry gas meter
13) Orifice tube
14) Incline manometer
15) Solenoid valves
16) Wet bulb thermometer
r \
14
Figure 3, Actual stack sampling flow system
-------
port, itself on top of the horizontally positioned duct. The inten-
tion was for the filter to trap all sizes of particles reaching it,
differing from the arrangement in Figure 2 whereby the first impinger
collects larger particles. A second difference in procedure was in
the method of moisture measurement. A wet bulb thermometer was
placed in the duct to determine the dew point. This was preferred
to the method of monitoring the water levels in the impingers and
measuring the weight gain of the silica gel impinger, as small incre-
mental increases were anticipated.
Description of Sampling Components
The sampling train (Figure 3) consisted of the filter holder,
filter cord to impingers, impingers, cord to the sampler, and the
RAG stack sampler. Manometers were placed at the HEPA filter, venturi
meter, and baghouse, in addition to the two contained in the sampler
for monitoring orifice pressure drop and stack-velocity pressure.
Also, thermometers are placed on the barrel lid near the spray head,
2 inches downstream of the sampling port, and at the baghouse.
Sampling Port. A % inch opening, 11 diameters from exhaust
port on lid.
Probe. The S shaped straight-inlet nozzle was inserted into
the air stream 1/3 diameter vertically from the port entrance.
Filter holder. The Gelman filter holder accepts 49 mm filters.
The nozzle assembly is directly connected to the inlet of the holder.
Filters. Millipore filters, type HA, .45 (jtm pore diameter,
were used to determine the mass concentrations which required collection
2-11
-------
times of several minutes. General Electric Nuclepore filters were
employed for short duration particle sizing sampling. The filters
are discussed further in the next sections.
Impingers. Moisture in the sampled air was removed by the
series of four 1 liter impingers, chilled by an ice bath. The first
two impingers each had 100 ml distilled water, the third was dry
and the fourth 200 grams of dry silica gel.
RAG Stack sampler. The RAG Stack sampler Model 2343(5»^
consists of all elements downstream of the' umbilical eord in
Figures 2 and 3. It satisfies the specifications of the EPA
Method V. '
Venturi Flow meter. A constriction was built into the ductwork
about 5 diameters downstream from the sampling port (Figure 4).
(9)
By a generally accepted stack sampling procedure, velocity tra-
verses along two perpendicular diameters at the sampling port loca-
tion were tabulated as a function of the pressure drops measured by
the manometer across the venturi. The velocities were averaged and
corresponding flow rates were calculated. The data and calculations
are presented in Appendix B. A plot of exhaust flow versus venturi
pressure drop is shown in Figure 5.
Filters
Nuclepore filters (0.2 (jjn pore diameter) were selected
for determining the distributions of particle diameters and for
identifying the particles for the following reasons. Their smooth
surface and better controlled pore size and frequency make them
2-12
-------
:M
D=6 inches
J
11D
•barrel
1) Direction of air stream from chamber
2) Sampling nozzle and filter holder with cord to impinger
3) Thermometer
4) Venturi constriction with taps for static pressure
5) Incline manometer
Figure 4. Schematic of sampling point and venturi meter
2-13
-------
450
400
350 -
o
300 -
s:
x
UJ
250 -
200
.1
.2
.3
.4
.5
Venturi pressure drop (inches water)
Figure 5. Exhaust flow versus venturi pressure drop
2-14
-------
Ideal for analysis by electron microscopy and electron dlffrac-
tlon- ' ' ' The Millipore (Type HA, 0.45 (j,tn pore diameter)
filters were selected for determining the mass concentrations of
the particles sampled. The Millipore filter's greater thickness
(150 urn) than the Nude pore filters (12 ^m) and its random fibrous
structure allows the greater retention volume needed for several
minutes of exposure in the particle stream.
The added advantage of the Nuclepore filters, which have
simple straight through pores compared to the more "tortuous path"
configuration of the Millipore membranes, is that they most closely
approximate the popular model of filters used to predict particle
collection efficiency. '
Calculation of Nuclepore Efficiency
One mathematical model for the study of membrane filters
assumes a flat surface with equidistant parallel straight through
pores. Since Nuclepore filters more closely resemble this model
than other membrane filters the theory has been applied and corrected
(14)
by Spurny, et. al.
Three mechanisms: interception, diffusion, and impaction,
contribute to the overall efficiency Efc. If the impaction efficiency
is labeled as E., the diffusion efficiency as Ed, and the interception
as E , then the respective efficiences are individually calculated
and combined as follows:
Et =E.+Ed+0.15 Er -EiEd - 0.15 EI Er
In Appendix C, the individual efficiencies are defined, and appropriate
2-15
-------
values assumed are listed and efficiences are calculated. The
following table summarizes the computed relationship between effi-
ciences and particle size.
Table 1. Summary of Nuclepore Impaction, Interception, Diffusion
and Total Collection Efficiencies (Pore Diam. = 0.2 (j,m)
Part. Dia.
( fim)
.02
.04
.10
.15
.20
.30
E.
impact.
.00
.00
.012
.028
.05
.447
.15E
r
intercept .
.00
.054
.11
.13
.15
1.00
Ej
d
diffusion
1.00
1.00
1.00
.993
.962
.853
E
t
total
1.00
1.00
1.00
1.00
1.00
1.00
Sampling Procedure for Mass Concentration
The following steps outline the method followed by the experi-
menters when sampling for mass concentration. Before sampling:
Measure atmospheric pressure at the equipment location. Oessicate
the Millipore filters for two days and obtain tare weight to nearest
0.1 mg.
1) Equipment should be set up as in Figure 1. Metallizer
has been running at a specific condition for at least two minutes
and baghouse has been turned on. Millipore filter has been placed
into holder.
2) Adjust flow at venturi meter. For a desired flow of 250
acfm (7.08acmm), for example, the manometer should indicate a pressure
difference of 0,2 inches according to Figure 5.
3) Measure the velocity pressure of the exhaust stream at
2-16
-------
the sampling nozzle location. Continuing the above example, this has
been measured as 0.12 inches for 250 scfm. Compute stack velocity by
Vstack= 234 (298 x °' 12>^fPm - 140° fPm (see Appendix B) . Convert this to
a nozzle flow yielding same velocity. Since the nozzle has an 1/8 inch dia-
meter, with area of 0-000085 ft2 the required nozzle flow is 0.12 scfm.
4) Make a preliminary run. Since isovelocity sampling is
planned, the ratio of inlet nozzle velocity to duct velocity should
be in the range of 0.9-1.1 (4). A preliminary run is necessary to
check the rate of sampler pumping. Consult Figure 6 for an approxi-
mate setting for the orifice pressure drop. For .12 scfm, k, (orifice) =
0.07 inches. This setting should be checked and manipulated after the
probe has been inserted into the gas stream, described in step 6.
5) On a form similar to Figure 7, record HEPA filter APQ ,
barrel (chamber) T, Average dry gas meter T at start, Meter CF start,
and venturi AP-
6) Two to three persons are necessary. One person manually
inserts probe into sampling port on a signal. A second person turns
on sampling pump and timer at same signal. A third person gives the
signal. All three monitor and note stack temperature, baghouse
temperature, the temperatures and pressures in step 5 as well as
the metallizer load voltage response, amperage, air pressure and
quality of generation.
7) Upon signaling the end of the preliminary sampling period,
about 2 minutes, remove the probe and turn off the pump.
8) Remove the filter from the holder and store in a covered
2-17
-------
E
>»—
u
N
N
O
.20
.19 -
.18 -
.17-
.16-
.15 -
.14 -
.13-
.12-
.1 1 -
.10 -
.09 -
.08 -
.07 -
.06 -
.05
i i i I I i i i i
.01 .02 .03 .04 .05 .06 .07 .08 .09 .10
AH of orifice (inches water)
Figure 6. Approximate relationship of sampling nozzle
flow to AH orifice of sample
2-18
-------
Wire type p
• • bar-
Wet bulb Temp.
1. Sample #
2. Barrel T(°C)
\234
25. I ^
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19-
20.
21.
22.
23.
24.
HEPA AP
Pduct=Pbar-@/13.6
Venturi AP
Duct T (°C)
Bag T
Pitot AP
Wire feed rate
Amps
open
Vload
MAP
SAP
AH orifice
CF start
CF end
ACF
ATIME
CFM = @/@
Tmeterstart
T
meter ,
end
Nozzle Vel.
Vduct
Figure 7. Typical form for Recording
Mass Concentration Sampling Data
2-19
-------
petri dish. With compressed air, blow the probe and holder inlet free
of settled dust.
8) Note CF end. Approximate the isovelocity parameter by
performing following computations.
Nozzle vel. = ACF/ATIME •(!/.000085 fpm)
Stack Vel. = 234 (•12x29-92 xTstaclloK^/Pbar' ^See APPendix B)
P, is the absolute total pressure at the sampling port.
D fit IT
I =Nozzle Vel/Stack Vel.
This is only approximate since it ignores moisture and meter temper-
ature. But it is adequate for sampling. If the computed I is within
the range of .9-1.1 then assume the sampler is properly adjusted and
a sample may be taken by repeating the above steps. The sampling
time should be designed to yield approximately a lOmg to 20mg
deposit on the filter.
If the computed I is outside of the range, the $H on the pump
must be changed in a direction that would bring the I ratio closer
to unity. To achieve the acceptable ratio, another preliminary
sample should be taken if the deviation is significant, otherwise
take a sample and adjust the AH as necessary.
9) During one of the sampling periods, the wet bulb tempera-
ture should be measured and recorded. As a computation in Appendix D
shows, the maximum moisture influence at the range of operating
conditions was 4.4%, and thus this factor is not important in esti-
mating the isovelocity factor.
10) At the end of all sampling for the day, place filters
with samples that are considered satisfactory in a dessicator, laying
2-20
-------
in the petri dish with the lid removed. After about 2 days, weigh
them to the nearest 0.1 mg. (Weighing was not necessarily done in
a dry environment.)
Sampling Procedure for Electron Microscopy
The goal in sampling with the Nuclepore filters was to obtain
a deposit with a countable number of particles. When the loading
is too intense, individual particles are not distinguishable. When
loading is too light, the array of particles is too low to provide
statistically satisfactory results. Isovelocity conditions are
strived for but are difficult to confirm due to the short sampling
time. Particle size sampling is usually done concurrently with mass
sampling. Nuclepore filters are not dessciated and tared. The
following steps outline the method followed.
1) Satisfy the first 3 steps on page 2-16. Place the Nuclepore
filter into holder with glossy side toward sample stream.
2) Start the pump with the probe not in the exhaust stream.
After about 5 seconds, insert the probe and direct the
nozzle toward the stream and hold for an interval of 1
second to 4 seconds.
3) Repeat the above steps, using different holding times.
4) After collecting several filters, examine them visually and
with ari optical microscope. The filters should show a
slight shadow or darkening due to the deposit. At a
magnification of several hundred times, distinct particles
should be apparent along with ample clean area.
2-21
-------
Sensitivity of Results to Deviations from Isovelocity Sampling
According to Peterson, aerosol sampling is subject to two
biases: velocity and inertial impaction. Bias is defined as
(13, pg F-4).
"Volumetric particle concentration or size distri-
bution determined from probe sample divided by
actual array of particles which existed in aerosol
cloud."
He described work by Lundgren and Calvert in which the two
sampling biases are referred to as R and K.
R s V /V
o c
K = (Cp V D )/(18|j,D )
where C = the Cunningham Correction factor =
1 + 1.62xlO"5/D +((.55) xlO"5/D ) EXP(-.67 x 10-+5 D )
P P- P
V = Duct Velocity = (1400 fpm) = 711 cm/sec
V = Probe inlet velocity
D = particle diameter (cm)
P
p = particle density,
p, = viscosity = 185 x 10" poise (at 25°C)(ref. 16)
D = probe inlet diameter = 1/8 inch = .32 cm
c
Peterson developed the following graph, modified from Lundgren
and Calvert data, that indicates "typical experimental sampler inlet
bias for a straight inlet probe" plotted as a function of R and K.
Peterson said, (13, p. 1-4)
"As is quite often true in aerosol studies the . . .K
is of primary importance. The range of maximum interest
is the K value range from0.01 to 1.0. Below0.01 the
sampling bias is low enough to be relatively unimpor-
tant. Above 1.0 the sampling bias is equal to the
velocity ratio (R)."
2-22
-------
R 1.0
0.001 0.01 0.1
Value of K, Dimensionless
1.
Figure 8. Plot of lines of equal bias as function of
R and K (13, pg. L-5)
The above formula for K and graph of R and K are applied to
predict the biases to which the particle sampling is subject. Since
there are small fluctuations in the duct velocity pressure, after
sampling has commenced, it is assumed the actual R could vary from
0.8 to 1.2, The following table summarizes the bias study for which
computations are given in Appendix D.
Table 2. Summary of Possible Sampling Biases Duct
Velocity - 711 cm/sec
Particle
Diameter
cm
1 x 10~7
•*• ** _/i
i x 10 7
—A
i x 10 7
—A.
i x 10 7
-6
1 x 10 ?
-6
1 x 10 ?
1 x 10"1
1 x 10 -6
|jiin
1.00
1.00
1.00
1.00
0.01
0.01
0.01
0.01
c
1.16
1.16
1.16
1.16
22.34
22.34
22.34
22.34
PP
5
5
1
1
5
5
1
1
R
1.2
0.8
1.2
0.8
1.2
0.8
1.2
0.8
K
.039
.039
.0077
.0077
.0001
.0001
.0001
.0001
Bias (1.00
is left of
^ sign)
< 1.03
> .97
< 1.01
> .99
= 1.00
- 1.00
= 1.00
= 1.00
2-23
-------
For the most conservative assumptions, the maximum error,
according to the Peterson article, would be 3%. If discrete parti-
cles on the order of 0.01(j,m (lOntn) are present, the sampling probe
would show no significant bias to them.
Another bias that might occur during the particle sizing
sampling would be a misalignment of the probe with respect to the
exhaust stream. The filter is exposed while the probe is being posi-
tioned during which some misorientatien necessarily occurs, Limited data
by Lundgren and Calvert indicate an angular difference of 90 between
inlet and stream flow for 2 |j,m particles gives a bias of less than
(13")
5%. Therefore, the placing and removing of the probe probably
causes little bias in the observed particle sizes and distribution.
This problem could be expected to be insignificant in the mass con-
centration samples. An additional bias occuring is due to the de-
positing of particles along the walls of the probe and filter holder
upstream of the filter.
ANALYTICAL TECHNIQUES
Mass Concentration Analytical Techniques
Compute the mass concentration of the sampled air by dividing
the weight gain of a Millipore filter by the standardized volume of
air pumped by the sampler, Dessicate the filters before and after
use for approximately 48 hours. Weigh the filters on an analytical
electro balance (approximate precision of +0.5mg). The balance
scale indicates weights to the nearest 0.1 mg.
The mass concentration analyses are done by the experimenters.
2-24
-------
Particle Sizing and Identification Techniques
Store the filtered samples individually in labeled, capped petri
dishes. Send to Walter McCrone Associates, Chicago, Illinois, the
contractor for microscopy analysis. McCrone examines the samples by
transmission electron microscopy (TEM), scanning area electron micros-
copy (SEM), and selected area electron diffraction (SAED),
McCrone prepares samples for scanning electron microscopy by
a replication process. See Appendix E. All of the original features
on the filter surface, i.e., pore openings, and aggregates are pre-
served.
SEM photographs are most informative because of their three
dimensional representation. Objects closer to the electron beam
collector appear brighter than those farther away. The depth of
field of the SEM is 300 to 500 times that available in a light micro-
(12)
scope of the same magnification. Photographs from McCrone indi-
cated magnifications ranging from 5,000 x to 11,000 x. SEM is thus
useful for studying particle shapes, porosity, aggregation and surface
features.
The TEM allows inspection of individual particles by magni-
fications of up to several hundred thousands times but sacrifices
the SEM's depth of field.
Using the TEM, particles are counted and sized by "hand" by
(14)
observing about six areas of the deposit.
High energy electron diffraction of an individual particle
yields photographs showing a set of dark and light concentric bands,
whose spacings and relative intensities depend upon the structure
2-25
-------
uncl composition of the specimen. The pattern formed by scattered
electrons is Mcanned by a recording microphotometer. The calibrated
output shows experimental diameters of circles from specific reflect-
ing planes. These diameters are compared with the ASTM diameter,
(14)
for the specified plane, to identify the sample material.
SELECTION OF EQUIPMENT OPERATING CONDITIONS FOR SAMPLING
The metallizer manufacturer recommends that the electric arc
be kept as small as possible without 'spitting1 to achieve the most
efficient metallizing. The experimenters consider this condition
as stable and observe that stability results in the following charac-
teristics :
1. Particle generation is smoother and quieter.
2. The indicated amperage remains steady.
3. The load voltage is either equal to or a few volts
less than the open circuit voltage setting, and does
not oscillate about this level by more than one volt.
The experimenters assume that smaller particles and repro-
ducibility will result from stable operating conditions.
The metallizer can be readily adjusted for rates of wire feed,
air pressures to the arc (MAP and SAP) and open circuit voltage.
The exhaust flow rate can also be adjusted by a damper and valve in
the exhaust ductwork. It should be noted that the open circuit
voltage is considered a dependent variable of the wirefeed rate.
The wirefeed rate establishes the amperage across the arc.
With due consideration for the funds available for electron
2-26
-------
microscopy analysis, the following conditions were selected to examine
the zinc wire fume.
Table 3. Zinc Wire Sampling Conditions
Wirefeed
Condition Rate open load Amps MAP Flow SAP
1
2
3
4
30
30
50
50
21
21
23
23
21
21
22
22
30
30
70
70
70
50
70
50
250
250
250
250
20
20
20
20
Notes: Wirefeed rate- percent of full rate, V0 - open circuit
voltage, V1 ,- voltage across arc, MAP - main air pressure-psi,
SAP- secondary air pressure, Flow- Exhaust flow in duct-scfm.
2-27
-------
RESULTS
BRIEF EVALUATION OF EQUIPMENT OPERATION
As of April 1976 the experimental apparatus was considered
moderately reliable by the experimenters. The equipment would not
operate consistently satisfactory at established stable conditions.
Specifically, many runs of zinc wire were interrupted when either
of the wires jammed in the flexible conduit due to various friction
affects. Also, the observed stabilities of steel wire at several
operating conditions and of the zinc at condition four (Table 3)
were not repeatable on different days.
It is important to note that the metallizer by itself is
noisy and the use of the large particle collection drum adds to
the noise. This noise, while not measured, interferred with communi-
cation between research personnel in a high ceiling room. If not
corrected, it would reduce research work effectiveness, possibly
interfere with the other research work nearby and could be expected
to create stress in animals if the particles were used in animal
quarters. The noise can undoubtedly be corrected by inherent design
changes and by additional noise control materials.
2-28
-------
SAMPLE NUMBER IDENTIFICATION
Mass Concentration
The zinc oxide samples shown in Table 4 were analyzed for
mass concentration.
Table 4. Identification of Mass Concentration Samples
Sampling Condition Sample Numbers
1 13, 19, 34, 45, 46
2 31, 32, 33, 44
3 15, 18, 35, 36
4 37, 38, 40
Electron Microscopy and Electron Diffraction
The zinc oxide samples shown in Table 5 were sent to Walter
McCrone Associates for SEM, TEM and SAED analysis.
Table 5. Identification of McCrone Samples
Sampling Condition Sample Numbers
1 101, 125
2 133, 134
3 104, 128
CHEMICAL COMPOSITION
Electron Diffraction Data
Selected area electron diffraction results, as received from
McCrone Associates, are given in Tables 6 and 7. According to Richard
Shimps of McCrone Associates, these tables indicate that the particles
generated were ZnO.
2-29
-------
Table 6. Electron Diffraction Data -- Sample 104
SAED Calculated diameters ASTM Card S-0664
of Circles from Specific (^nO) diameters
Reflecting Planes of Circles from Specific
Reflecting Planes
Angstrom (0.1 nm)
2.814
2.592
2.463
1.903
1.622
1.498
1.378
Angstrom (0.1
2.816
2.602
2.476
1.911
1,626
1.477
1.379
nm)
Table 7. Diffraction Data -- Sample 133
SAED Calculated diameters *>™ C*f S:°664 _
of Circles from Specific ^n0 diameters of
Reflecting Planes Circles from Specific
_ Reflecting Planes
Angstrom (0.1 nm) Angstrom (0.1 nm)
2.82 2.816
2.60 2.602
2.50 2.476
1.92 1.911
1.635 1.626
1.483 1.477
1,392 1.379
Assay of Wire
A typical assay of the Wall Colmonoy zinc wire (Platt #302)
15 gauge (.145 cm diameter)is
Zinc 99.99$ +•
Iron .0015%
Cadium .0015%
Lead .002% Maximum
2-30
-------
MASS CONCENTRATION
Mass concentration sampling data as observed are presented
in Appendix F. Concentrations expressed as mg/acf are converted to
mg/scf at the standard conditions of 760 mm mercury and 25° Centi-
grade. Table 8 summarizes the standardized concentrations expressed
/ 3
as g/sm .
Table 8. Standardized Zinc Oxide Mass
Concentrations for all Samples
CONDITION
REPLICATE
1
#
13
19
34
45
46
3
g/sm
0.87
1.26
1.00
1.20
0.80
2
#
31
32
33
44
g/sm
0.74
0.67
0.62
0.63
3
#
15
18
35
36
g/sm
1.91
2.04
2.02
2.25
4
#
37
38
40
g/sm
1.03
1.09
1.23
The concentrations in Table 8 are more meaningful when the
statistical parameters -- mean, standard deviation, and coefficient
of variation (C.V.) are calculated. The coefficient of variation
is one indicator of reproducibility or precision of concentration
generation. These parameters are summarized in Table 9A.
Table 9A. Zinc Oxide Mass Concentration
Mean and Variation Summary
Condition 1 2 3 4
(1) Mean (g/sm3) 1.03 0.68 2.05 1.12
(2) Mean (grain/scf)©x 2.29 2.36 1.56 4.69 2.56
(3) Stand, dev. (g/sm3) 0.20 0.05 0.14 0.10
(4) Coefficient (3, /(J)x 100) 19% 8.4% 6.9% 8.7%
of variation
2-31
-------
Table 9B. Mean Mass Concentration Matrix
Wirefeed 30 Wirefeed 50
MAP 50 0.68 cond. 2 1.12 cond. 4
MAP 70 1.03 cond. 1 2.05 cond. 3
Row 4 of Table 9 indicates that the C.V.'s of conditions 2,
3, and 4 are less than 9% while condition 1 has a C.V. of 19%.
Condition 1 is apparently much less reproducible.
Means Testing
A question relative to the values in Table 9B is whether the
means are significantly different. The mean of condition 1 (X..)
is fairly close to X, but two operating variables (wire feed rate
and MAP) are different. A statistical method to test the other
C\g\
pairs of means is the Smith Satterthwaite testv . This method
is utilized by hypothesis tests applied (see Appendix G-l).
These hypothesis tests support the conclusion at the 5%
level of significance that X^ > X^, X- > iL and X- > X,. This
implies that the mass concentration of zinc oxide particles is
increased when air pressure or wirefeed rate or both are increased
Analysis of Variance
Another statistical analysis applicable to the mass concen-
tration data is the analysis of variance. This is generally a more
efficient and more powerful systematic approach of distinguishing
experimental (random) error from variation due to different operating
(19)
conditions. ' However, appropriate experimental design to facili-
tate the analysis of variance would have required samples from
2-32
-------
several more wirefeed rates and MAP's instead of several replicates
for each condition. Application of the two way classification
analysis of variance as described on pp. 274-278 of ref. 19 yields
results in contradiction to the hypothesis tests. But the contra-
diction is presumably caused by the insensitivity of the test at
the minimum degrees of freedom of the two variables.
PARTICLE SIZING BASED ON ELECTRON MICROSCOPY
The electron microscopy photographs of the zinc oxide
samples were all similar in terms of particle appearance. SEM
photographs in Appendix H of sample 133, show agglomerates of dif-
ferent shapes and sizes. The agglomerates range in size from less
than 0.1 p,m to over 2 p,m and appear to be interconnected. The
shapes vary from spherical to chainlike. These agglomerates are
composed of very fine particles as shown in the TEM photos for
sample 101 (Fig. H-3). These particles, approximately circular
and presumably spherical, have diameters of about 10 nm (.01 fj,m).
These were the particles sized and counted by the McCrone organiza-
tion. Since the agglomerates were interconnected, McCrone's group
made the assumption that there was no meaningful method to count
them for some appropriate dimension. The agglomeration is dis-
cussed more extensively later in the report.
Table 10 is a modified form of the zinc oxide particle
sizing and count data as it was received from McCrone Associates.
This table indicates the number of observed particle diameters in
each one nanometer wide interval. Particles were not counted by
2-33
-------
Table 10A. Zinc Oxide Particle Size Data
N3
I
Size Range
nm
20 -» (a)
19-20
18-19
17-18
16-17
15-16
14-15
13-14
12-13
11-12
10-11
9-10
8-9
7-8
6-7
5-6
4-5
3-4
2-3
1-2
0-1
Total
Sample
number
2
0
0
1
1
2
3
2
9
13
8
13
23
32
36
41
15
15
26
23
32
297
101
percent
.67
-
-
.33
.33
.67
1.01
.67
3.03
4.37
2.69
4.37
7.74
10.77
12.12
13.80
5.07
5.07
8.75
7.74
10.77
99.93%
Sample
number
24
2
7
17
18
24
16
22
34
22
22
20
35
19
16
12
17
12
20
11
6
376
104
percent
6.36
.53
1.86
4.52
4.78
6.38
4.25
5.85
9.04
5.85
5.85
5.31
9.30
5.05
4.25
3.19
4.52
3.19
5.31
2.92
1.59
99.90%
Size Range
nm
20 •*
19-20
18-19
17-18
16-17
15-16
14-15
13-14
12-13
11-12
10-11
9-10
8-9
7-8
6-7
5-6
4-5
3-4
2-3
4-2
Sample
number
26
7
3
1
11
9
9
9
29
5
13
9
31
20
31
20
8
36
2
1
280
125
percent
9.29
2.50
1.07
1.07
3.93
3.21
3.21
3.21
10.36
1.79
4.64
3.21
11.07
7.14
11.07
7.14
2.86
12.86
0.71
0.36
99.99%
(a) In samples 104, 125, 128, 133, 134 there were no more than 3 particles per 1 nm interval > 20nm.
Particles were observed up to 50 nm. In this report particles > 20 nm are treated mathe-
matically as belonging to an interval from 20-22nm. The investigator realizes that this
assumption significantly influences the computation of statistical parameters.
-------
Table 10B. Zinc Oxide Particle Size Data
I
U>
Ul
Size Range
nm
20 -» (a)
19-20
18-19
17-18
16-17
15-16
14-15
13-14
12-13
11-12
10-11
9-10
8-9
7-8
6-7
5-6
4-5
3-4
2-3
*2
Total
Sample 128
number percent
19
2
5
2
4
7
9
13
43
17
33
14
11
25
19
17
23
17
1
0
281
6.76
0.71
1.78
0.71
1.42
2.49
3.20
4.63
15.30
6.05
11.74
4.98
3.91
8.90
6.76
6.05
8.19
6.05
0.36
0.00
99,99
Sample
number
25
5
6
2
11
12
8
17
27
9
29
9
21
15
29
45
21
41
13
5
350
133
percent
7.14
1.43
1.71
0.57
3.14
3.43
2.29
4.86
7.71
2.57
8.29
2.57
6.00
4.29
8.29
12.86
6.00
11.71
3.71
1.43
100.00
Sample
number
31
8
6
11
9
13
3
10
8
15
24
5
8
11
4
11
25
49
25
4
134
percent
11.07
2.86
2.14
3.93
3.21
4.64
1.07
3.57
2.86
5.36
8.57
1.79
2.86
3.93
1.43
3.93
8.93
17.50
8.93
1.43
280 100 . 01
(a) In samples 104, 125, 128, 133, 134 there were no more than 3 particles
per 1 nm interval > 20 nm. Particles were observed up to 50 nm. In
this report particles > 20 nm are treated mathematically as belonging
to an interval from 20-22 nm. The investigator realizes that this
assumption significantly influences the computation of statistical
parameters.
-------
mass or surface area.
Analysis
It was intended that the analysis of the particle size data
would answer the following questions.
«
1) Are the particle size statistical parameters (mean,
variance, distribution) from each replicate sample
significantly different?
2) Are the statistical parameters from samples differing
by one variable significantly different?
3) Do any theoretical statistical distributions reasonably
fit the observed distributions?
Such an analysis would give a quantitative measure of the repro-
ducibility potential and effect of the operating variables in addition
to a characterization of the generated particles.
Computation of Means and Variances
The mean diameter for each sample can be calculated by
_ n
D - S d.f
i-1 1
where d. = midpoint of size interval
n = number of intervals
fi = decimal fraction of total within interval
n
E f, = 1.00
2-36
-------
The variance of the sample diameters can be computed from
2 n 2 -2
S = S x/f - D^
L i
Table 11. Means and Variances of zinc Oxide Particle Size Data
Sample #
Condition II
D (nm)
S2 (nm2)
S (nm)
Sample Size
101
1
6.0
14.6
3.8
297
125
1
10.2
29.2
5.4
280
133
2
9.1
35.6
6.0
350
134
2
10.0
41.1
6.4
280
104
3
10.8
28.1
5.3
376
128
3
10.4
22.0
4.7
281
Reproducibility Among Replicates
The mean diameters for samples 101 (D,-,,) and 125 (D125) of
condition 1 are significantly different by inspection.
The values of the pairs D^- and D, •,* f°r condition 2, and
^in4 an(* ^128 ^or con<*ition 3 are closer and are analyzed by a
statistical method for significant difference in Appendix G-2.
This analysis indicates that the two replicates for condi-
tion 2 have diameter means and variances that are not significantly
different. The analysis also indicates that the means and vari-
ances for condition 3 replicates are not significantly different.
Differences Between Conditions
An important question is whether the means and variances for
particle diameter populations generated under different conditions
are different. Since the mean of condition 1 cannot be estimated
within a relatively tight range, only the relationship of condition
2-37
-------
2 to condition 3 is examined. However, conditions 2 and 3 differ by
two variables, and if the mean sizes are different there is a lack
of data to distinguish variable influence.
Hypothesis testing in Appendix G-3 demonstrates that the
mean particle diameters for conditions 2 and 3 are significantly
different. Thus changing the operating conditions apparently
changes the mean particle diameters.
Frequency Distributions
Another statistical parameter that provides useful informa-
tion is the frequency distribution. The observed distributions are
presented above in Table 10. The following two tables and two
graphs indicate the cumulative frequency for conditions 2 and 3.
In addition typical histograms representing observed frequencies
and frequencies expected from the appropriate normal distribution
are plotted.
The normal distribution provided a better fit to the observed
distributions than did the log normal. This was found by calcu-
lating the maximum differences between the observed and expected
(theoretical) distributions. This is examined statistically in
the next section.
It was also attempted to decide tfhich of the two fit better
by plotting the data onto log normal and normal probability paper.
By this method, data should plot as a straight line. However,
this method was inconclusive as indicated by Figures 13 and 14.
2-38
-------
Table 12, Cumulative Observed and Expected Frequencies
for Condition 2
Maximum Diff. (a)
(133) (134)
d
ran
: 2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
z
-1.23
-1.06
- .90
- .74
- .58
- .42
- .26
- .10
+ .06
+ .23
+ .39
+ .55
+ .71
+ .87
1.03
1.19
1.35
1.52
1.68
(J, « 9.6
F(d)133
.014
.051
.168
.228
.357
.440
.483
.543
.569
.652
.678
.755
.804
.827
.861
.893
.898
.916
.930
a - 6.2
F(Z) :
.109
.145
.184
.230
.280
.337
.400
.46
.52
.59
.65
.71
.76
.81
.85
.88
.91
.94
.95
F(d)1;
.014
.104
.279
.370
.407
.42
.46
.49
.51
.59
.65
.67
.71
.72
.77
.80
.84
.86
.89
.140
,103
(a) The maximum absolute difference between observed (F(d)) and
expected frequency (F(Z)).
2- 39
-------
Table 13. Cumulative Observed and Expected Frequencies
for Condition 3
j, * 10,44 a - 4.9
Maximum Diff. (a)
(104) (128)
di
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Z
-1.93
-1.72
-1.52
-1.31
-1.11
- .91
- .70
- .50
- .29
- .09
+ .11
+ .32
+ .52
+ .73
+ .93
+1.13
+1.34
+1.54
+1.75
+1.95
F104
.016
.045
.098
.130
.175
.207
.250
.300
.393
.446
.505
.563
.654
.712
.755
.818
.866
.911
.930
.935
F(Z)
.0268
.043
.064
.095
.134
.181
.242
.309
.386
.464
.544
.626
.698
.767
.824
.871
.910
.938
.960
.974
F (d ) .. ,
-
-
.004
.064
.146
.206
.274
.363
.402
.452
.569
.630
.783
.829
.861
.886
.900
.907
.925
.932
.0844
.0688
(a) The maximum absolute difference between observed (F(d)) and
expected frequency (F(Z)).
2-40
-------
«0
i_
(D
E
o
\l^
c
rt
c
i_
/«)
VL*
+•>
(U
E
rd
T>
/ii
vU
lj
-H
C.
rt
IU
Q.
,
cv -
19 -
1S -
1 7 -
16 -
15 -
14 -i
13 '
12 -
11 -
10 -
9 -
S -
7 -
6 -
5 -
4 -
3 '
— — . o-^a— i
D Cum. freq. O AD
Normal dist.
O Cum. freq. ° ^
Sample 134
O IA
A Cum. freq.
Sample 133 o ^
O C&
O Q A
O D A
a\
D A
CQ A
DO A
D OA
D Gk
f
a A o
Q 0
za o
A oa
\ n • — —
O LJ 1 1 1 1 1 T i '
O 10 20 30 40 50 60 70 SO 90 10
: Cumulative percent
Figure 9. Comparison of cumulative frequencies for
condition 2 (table 12)
2-41
-------
'in
I
O
c
cti
* w
c
c_
"£
E
Til
"y
OJ
Q_
20 -
19 -
18 -
17 -
*
16 '
1 5 -
14 -
13 -
12 -
11 -
10 -
9 -
8 -
7 -
6 -
5 -
4 -
3 (
ZJLJ
D Cum. freq. CD
Normal dist.
O Cum. freq.
Sample 12fc
A CD
A Cum. freq.
Sample 104 ADO
ADO
AD 0
AD 0
A D
ADO
O
DO
tE O
CZLO
DO
DDA
CD A
1 DA i i
J LJ*-i .1 i l i i i i t
0 10 20 30 40 50 60 70 S»0 90 10
Cumulative percent
Figure 10. Comparison of cumulative frequencies for
condition 3 (table 13)
2- 42
-------
to
I
18 -
17 -
16 -
4) 15 -
T)
^ 14
13 -
E 12 -
C
v- 11 -
Percent in interva
-*t3(t>b-(fl®^CD(9O
•••^M^
— — Normal
distribution
f 17
- 16
- 15
• 14
- 13
• 12
- 11
• 1O
• 9
8
7
6
5
4
3
2
1
2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 1O.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5
Particle diameter, midpoint of interval (nm.)
Figure 11. Observed frequency of sample 134 and normal frequency
-------
NJ
JS
-P-
E
c
(U
c
c
-H
c
(D
Q_
14 -
13 -
12
11 -\
10
9 -
7 -
6
5 -
4 -
3 -
2 -
1 -
Normal distribution
-14
-13
-12
- 11
-10
- 9
- 8
- 7
6
- 5
- 4
- 3
- 2
1
2.5 3.5 4.5 5.5 6.5 7,5 8.5 9.5 10.5 11.5 12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5
Particle diameter, midpoint of interval (nm.)
Figure 12. Observed frequency of sample 133 aad normal frequency
-------
c
4)
U
£_
CD
Q.
3J
E
rj
U
95
90 -
80
70
60
50
40
3O
2O J
10 -
5 -
024
T 1 1 1 1 1 1 1
6 8 10 12 14 16 18 20 22
Particle diameter (nanometers)
Figure 13. Arithmetic probability plot of sample
128 diameters
2-45
-------
c
(D
U
t_
U
95
90 -
80
70
60
50
40
30
20 -
1 0 -
5 -
2 ~
1 -
0.5 -
0.2 -
0.1
—T~
3
-i 1 1 1—i—i—
5 6 7 S 9 10
20 30
Particle diameter (nanometers)
Figure 14. Log probability plot of sample 128 diameters
2-46
-------
Computation of Normal Distribution
The cumulative normal distribution can be constructed from
statistical tables by assuming values of jj, and a for the population
of diameters (d^. Tables ^ 9^ of the standard normal distribution
d. - n
give values of F(Z). F(Z) is the probability that (Z. = — ) < Z
1 C7 —
Goodness of Fit
Figures 9 and 10 show by inspection that the normal distri-
butions provide an approximate fit to the observed data. Two of the
statistical methods for testing goodness of fit are the Kolmogorov-
Smirnov (K-S) test(19'20^ and the chi square test.(19'21^ The
investigator used both methods and selected the K-S test for this
report. The K-S test is a more powerful test for continuous dis-
(25}
tribution fitting. '
The one sample K-S test concerns the agreement between an
observed distribution and a specified continuous distribution. The
two sample K-S test concerns the agreement between two observed
distributions. The test is (19, p. 222) "sensitive to population
differences with respect to location, dispersion or skewness."
Both tests are based on the maximum absolute difference between
the two cumulative distributions.
In Appendix G-4 the one sample test is applied to samples
133, 134, 104, and 128. Only the distribution of diameters from
sample 104 statistically fit an appropriate normal distribution.
The two sample test is also applied to the replicates. The dis-
tributions of diameters for samples 133 and 134 of condition 2 are
2-47
-------
not significantly different, but the distributions for samples 104
and 128 are different.
AGGLOMERATION
An important question is where do the agglomerates originate?
The available information does not directly answer this question.
One reason why this question is important is that the agglomeration
might be a peculiarity of only the generating equipment, the drum
and duct system, the sampling processes, or of some combination of
them.
Particle agglomeration is discussed in references 2, 7, 22
(22)
and 23. According to Green and Lane the general equation "for
the rate of coagulation (agglomeration) of an aerosol" is
-2
where
A
1 + = Cunningham correction factor
A = .9 for smokes
Ji = mean free path
Jl = 9 x 10~6 cm
r = particle radius = mean value of r, and r2 (cm)
r^ = radius of particle 1 (cm)
•^2 = radius of particle 2 (cm)
3
n = number of particles per cm
23
N = Avogadro's number = 6.0 x 10
R = gas constant = 8.3 x 10 for
2-48
-------
T = 293° K
T) - 1.82 x 10 poise
S * sphere of influence/particle jradius
(r,+r )2
Green indicates that %[— — - ], wMoh has a minimum of 1, has an average
12 r
value of 1.27 when the range of radii ~- in a uniformly polydiaperse aerosol is 1-8 .
Then ^=ffL(l+f)n2 (1.27) (2)
= K (l+^-)n2 (1.27) (3)
Green lists a table giving values of K for ferric oxide and magnesium
-9 3
oxide, among others, which are .66 and .83 x 10 cm /sec, respectively.
By integrating the above equation, it follows
^ - ^ = 1.27 K (1 + ^)t (4)
O _y
For the zinc oxide particle we can let r = 5x10 cm (5nm)
(1 -H-) -1+18-19
-9
and assume K = .7 x 10 . To estimate n , we can assume concentration^
o ^f O
1 g./m = 1 x 10 g/cm and
43
S.G. =5; C = nQ 5.|- nr g/cm
nQ = 3.83 x 10 particles/cm
3
Computing the time to reduce the initial number of particle n , to
-4
tj, = 1.55 x 10 sec
This implies that the agglomeration takes place rapidly and well
before the particles reach the sampling probe, an elapsed time
of about 0,3 seconds. The agglomeration apparently commences close
to the arc.
2-49
-------
The agglomeration rate is substantially dependent upon n and
particle size. If either variable were changed to give a combined
3
change of 10 (toward smaller concentration and larger size) the
t, would be altered to approximately the elapsed time.
%
Another factor in the agglomeration is that the particles
are initially charged. (2,9) Aggregates of charged particles
show pronounced tendency to be oriented in a chainlike fashion,
indicating polarization. Since particles in arc smoke are highly
charged, chain formation is not unexpected.
COMPARISONS WITH HEDDEN'S RESULTS
Although earlier particle generator work at W.V.U. by Redden*
involved iron oxide, it is nevertheless interesting to compare some
of his results to the zinc oxide results contained in this report.
The following table compares mass concentration and particle
size data.
Table 14. Hedden-Naylor Data
Hedden Naylor
(iron oxide) (zinc oxide)
Limit of microscope resolution (nm) 5 1
Width of diameter count interval (nm) 10 1
3
Mass concentration (gm/m ) .63 .67-2.05
Mean particle diameter (nm) 22 6.0-10.8
Range of diameters (nm) 5->100 1-50
Standard deviation (nm) 16.3 3.8-6.4
Cumulative Frequency Distribution log normal normal'
2-50
-------
Hedden apparently did not observe agglomerated particles.
This was presumably caused by his procedure of resuspending the
samples in solution and dispersing, rather than due to the fact
that the metal oxides examined were different.
2-51
-------
CONCLUSIONS
By a method described above, very fine zinc oxide particles were
generated by a commercial electric arc metallizer fed by two wires
and sampled using out-of-stack membrane and Nuclepore filters.
At four different combined settings of air pressure to the arc and
rate of wire consumption, mass concentrations were measured. At
three of these settings particle diameters by count were measured
by electron microscopy. The mass concentration means ranged from
3
0.67-2.05 g/sm (1.5-4.7 grain/scf). The mean particle diameters
ranged from 6.0-10.8 nm (0.006-0.0108 |im).
The generation of mass concentration was more precisely re-
produced at three of the conditions. The mass concentration was
sensitive to changes of the operating conditions.
Two samples from each of three conditions were observed by
electron microscopy. At the one condition for which mass concentra-
tion was not reproducible, the mean particle diameter was and is
not considered to be reproducible. At a second condition, the
particle diameters' mean and standard deviation were reproducible,
while at a third the mean, standard deviation and cumulative fre-
quency distribution were reproducible.
The mean particle diameters of the second and third condi-
tions were significantly different. The distribution of particle di-
ameters from each sample is apparently approximated by an appropriate
cumulative normal distribution. However, only one sample distri-
bution, from the third condition, has a good fit, statistically.
2-52
-------
Virtually all particles were agglomerated. The agglomeration
initiates immediately after particle generation at the electric arc.
Its fast rate results from a high initial number of particles per
volume and very fine particle size.
RECOMMENDATIONS
The Flamespray model VT500 electrospray metallizer has
demonstrated several important capabilities. For at least one
operating condition, it can generate very fine agglomerated zinc
oxide particles that are reproducible with respect to both the
distribution of particle diameters and mass concentration.
Therefore, this metallizer is considered worthy of continued
attention by researchers as a device for generating very fine par-
ticles. This attention would include, but not be limited to:
A. Improving the research tool.
1) Further development of equipment to improve
operational quality
a. Improve the wire feeding of the softer metals
to prevent interruptions of particle genera-
tion.
b. Determine requirements for maintaining con-
sistently stable operating conditions.
c. Reduce the ambient noise levels caused by
this metallizer.
2) Investigation of other metals and expansion
of generating and sampling program.
a. Utilize different wire types and diameters.
b. Determine the sensitivity to additional variables
such as exhaust flow and secondary air pressure.
2-53
-------
c. Redesign the sampling program to facilitate
an analysis of variance of mass concentration.
Instead of two different wirefeed rates and
MAP (main air pressure), these could be in-
creased to 6 different ones. However, retain
the present practice of collecting two repli-
cate samples for electron microscopy analysis.
3) Investigation of methods of sampling and measure-
ment compatible with preserving freshness of
particles. This would necessitate a definition
of freshness and a method for preserving the
sample during collection, during storage and
transportation, and during analysis by electron
microscopy, electron diffraction, x-ray diffrac-
tion, etc.
4) Review of present sampling and measurement methods
to determine if they reasonably utilize the best
available technology.
5) Further study of the agglomeration of very fine
particles.
a. Define appropriate dimensions for describing
the agglomerates.
b. Based on these dimensions, determine whether
or not the metallizer generates predictably
reproducible distributions of sizes.
c. If predictable reproducibility is established
determine if there is a significant correla-
tion between agglomerate size and such variables
.as chamber-duct retention time, number of very
fine particles per unit volume, and diameters
of the very fine particles.
d. Related to c, determine if there is a signifi-
cant correlation between very fine particle
diameter, agglomerate size and mass concen-
tration.
B. Using the research tool.
1) When a better understanding of this agglomeration
is reached, modify the experimental equipment
design to achieve predictable levels of exposure
to mass concentrations, particle diameters, and
agglomeration sizes found in various types of
2-54
-------
occupational and community atmospheres.
2) Modify the experimental equipment design to achieve
effectiveness of small particle control-equipment.
3) Investigation of the respiratory effects of fresh
particles.
a. What is the fate of agglomerated particles
collected at various locations in the lung?
b. What is the relationship between respiratory
effects and particle age?
2-55
-------
LIST OF REFERENCES
1. Wall Colmonoy Corporation, "Directions for Operating the Model
VT-500 Colmonoy Electrospray Metallizer," Detroit, Michigan, 1973.
2. Green, H. L. and Lane, W. R., Particulate Clouds; .Dusts, Smokes
and Mists, 2nd edition, Van Nostrand Company, Co., Princeton,
N. J., 1964.
3. Linsky, B., Smith, G., "Sources of Information on Air Pollution,
Proceedings National Conference on Air Pollution, Washington,
B.C., Nov., 1958.
4. Hedden, Robert, Electric Arc Generation of Polydispersed Iron
Oxide Aerosol in an Air Stream,Problem Submitted to WVU,
Morgantown, West Virginia, 1972.
5. Johansen, Arland, Unpublished Report, WVU, 1974.
6. Sears, D. R., "Fresh Metal Fumes Project -2," Laboratory Log
at WVU, 1975.
7. Brain, Valberg, Sorokin, "An Iron Oxide Aerosol Suitable for
Animal Exposures," Environmental Research, Vol. 7, No. 1, Feb.,
1974.
8. "Method 5 - Determination of Particulate Emissions from Sta-
tionary Sources," Federal Register. Vol. 36, No. 27, 12123, 1971
pp. 24888-24890.
9. Research Appliance Company, "Instructions for RAG Stack Sampler,"
Gibsonia, Pa., 1971.
10. Joy Mfg. Co., "Methods for Determination of Velocity . . .
Volume of Gases," Bulletin No. W.P. 50, 1970.
11. Davies, C. N., Air Filtration. Academic Press, N. Y., 1973.
12. Silverman, L., Billings, C., First, M., Particle Size Analysis
in Industrial Hygiene, ACGIH and USAEC, Academic Press, 1971.
13. Air Sampling Instruments Committee, Air Sampling Instruments for
Evaluation of Atmospheric Contaminants, ACGIH, 4th Edition, 1972.
14. Shimps, Richard, Walter McCrone Associates, series of Private
Communications, Jan. 1976-Mar. 1976.
15. Spurny, Lodge, ejt.al., "Aerosol Filtration By Nucleopore Filters,"
Environmental Science and Technology. Vol. 3, 1969, pp. 453-64.
2-56
-------
16. CRC, Handbook of Chemistry and Physics, Van Nostrand, 1975.
17. "Electron Diffraction," Encyclopedia of Science and Technology.
McGraw Hill, 1974.
18. Miller, Richard, Platt Industries, Private Communication,
January 23, 1976.
19. Miller, I., Freund, J., Probability and Statistics. 12th
Printing, Prentice Hall, Englewood Cliffs, N.J., 1965.
20. Kim, P. J. and Jennrich, R. I., "Tables of the Exact Sailing
Distributions of the Two Sample Kolmogorov-Smirnov Criterion,"
Selected Tables in Mathematical Statistics, Barter and Owen, ed.,
Vol. 1, Markan Pub. Co., Chicago, 1970.
21. Mercer, Thomas, Aerosol Technology in Hazard Evaluation, ACGIH,
Academic Press, N. Y., 1973.
22. Herdan, G., Small Particle Statistics, 2nd Edition, Academic
Press, N. Y., 1960.
23. Hidy and Brock, ed., Topics in Current Aerosol Research,
Part 2, Vol. 3, 1st edition, Pergamon Press, N. Y., 1972.
24. Danielson, John A., Air Pollution Engineering Manual. U. S.
Department of H.E.W., No. 999-AP-40, Cincinnati, Ohio, 1967.
25. Damewood, E. Z., West Virginia University Industrial Engineering,
Private Communication, April 16, 1976.
2-57
-------
APPENDIX A
COLLECTION EFFICIENCY OF THE BARREL
An empirical formula by Lapple^ ' p^ ' is "sufficiently
accurate for an engineering estimation of many cyclone applications."
He defines D as the diameter of those particles collected with 50
percent efficiency. After determining D by the formula following,
the collection efficiency of particles D can be determined by a
graph showing cyclone efficiency versus D /D
p pc
J) ;s I i i ii mi rfA»-»n»-i«^ I
where p, = gas viscosity = (183x10 x.672x10~3)lb mass/ sec ft.
b = cyclone inlet width = 0.08 ft
N = effective turns = % (For cyclones may vary from %-10)
e
V. = gas inlet velocity = 44 fps
p = particle density = 5.6(s.g.) x62.4
= 349 lb/ft3
b, V., are estimated as spread and velocity of
compressed air stream reaching arc
fe x (183 x 10"4)(.672 x 10"3) (.08))^ , . .Q-5 ,
= V 2(%)44(349)n 7"
V
= 7.31 x 104 cm = 7.3 |j,m
50% by weight of the 7.3 urn particles are collected.
For 95% efficiency D /D c = 4 , according to Figure 48, pg. 95,
re£. 24.
D = 4 x 7.3 tun = 29 JimJSr 30
P
2-A-l
-------
APPENDIX B
COMPUTATION OF STACK GAS VELOCITY.
VELOCITY TRAVERSES, AND STACK FLOW
According to reference 10, the velocity at a stream line
in a flue V , can be determined from the following
s
Using standard pitot tube
V (fps) = 3.90 ^•^•xi^xHT
s
P = absolute pressure in flue in inches mercury
s
Gd = specific gravity of gas referred to that of air
H = velocity head in inches of water = AH
T = flue gas temperature in degrees Kelvin assuming Gd - 1
s
= 3.90 (^~- £HTgJ *fps - 234
2-B-l
-------
APPENDIX B
Table B-l. Velocity Traverses for Different
Venturi Pressure Drops (6)
Venturi AP .20 inches water
% of duct
diameter
6.2%
25
75
93.8%
6.2%
25
75
93.8%
Pitot AP
in water
.09
.12
.12
.09
•10
.12
.12
.08
v
fpm
1201
1387
1387
1201
1266
1387
1387
1133
Pitot AP
in water
.10
.12
.12
.09
.11
.12
.11
.08
v
fpm
1266
1387
1387
1201
1328
1387
1328
1133
(1) v = 1298 fpm 2
for 6" duct Area » .197 ft
(2) Q = 256 cfm
Venturi AP =
% of duct
diameter
6.2%
25
75
93.8%
6.2%
25
75
93.8%
Venturi AP =
% of duct
diameter
6.2%
25
75
93.8%
.31 inches
Pitot AP
in water
.11
.17
.18
.14
.15
.18
.18
.12
v = 1596
.40 inches
Pitot AP
in water
.14
.20
.24
.18
water
v
fpm
1328
1651
1699
1498
1551
1699
1699
1387
fpm Q = 314
water
v
fpm
1498
1791
1962
1699
Pitot AP
in water
.17
.18
.19
.14
.17
.19
.17
.12
cfm
Pitot AP
in water
.17
.23
.24
.18
v
1651
1699 vertical
1746 traverse
1498
1651
1746 horizontal
1651 traverse
1387
v
fpm
1651
1921 vertical
1962 traverse
1699
(continued)
2-B-2
-------
Venturi ftp = .40 inches water (continued)
% of duct
diameter
6.2%
25
75
93.8%
Pitot AP
in water
.22
.24
.22
.15
V
fgia
1879
1962
1879
1551
Pitot AP
in water
.23
.24
.21
.15
V
fpm
MBlriMMIIBH
1921
1962
1835
1551
horizontal
traverse
v = 1795 Q - 354 cfm
(1) Velocities are calculated by using "standard air tables" and
assuming standard conditions of 760 mm and 70 F. This intro-
duces only a small error.
(2) Flows are calculated by the formula
Q = v • A where A = n(3/12)2 = .197 ft2
The values of Venturi AP and Q are plotted in Figure 5.
2-B-3
-------
APPENDIX C
EFFICIENCY OF NUCLEOPORE FILTERS
Definitions of Individual Efficiencies:
1) Partial efficiency of impaction
Ei
where
E'± = 2 STK/T + 2
e =
P = filter porosity - .094 (ref. 21)
STK ( Stokes number) = m^/(6T|rRo)
m = mass of single particle (gm)
r - radius of particle (cm)
44
Assume specific gravity - 1 (conservative)
m = 4/3 nr3
2-C-l
-------
Table C-l. E. for Values of Particle Diameter
STK
D
urn
.1
.15
.2
.3
r
-4
xlO cm
.05
.075
.10
.15
m
gm
1.25x10
.42x10
1.00x10
1.41x10
-16
-15
-15
-14
6.74x10
1.51x10
2.7 xlO
2.43x10
-3
-2
-2
-1
.0089
.020
.0358
.322
.012
.028
.0497
.447
2) Partial efficiency of diffusion
ED = 1 - .81904 exp [-3.6568 ND
where
N -
-
L = filter thickness (cm) = 12 x 10 cm
2
D = aerosol particle diffusivity (cm /sec)
P = filter porosity = 0.094
R = pore radius =0.1x 10
= 3 cm/sec
cm
Table C-2
E
for Values of Particle Diameters
Diam.
urn
.02
.04
.10
.15
.20
.30
D
(ref . 11)
-4
1.4 x 10
3.6 x 10 "5
-6
6.8 x 10
-6
3.4 x 10 °
-6
2.2 x 10
-'4
1.24x 10
ND
52.6
13.7
2.58
1.29
.84
.47
ED
1.00
1.00
1.00
.993
.962
.853
2-C-2
-------
3) Partial efficiency of Interception
ER =
NR = r/RQ if NR> 1 efficiency is 1
r = particle radius
-4
R =0.1 x 10 cm
o
Table C-3. E for Values of Particle Diameter
E
D
urn
.04
.10
.20
.21
4.
10 "^cm
.02
.05
.10
.105
NR
.2
.5
1.00
1.05
ER
.36
.75
1.00
1.00
.15ER
.054
.11
.15
1.00
The partial efficiencies are tabulated and combined in Table 1.
2-C-3
-------
APPENDIX D
COMPUTATIONS FOR K IN TABLE 2
D C
P
1 x 10"4 1 + 1.62 + .0000677 = 1.16
1 x 10~6 1 + 16,2 + 5.14 = 22.34
2-D-l
-------
APPENDIX E
PROCEDURE FOLLOWED BY McCRONE ASSOCIATES
FOR ELECTRON MICROSCOPY AKD ELECTRON DIFFRACTION
Shimps outlines analytical procedures practiced by McCrone
Associates in the following excerpted letter. (14)
"The sampling procedure is our standard one for
both the SEM and TEM examinations.
For the TEM, a square of filter 3 mm on a side
is cut out and placed sample side down on a TEM grid
with a. carbon substrate. This preparation is then
placed on a stainless steel mesh support on a cold
finger in a condensation washer where the filter
dissolves and thus deposits the sample material on
the carbon substrate.
For the SEM, a square of filter is cut and placed
on adhesive on an SEM stub. First carbon and then
gold is vacuum evaporated on the specimen so that it
may be examined in the SEM."
2-E-l
-------
APPENDIX F
MASS CONCENTRATION DATA
The following -1 tables represent a condensed tabulation of
most of the data observations for each sample during the sampling.
Each -1 table is followed by a -2 table which includes net sample
weight, mass concentration at actual conditions (gm/acf) and mass
concentrations at standard conditions (gm/scf ) .
Although moisture in the air stream was monitored, sampled
volumes were not corrected for it as the correction was small. The
following calculations give conservative estimates of the corrections.
a. given: T = 30°C
Rel Humid - 25%
p. = 28.4 in.
bar
Vapor Pressure = 31.824 mm= 1.25 in (ref. 16)
V P x(R.H )
Percent Moisture in air '•—'• -v-•'—1Z = R = 1.10%
bar
b. given: Same as a) except Rel. Humid - 100%
», ., <. 1.25 x 1007, . ,„
Percent Moisture = 2874 = 4<4'°
Condition a was probably not exceeded during the sampling. The
equation for the isovelocity factor I is
I =
ATime VSTACK
STACK
( IX ) x ( C )
2-F-l
-------
where P__ = 29.92 inches
O ±.U
- 298°K
V - standardized meter dry volume
V - volume of water vapor
WSTD
VSTACK = STACK ga
Anozzle - •°00085
P, = Atm P at dry gas meter
T = average dry gas meter temperature
mAVG
R = V /V =0
WSTD "^TD
I1 = approximate I
C = Correction factor
Tables -1 also provide the I' and C numbers. In tables -2, actual
concentrations are standardized by multiplying by 1/SC where
ACF x x . = ^ x sc
STD meter
2-F-2
-------
Table F-1A. Zinc Sampling Data as Observed
2
3
4
5
Pitot AP = .12
Sample #
Date
Pbar
Chanib T
HEPA AP
PSTACK=©-@/13.6
Venturi AP
STACK T(°C)
Wirefeed
Amps
MAP
AH or if
ACF
A Time(min.)
acfm
T (°F)
mavev '
Nozzle vel.
Stack vel.
l' - @/
-------
Table F-1B. Zinc Sampling Data as Observed
Pitot AP = .12 SAP = 20 psi
Sample #
Date
Pbar
Chamb T
HEPA AP
PSTACK in'
Vent. AP
Stack T(°C)
Wirefeed
Amps
MAP
AH orif
ACF
A Time (rain. )
acfm
mave
Nozzle vel.
Stack vel.
l'-
Rel. Humid
C
33
2/14
29.51
20°
1.09
29.43
.20
20
30
30
50
.06
.816
6.0
.136
75
1600
1401
1.15
20%*
.99
34
2/14
29.51
20°
1.09
29.43
.19
23
30
30
70
.058
.779
6.0
.129
77
1527
1409
1.09
20%*
.99
35
2/14
29.51
28
1.09
29.43
.21
25
50
70
70
.055
.476
4.0
.119
79
1400
1416
.99
20%*
1.00
36
2/14
29.51
28
1.09
29.43
,21
26
50
70
68
.055
.487
4.0
.122
78
1432
1418
1.01
20%*
1.00
37
2/18
28.54
31
1.0
38.47
.20
28
50
70
50
.055
.395
3.0
.132
66
1549
1439
1.07
20%*
1.03
38
2/18
28.54
29
.8
28.49
.20
24
50
70
50
.052
.539
4.0
.135
64
1585
1431
1.11
20%*
1.00
Table E-2B. Standardized Mass Concentration Data
I = l' x C
Net wt. gain
gm/acf
1/sc
gm/scf
1.14
.0143
.0175
1.013
.0177
1.08
.0218
.0280
1.013
.0284
.99
.0268
.0563
1.018
.0573
1.01
.0305
.0626
1.018
.0637
1.10
.0113
.0286
1.025
.0293
1.11
.0163
.0302
1.025
.0310
2-F-4
-------
Table F-1C. Zinc Sampling Data as Observed
Pitot AP = .12 SAP = 20 psi
Sample #
Date
Pbar
Chamb T
HEPA AP
PSTACK
Vent AP
Stack T
Wirefeed
Amps
MAP
AH orif
ACF
A Time
acfm
T
mav
Nozzle vel
Stack Vel.
I'
Rel. Humid
C
Table E-2C.
40
2/18
28.54
30
.8
28.48
.20
25
50
70
50
.05
.478
4
.120
63.5
1406
1434
.98
20%
1.01
Standardized Mass
I = I'x C .99
Net weight
6) gm/acf
7)> 1/sc
8) gm/scf
gain .0162
.0339
1.025
.0347
44
2/18
28.54
23.5
.8
28.48
.20
22
30
30
50
.05
.471
4.11
.115
63
1348
1424
1.06
20%
1.00
45
2/25
29.11
27
.70
29.06
.20
28
30
25
70
.04
,766
6.0
.128
79
1502
1427
1.05
25%
1.01
46
2/25
29.06
30
.70
29.06
.215
28
30
25
70
.355
2.79
.127
80
1497
1427
1.05
25%
1.00
Concentration Data
1.06
.0082
.0174
1.025
.0178
1.06
.0260
.0339
1.0056
.0341
1.05
.0081
.0228
1.0056
.0229
2-F-5
-------
APPENDIX G
C-l. HYPOTHESIS TESTS OF THE MEANS
OF THE MASS CONCENTRATIONS
The Smith Satterthwaite test requires no assumption in
equality of variance, which makes it a powerful test. It does
assume that the observed values in a given sample are normally
distributed, however.
Generally:
0* ^1 = ^"2
Reject HQ if t' > ta
where ^- = true but unknown condition #1 population of mass concentrations
HQ = Null hypothesis
H, = First alternative hypothesis
X, - Xo
t'
2 S
nl °2
vi = degrees of freedom = « •> o •?
(S2/n )Z (S^ In2r
t = value in tables for t as function of v and a
a
a = significance level = probability of rejecting the
null hypothesis when it is true.
X, = observed mean of sample
standard deviation from
number of observations in sample #1
S = standard deviation from sample #1
2-G-l
-------
Specifically, for X;L and KZ (see Table 9A)
Ho:
H^ p^ > |j,2
a- = 0.05
t* =3.2 v = 2.2
* * 2"8
(Table IV, ref . 19)
**>*.
H is rejected. Therefore we conclude jj,- >
Considering X, and X_
Ho:
a = 0.05
t* - 7.43 v = 6.94
t - 1.9
o;
t' > t
at
H is rjected. Therefore we conclude p,_ > (j,
Continuing with X2 and X,
HQ:
01 = 0.05
t' = 18.4 v = 3.8
t - 2.15
HQ is rejected. Therefore we conclude |ju >
2-G-2
-------
G-(2. HYPOTHESIS TESTS OF THE MEANS AND VARIANCES
OF REPLICATE PARTICLE SIZE DATA
(19)
This methodv ' employs a simpler hypothesis procedure than
described in G-l because the sample sizes are well over 30, thus
eliminating the need for a degrees of freedom calculation. The
first criterion for replicate reproducibility is equality of the
true mean diameters.
Generally HQ: jj^ = |j,2 (DI > D2)
^1" ^1 ^ ^2
|j,1 = true but unknown diameter of replicate 1 >
population
D- = mean of replicate 1 (best estimate of (j,^)
* 1 9
Z =
Reject H if Z* >
For condition 2 (see Table 11);
D1 = 10.8 Sx2 - 35.59 nL = 350
D2 = 9.1 S22 = 41.14 n2 = 280
z* = -
/35.6 . 41.1 )
» /
350 280
For a - 0,05 Z = 1.96
Hn cannot be rejected. Therefore we conclude ^ =
2-G-3
-------
For condition 3:
Dl =
D2 =
*
Z =
*
Z >
10
10
0.
z
.8
.4
92
In
Sl
2
S2
Z /o
a 12
= 28.
= 21.
= 1.
14
96
96
nx = 376
n2 = 281
a = 0.05
Hn cannot be rejected. Therefore we conclude (j,, = M-O-
A second criterion for replicate reproducibility is equality
of the true variances. Generally
2 2
H0:
2 2
j4 a
Reject HQ if F > F^/2 (n^l,
2 2
where F = S- /S2
n = sample size
Applying this to conditions with means that are not signifi-
cantly different, for condition 2
Sj2 = 41.14 nx-l - 279
S22 - 35.59 n^J = 349
v - 41.14 _ i 16
F " 35.59 ~ 1'16
for a =0.02 F ._(279, 349) - 1.32
a/2
2 2
F < Py/2 and H~ cannot be rejected. Therefore we conclude cr, =
-------
G-3. HYPOTHESIS TESTS OF THE MEAN
DIAMETERS OF CONDITIONS 2 AND 3
Generally, the analysis is similar to the ones for replicate
reproducibility. The means and variances for each condition are
first averaged.
For condition 3
T *. n 376(10.8). + 281(10.4)
Let Dl 376 + 281
DX = 10.6
2 376(28.14) + 281(21.96)
1 = 376 + 281
S-j2 = 25.50
Sl = 5.05
For condition 2
n - 9.13(280) + 9.99(350) _ g 6
U2 ~ 280 + 350
<, 2 280(41.14) + 350(35.59) _ ,«
S2 - - * - 280 + 350 - 38'
S2 = 6.17
HQ: nx =
DD
* 2 _ i
- 0 32
» _ . _ 3 17
Z =— - - ~
For a = 0.05 Za/2 = 1-96
Z* > Z ,2 and HQ is rejected . Therefore we conclude ^ #
The mean sizes of particles from conditions 2 and 3 are therefore
considered to be significantly different.
2-G-5
-------
G-4. KQMOGOBOV - SMIRNOV FITNESS TESTING
Applying the one sample test to the samples in condition 2
(Table 12)
D (sample 133) =0.103
max r
for a = 0.05 Da = 1.36//n n = sample size = 350 D = 0.073
a
D > Dry Sample 133 does not fit the given distribution
max -« r
at 5% level of significance
D (sample 134) = 0.140
for (X = 0.05 n = 280 % = 0-081
D > Dry- Sample 134 does not fit the given distribution
max
at the 5% level of significance.
The two sample test is also applied at a - 0.05.
D =0.12 reference (20)
D = (0.37-0.23) = 0.14
max x
D > D . Therefore Samples 133 and 134 have significantly
max ot
different distributions .
Applying the one sample test to the condition 3 samples
(Table 13)
D =0.0688 n - 376 a =0.05 D =0.070
raax!04
D < D
max a
Sample 104 fits the normal distribution
D =0.0844 n = 281 D =0.081 a =0.05
max128 a
> D ; Sample 128 does not fit the distribution.
2-G-6
-------
Applying the two sample test to condition 3 samples
D^ =0.12 at a =0.05 (19)
Dmax = (°-83
Dmax < Da» SamPles 104 and 128 do not have significantly
different distributions .
G-5. SUMMARY OF STATISTICAL TESTS
The above statistical tests, applied to mass and size count
data are summarized.
Mass Concentration-Comparing Conditions
M-! > M-2 M-3 > 1*2 ^3 > ^1
Particle Sizing
Replicate Reproducibility
Condition 1 Condition 2 Condition 3
a!33 = CT134 a!04 =
Differences between conditions
fi2 * M-3
Frequency Distribution
a) Closeness of fit to normal distribution
133 and 134 do not fit
104 fits but 128 does not
b) Closeness of fit to each other - Reproducibility
133 and 134 are different
104 and 128 fit
2-G-7
-------
APPENDIX H
ELECTRON MICROSCOPY PHOTOGRAPHS
The following photographs were selected by the investigator
as typical of all the photographs of the 6 samples analyzed by
McCrone Associates.
The reproductions are not altered with regard to magnifica-
tion. The prints are reproduced from negatives made by Copycat,
Inc., Morgantown, using the McCrone photographs.
Figures H-l and H-2 indicate the light colored chain like
agglomerates against the porous Nucleopore filter background. The
"dark holes" represent the 0.2 |j,m pores.
Figure H-3 represents the composition of part of an agglomerate.
The particles here are dark and are distinguished by light borders.
2-H-l
-------
Figure H-l. SEM Photograph of Sample 133
(5000X). 1 cm. = 2 urn.
Figure H-2. SEM Photograph of Sample 133
(10000X) . 1 cm. - 1 \im.
2-H-2
-------
Figure H-3. TEM Photograph of Sample 101
(300,OOOX)
2-H-3
-------
ELECTRIC ARC GENERATION
OF
FRESH IRON OXIDE AEROSOLS
THESIS
Submitted to the Graduate School
of
West Virginia University
In Partial Fulfillment of the Requirements For
The Degree of Master of Science in Engineering
by
William Fred Dimmick II, A.B.
Morgantown,
West Virginia
1977
3-i
-------
ABSTRACT
ELECTRIC ARC GENERATION OF FRESH IRON OXIDE AEROSOLS
The development of a fresh metal oxide fine particle aerosol
generator was needed to provide necessary true-to-life conditions for
testing air pollution control and measurement equipment. A generator
of fresh aerosols was considered to be applicable to toxicological
studies. An electric arc metallizer employing dual consumable electrode
wires was developed for fresh metal oxide aerosol generation with the
guidance of Professor Benjamin Linsky. The present engineering problem
consisted of characterizing the aerosols produced by the generator when
metallizing steel feed stock and investigating the influence of operation
variables on the conversion of mass from solid wire to aerosol particles.
Generation of fine particle aerosols was done using a commercially
available electric arc metallizer to spray melted and vaporized feed stock
electrode wires into a 55 gallon barrel. The spray generated aerosol
was exhausted through a duct work.
The generated aerosols were sampled on membrane filters for sub-
sequent particle size count analysis using scanning electron microscopy
and transmission electron microscopy. The same samples were analyzed
for chemical composition by selected area electron diffraction.
Membrane filters were also used to capture mass samples for deter-
mining the mass concentration of the aerosol in the exhausting duct works.
The mass concentration was used as the dependent variable in a Latin
Square designed experiment where operation adjustments of the equipment
as manufactured were the independent variables. These independent var-
iables were the wire feed rate, the open voltage across the arc, and the
3-ii
-------
pressure of the compressed air jet used to form the spray.
The data shows the aerosols to be composed of agglomerated particles
formed from very fine magnetite particles. The agglomerated particles
had a mean size of 0.4 ym by count whereas the very fine particles
have mean count diameters ranging from 0.0054-0.009 ym (5.4-9nm). At
three different operating conditions three different very fine particle
means were found. Each of these conditions produced an unique particle
size distribution.
The mass concentration Latin Square experimental design found the
wire feed rate and the open voltage to influence changes in the mass
concentration of the aerosol. A maximum mass concentration was found to
3
be approximately 2 gm/m . The variability produced by replication of a
data set was found to be significant.
In conclusion, the generator was found to produce very fine ferrous
particles that agglomerated into submicron sized particles. Evidence
suggests that, with further characterization of the agglomerated very
fine particle aerosols, the generator will produce true-to-life fresh
aerosols with reproducible properties. The aerosol mass concentration
was considered to be adequate for testing of air pollution control and
measurement equipment.
3-iii
-------
ACKNOWLEDGMENT
The author wishes to graciously thank everyone involved in
the completion of the research. His major professor and advisor,
Professor Benjamin Linaky, introduced him to the field and concepts of
fine particles. His colleague Mike Naylor taught him the set up and
operation of the aerosol generating apparatus. Don Stone and Craig
Repp, fellow graduate students, helped in the collection of samples.
Craig Repp and the author designed and built a metallizing wire dispenser
to deliver feedstock from 50 pound wire coils to the generating apparatus.
The author expresses gratitude for helpful suggestions from
Mr. Charles Ghastin of the Wall Colmonoy Corporation and Mr. Richard
Shimps of McCrone Associates. Suggestions and questions from Dr. S. R.
Borbash of the Industrial Engineering Department at West Virginia
University guided the researcher's statistical analysis.
The author was a graduate air pollution trainee under Environmental
Protection Agency grant numbers T900567010 and T900567020. The financial
support for this research was funded as part of Environmental Protection
Agency grant R-801858-01-1.
The author thanks his wife, Linda, for her patience during the
time of his presently completed studies.
3-iv
-------
TABLE OF CONTENTS
Page
ABSTRACT 3_i]L
ACKNOWLEDGEMENT 3_iv
LIST OF TABLES >vll
LIST OP FIGURES 3_ix
LIST OF SYMBOLS, ABBREVIATIONS, AND CONSTANTS .... 3_x
INTRODUCTION 3-1
STATEMENT OF ENGINEERING OBJECTIVES 3-2
REVIEW OF PERTINENT LITERATURE 3-3
PROCEDURE 3-6
Equipment and Materials 3-6
Operation of Generator Apparatus 3-10
Sampling 3-11
Phenomenon Characterization Techniques 3-11
Experiment: Analysis of Phenomenon 3-13
Data 3-16
ANALYSIS OF DATA 3-30
Agglomerates 3-30
Fine Particles 3-33
Mass 3-46
DISCUSSION OF RESULTS 3-51
Agglomerates 3-51
Fine Particles 3-51
Chemical Composition 3-53
Mass Analysis 3-54
CONCLUSIONS 3-57
BIBLIOGRAPHY 3-60
APPENDICES 3-62
Appendix A 3-63
Appendix B • 3-64
Appendix C-l Particle Sizing Pictures 3-65
Appendix C-2 Sizing of the Agglomerates . . . .3-66
3-v
-------
TABLE OF CONTENTS (continued)
Appendix C-3
Appendix D-l
Appendix D-2
Appendix D-3
Appendix D-4
Appendix D-5
Mass Concentration Data
Latin Square ....
T Test Procedure
ANOVA
.... 3—68
.... 3-81
. . . . 3-84
. . . . 3-88
Stepwise Procedure 3-89
GLM and Residual Plots 3-92
Appendix E 3-100
3-vi
-------
LIST OF TABLES
Page
1. Hobart steel metallizing wire (HB 18) Assay . . . .3-10
2. Aerosol generator operating conditions
for particle size analysis 3-14
3. Comparison study operating conditions 3-14
4. Determination of orthogonal data
collection points 3-15
5. Agglomerated data corrected for magnitude
of magnification: sample 715 3-17
6. Agglomerated data corrected for magnitude
of magnification: sample 714 3-17
7. Very fine particle distributions: set 1 3-18
8. Very fine particle distributions: set 2 3-20
9. Very fine particle distributions: set 3 3-22
10. Very fine particle distributions: set 4 3-23
11. Very fine particle distributions:
samples 712 and 716 3-24
12. Selected area electron diffraction data ..... 3-25
13. Diffraction lines for magnetite
and maghemite 3-25
14. Latin square mass concentration
data summary (gm/m3) 3-26
15. Orthogonal mass concentration
data summary (gm/nr) 3-27
16. A listing of all the mass concentration data . . • 3-28
17. Two means equality test hypothesis
decision-making criteria 3-31
18. Statistics for agglomerate samples 3-33
3-vii
-------
LIST OF TABLES (continued)
19. Statistics for fine particle distributions . . . 3-40
20. Within group t-statistics for
fine particle samples 3-41
21. Between group t-statistics for
fine particle samples 3-42
22. Within group chi-square statistics 3-44
23. Between group chi-square statistics 3-45
24. Analysis of variance for latin square
experimental design 3-47
25. SAS ANOVA for latin square experimental
design data 3-48
26. SAS GLM table for orthogonal data 3-49
27. SAS GLM table for all of the data 3-50
28. Raw data for agglomerated particle count .... 3-67
29. Mass data for first latin square experiment . . . 3-71
30. Mass data for second latin square experiment . . . 3-75
31. Mass data for orthogonal experiment 3-79
32. Formula and definitions for latin square
analysis of variance 3-82
33. Calculation table for latin square design .... 3-83
34. SAS T Test table 3-87
35. Stepwise regression procedure for
dependent variable Y 3-90
36. SAS GLM table including residuals 3-93
37. Relating wire feed rate and percent
mass converted 3-100
3-viii
-------
LIST OF FIGURES
Page
1. Equipment set up with close up of
arc gun assembly 3-8
2. Example of data sheet 3-12
3. Set 1 frequency distributions 3-35
4. Set 2 frequency distributions 3-36
5. Set 3 frequency distributions 3-37
6. Set 4 frequency distributions 3-38
7. Sample 712 and 716 frequency distributions . . . 3-39
8. Histograms of agglomerated particles 3-52
9. SEM picture sample 715 3-65
10. SEM picture sample 714 3-65
11. Explanation of data sheet and mass
calculation symbols 3-70
12. Graph: wire feed rate versus standard
residual 3-96
13. Graph: open voltage versus standard
residual 3-97
14. Graph: predicted outcome (Yhat)
versus standard residual 3-98
15. Graph: order of presentation versus
standard residual 3-99
16. Wire feed rate calibration 3-101
3-ix
-------
LIST OF SYMBOLS, ABBREVIATIONS AND CONSTANTS
ANOVA
GLM
T Test
SS
df
Yhat
SEM
TEM
SAED
lines
d A
HEPA
R#
Pbar
P
duct
HEPA A P
Venturi ^ P
Pitotflp
Duct T
Bag T
V
open
AMPS
= analysis of variance
• general linear model
= a statistical test based on the t-distribution
= sum of squares
= degrees of freedom
= predicted outcome from an estimated descriptive
equation
= scanning electron microscopy
= transmission electron microscopy
= selected area electron diffraction
* diffraction lines in angstroms (A)
= high efficiency particulate air
= reynolds number
« ambient barometric pressure
= pressure in exhaust duct
= pressure drop across HEPA filter
= pressure drop across calibrated venturi
= velocity pressure differential as measured by a
standard pitot tube
= temperature in exhaust duct
= temperature within bag house
= open voltage across electrodes
= electrical current in amperes
3-x
-------
LIST OF SYMBOLS, ABBREVIATIONS AND CONSTANTS (continued)
MAP
SAP
AE orifice
CF start
CF end
4CF
dT
T
meter-start
T
meter-end
duct
STP
3 3
gm/m = gm/sm
acf
fpm
8
P
P
d
main air jet pressure at machine
secondary air supply pressure
manometer differential on RAC sampler
dry gas meter reading before sampling
dry gas meter reading after sampling
sample volume
time of sampling in minutes
temperature of RAC thermometer at beginning of
sampling
temperature of RAC thermometer at end of sampling
velocity of gas stream in exhaust duct
standard temperature (273 °K) and pressure
(1 atmosphere)
grams per standard (STP) cubic meters
grams per actual cubic feet
feet per minute
acceleration due to gravity
density
dynamic viscosity
circular diameter
3-xi
-------
LIST OF SYMBOLS, ABBREVIATIONS, AND CONSTANTS (continued)
lA = 10 10m - 10 8cm
grains/ft3 x 2.288
(*F-32) x 5/9
ft3/mlnute x 0.0283
diameter
-7 -4 -1
10 mm =10 ym = 10 nm
= grams/m
-*C
3
= m /minute
= a physical quantity
3-xii
-------
INTRODUCTION
The generation of fresh dry metal aerosols for experimental
analysis, whether it is testing fabric filters for particle penetration
or examining the effect of inhaled cadium oxide particles on hamster
kidneys, is a necessity for proper similitude to many industrial events.
The known discrepancy between the occurrence of brass fever after ex-
posure to "fresh" zinc oxide fumes and the lack of occurrence with
exposure to stale, redispersed zinc oxide fumes (14) expresses the
validity of discriminating between fresh and stale particles. Again,
the fact that the emission of fine particles as air pollutants has re-
ceived increasing attention in the past nineteen years (7) implies the
concurrent need of fine particle aerosol generators for evaluation of
engineering designs. Eugene Grassel (_4_) states recent scrutiny of fine
particles in the atmosphere is producing an aerosol generator market.
A characteristic of fine particulate aerosols is the small rate
at which the particles settle due to the force of gravity. In Appendix A,
it is shown that a 0.5 ym diameter solid spherical particle with a unit
density of one gram per cubic centimeter falls less than one-half cen-
timeter in one hour (approximately four and a half inches per day).
This property virtually eliminates the possibility of fine particles
settling quickly by gravity to earth's surface. Presumably, the par-
ticles are blown about the atmosphere by winds.
Fine particles can absorb and scatter light. Aerosols with par-
ticles in the size range of visible light extinguish light (8).
Waggoner and Charlson (18) state that blue hazes in mountainous areas
may or may not be due to scattering by particles. There should be no
3-1
-------
haze unless there are particles between the observer and the mountain
If the mountain is closer than 10 km (7 miles). They also say that
light scattering increases with humidity, but for relatively hygrophobic
systems the increase may be very slight. Certainly extinction increases
with an increase in the light extinguishing particle number in the
atmosphere.
As shown in Appendix B, a unit density spherical particle of diam-
eter 0.1 ym weighs one thousand times greater than a unit density spher-
ical particle with a 0.01 ym diameter. Any air pollution emission cri-
terion that relies only on measurement by weight is not including all
the proper physically descriptive information. The importance of char-
acterizing an aerosol by its particle size distribution has been known
in industrial hygiene for decades. Measurement of a fine particulate
aerosol without considering the size distribution is incomplete.
STATEMENT OF ENGINEERING OBJECTIVE
Further research with the Electrospray aerosol generator (as de-
scribed in Equipment and Materials) was needed to provide more infor-
mation leading to a better engineering characterization for the uses
of the equipment and its output. A statistical-design approach was
incorporated with the object of furthering the knowledge of the var-
iables of solid feed material to particle mass concentration conversion.
These variables as represented in this research were (1) the open
voltage across the arc, (2) the wire feed rate, and (3) the pressure
of main gas jet which impinges upon the arc to form the metallizing
spray. All of these variables were obtainable from adjustment mech-
anisms routinely supplied as part of the metallizer. The dependent
3-2
-------
variable (the outcome) was the aerosol mass concentration.
Further characterization of the aerosol produced by the Electro-
spray aerosol generator was also an objective. Naylor (10) reported
on the results of Electrospray production of zinc oxide particle aerosols
from zinc consumable electrodes. In the current research, carbon steel
wire was used to produce iron oxide aerosols for fine particle size
distribution analysis. Both the aerosol characterizing analysis and
the mass conversion analysis are analyses in the sense that they examine
a system by looking at some of its elements and their relations. A
simple comparative experiment also observed the effect on aerosol char-
acteristics of faster gas and particle removal from the chamber (barrel)
where the gas and the particles received direct line-of-sight radiation
from the arc.
REVIEW OF PERTINENT LITERATURE
Holmgren et_.al_. (6) reported a technique for production of fine
particles in a high intensity arc. The feed material was incorporated
in the anode and was vaporized as a consequence of the extreme temper-
atures produced by the arcing. They explained that due to a sudden
drop in temperature of the vapor, as it was carried away from the arc,
fine particles were formed by condensation.
Particles that ranged from 100 A to 1000 A (as measured by elec-
tron microscopy) were produced. Mass production rates of 10-20 Ib/hr
(76-152 gm/min) under a practical running condition were reported for
generation of oxide-type materials.
Holmgren et.al. reported the production of Fe203 particles with
an 'equivalent sphere diameter1 of 658 A (65.8 nm, 0.0658 ym). They
3-3
-------
described the aerosol particles us roughly spheroidal with a distinct
trend toward a hexagonal outline. They stated that dispersion about
the mean varied with the physical properties of the feed material.
Pfender and Boffa (12) described a method of generating fine
particle aerosols in a high intensity arc. Their apparatus included
transpiration cooling of the anode to facilitate the production of
smaller particles. They stated that with low intensity arcs coagulated
and chainlike particles with poorly controlled size distributions are
produced. They referenced Holmgren .et.al^ to state that better re-
sults are obtained with high intensity arcs.
Pfender and Boffa (12) reported generation of monodisperse
spheroidal shaped particles with a diameter ranging from approximately
100 to 1000 A as measured by a Whitby analyzer. They did not report
the criterion used to determine monodispersity nor the mass production
rate.
Formenti et^.a.1^. (3) reported the production of fine metal oxide
particle aerosol from diffusion burning of volatile metallic chlorides
in a hydrogen oxygen flame. They reported generating particles with
a range of diameters pf 100-2000 A. They stated that the diameter
of the particles and their morphology are controlled by (1) the tem-
perature of the flame, (2) the concentration of chloride injected into
the burner, and (3) the residence time of the chloride vapor in the
flame. They reported production of Fe~0. particles in a mixture of alpha
and gamma phases. They did not report aerosol size distribution dis-
persity or mass production rate.
Grassell (4) discussed the development of aerosol generators for
3-4
-------
Industrial research with respect to fine particles aerosols. Two basic
criteria were used to evaluate the propriety of an aerosol for filter
medium testing. These criteria were (1) sufficient quantity of the
aerosol for testing a 1000 cfm system and (2) the aerosol should re-
flect problems of both penetration and cleanability observed in real
life situations. Grassell reported the need and development of an
aerosol generator for testing air cleaners exposed to diesel exhaust.
The generator built from an acetylene burner was relatively stable with
day to day reproducibility of filter media loadings varying by a. factor
of 10 when using repetitive settings of controls.
Robert Hedden (5) working with Linsky developed an arc generator
of a fine particle aerosol. A single consumable electrode functioning
as an anode was vaporized by the heat generated by the arc. The
vaporized metal condensed as it was quenched by a passing air stream
to form fine particles. The iron oxide particles were sized by electron
microscopy to have an average diameter of 22 nm (220 A) . The mass
production rate was found to be 1.95 gm/minute. Hedden characterized
by weight the aerosol as containing 75% particles with effective diameter
less than 0.1 ym.
Mike Naylor (10) working with Dr. Richard Sears under Linsky
reported experimental results on a two consumable wire electrode,
electric arc aerosol generator. Naylor produced zinc oxide aerosols
with the generating apparatus. The apparatus consisted of a commercially
available electric arc metallizer, a settling barrel and an aerosol
exhausting system. The same generator unit was employed to produce
aerosols for this report.
3-5
-------
The consumable electrodes were melted and vaporized in the intense
heat of the arc and subsequently the vapor condensed into fine particles
in an air stream. These fine particles were found in agglomerated
chains as well as individually. Naylor reported a range of average
diameter sizes from 6-11 nm as measured by transmission electron
3
microscopy (TEM). A mass concentration of 0.65-2.0 gm/m was found by
3
Naylor at approximately 7.8 m /minute. This is a production rate of 5-
15.6 gm/minute.
Amson (1) stated that even though erratic behavior of the con-
sumable-electrode arc system is not rare, choosing a suitable power
source and ambient gas with suitable settings for the open-circuit
voltage and the wire electrode feed speed often leads to the system
operating in a 'quasi-steady state' mode. He reported that the elec-
trode voltage drop plays a very small part in the behavior of the con-
sumable electrode system. A decisive part is played by the electrode
stick out only when high resistivity electrodes are employed. The
electrode stick out is the portion of the electrode wire from the current
transfer tip to the arc.
PROCEDURE
Equipment and Material
The aerosol generating apparatus consisted of an Electrospray
metallizer that used an electric arc and a compressed air jet to spray
melted and vaporized consumable wire electrodes, a 55 gallon oil drum
connected to a supplemental air supply cleaned by a HEPA (High Efficiency
Air Particulate) filter, a draft fan with accompanying duct works from
3-6
-------
the oil drum, and a cloth filter particle collector (See Figure 1).
The metallizer system supplied the necessary power and material feed
delivery. The duct work and fan provided for exhausting and sampling
the aerosol. The baghouse cleaned the particle laden gas stream of
most of the particulates.
The metallizer (19) consisted of a power source, dual wire feeding
system, and an electric arc "gun." The metallizer power supply was a
three phase, 230 volt, 60 cycle AC to DC solid state full wave rectifier.
The DC supply could be varied from 20 to 40 volts. The wire feeding
system consisted of a reproducibly variable drive motor connected to
a pair of interlocking pinch rollers (with tension adjustments), wire
straighteners, and wire conduits for simultaneous delivery of the dual
feedstock wire electrodes to the arc gun.
The electric arc gun provided the transfer of electric current to
the consumable feedstock electrodes and atomization and quenching of
the products of the arc. The gun also provided a stationary structure
for proper electrode alignment and consistent impingement of the main
gas (compressed air) jet on the arc zone by the spray nozzle.
The arc gun was mounted through a cutaway in the lid of the oil
drum. A tip shroud which reduced entrainment of particles through the
arc zone was mounted on the arc side of the lid. A spacer bar between
the gun assembly and the barrel lid permitted connection of a second
air supply.
The 55 gallon barrel was set horizontally in a rack to facilitate
connection with the metallizer. The barrel's lid clamp was used to hold
the lid "air tight" against a tape gasket. An arc gun opening, a
3-7
-------
Wire
dispenser
(top view)
Electrospray
Secondary
Arc gun air
head adaptor
Secondary air
connection
Electrode tip
Main gas jet
Electrode wire
Shroud
Lid
ttop view)
Figure 1. Equipment set-up with closeup of arc gun assembly.
3-8
-------
filtered diluting make-up air inlet, and an exhaust air outlet were
provided in the lid.
The barrel functioned as a settling chamber for the larger particles.
Particles larger than 30 ym in diameter were collected with an efficiency
of approximately 95% by weight (10). The separation occurred from
gravity and a centrifugal force created by the 180 degree rotation of
the gas stream upon entering the exhaust duct.
The exhaust duct began one third the distance from the bottom of
the barrel. After leaving the barrel, the duct was immediately directed
for a straight run of at least 20 pipe diameters to the exhaust system
to the baghouse. At a point 10 pipe diameters from the last duct
disturbance, the sampling port was located. A calibrated venturi flow
meter was placed between the sampling port and the exhaust pipe system
to the baghouse. After the venturi, a damper was placed for adjustment
of the total air stream flow rate. The cloth filter and fan completed
the exhaust duct system.
The sampling train consisted of an RAC stack sampler with Greenburg-
Smith impinges connected to the sampling probe. A S-hook nozzle
was connected to a membrane filter holder. The filter holder retained
an appropriate filter for collecting the aerosol samples.
The stack sampler provided for adjustment of probe tip entrance
velocity as well as a measurement of dry gas volumetric flow. The
impinger system cooled and then dewatered the gas for measurement by
the dry gas meter.
Membrane filters were used to collect samples for subsequent
particle count or mass analysis. Membrane filters with 0.45 ym pore
3-9
-------
sizes were used for mass analysis and filters with 0.20 ym pore sizes
were used for particle count analysis. The 0.20 ym membrane filters
were also chosen because they provided good surface properties for elec-
tron microscopic particle counting.
The wire used in this research was stainless steel welding or
metallizing wire. Hobart steel welding wire (HB #18) was used for pro-
duction of fine particle aerosols for particle count and morphologic
analysis. Wall Colmonoy wire #10 was used to produce fine particle
aerosols for mass analysis. The following was an assay of the Hobart
steel wire (16):
Table 1. Hobart steel metalizing
wire (HB 18) assay
Fe 96.84%
Mn 1.90%
Si 0.65%
Mo 0.45%
C 0.12%
P 0.02%
S 0.02%
Operation of Generator Apparatus
The Electrospray control unit was operated with spray settings
within the range recommended by Wall Colmonoy (19). The settings for
the wire feed rate varied within 30-60 (see Appendix E for units), the
open voltage ranged from 20-31 volts, and the main jet pressure ranged
from 40-60 psi. The experimenter set the control variables and began
operation. If the apparatus ran without sputtering (an audio character-
istic of inefficient operation) for two minutes, the apparatus was con-
sidered operating stably. The closed voltage (voltage across arc during
operation) was not used as a criterion for stability.
3-10
-------
Sampling
After generation for at least two minutes, samples were taken
either for mass conversion analysis or particle size distribution
analysis. Consecutive randomly ordered samples were taken with a
two minute shut down between each sampling run.
Sampling for mass analysis consisted of pre-trial determination
of isokinetic sampling conditions and then the actual sampling. After
adjusting the stack sampler for isokinetic sampling as suggested by the
accompanying operation manual (13) , samples were taken such that at
least 10 mg of mass was deposited on the tared filter. All necessary
data was recorded before and during the trials on a form such as in
Figure 2.
Sampling the aerosol for subsequent particle size distribution
analysis consisted of estimating isokinetic conditions and then collection
on appropriate membrane filters with sampling times of three, five and
seven seconds. Then, the filters were enclosed in a plastic filter
holder and were shipped by United Postal Service to McCrone Associates,
Inc. in Chicago for analysis by scanning electron microscopy (SEM), TEM,
and selected area electron diffraction (SAED).
Phenomenon Characterization Techniques
The sizing of the fine particles was done by McCrone Associates
through TEM with a particle size count of an enlarged photomicrograph.
The fine particles counted by size were believed to agglomerate into
larger particles prior to sampling. The size limits that the McCrone
laboratory considered appropriate for adequate TEM characterization
ranged from less than 2 nm to greater than 20 nm by increments of 1 nm.
3-11
-------
Wire Type: Pbar- Dry Bulb TemPerature:
Date: ' Wet Bulb Temperature:
Sample #
Barrel Temp CO
71 HEPA Ap
p — p _ *•*•«
duct bar 13.6
Venturi A p
Duct T CO
Bag T CO
Pitot A p
Wire Feed Rate
Amps
V
open
MAP
SAP
A H orifice
CF start
CF end
4 I A CF
A Time
5
T
meter-start
T
meter-end
Nozzle Velocity
CFM ( /
V - 234
Vduct "*
V duct
Figure 2. Example of data sheet
3-12
-------
SEM was used to check samples for the proper laoding (particles per
area) necessary for TEM particle sizing. SAED analysis was performed
to determine phase characteristics of the particulate matter.
Sizing the agglomerates of fine particles characteristic of the
aerosol was done by the experimenter with SEM photomicrographs. SEM
negatives were enlarged noting the actual enlargement onto Kodak RC
paper as suggested by Mr. Richard Shimps of McCrone Associates (17).
The particles were sized with an uncalibrated clear plastic metric ruler.
A minimum of 150 particles were counted by size and then corrected to
actual size with the known magnitude of magnification.
Mass determination analysis was done by the experimenter weighing
the exposed sample membrane filters on an analytical balance accurate
to + 0.5 mg. With the mass determined by the TARE method and sampling
volumetric flow (STP) determined, an iron oxide mass concentration was
found.
Experiment: Analysis of Phenomenon
Four sets of samples were taken at four different operating con-
ditions for particle size analysis. The conditions were shown in
Table 2.
Two sets of samples were taken under identical conditions with the
exception that the gas stream velocity within the duct work was varied.
These conditions were shown in Table 3.
Mass samples were taken to conform to a Latin Square experimental
design. The Latin Square design is explained in Appendix D-l. Two
Latin Square designs were randomly chosen and operating conditons
(variables) were designated to the statistical design structure. The
3-13
-------
Table 2. Aerosol generator operating conditions
for particle size analysis
Sample
Numbers
600
601
602
603
604
605
607
611
609
612
Condition
Set Number
Wire Feed
Rate
25
25
25
25
25
25
50
50
50
50
Open
Voltage
20
20
20
20
20
20
22
22
22
22
Main Jet
Pressure
(psi)
60
60
60
45
45
45
60
60
45
45
Table 3. Comparison study operating conditons
Sample
Number
715
716
712
714
Condition
Set Number
Wire Feed
Rate
50
50
50
50
Open
Voltage
(volts)
22
22
22
22
Main Jet
Pressure
(psi)
60
60
60
60
Velocity of
Gas Stream
(fpm)
1300
1300
1850
1850
3-14
-------
samples for the structure were randomly ordered and then the samples
were taken. Two trials, the second trial a replicating trial, were
performed; the trials occurred on different days.
Another set of mass samples were taken to supplement the Latin
Square design. A collection of orthogonal points were selected about
an observed maximum point in each of the operating variables; a wire
feed rate of 60, an open voltage of 29, and a main jet pressure of 45
were considered midpoints. A unit "distance" was added to and subtracted
from each midpoint to develop a resulting eight orthogonal samples as
follows in Table 4.
Table 4. Determination of orthogonal
data collection points
Wire Feed Open Voltage
Rate (Volts)
Midpoint value 60
Unit distance 5
Resulting points
Additive 65
Subtractive 55
Resulting samples
Condition
29
1
30
28
Main Jet
Pressure (psi)
45
5
50
40
1
2
3
4
5
6
7
8
65
55
55
65
65
65
55
55
28
28
30
30
30
28
30
28
40
50
40
40
50
50
50
40
These resultant sampling conditions were randomly order and then samples
were taken.
3-15
-------
Data
The aerosol size distributions found for the agglomerated fine
particles of sample 714 and 715 were presented in Tables 5 and 6. The
percentages of the total number of particles and the cumulative per-
centages were also given. The raw data uncorrected for the different
magnifications were presented in Table 27.
The fine particle size distributions found for the various sets
of operating conditions were presented in Tables 7, 8, 9, 10, and 11.
The percentages and cumulative percentages were also given.
SEM pictures of the agglomerated very fine particles were shown in
Appendix C-l.
In Table 12, the results of the SAED analysis were presented. In
Table 13, ASTM electron diffraction lines were given for magnetite
(Fe-,0.) and maghemite (r-Fe^,).
The mass concentration data from the two Latin Square designs
was presented in Table 14. The mass concentration data from the ortho-
gonal collection of data was presented in Table 15. Table 16 contains
a listing of all the mass concentration data. The data and method
of mass concentration calculation for the Latin Square and the orthogonal
designs were presented in Appendix C-3.
3-16
-------
Table 5. Agglomerated data corrected for
magnitude of magnification: sample 715
Class Frequency Cumulative Percent Cumulative
Mark Frequency Percent
(pm)
0.09 141 141 47.0 47.0
0.28 56 197 18.7 65.7
0.43 31 228 10.3 76.0
0.61 23 251 7.7 83.7
0.78 10 261 3.3 87.0
0.96 7 268 2.3 89.3
1.1 11 279 3.7 93.0
1.3 4 283 1.3 94.3
1.5 4 287 1.3 95.7
1.7 3 290 1.0 96.7
1.8 1 291 0.3 97.0
2.3 1 292 0.3 97.3
2.6 8 300 2.7 100.0
Table 6. Agglomerated data corrected for
magnitude of magnification: sample 714
Class Frequency Cumulative Percent Cumulative
Mark Frequency Percent
0.09 156 156 55.3 55.3
0.27 41 197 14.5 69.9
0.46 19 216 6.7 76.6
0.64 17 233 6.0 82.6
0.82 14 247 5.0 87.6
1.0 8 255 2.8 90.4
1.2 7 262 2.5 92.9
1.4 3 265 1.1 94.0
16 3 268 1.1 95.0
17 6 274 2.1 97.2
23 2 276 0.7 97.9
2*5 1 277 0.4 98.2
26 1 278 0.4 98.6
27 4 282 1.4 100.0
3-17
-------
Table 7. Very fine particle
distributions set 1
Class
Mark
(nm)
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16
17
18
19
21
,5
,5
,5
,5
Class
Mark
(nm)
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
Frequency
18,
19.
21
13
53
53
37
33
23
18
10
18
19
19
17
21
4
7
7
9
3
6
16
Frequency
8
37
55
65
45
20
15
25
55
16
40
7
9
13
13
2
2
3
6
10
Sample 600
Cumulative
Frequency
13
66
119
156
189
212
230
240
258
277
296
313
334
338
345
352
361
364
370
386
Sample 601
Cumulative
Frequency
8
45
100
165
210
230
245
270
325
341
381
388
397
410
423
425
427
430
436
446
Percent
Percent
1.8
8.3
12.3
14.6
10.1
4.5
3.4
5.6
12.3
3.6
9.0
1.6
2.0
2.9
2.9
0
0
0.7
1.4
2.2
.5
.5
Cumulative
Percent
3.4
13.7
13.7
9.6
8.6
6.0
4.7
2.6
4.7
4.9
4.9
4.4
5.4
1.0
1.8
1.8
2.3
0.8
1.6
4.1
3.4
17.1
30.8
40.4
49.0
54.9
59.6
62.1
66.8
71.8
76.7
81.1
86.5
87.6
89.4
91.2
93.5
94.3
95.9
100.0
Cumulative
Percent
1.8
10.1
22.4
37.0
47.1
51.6
54.9
60.5
72.9
76.5
85.4
87.0
89.0
91.9
94.8
95
95
96
97.8
100.0
3
7
.4
3-18
-------
Table 7. Continued
Sample 602
Class Frequency Cumulative Percent Cumulative
Mark Frequency Percent
(nm)
1.5 24 24 5.5 5.5
2.5 79 103 18.2 23.7
3.5 57 160 13.1 36.8
4.5 21 181 4.8 41.7
5.5 25 206 5.8 47.5
6.5 43 249 9.9 57.4
7.5 20 269 4.6 62.0
8.5 20 289 4.6 66.6
9.5 22 311 5.1 71.7
10.5 13 324 3.0 74.7
11.5 33 357 7.6 82.3
12.5 16 373 3.7 85.9
13.5 2 375 0.5 86.4
14.5 6 381 1.4 87.8
15.5 9 390 2.1 89.9
16.5 9 399 2.1 91.9
17.5 12 411 2.8 94.7
18.5 2 413 0.5 95.2
19.5 9 422 2.1 97.2
21 12 434 2.8 100.0
3-19
-------
Table 8. Very fine particle
Class
Mark
(nm)
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.
9,
10.5
11.5
12.5
13.5
14.5
15.5
16,
17.
18,
19.
21
Class
Mark
(nm)
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10.5
11.5
12.5
13.
14.
15.5
16.5
17.5
18.
19.
21
.5
.5
,5
,5
Frequency
2
45
58
34
29
18
13
18
30
11
16
14
12
19
7
7
11
4
7
14
Frequency
3
51
66
18
33
19
13
14
39
8
21
12
9
10
5
7
7
5
9
19
distributions set 2
Sample 603
Cumulative Percent
Frequency
2
47
105
139
168
186
199
217
247
258
274
288
300
319
326
333
344
348
355
369
Sample 604
Cumulative
Frequency
3
54
120
138
171
190
203
217
256
264
285
297
306
316
321
328
335
340
349
368
0.5
12.2
15.7
9.2
7.9
4.9
3.5
4.9
8.1
3.0
4.3
3.8
3.3
5.2
1.9
1.9
2.9
1.1
1.9
3.8
Percent
Cumulative
Percent
0.
12.
28.4
37.6
45.5
50.4
53.9
58.8
66.9
69.9
74.2
78.0
81.3
86.5
88,
90,
93-
94,
96
100.0
Cumulative
Percent
8.8
13.8
17.9
4.9
9.0
5.2
3.5
3.8
10.6
2.2
5.7
3.3
2.4
2.7
1.4
1.9
1.9
1.4
2. A
5.2
0.8
14.6
32.6
37.5
46.5
51.6
55.2
59.0
69.6
71.7
77.4
80.7
83.2
85.9
87.2
89.1
91.0
92.3
94.8
100.0
3-20
-------
Table 8. Continued
Sample 605
Class Frequency Cumulative Percent Cumulative
Mark Frequency Percent
(nm)
1>5 4 4 1.1 1.1
2'5 41 45 11.5 12.6
3.5 56 101 15.7 28.3
4-5 42 143 11.7 40.1
5-5 21 164 5.8 46.0
6-5 19 183 5.3 51.4
7-5 17 200 4.7 56.1
8-5 17 217 4.7 60.9
9-5 29 246 8.1 69.1
10-5 9 255 2.5 71.6
H-5 17 272 4.7 76.4
12.5 12 284 3.3 79.7
13.5 12 296 3.3 83.1
14.5 11 307 3.0 86.2
15.5 5 312 1.4 87.6
16.5 8 320 2.2 89.8
17.5 8 328 2.2 92.1
18.5 4 332 1.1 93.2
19.5 8 340 2.2 95.5
21. 16 356 4.4 99.9
3-21
-------
Class
Mark
(nm)
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10.
11.
12.5
13.5
14.
15.
16.
17.
18.
19.
21
.5
.5
,5
.5
.5
.5
,5
,5
Table 9. Very fine particle
distributions set 3
Sample 607
Frequency Cumulative Percent
Frequency
Class
Mark
(nm)
1
16
47
40
20
24
6
12
11
9
19
18
10
12
11
9
5
6
6
14
Frequency
1
17
64
104
124
148
154
166
177
186
205
223
233
245
256
265
270
276
282
296
Sample 611
Cumulative
Frequency
0.3
5.4
15.8
13.5
6.7
8.1
2.0
4.0
3.7
3.0
6.4
6.0
3.3
4.0
3.7
3.0
1.6
2.0
2.0
4.7
Percent
Cumulative
Percent
0.3
5.7
21.6
35.1
41.8
50.0
52.0
56.0
59.7
62.8
69.2
75.3
78,
82,
86,
89,
91
93
95
99.9
Cumulative
Percent
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10.5
11.5
12,
13,
14,
15,
16,
17.
18.
19.
21
13
54
46
12
25
8
4
7
7
28
14
10
7
8
5
5
1
5
15
13
67
113
125
150
158
162
169
176
204
218
228
235
243
248
253
254
259
274
4.7
19.7
16-. 7
4.3
9.1
2.9
1.4
2.5
2.5
10.2
5.1
3.6
2.5
2.9
1.8
1.8
0.3
1.8
5.4
4.7
24.4
41.2
45.6
54.7
57.6
59.1
61.6
64.2
74.4
79.
83.
85,
88.
90,
92,
92
94
5
2
7
6
5
3
,7
,5
99.9
3-22
-------
Table 10.
Class
Mark
(nm)
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10.
11,
12,
13,
14,
15,
16,
17.5
18.
19.
21
Class
Mark
(nm)
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18,
19,
21
Frequency
18
55
22
16
26
22
20
18
18
7
12
9
8
6
8
5
6
8
3
11
Frequency
16
47
23
19
22
21
26
17
13
11
12
12
11
8
5
4
7
8
6
9
Very fine particle
distributions set 4
Sample 609
Cumulative Percent
Frequency
18
73
95
111
137
159
179
197
215
222
234
243
251
257
265
270
276
284
287
298
Sample 612
Cumulative
Frequency
16
63
86
105
127
148
174
191
204
215
227
239
250
258
263
267
274
282
288
297
6.0
18.4
7.3
5.3
8.7
7.3
6.7
6.0
6.0
2.3
4.0
3.0
2.6
2.0
2.6
1.6
2,0
2.6
1.0
3.6
Percent
5.3
15.8
7.7
6.3
7.4
7.0
8.7
5.7
4.3
3.7
4.0
4.0
3.7
2.6
1.6
1.3
2.3
2.6
2.0
3.0
Cumulative
Percent
6.0
24.4
31.8
37.2
45.9
53.3
60.0
66.
72,
.1
.1
74.4
78,
81,
84,
86,
88,
90,
92.6
95,
96,
100.0
Cumulative
Percent
5.3
21.3
28.9
35.3
42.7
49.8
.5
.3
.6
.4
.1
58
64
68
72.3
76.4
80
84
86.8
88.5
89.8
92.2
94.9
96.9
99.9
3-23
-------
Table 11. Very fine particle distributions
for samples 712 and 716
.5
.5
Clans
Mark
(nm)
1.5
2.5
3.5
4-5 i
5.5
6.5
7.5
8.5
9-5
10.5
11.5
12.5
13.5
14.
.15.
16.5
17.5
18.5
19.5
21
Class
Mark
(nm)
1.5
2.5
3.5
4.5
5.5
6.5
7.5
8.5
9.5
10.5
11.5
12.5
13.5
14.5
15.5
16.5
17.5
18,
19,
21
Frequency
26
148
63
39
24
19
23
24
21
20
14
11
13
10
11
11
10
4
2
4
Frequency
17
49
66
47
35
28
19
23
18
18
15
19
23
17
12
9
6
11
9
8
Sample 712
Cumulative
Frequency
26
174
237
276
300
319
342
366
387
407
421
432
445
455
466
477
487
491
493
497
Sample 716
Cumulative
Frequency
17
66
132
179
214
242
261
284
302
320
335
354
377
394
406
415
421
432
441
449
Percent
5.2
29.7
12.6
7.8
4.8
3.8
4.6
4.8
4.2
4.0
2.8
2.2
2.6
2.0
2.2
2.2
2.0
0.8
0.4
0.8
Percent
Cumulative
Percent
5.2
35.0
47.6
55.5
60.
64.
68.8
73.6
77.8
81.8
84.7
86.9
.3
.1
89
91
93
95.9
97.9
98.7
99.1
.5
.5
.7
99.9
Cumulative
Percent
3.7
10.9
14.6
10.4
7.7
6.2
4.2
5.1
4.0
4.0
3.3
4.2
5.1
3.7
2.6
2.0
1.3
2.4
2.0
1.7
3.7
14.6
29.3
39.8
47.6
53.8
58.1
63.2
67.2
71.2
74.6
78.8
83.9
87.7
90.4
92.4
93.7
96.2
98.2
100.0
3-24
-------
Table 12. Selected area electron
diffraction data
Sample 602 604 611 609
lines
d A
^» **
Sample
lines
d A
v* c\
1.478
1.615
2.094
2.526
2.97
712
1.257
1.320
-
1.628
1.721
2.106
2.54
2.96
1.485
1.615
1.710
2.10
2.526
2.97
716
-
-
1.487
1.618
1.713
2.101
2.53
2.97
1.489 1.476
1.615 1.615
1.706
2.10 2.11
2.526 2.517
2.962 2.961
Table 13. Diffraction lines for
magnetite and maghemite
Magnetite, Fe-jO^ Maghemite, y-Fe
ASTM Card 19-629 ASTM Card 15-615
lines
d A
•
1.266
1.328
1.485
1.616
1.715
2.099
2.532
1.008 s
1.702 s
2.089 s
2.521 s
2.950 s
s = lines marked with s
2.967 include those of the
spinel structure
3-25
-------
Table 14. Latin square mass concentration
data summary (gm/sm3)
Main Air Pressure (psi)
25
8> 27
ca
•M x-\
r-l CO
O 4J
0)
p.
O
31
40
45
50
55
0.92
A
1.19
B
1.51
C
1.70
D
1.11
B
1.28
C
1.77
D
1.05
A
1.37
C
1.74
D
1.07
A
1.31
B
1.03
D
0.81
A
1.48
B
1.18
C
Replicate 1
Wire Feed Rates:
A - 30
B - 40
C - 50
D - 60
Main Air Pressure (psi)
40
45
50
55
25
0)
00
2~ 27
rH CO
O -U
> H
gS 29
31
1.00
A
1.50
B
1.43
C
1.67
D
1.22
B
1.55
C
1.38
D
1.19
A
1.46
C
1.83
D
1.13
A
1.92
B
1.81
D
1.15
A
1.38
B
1.26
C
Replicate 2
3-26
-------
Table 15. Orthogonal mass concentration
data summary (gm/m3)
Main Air Pressure (psi)
-------
Table 16. A listing of all the mass concentration data
Sample
Number
Mass Concen.
(gm/m3)
Temp . in
Exhaust
Duct (°C)
Wire
Feed
Rate
Open
Voltage
(Volts)
Main Air
Pressure
(psi)
563
702
562
566
581
554
558
705
550
707
564
565
704
529
703
561
560
576
701
706
573
700
571
575
2.16
2.13
1.92
1.83
1.81
1.77
1.74
1.71
1.7
1.7
1.67
1.55
1.52
1.51
1.51
1.5
1.48
1.46
1.44
1.44
1.43
1.41
1.38
1.38
82
90
62
80
80
85
80
95
72
85
80
77
80
68
72
71
70
76
72
85
68
70
58
76
50
55
40
60
60
60
60
55
60
55
60
50
45
50
40
40
40
50
45
55
50
45
40
60
31
30
31
27
25
29
27
30
31
28
31
27
28
29
30
27
29
25
28
28
29
30
29
29
. 55
50
50
50
55
45
50
40
40
50
40
45
50
40
40
40
55
50
40
40
40
50
55
45
3-28
-------
Table 16. Continued
Sample
Number
Mass. Concen.
(gm/m3)
Temp . in
Exhaust
Duct (°C)
Wire
Feed
Rate
Open
Voltage
(volts)
Main Air
Pressure
(psi)
555
546
548
572
553
568
549
569
570
547
556
552
559
582
557
551
1.37
1.31
1.28
1.22
1.19
1.19
1.18
1.15
1.13
1.11
1.07
1.05
1.03
1.0
0.92
0.81
76
73
74
65
73
57
75
58
54
64
60
53
70
59
57
60
50
40
50
40
40
30
50
30
30
40
30
30
60
30
30
30
25
31
27
25
27
31
31
27
29
25
29
31
25
24
25
27
50
50
45
45
40
45
55
55
50
45
50
45
55
40
40
55
3-29
-------
ANALYSIS OF DATA
Agglomerates
In the comparative experiment samples 715 and 714 were chosen as
representative of the agglomerates found in the samples. The agglomerates
measured by the experimenter were found to have means of 0.41 ym
(S=0.53 ym) and 0.40 ym (S=0.54 ym) as shown in Appendix C-2. The follow-
ing analysis used the student t-test for determining the equality of ex-
pected populations of two different samples. The implication was that
the means were equal if the difference between the two means was equal
to zero.
The null hypothesis that the difference between the two means was
equal to zero:
V Uru2=0
was tested against the alternative hypothesis that the difference was
not equal to zero,
HA: uru2XO.
To determine which hypothesis was true, the t-test statistic was used.
This statistic was determined by the following general formula:
cal S-
xp
where x..-x_ was greater than or equal to zero and S- was the pooled
standard deviation as defined in Appendix D-2. The t 1 value was com-
pared to a tabulated t .^ (v) statistic value where the a- error was
equal to 0.10 and the degrees of freedom (v) were set equal to infinity
because the sample sizes were much larger than 30. As shown in Appendix D-2,
3-30
-------
the hand calculations showed the t value to be 0.194. The t (v-«)
value was found to be 1.645.
The criteria for hypothesis test was as follows:
Table 17. Two means equality test hypothesis decision-making criteria
If tcal - ta/2> then Ho is accePted
and H is rejected.
A
If 'cal > '0/2 f then Ho ls reJected
and H is accepted .
Because tcal < t ,2,the null hypothesis, H : u -u =0, is accepted. This
implied that the two sample means came from the same population; this was
tested for 90% confidence.
The above student t-test was only applicable if the variances of
the samples came from the same population. The F-statistic was used to
determine if the variances came from the same population. Where
cal 2
S2
2 •
with S* equal to the variance of sample 1 and F , having an F-distri-
1 cax
bution with degrees of freedom v.. for s- and v_ for s_ was compared to a
tabulated F (v, ,v_) and found to be less than or equal to the F value,
a 1 2 ex
the variances were considered to be from the sample population. If the
variances were different, the Smith-Satterthwaite test (9) was used to
determine the equality of two means .
The F was found to be 1.05 and F (v. ,v_) was not found in any
cal a 1 2
3-31
-------
statistical table. But, if both of the degrees of freedom (v., and v_)
approach large values (> 100), values of Fa(vi»v2) approach 1.00. It was
assumed that F n was greater than F and therefore the Smith-Satterthwaite
cal ct
test was needed for the t-statistic test for equality of means.
The hypothesis testing using the Smith-Satterthwaite t-statistic is
the same as the equivalent variance t-eest with the exception that the
t-value was found by the following equation:
Xrx2
"cal
nl n2
and that the degrees of freedom (df) were found by the following equation:
df
s 2 2
!i_ + !2_
nl n2
V1
V1
The
t 1 value was found to be 0.225 with degrees of freedom 576.
The t ,„ value was still 1.635. The outcome of the hypothesis that the
means were equal was not changed by the fact that the variances were not
equal.
The coefficient of variance was also calculated for each sample dis-
tribution and was presented in Table 1. along with the means (x), standard
deviation (S), and standard errors of the mean for each sample. The co-
efficient of variance was an indicator of the monodispersity of a size
distribution (2) and is given by the following formula:
3-32
-------
C.V. - S/x
The coefficient of variance for the agglomerate size distributions of
samples 714 and 715 were 1.35 and 1.29, respectively. Any sample size
distribution with a coefficient of variance less than 0.10 (2_) was con-
sidered raonodisperse.
Table 17. Statistics for agglomerate samples
mple No.
714
715
Mean
(ym)
0.40
0.41
Standard
Deviation
(ym)
0.54
0.53
Standard Error
of the Mean (ym)
O..Q3
0.03
Coeffi
Varian
135
129
Fine Particles
The fine particle size distribution data was broken down into six
sets of data contingent on the six sets of conditions. Two basic tests
were performed on the fine particle size distribution data: (1) within
group t-test for equality of means and between group t-test for equality
of means and (2) within group and between group chi squared test for
equality of frequency distributions. Plotted frequency distribution were
shown in Figures 3, 4, 5, 6, and 7.
The SAS T-TEST and MEANS (15) procedures were employed to determine
relevant statistics for the data. The MEANS procedure calculated the means,
standard deviations, standard errors of the mean, and the coefficients of
variance as presented in Table 19. The T-TEST procedure calculated the
F-statistic for comparison of two sample variances and the t-statistic for
determination of mean equality. A summary of the t-test statistics for the
within group analysis was presented in Table 20. A summary of t-test
3-33
-------
statistics for between group analysis was presented in Tables 20 and 21.
The tables were structured for presentation of statistics concerning the
intersecting samples.
The statistic in the middle of each block was the t-value calculated
for the two mean t-test where
s-
xp
with the hypothesis that the means of the intersecting samples were equal.
The box above the t-value was checked if the variances of the two samples
were considered to be unequal; this implied that the t-value had been
determined by the Smith-Satterthwaite t-statistic. The value beneath
the t-value was the probability that the t-value could be exceeded due
to random error in the sampling.
The equality of the frequency distributions were tested by the use
of the SAS FREQUENCY (15) procedure. Chi-squared statistics were determined
by the FREQUENCY procedure and presented for within and between groups
in Tables 22 and 23 respectively.
3-34
-------
U)
150'
F 100
R
E
Q
u
c
Y
50
A
Q
I
©
1.5
3.5 H'
Sample
•600
©601
£602
7-5 *-5 "?•? »-5 U-S 12.5
Class Mark (nm)
17.5
Figure 3. Set 1 frequency distributions.
-------
u>
150 1-
F 100
R
E
Q
u
E
N
504
Sample
•603
©604
& 605
« « *
f •
» f A
2.1
Class Mark (nm)
Figure 4. Set 2 frequency distributions.
-------
10
I
CO
150
F 100
R
E
Q
u
E
N
C
Y
50
Sample
•607
O611
<»
•
o
3-5 4-5
7-ST 8.5 <».f
Class Mark (nm)
l&S
17.5
Figure 5. Set 3 frequency distributions.
-------
LJ
OJ
c»
150
F 100
R
E
Q
u
E
N
50
Sample
• 609
0612
O
1-5 3L-5 3-5
5.5 fe-5
13.5 if,5 i£*
Class Mark (nm)
Figure 6. Set 4 frequency distributions.
17-5
-------
150-
F 100
R
E
Q
u
E
N
C
Y
50
Sample
• 712
©716
Q
0
•
4
©
•
1
1 - 1
o
1
-5 *S \tS U.S
Class Mark (nm)
Figure 7. Samples 712 and 716 frequency distributions.
-------
Table 19. Statistics for fine particle distributions
Set Number
1
2
3
4
5
6
Sample No .
600
601
602
603
604
605
607
611
609
612
716
712
Number of
Particles
386
446
434
369
368
356
296
274
298
297
449
497
Mean
(nm)
8.00
7.83
7.53
8.45
8.32
8.31
9.05
8.57
7.86
8.13
8.05
6.42
Standard
Deviation
(nm)
5.35
4.56
5.23
5.37
5.51
5.41
5.49
5.40
5.40
5.37
5.21
4.79
Standard of Error
of the Mean (nm)
0.27
0.22
0.25
0.28
0.29
0.29
0.32
0.33
0.31
0.31
0.25
0.21
Coefficient of
Variance (%)
66.9
58.2
69.5
63.5
66.3
65.1
60.5
62.9
68.8
66.0
64.7
74.6
u>
-p-
o
-------
Table 20. Within group t-statistics for fine particle samples
600
601
602
_
^
0.496
0.62
D
1.28
0.20
0.919
0.36
-
603
D
1.04
0.30
611
609 Sample Number
Q
0.628
0.53
612
604
605
-
[umber
D
0.302
0.76
D
0.328
0.74
a
-0.026
0.98
-
716
-5.02
0.001
Sample Number
600
601
602
Sample Number
603
604
605
Sample Number
712
3-41
-------
Table 21. Between Group t-statistics for fine particle samples
i
-P-
Sample
Number
600
601
602
Sample
Number
603
604
605
603
607
604
605
607
611
609
612
716
611
609
612
716
712
a
-1.43
0.15
a
-1.71
0.09
a
-1.74
0.08
a
-0.30
0.76
a
-0.59
0.55
a
-0.62
0.54
a
1.40
0.16
a
-1.08
0.28
P
1.06
0.29
n
0.74
0.46
a
0.43
0.67
a
0.40
0.69
a
1.05
0.29
a
0.70
0.49
a
0.67
0.50
5.76
0.001
5.29
0.001
5.27
0.001
712
a
-1.14
0.25
-l775
0.08
a
-2.46
0.02
a
-0.80
0.42
-1.36
0.18
a
-2.03
0.04
a
0.78
0.44
GZf
-1.33
0.19
a
-2.06
0.04
a
-2.52
0.12
Of
-3.17
0.001
a
-3.80
0.002
a
-1.36
0.18
-1.91
0.06
a
-2.57
0.01
a
0.34
0.73
3f
-0.07
0.94
a
0.83
0.41
a
-0.33
0.74
GZf
0.80
0.42
a
-1.53
0.13
a
-0.15
0.88
-0.69
0.49
0
-1.51
0.13
sf
4.56
0.001
a
4.63
0.001
CS'
3.35
0.008
-------
Table 21. Continued
Sample
Number
716
712
600
601
602
603
604
605
607
611
609
612
-0.15
0.88
4.56
0.001
-0.69
0.49
a
4.13
0.001
a
-1.51
0.13
3.35
0.008
a
1.05
0.29
5.76
0.001
a
0.70
0.49
5.29
0.001
a
0.67
0.50
5.27
0.001
a
2.50
0.01
6.86
0.001
a
1.29
0.20
5.53
0.001
a
-0.50
0.62
3.79
0.001
a
0.20
0.84
4.54
0,001
u>
I
Sample
Number
607
611
609
612
716
712
a
2.67
0.008
a
1.59
0.11
a
2.06
0.04
a
0.98
0.33
a
2.50
0.01
a
1.29
0.20
6.86
0.001
5.53
0.001
Sample
Number
609
612
716
712
a
-0.50
0.62
a
0.20
0.84
3.29
0.001
G3f
4.54
0.001
-------
Table 22. Within group chi-squared statistics
Sample
Number
600
601
602
600
601
602
—
67.2
0.0001
-
44.6
0.0008
99.98
0.0001
-
Sample
Number
607
Sample
Number
603
604
605
603
604
611
605
—
14.8
0.733
-
Sample
Number
609
6.98
0.994
17.93
0.527
-
612
7.154
0.993
Sample
Number
712
716
71.77
0.001
3-44
-------
Table 23. Between group chi-squared statistics
Sample
Number
603
604
605
607
6L1
609
612
29.8
0.055
51.7
0.0001
28.5
0.074
56.1
0.0001
73.9
0.0001
45.6
0.0006
49.8
0.0001
47.9
0.0003
46.3
0.0005
45.1
0.0007
51.8
0.0001
39.7
0.0036
Sample
Number
609
612
607
611
77.7
0.0001
67.7
0.0001
109.9
0.0001
96.6
0.0001
3-45
-------
Mass
The Latin Square data was organized and presented In the Latin
blocks in Table 14. The Latin Square analysis of variance calculations
were carried out in Appendix D-l with a summary presented in Table 24.
The data was also analyzed using the SAS ANOVA procedure. This data
was shown in Table 25. This table was explained in Appendix D-3.
The data taken about an expected midrange of operating variables
was analyzed by the SAS GLM (15) procedure and was presented in Table 26.
All of the mass data, that is, both the Latin Square and the
orthogonal data sets were analyzed by the SAS GLM and STEPWISE pro-
cedure. The SAS GLM table was presented in Table 27. The STEPWISE
procedure used a statistical method to develop a "best-fit" linear
model for the prediction of the dependent variable-mass concentration
of aerosol within the gas stream. In Appendix D-4, this procedure was
described.
The "best-fit" linear model was analyzed for its aptness with
respect to its predictive ability using the SAS GLM procedure and
associated calculations for model residual analysis. This procedure
was described in Appendix D-5.
3-46
-------
Table 24. Analysis of variance for latin square experimental design
mass data
Variable
Wire feed
rate
Open Voltage
Main air-jet
pressure
Replication
Experimental
error
Total
df
o
3
*
3
1
21
31
Sum of
Squares
1.473
0.319
0.106
0.335
0.952
3.185
Mean
Square
0.4908
0.1065
0.0352
0.3349
0.0453
F
cal
10.86
2.35
0.77
7.39
Fa-0.05
3.07
3.07
3.07
4.32
Fo-0.01
4.87
4.87
4.87
8.02
3-47
-------
Table 25. SAS ANOVA for latin square experimental design mass data
Dependent Variable: Y, Mass Concentration
Source Sum of Squares df Mean Square
Model
Error
Corrected Total
F Value
2.23 10 0.22
0.95 21 0.045
3.18 31
R-Square C.V. *
0.70 15.4
Std. Dev. Y Mean
0.21 1.38
4.94
PR>F
0.0010
Source
df
ANOVA SS
F Value
PR>F
Wire feed
rate
Voltage
Air Pressure
Replication
3
3
3
1
1.47
0.32
0.11
0.33
10.86
2.36
0.81
7.35
0.0002
0.1004
0.5042
0.0131
3-48
-------
Table 26. SAS GLM table for orthogonal data
Dependent Variable: Y, Mass Concentration
Source df Sum of Squares
Model 3 0.26
Error 4 0.14
Corrected Total 7 0.40
Mean Square F Value
0.08
0.03
2.41
PR>F
0.2079
R-Square
0.64
Std. Dev.
0.19
C.V.
11.8
Y Mean
1.61
Source
Wire feed
rate
Voltage
Air Pressure
df
Type I SS
F Value
PR>F
1
1
1
0.15
0.054
0.054
4.20
1.51
1.51
0.1099
0.2864
0.2864
3-49
-------
Table 27. SAS GLM table for all of the data
Dependent Variable Y, Mass Concentration
Source df
Model 13
Error 26
Corrected Total 39
Source
Wire feed
rate
Voltage
Air Pressure
Replication
Sum of Squares Mean Square
2.81 0.22
1.09 0.042
3.90
R-Square C.V.
0.71 14.4
Std. Dev. Y Mean
0.21 1.43
f Type I SS F Value
5 1.94 9.21
4 0.37 2.22
3 0.16 1.27
1 0.33 7.87
F Value
5.12
PR>F
0.0002
PR>F
0.0001
0.0946
0.3063
0.0094
3-50
-------
DISCUSSION OF RESULTS
Agglomerates
The size means of the agglomerates were found to be 0.41 ym
(S=0.53 ym) for sample 715 and 0.40 ym (S=0.54 ym) for sample 714.
These means were also found to be equal; this implied that they were
from the same population. The coefficient of variances were rather
large, thus the agglomerated particle distributions representative of
the aerosols were not "monodisperse."
The simple comparative analysis of increasing the duct exhausting
velocity from 1300 fpm to 1850 fpm showed only a decrease in particle
size from 0.41 ym to 0.40 ym. This difference was not significant at a
0.1 level of significance.
The distribution of the agglomerates was not normally distributed
about the mean as shown in Figure 8. This fact indicated an inadequacy
of this analysis because the t-test for equality of means assumed the
sample distributions to be normally distributed. Therefore, the analysis
of the agglomerate distributions was, at least, a characterization of
the aerosol. While this was true, the analysis still served to illus-
trate the t-test for equality of mean technique.
Fine Particles
The within group t-test analysis showed that the sample means in
all the sets were equivalent within groups. T-test hypothesis and
implications were outlined in Appendix D-2. Set 2 t-test analysis
showed a very small difference within this group of operating conditions.
The variance of this set of samples were also considered equal by the
F-test as set forth in Appendix D-2. Certainly, these three sample
3-51
-------
\s»
1*0
Particle
Count
Sample 714
o a. 4 fe 8 to 12 N \fe 10 20 *# #a 21 30 37.
Size range
(So
loo
Particle
Count
Sample 715
n.
9 2. H
Size range
Figure 8 Histograms of agglomerated particles
3-52
-------
populations (set 2) were representative of the same aerosol parent
population.
The between group t-test analysis showed a variety of equal and
unequal means. Table 21 showed that the means of set 1 were all different
than the means in set 3. Three of the four means in the combination of
set 3 and set 4 were .different. Five of the six means in the combination
of set 1 and set 4 were equal. The majority of the means of the combi-
nation of set 1 and set 2 were unequal.
The t-test analysis between the two comparative samples 712 and
716 showed sample 712 to have a significantly smaller very fine particle
size than sample 716. In fact, sample 712 was found to have a mean
much smaller than other samples taken. The increase in the exhausting
rate of the barrel was the only known difference in the operating con-
ditions and most likely had the observed effect on the means and variances
of the fine particle size distributions. Even with this information,
the simple comparative experiment was not considered to be conclusive
because of the small number of samples.
The within group very fine particle size distributions for sets
2, 3, and 4 were found to be equivalent. Size distributions between
groups were found to be all different at an alpha level of 0.05. This
implied that at three of four different operating conditions three
different fine particle size distributions were generated.
Chemical Composition
The SAED analysis showed two possible resultant materials. Two
different phases of iron oxide were possible: magnetite and maghemite.
Mr. Richard Shimps of McCrone said: There is difficulty in deciding
3-53
-------
which one may truly be present, or if both are present. This stems
from the fact that the strongest diffraction lines of each material
are the same since the strongest lines of maghemite are those which
include the spinel structure. (17) Magnetite was considered to be present
with a possibility of maghemite being present.
Mass Analysis
The Latin Square experimental design was analyzed by two analyses
of variance methods; as expected, the Latin Square hand calculation and
the SAS ANOVA procedure resulted in the same values for the statistics.
The F statistic in ANOVA was used to determine if the particular variable
had any effect on phenomenon of gas stream particulate mass concentration.
The test was explained in Appendix D-3.
The Latin Square analysis found the wire feed rate to be significant
at the 0.01 significance level. This implied the wire feed rate had an
effect on the resultant mass concentration. The variables of open
voltage and jet pressure were not found to have any effect on the mass
concentration of the aerosol even at the 0.1 significance level. The
analysis of variance statistics for the effect of replication was also
calculated and found to indicate a significant difference between trials
at the 0.05 significance level but not the 0.01 significance level.
As explained in Appendix D-3, the SAS ANOVA procedure also analyzed
the variance of the data to determine if the variability of the data
was a consequence of random error (noise) in the collection of the data
or a consequence of the changing of the experimental variables. The
F test value for this test was found to conclude that the variability
was a consequence of the experimental variables at the 0.01 significance
level.
3-54
-------
The SAS GLM procedure found this particular set of operating points
to insignificantly describe the variability of the orthogonal data. Even
though this was found, the wire feed rate was found to be almost signi-
ficant at the 0.1 significance level.
The SAS GLM analysis of the combination of the Latin Square and
Orthogonal data found the data to significantly (a = 0.01) describe the
variability of all the data. With this total set of 40 observations,
the SAS GLM procedure calculated a F statistic for determining the effect
or noneffect of the independent variables in the experimental model.
The procedure found the variables included in the model to have a signi-
ficant effect on the outcome: mass concentration of the gas stream.
That was, the GLM procedure determined that the wire feed rate, the open
voltage across the arc, and the sample replication variables affected the
results of the experiment.
The SAS STEPWISE procedure determined a "best-fit" linear equation
for the prediction of mass concentration in the gas stream. The equation
was
Y = -0.81 + 0.01 (WP) + 0.048 (V)
3
where Y is the mass concentration (gm/m ), W is the wire feed rate
(range 30-60), and V is the open voltage (range 25-31 volts). By the
SAS GLM procedure the intercept beta value (-0.81) was significant at
the 0.11 level. The wire feed rate beta value (0.019) was significant at
the 0.001 level and the open voltage beta value (0.048) was significant
at the 0.01 level. It was found that the prediction of mass concentration
was not to be predicted by the gas jet pressure variable. The variability
in the data brought about by the replication factor was ignored in this
3-55
-------
analysis because it was not a controlled independent variable.
The analytical method of investigating the aptness of the model
developed by the SAS GLM procedure by plotting a standardized residual
was presented in Appendix D-5. The difference between the observed out-
come and the predicted outcome - the residual - at a collected data point
was expected to be zero assuming a perfect fit of the model. A real
model was expected to have residuals scattered evenly about zero for
all the data points used to determine the model.
Plots of the independent variables against the standard residual
were expected to evaluate nonlinearity of the regression function: the
model. Plots were shown in Appendix D-5 (Figures 12 and 13) and were
interpreted as showing that the model, indeed, fitted a linear function.
A plot of the predicted values of the model against the standard
residual was expected to evaluate nonconstancy of error variance. If
the standard residual increased or decreased as the predicted value
changed, the error variance was considered not to be constant through*-
out the model's predictive range. As shown in Appendix D-5 (Figure 14),
this was the case with the predictive model determined by the SAS GLM
procedure. The implication of this fact was that the error variance was
larger when the predicted value was larger; this meant that the variability
of gas stream mass concentration can be expected to be greater when
larger mass concentrations are being considered.
A plot of the collection order with respect to the time sequence
of the data against the standard residuals was expected to evaluate
nonindependence of the error terms. The implication of this was that
the ordering of the sample collection was not random. The residual plot
3-56
-------
in Appendix D-5 (Figure 15) showed the sampling order to be random.
CONCLUSION
The Iron oxide aerosol has been characterized by count with a particle
size mean of 0.4 ym. The inherent limitations in this analysis, e.g. choice
of magnification, nonnormality of sample distribution, indicates that the mean
is probably less than 0.4 ym. It should be noted that most of the particles
were measured to be less than 0.1 ym.
The iron oxide fine particles that formed by agglomeration into the
aerosol particles have been characterized to have means ranging from 5.4nm
to 9.1 nm (0.0054 ym to 0.0091 ym) . The size means and size distributions
of samples at given operating conditions were found to be equal. This im-
plied that a specific operating set up generated a particular population of
fine particles. Differences in size means and size distributions between
sets of operating conditions indicated that changing operation variables
changed the characteristics of the population of fine particles generated.
The simple comparative test for discerning the effect on size of
changing the exhausting rate of the arc chamber was inconclusive. The ag-
glomerated fine particle aerosol size was found not to change with an increase
in exhausting rate. However, the fine particle mean diameter size was
found to be greatly affected by the increase in exhausting rate. The evi-
dence of this change in generating properties indicated a need for further
experimentation.
The analysis of the production of the mass concentration of the aerosol
showed that a linear expression aptly fitted the data. The predictive model
showed the mass concentration to be a function of the wire feed rate and the
open voltage across the electrodes. The model predicted a range of 1.0 to
3-57
-------
1.8 gm/m3 with a concurrent increase of wire feed and voltage. The gener-
ation of a particular mass concentration should therefore be predictable
given adequate preliminary characterization for at least this material.
With respect to possible suggestions for further development of the
Electrospray aerosol generator, there are two areas that require further
work. The first area coneerns the control of variables, such as the wire
feed rate, the open voltage, and the main comressed air-jet. The second
area concerns application precautions.
The wire feed delivery should be smooth - free of delivery impedance.
The wire straighteners should control the curvature of the wire for con-
sistent contact area at the arc. It was felt that the straighteners and
;/•
the pinch rollers imparted an oscillating tension to the wire and conse-
quently an oscillating arc contact area. This behavior was observed as a
sound changing in pitch during operation and as an oscillating - up and
down - motion of the wire as it left the electrode tips.
The open voltage would require very precise control if its effect
on the aerosol's particle mean size or size distribution were to be inves-
tigated. The compressed air-jet that atomizes and quenches the melted and
vaporized metal should be studied to ascertain its effect on the aerosol's
particle size mean and distribution. Possibly, variables related to the
agglomeration of the very fine particles could be studied. Also the effect
of the temperature of the particle laden gas stream from the barrel to the
sampling port could be studied.
In the application of this aerosol generator, an on-line continuous
mean size and size distribution analyzer, such as a Whitby-Liu mobility
analyzer, would provide the necessary before control equipment and after
3-58
-------
control equipment data in an engineering evaluation of a piece of air
pollution control equipment. Control of the separation of the larger
particles within the barrel could also be used to help approximate real
life circumstances.
In conclusion, the electric arc generator produced an iron oxide
aerosol of agglomerated fine particles in mass concentration averaging
3
1.4 gm/m . The generator showed controllability of mass concentration and
variability of fine particle populations with respect to operating variables.
The resultant conclusion of this report is that the generator is controllable
and, with detailed characterization of the aerosol below 0.5 ym that the
generator would fill a much needed service in production of standard, fresh
aerosols for engineering evaluations of bench or pilot scale air pollution
control equipment and source sampling (stack testing) equipment.
3-59
-------
BIBLIOGRAPHY
1. Amson, .1. D.; "Electrode Voltage in the Consumable-Electrode
Arc System." Journal of Physics D; Applied Physics. Vol. 5, 1972.
2. Davies, C. N.; Aerosol Science, Academic Press, 1966
3. Foment!, M., Julliet, F., Meriavdeau, P., Teichner, S. J.,
and Vergnon, P.; "Preparation in a Hydrogen-Oxygen Flame of Ultrafine
Metal Oxide Particles" in Aerosol and Atmospheric Chemistry, edited by
G. M. Hidy, Academic Press, 1972, pp. 45-50.
4. Grassel, Eugene E.; "Aerosol Generation for Industrial. Research
and Product Testing," in Fine Particles; Aerosol Generation, Measure-
ment, Sampling, and Analysis, edited by Benjamin Y. H. Liu, Academic
Press, 1976, pp. 145-173.
5. Hedden, Robert; Electric Arc Generation of Polydispersed
Iron Oxide Aerosol in an Air Stream, problem report submitted to West
Virginia University, Morgantown, West Virginia, 1972.
6. Holmgren, J. D., Gibson, J. 0., and Sheer, C.; "Some Character-
istics of Arc Vaporized Submicron Particles, " Journal of the Electro-
chemical Society, Vol. Ill, No. 3, March 1964, pp. 362-369.
7. Linsky, B., and Smith, G,; "Sources of Information on Air
Pollution," Proceedings, National Conference on Air Pollution, Washington,
D. C., November 1958.
8. Magill, P. L., Holden, F. R., and Ackley, C; Air Pollution
Handbook, McGraw-Hill Book Company, 1956, section 6.
9. Miller, I., and Friend, J. E.; Probability and Statistics for
Engineers, Prentice-Hall, Inc., 1965.
10. Naylor, Mike; Very FineParticle Generation by Electric,Arc;
Sampling and Analysis, problem report submitted to West Virginia University,
Morgantown, West Virginia, 1976.
11. Neter, J. and Wasserman, W.; AppliedLinear Statistical Models,
Richard D. Irwin, Inc., 1974.
12. Pfender, B. and Boffa, C. V.; "Generation of Ultrafine Aerosols
with a Transpiration Cooled Anode in a High Intensity Arc," The Review
of Scientific Instruments, Vol. 41, No. 5, May, 1970, pp. 655-657.
13. Research Appliance Company; "Model 2443 Staksampler: Description,
Operating Instruction, and Procedures," Gibsonia, Pa., 1972.
14. Rohrs, Lloyd; "Metal-Fume Fever from Inhaling Zinc Oxide," AMA
Achieves of Industrial Health. Vol. 16, No. 1, July 1957, pp. 42-47.
3-60
-------
15. SAS INSTITUTE, INC.; A USER'S GUIDE TO SAS 76. Raleigh, North
Carolina, 1976.
16. Sears, D. R.; "Fresh Metal Fume Project-2," Laboratory Note
Book at West Virginia University, Morgantown, West Virginia 1975.
17. Shimps, Richard; Walter McCrone Associates, private communi-
cations from April to December 1976.
I
18. Waggoner, A. P., and Charlson, R. J.; "Measurements of Aerosol
Optical Parameters," in Fine Particles: Aerosol Generation, Measure-
ment, Sampling and Analysis, edited by Benjamin Y. H. Liu, Academic Press,
Inc., 1976, pp. 511-535.
19. Wall Colmonoy Corporation; "Model UT-500 Colmonoy Electrospray
Metallizer, Direction for Operating," Detroit, Michigan, 1973.
20. Western Precipitation Group - Bulletin WP-50; "Methods for
Determination of Velocity Volume, Dust, Nd Mist Content of Gases,"
Joy Manufacturing Company, Los Angeles, California.
3-61
-------
APPENDICES
3-62
-------
Appendix A
Assuming that a particle is spherical and that its Reynolds number is
small enough to indicate laminar flow, it is possible to calculate the
terminal velocity of the particle by the following expression:
d
Vt
where d is the diameter of the particle, p is the density of the particle,
Pf is the density of the fluid, g is the acceleration due to gravity, and
Vi is the dynamic viscosity of the fluid.
Assuming a particle of diameter 0.5 ym with a unit density of
3
(Igm/cm ) in an atmospheric fluid (20 *C,1 atmosphere), the formula yields
V = (5 x 10~7m)2(l x 103Kg/m3- 1.2kg/m3)(9.81m/s2)
t 1.82 x 10~3Kg/m s
V = 1.4 x 10~ m/s = 0.48 cm/hr
V =0.19 in/hr =4.58 in/hr
V • d p
DV
R | . (1.4 x 10"6m/s)(5 x 10"7m)(l x lQ3Kg/m3)
1.82 x 10~3 Kg/m s
R # = 3.8 x 10~7
A particle with above assumptions will have a terminal velocity of less
than one half centimeter per hour; this particle has a Reynolds number small
enough to indicate laminar flow; this satisfied the second assumption
necessary for use of this terminal velocity expression.
3-63
-------
Appendix B
The mass of an assumed spherical particle is given by the following
expression:
M = p (4Trr3/3)
3
M = pird /6
where r is "the radius of the particle, p is the density of the particle,
and d is the diameter of the particle.
The ratio of the masses of two spherical particles with different
diameter but equivalent densities is given by:
M = M1/M2
M - (Pird13/6)/(pTrd23/6)
M = dx3/d23
with the assumption of equal density spherical particles with diameters
of 0.1 pm and 0.01 ym, the larger particle is 1000 times heavier than the
smaller particle at the same point on the Earth's surface.
3-64
-------
Appendix C-l Particle Sizing Pictures
The following pictures were presented to show the SEM pictures
used to measure the agglomerated particles.
Figure 9. SEM picture sample 715
Figure 10. SEM picture sample 714
3-65
-------
Appendix C-2 Sizing of the Agglomerates
The agglomerated particles were measured with a clear plastic
metric ruler. The photomicrographs were enlarged such that a measure-
ment of 2 mm would approximately correspond to 0.2 ym. The maximum
horizontal - across the photograph - and the maximum vertical distance a
particle covered were measured and averaged to give an average particle
diameter. With known magnification, the actual size was determined by
the following expression:
,,,,,. Image Size
Real Size « e
Magnitude of Magnification
Groups ranging from less than 2 mm to greater than 30 mm by increments
of 2 mm were used in counting the number of particles within a range.
Class marks were assigned to each group interval. The raw data is pre-
sented in Table 28. The SAS MEANS procedure determined the statistics
presented in Table 18.
The photographs from the SEM negative were not exceptionally clear;
that is, the particles did not have sharp edges. This effect would be
a factor in both sample measurements; therefore it was felt the results
were not affected by this extraneous variable.
3-66
-------
Table 28. Raw data for agglomerated particle count
Sample #
Magnifier (1) Negative
(2) Enlargement
Less than 2mm
2 to 4 mm
4 to 6 mm
6 to 8 mm
8 to 10 mm
10 to 12 mm
12 to 14 mm
14 to 16 mm
16 to 18 mm
18 to 20 mm
20 to 22 mm
22 to 24 mm
24 to 26 mm
26 to 28 mm
28 to 30 mm
Greater than 30 mm
Class Mark
1 inni
3 mm
5 mm
7 mm
9 mm
11 mm
13 mm
15 mm
17 mm
19 mm
21 mm
23 mm
25 mm
27 mm
29 mm
31 mm
714
5000
23/10.5
Count
156
41
19
17
14
8
7
3
3
6
0
0
2
1
1
4
715
5000
23/10
Count
136
56
31
23
10
7
11
4
4
3
1
0
0
1
0
8
3-67
-------
Appendix C-3 Mass Concentration Data
The mass concentration data is presented in the following three
sets of tables. The top listings in each table are some of the data
taken during sampling. The bottom listings are the calculated values
for determining the mass concentration (STP) and indicator of isokenitic
conditions (10,13).
After finding the net weight of a mass sample, STP mass concen-
tration can be found by dividing the ratio of the mass collected to the
volumetric flow through the filter by a correction factor sc. This
factor compensates for the^ differences in temperature and pressure re-
lative to standard temperature (298*K) and pressure (29.92 in.) by the
following formula:
v v PMeter 298 °K
STD ~ Meter 29.92 in.X T = Meter x sc'
Meter
This calculation assumes the negligible influence of water in the duct
gas stream (10,20); the relative humidity was found to be less than 20%.
Symbols are explained in Figure 11.
The isokenitic factor was approximated by the ratio of the nozzle
velocity to the stack velocity:(!'). It was expected to provide an adequate
target for approximating isokenitic; as can be seen in the concentration
raw data tables, the samples were not taken isokenitically. The sample
nozzle velocity was mostly greater than the gas stream velocity. The
approximation (I1) is transformed into the isokenitic relation (I):
3-68
-------
l _ (TSTACK/TSTD)(l'STD/1>STACR)(Vm STD * Vw STD>
(VelST )(ATlme)(Area of Nozzle)
by multipling approximation (I1) (10) by the correction factor c which
is given in Figure 11.
3-69
-------
V is STP volumetric flow
o J.U
V is volumetric flow (cfm)
Meter
P is Atmospheric pressure at stack sampler (in.Hg)
Meter
T is Average temperature of stack sampler thermometer (*F)
Meter
ap HEPA is Ap across HEPA filter (in. H20)
Pilot Ap is standard pilot p at point of sampling (in. HJD)
SAP is secondary air supply (psi)
Stack T (TgTACK) is temperature of duct (*C)
P is gage pressure of sampling duct (in.Hg)
o JLA.wJx
Wire feed rate is 30-60 (corresponding to 37 to 86 gm/min)
MAP is main air jet pressure (psi)
Open Voltage is voltage across arc prior to operation (volts)
CF is dry gas meter volumetric flow (ft )
Time is time of sampling (sec)
Area of Nozzle is (1/4 • l/!2)2(1T/4) (ft2)
Nozzle Velocity is CFM area of nozzle (ft/min)
/29.!
V PSTJ
Stack Velocity is 234 / f^^ (273 + !__....) (PiloUp)
STACK blACK
Isokenetic approximation is N°zzl
vv
Stack Velocity
C = TSTACK Pbar „ _,_ „.
p - • ^ - • (1 + R)
barStack meters
R = % water in gas stream
Figure 11. Explanation of data sheet and mass calculation symbols
3-70
-------
Table 29- Mass data first latin square experiment
(1) Pitot AP - 0.065 Date 7/5/76
SAP = 20 psi (2) P, = 28.98 in
Sample #
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Ap- HEPA
Stack
Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
CF
Time
T
Meter
Nozzle Velocity =
Stack Velocity =
Isokenetic estimate (I1)
C
I = I' x C
gm/acf
gm/sm
ftr.
529
0.5
28.92
68
50
40
29
0.359
1 min
76
1053
1121
0.94
1.15
1.08
0.0428
1.51
0.97
546
0.5
28.92
73
40
50
31
0.412
1 min
77.5
1209
1129
1.08
1.16
1.24
0.0371
1.31
0.968
547
0.55
28.93
64
40
45
25
0.372
1 min
79.5
1091
1114
0.98
1.13
1.11
0.0315
1.11
0.964
548
0.6
28.93
74
50
45
27
0.399
1 min
80.5
1170
1131
1.03
1.16
1.19
0.0362
1.28
0.96
3-71
-------
Table 29i (Continued)
(1) Pltot AP • 0.065 Date 7/5/76
SAP - 20 psi (2) Pbar - 28.98 in Hg
Sample #
(3)Ap HEPA
<*> 'stack - (2) - (3)/13'6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
T
Meter
(8) Nozzle Velocity =
(9) Stack Velocity =
Isokenetic estimate (I1)
C
I = I1 x C
gm/acf
/ 3
gm/sm
sc
549
0.6
28.93
75
50
55
31
0.364
1 min
81.75
1068
1133
0.93
1.16
1.00
0.0334
1.18
0.96
550
0.65
28.93
72
60
40
31
0.383
1 min
83.5
1124
1127
1.00
1.15
1.15
0.0481
1.70
0.957
551
0.65
28.93
60
30
55
27
0.386
1 min
84
1132
1107
1.02
1.10
1.12
0.0228
0.81
0.956
552
0.7
28.93
53
30
45
31
0.401
1 min
87
1176
1096
1.07
1.08
1.16
0.0297
1.05
0.951
3-72
-------
Table 29. (Continued)
(1) Pitot AP = 0.065 Date 7/5/76
SAP = 20 psi
(2) p 28.98 in Hg
bar
Sample #
553
554
555
556
(3)
(4)
(5)
(6)
(7)
(8)
(9)
Ap HEPA
c»-™~i, = '*/ ~ \3J/13.6
otacK
Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
CF
Time
T ( F)
Meter v '
Nozzle Velocity =
Stack Velocity =
Isokenetic estimate (I1)
C
I = I' x C
gm/acf
/ 3
gm/sm
sc
0.7
28.93
73
40
40
27
0.389
1 min
88.5
1141
1129
1.01
1.14
1.15
0.0336
1.19
0.949
0.65
28.93
85
60
45
29
0.390
1 min
88.25
1144
1148
1.00
1.18
1.18
0.0502
1.77
0.950
0.65
28.93
76
50
50
25
0.382
1 min
88.75
1121
1133
0.99
1.15
1.14
0.0388
1.37
0.950
0.65
28.93
60
30
50
29
0.380
1 min
89
1115
1107
1.01
1.10
1.11
0.0303
1.07
0.947
3-73
-------
Table 29. (Continued)
(1) Pltot AP = 0.065 Date 7/5/76
SAP = 20 psi (2) Pfear = 28.98 in Hg
Sample #
557
558
559
560
(3) Ap HEPA
<4> PStack ' (2) - (3)/13-6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
T
Meter
(8) Nozzle Velocity «=
(9) Stack Velocity =
Isokenetic estimate (I1)
C
I » I' x C
gm/acf
/ 3
gm/sm
sc
0.65
28.93
57
30
40
25
0.370
1 min
89.75
1085
1102
0.98
1.08
1.06
0.0260
0.92
0.946
0.65
28.93
80
60
50
27
0.378
1 min
90
1190
1140
0.97
1.16
1.13
0.0493
1.74
0.946
0.65
28.93
70
60
55
25
0.400
1 min
90.25
1173
1124
1.04
1.13
1.08
0.0291
1.03
0.945
0.70
28.93
70
40
55
29
0.383
1 min
90.5
1124
1124
1.00
1.04
1.04
0.042
1.48
0.945
3-74
-------
Table 30. Mass data second latin square experiment
(1) Pitot AP = 0.06
SAP = 20 psi
Date 7/11/76
(2) P = 29.05 in
bar
Sample #
561
562
563
564
(3) Ap HEPA
(4) PStack = <2) ~ <3)/13-6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
T
Meter
(8) Nozzle Velocity =
(9) Stack Velocity =
Isokenetic estimate (I')
C
I - I1 x C
gm/acf
gm/sm
sc
0.5
29.01
71
40
40
27
0.389
1 min
80
1141
1089
1.06
1.15
1.22
0.0425
1.50
0.966
0.5
29.01
62
40
50
31
0.350
1 min
81
1026
1006
1.02
1.12
1.14
0.0543
1.92
0.964
0.5
29.01
82
50
55
31
0.340
1 min
82
974
1097
0.89
1.18
1.05
0.0611
2.16
0.962
0.5
29.01
80
60
40
31
0.352
1 min
82
1033
1094
0.94
1.17
1.10
0.0473
1.67
0.962
3-75
-------
Table 30. (Continued)
(1) Pitot AP * 0.06 Date 7/11/76
SAP - 20 psi (2) Pu - 29.05 in Hg
bar
Sample I 565 566 582 568
(3) Ap HEPA
(4) PStack * (2> - (3)/13'6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
T
Meter
(8) Nozzle Velocity =
(9) Stack Velocity =
Isokenetic estimated (If)
C
I = I1 x C
gm/acf
, 3
gm/sm
sc
0.5
29.01
77
50
45
27
0.369
1 min
80.5
1082
1089
0.99
1.17
1.16
0.0440
1.55
0.965
0.5
29.01
80
60
50
27
0.335
1 min
80
983
1094
0.90
1.18
1.06
0.0517
1.83
0.966
0.5
29.01
59
30
40
25
0.393
1 min
82
1153
1061
1.09
1.11
1.20
0.0283
1.00
0.962
0.5
29.01
57
30
45
31
0.375
1 min
82
1100
1057
1.04
1.10
1.14
0.0338
1.19
0.962
3-76
-------
Table 30. (Continued)
(1) Pitot AP = 0.06 Date 7/11/76
SAP - 20 psi (2) PI . 29.05 in Hg
Sample #
569
570
571
572
(3) Ap HEPA
(4) PStack = (2) - (3>/13'6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
Meter
(8) Nozzle Velocity -
(9) Stack Velocity =
Isokenetic estimated (I1)
C
I = I' x C
gm/acf
, 3
gm/sm
sc
0.5
29.01
58
30
55
27
0.368
1 min
82
1080
1059
1.02
1.10
1.12
0.0327
1.15
0.962
0.5
29.01
54
30
50
29
0.374
1 min
82
1097
1053
1.04
1.09
1.13
0.0319
1.13
0.962
0.5
29.01
58
40
55
29
0.357
1 min
82.25
1047
1059
0.99
1.10
1.09
0.039
1.38
0.961
0.5
29.01
65
40
45
25
0.361
1 min
83
1059
1070
0.99
1.12
1.11
0.0346
1.22
0.960
3-77
-------
Table 30. (Continued)
(1) Pitot AP = 0.06
SAP » 20 psi
Date 7/11/76
(2) Pbar - 29.05 in Hg
Sample #
573
581
575
576
(3) Ap HEPA
(4) PStack - (2> - (3>/13-6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
TMeter
(8) Nozzle Velocity «
(9) Stack Velocity =
Isokenetic estimated (I1)
C
I - I1 x C
gm/acf
/ 3
gm/sm
sc
0.5
29.01
68
50
40
29
0.360
1 min
83.5
1056
1074
0.98
1.13
1.11
0.0406
1.43
0.959
0.5
29.01
80
60
55
25
0.309
1 min
83.75
906
1094
0.83
1.17
0.97
0.0513
1.81
0.959
0.5
29.01
76
60
45
27
0.372
1 min
84
1091
1087
1.003
1.16
1.16
0.0392
1.38
0.958
0.5
29.01
76
50
50
25
0.356
1 min
84
1044
1087
0.96
1.16
1.11
0.0413
1.46
0.958
3-78
-------
Table 31. Mass data orthogonal experiment
(1) Pitot AP = 0.05
SAP = 20 psi
Date 8/9/76
(2) Pbar =28.91 in Hg
Sample #
700
701
702
703
(3) Ap HEPA
(4) PStack - (2> - <3>/13'6
(5) Stack T ( C)
Wire Feed Rate
Map
Open Voltage
(6) CF
(7) Time '
T
Meter
(8) Nozzle Velocity =
(9) Stack Velocity -
Isokenetic estimate (I1)
C
I = I' x C
gm/acf
/ 3
gm/sm
sc
0.4
28.88
70
45
50
30
0.236
77/100
72
986
899
0.91
1.16
1.06
0.04
1.41
0.975
0.4
28 . 88
72
45
40
28
0.236
72/100
72
989
961
0.97
1.17
1.13
0.0408
1.44
0.975
0.4
28.88
90
55
50
30
0.165
50/100
72
1015
968
0.95
1.23
1.17
0.0603
2.13
0.975
0.4
28.88
72
45
40
30
0.216
71/100
72
989
892
0.90
1.17
1.05
0.0428
1.51
0.975
3-79
-------
Table 31. (Continued)
(1) Pitot AP = 0.05
SAP - 20 psi
Sample //
(3) Ap HEPA
(4) PStack= (2) - (3)/13'6
(5) Stack T ( C)
Wire Feed Rate
MAP
Open Voltage
(6) CF
(7) Time
T
Meter
(8) Nozzle Velocity =
(9) Stack Velocity =
Isokenetic estimate (I1)
C
I = I' x C
gm/acf
. 3
gm/sm
sc
Date 8/9/76
(2) P = 28.91 in Hg
bar
704
705
706
707
0.4
28.88
80
45
50
28
0.217
60/100
72
948
1061
1.12
1.20
1.34
0.043
1.52
0.975
0.4
28.88
95
55
40
30
0.205
72/100
72
970
835
0.86
1.25
1.08
0.0485
1.71
0.975
0.4
28.88
85
55
40
28
0.236
67/100
72
1008
1033
1.03
1.21
1.25
0.0408
1.44
0.975
0.4
28.88
85
55
50
28
0.192
65/100
72
1008
867
0.86
1.21
1.04
0.0481
1.70
0.975
3-80
-------
Appendix D-l Latin Square
The Latin Square Design is considered an economic, balanced design.
Extraneous variables can be tested for any effect on the dependent
variable in this design. The variability produced by the extraneous
variables is separated from the variability produced by the main indepen-
dent variables (9). A thorough presentation is included in references
9 and 11.
A F-statistic is used to test the null hopothesis that the effect
of the variable is equal to zero. The independent, extraneous, and a
replicate statistic are found. Table 32 presents short cut formulae and
definitions of terms necessary for the calculation of the F-statistic (9).
This statistic is compared to a tabulated F-value with the specified
degrees of freedom and confidence limit. If the calculated F-value is
greater than the chosen tabulated F-value, then the effect of the variable
is not equal to zero.
Table 33 shows calculation table and Latin Square Summary table.
Tables 24 and 25 show the equivalence of the SAS ANOVA procedure for
analysis of variance and the short cut Latin Square procedure; the F values
are essentially equal.
3-81
-------
Table 32. Formula and definitions for latin square
analysis of variance
Source of
Variation
Treatments
(wire feed rate)
Rows
Columns
Replicates
Error
Total
Degrees of
Freedom
n-1
n-1
n-1
r-1
(n-1) (rn+r-3)
rn2-!
Sum of
Squares
SS(TR)
SSR
SSC
SS(Rep)
SSE
SST
Mean
Square
SS(TR)
n-1
SSR
n-1
SSC
n-1
SS(Rep)
r-1
F
MS(TR)
MSB
MSR
MSB
MSC
MSE
MS (Rep)
MSE
SSE
(n-1) (rn+r-3)
C - (T ...)2/r-n2
SS(V
- c
k=l
SSR = — J T2 - C
r«n E i..
r » Number of replicates
n = Number of factors
SSC = — J T2 - C
r«n Z j..
1=1
SST
-C
SSE = SST - SS(TR) - SSR - SSC - SS(Rep)
3-82
-------
Table 33. Calculation table for latin square design
Replicate 1 (gm/sm )
MAP (psi)
Replicate 2 (gm/sm )
MAP (psi)
40
45
50
55
40
45
50
55
to
I
oo
01
4-1
o
>
25
27
29
31
A
0.92
B
1.19
C
1.51
D
1.70
B
1.11
C
1.28
D
1.77
A
1.05
C
1.37
D
1.74
A
1.07
B
1.31
D
1.03
A
0.81
B
1.48
C
1.18
4.43
5.02
5.83
5.24
w
4->
c-H
O
O
>
25
27
29
31
A
1.00
B
1.50
C
1.43
D
1.67
B
1.22
C
1.55
D
1.38
A
1.19
C
1.46
D
1.83
A
1.13
B
1.92
D
1.81
A
1.15
B
1.38
C
2.16
5.49
6.03
5.32
6.94
5.32
5.21 5.49 4.50
5.60 5.34 6.34 6.50
Wire Feed (Rate %)
A-30
B-40
C-50
D-60
Rep 1
3.85
5.09
5.34
6.24
Z Rep 2
4.47
6.02
6.60
6.69
ET... 44.30
-------
Appendix D-2 T Test Procedures'
The student t-test is based on a random variable having the student-t
distribution with n-1 degrees of freedom. Where n is the sample, x is
the sample mean, s is the sample standard deviation, and >n is the expected
mean of a normal population, the value of the t-statistic is given by
The testing of the hypothesis concerning two means with unknown population
variances is based on the t-statistic. If the variances are equal by the
F-test, the t-statistic that is used for decision-making is the Student-t.
The t-statistic for comparing two means is found with the formula
x \ »
C3.X S"~
xp
where
i Kn2(ni+ v 2)
Sxp (n + n )[(n -1)S2 +
(1)
£
|
2 | m
1 (n ~1)S 1 J
The value of 6 is set equal to zero when testing the hypothesis that
the means are not different. If the variances are unequal, then the Smith-
Satterthwaite t-statistic is used (9) where
(3)
3-84
-------
and
S2 S2 (S2/n )2 (S2/n )
nl U2 ' V1 + V1
The equivalent mean hypothesis tests in this report are based on
the assumption that 6=0.
V V v2 - « - o
H. : V.- Vn X 0
A 12
The criterion for rejecting H , where ta/2 is equal to the tabulated
t-value with (n..+ no~2) degrees of freedom and Type I error equal to a,
is given by
Reject H : V-- V * 0
if t < - t a/2
cal
or t , > t a/2 (5)
cal
2 2
The F-statistic given by F = 8.^82 is used to determine equality
of sample variances. If F ..is less than a tabulated F statistical, then
the variances are equal. If the variances are unequal, then the Smith-
Satterthwaite t-statistics are used in the hypothesis test.
The following are example Agglomerate T Test Calculations.
Given: x - 0.41 ym s = 0.53 jam ^ = 300
x - 0.40 ym s - 0.54 ym n = 282
£• £* £•
3-85
-------
with equation (1)
0.41 ym - 0.40 urn
f
»
(300) (282) (300+282-2) _ 11/2
(300+282)[ (294) <0.53)2+(281)(0.54)2] J
t . = 0.225 df - 300 + 282 - 2 - 580
cal
with equation (3)
, _ 0.41 ym ~ 0.40 ytm
cal '
./(0.53)2 (0.54):
V 300 282
t' = 0.225
cal
2 2 22 22
from equation (4)
... f(0.53)2 (0.54) . r (0.52)/300) (0.54)/282)
dt ~ \ 300 * 282 / ' l 299 * 281 J
df - 576.3
In either case, equivalent variances or not, the t , value is 0.225.
cal
The df are approximately 580; this degree of freedom for all practical
calculations is equal to infinity. With the tabulated t value (alpha
equals 0.10 and df equals infinity) being 1.645, the hypothesis test (5)
shows the means not different.
In Table 34, the SAS T Test Calculates both the general t-statistic
and the Smith-Satterthwaite t-statistic. The differences in the calculated
by hand and calculated by computer t-values are a consequence of rounding
and truncating errors.
3-86
-------
Table 34. SAS T test table
Variable: X, agglomerated particle diameter.
REPP
1
2
N
300
282
Mean
0.41
0.40
STD Dev.
0.53
0.54
STD Error
0.036
0.032
Minimum
0.087
0.091
Maximum
2.61
2.74
Variances
Unequal
Equal
DF
PROB>:T!
0.1938 575.5 0.8464
0.1939 580.0 0.8463
For HO: Variances are equal. F1 - 1.05 with 281 and 299 DF
PROS > F' = 0.565
3-87
-------
Appendix D-3 ANOVA
The SAS ANOVA procedure uses a matrix manipulation technique to
determine the analysis of variance in a collection of samples. The
procedure finds the analysis of variance sum of squares for each variable
in the selected model as well the sum of squares of the error in the data.
A comparison of the average drop in the sequential sum of squares to the
average drop in the sum of squares of the error terms is used to determine
if the independent variable helps explain the variability of the data.
The F-statistic in formula expression is
F - SSR/p-1
SSE/n-p
where SSR is the sequential sum of squares for the particular variable,
SSE is the error sum of squares for the data, p is the number of variables
in the selected model, and n is the number of samples. This statistic
is shown to the right of the variable names in Table 25. The ANOVA pro-
cedures also calculates the F-statistic associated with examining the
data fit of the model. This statistic is shown to the right of the model
source in Table 23. The term below this statistic, PR>F, gives the
probability that the model is fitting noise. Other statistics are shown
in the ANOVA table.
3-88
-------
Appendix D-4 Stepwise Procedure
One of the search methods for arriving at a "best" set of inde-
pendent variables from the total variables available is the stepwise
regression method (15,11). This set of variables will best describe
as well as predict the relation between the independent and dependent
variables. The stepwise procedure determines the variables which have
an observed effect on the outcome at a certain level of significance.
The variable which explains most of the variability in the data is first
added to the set of best independent variables.
After the first variable is chosen, a second best variable is chosen
(with the assumption that the first variable explained a portion of the
variance in the data), and then a statistical test determines if the
first variable (actually, any and all prior variables chosen) should be
eliminated from the set of "best" variables. If both of the choices are
kept in the set, the procedure continues to add to and check for deletion
of best variables until all the variables in the set are "best" with a
certain confidence. The SAS default confidence is 90%. The deleted
variables do not significantly explain the outcome therefore they are
dropped.
The SAS STEPWISE procedure calculated the best variables that des-
cribe (predict) the collection of mass concentration data. The analysis
printout format is presented in Table 35. This stepping procedure took
four steps. The wire feed rate variable was added first and the best
variable set contained the wire feed rate and open voltage variables.
3-89
-------
Table 35. Stepwise regression procedure for dependent variable Y, mass concentration
STEP 1 Variable WF Entered
R Square =0.43
u>
VO
o
Regression
Error
Total
Intercept
WF
STEP 2 Variable V Entered
Regression
Error
Total
Intercept
WF
V
DF
1
38
39
B Value
0.52
0.02
R Square
DF
2
37
39
B Value
-0.81
0.019
0.048
Sum of Squares
1.69
2.21
3.91
STD Error
0.0037
= 0.54
Sum of Squares
2.09
1.81
3.91
STD Error
0.0034
0.017
Mean Square
1.69
0.06
Type II SS
1.69
Mean Square
1.05
0,05
Type II SS
1.63
0.4
F PROB>F
29.1 0.0001
F PROB>F
29.1 0.0001
F PROB>F
21.4 0.0001
F PROB>F
33.3 0.0001
8.1 0.0070
-------
Table 35. (Continued)
STEP 3 Variable P Entered
R Square = 0.543
vo
Regression
Error
Total 39
Intercept
WF
V
P
STEP 4 Variable P Removed
Regression
Error
Total
Intercept
WF
V
DF
3
36
39
B Value
-1.05
0.019
0.048
0.0047
R Square
DF
2
37
39
B Value
-0.81
0.01
0.04
Sum of Squares
2.12
1.79
3.91
STD Error
0.003
0.017
0.006
- 0.53
Sum of Squares
2.09
1.81
3.91
STD Error
0.003
0.017
Mean Square
0.71
0.090
Type II SS
1.6
0.40
0.03
Mean Square
1.05
0.04
Type II SS
1.6
0.40
F PROB>F
14.3 0.0001
F PROB>F
-
33.1 0.0001
8.2 0.0070
0.56 0.4605
F PROB>F
21.4 0.0001
F PROB>F
33.3 0.0001
8.1 0.0070
All variables in the model are significant at the 0.1000 level.
-------
Appendix D-5 GLM and Residual Plots
The SAS GLM (General Linear Model) procedure functions much as the
SAS ANOVA procedure. The GLM procedure is more powerful than the ANOVA
procedure because it can be applied to virtually any data set; the ANOVA
procedure requires the data to be balanded. Since the GLM procedure
outputs the needed statistics, the GLM procedure was used to calculate
residuals for standardized residual plots.
A residual is the difference between the predicted out come from
the regression model developed from the observations and the observed
out come at the same point. A standardized residual is the residual at
a point divided by the square root of the mean sum of squares (the
standardized deviation) of the error. The standardized residuals from
a well fitting model should be evenly dispersed about the zero residual
line because they are normally distributed variables if the model ad-
equately describes the independent-dependent variable relationship (11).
The GLM table is presented in Table 36. The model F-statistic
value indicated that the model is fitting valid independent variable
induced variability. The F-value for the wire feed rate and open voltage
indicates that these variables significantly influenced the outcome. The
next table presents the parameter (Bi) estimates for the linear model:
Y = BQ + Bj^ (WF) + B2 (V). The t-statistic for acceptance of the beta
parameter is indicative of the confidence in the predictive ability of
the variable. The next table shows the observed data, the predicted data
using the model and beta parameters, and the residuals. The standardized
residuals are plotted against various variables in Figures 12, 13, 14, and
15.
3-92
-------
Table 36. SAS GLM table including residuals
Dependent Variable Y, mass concentration
Source
Model
Error
Corrected Total
CO
VO
CO
Source
WF
V
DF
1
1
Type I SS
1.69
0.40
Parameter
Intercept
WF
V
DF Sum of Squares Mean Square
2 2.09 1.05
37 1.81 0.04
39 3.90
R-Square C.V.
0.54 15.5
STD Dev. Y Mean
0.22 1.43
F Value
21.4
PR > F
0.0001
F Value
34.6
8.1
PR > F
0.0001
0.0070
DF
1
1
Type IV SS
1.63
0.40
F Value
33.3
8.1
PR > F
0.0001
0.0070
timate
-0.81
0.02
0.048
T for HO:
Parameter = 0
-1.64
5.77
2.85
PR > !T!
0.1085
0.0001
0.0070
STD Error
Estimate
0.49
0.003
0.017
-------
Table 36. (Continued)
5ERVATION
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
OBSERVED
VALUE
1.51
1.31
1.11
1.28
1.18
1.70
0.81
1.05
1.19
1.77
1.37
1.07
0.92
1.74
1.03
1.48
1.50
1.92
2.16
1.87
1.55
1.83
1.00
1.19
1.15
1.13
1.38
1.22
1.43
1.81
PREDICTED
VALUE
1.54
1.44
1.15
1.44
1.64
1.83
1.06
1.25
1.25
1.73
1.35
1.15
0.96
1.64
1.54
1.35
1.25
1.44
1.64
1.83
1.44
1.64
0.96
1.25
1.06
1.15
1.35
1.15
1.54
1.54
RESIDUA
-0.03
-0.13
-0.04
-0.16
-0.460
-0.13
-0.25
-0.20
-0.06
0.03
0.02
-0.08
-0.04
0.10
-0.51
0.13
0.25
0.48
0.52
-0.16
0.11
0.19
0.04
-0.06
0.09
-0.02
0.03
0.07
-0.11
0.27
3-94
-------
Table 36. (Continued)
liKVATJLUN
31
32
33
34
35
36
37
38
39
40
OBSERVED
VALUE
1.38
1.46
1.41
1.44
2.13
1.51
1.52
1.71
1.44
1.70
PREDICTED
VALUE
1.73
1.35
1.49
1.40
1.68
1.49
1.40
1.68
1.59
1.59
RESIDU
-0.35
0.11
-0.08
0.04
0.45
0.02
0.12
0.03
-0.15
0.11
Dependent Variable: Y
Sum of Residuals 0.00
Sum of Squared Residuals 1.81
Sum of Squared Residuals - Error SS -0.00
First Order Autocorrelation 0.15
Durbin-Watson D 1.69
3-95
-------
U)
1
2
S
JF
N i
D
A
R
D
R
E
S
I
•D
u
A
L
O
o
© ©
O
° • & O
-------
3
2
S
T
D L
A
R
D
w 0
i n
vo «•
"•si tj1
S
I
D -1
U
A
L
-2
-3
*
Q LEGEND:
Q O One
O Observation
4. Two
Observations
i
° 0
*" -
(D
0 © A ®
§ S .,
O O ^ ^
^ S
e o °
^ o ©
r
O
0
* 9
e
* A » •• • t * •
25
26
27 28
OPEN VOLTAGE
29
30
31
Figure 13 Graph: Standard Residual versus Open voltage
-------
u>
1
VO
oo
3
2
S
T
N
D
A
R
D
R °
E
S
I
D1
-L
U
A
L
-2
-3
0 LEGEND:
0 One
- - - -• ® O Observation
& Two
Observations
©
o
" - "•*•
o
o o © & ®
© ° <* ^ ® « «
& . Q . . P. .O. . . .
®® /a
® O * 6> S
© ° °
©
O
-
O
1* . * - • 1 . 1 II 1
1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
PREDICTED OUTCOME (YHAT)
1.9
Figure 14. Graph: Standard Residual versus Predicted Outcome(Yhat)
-------
3
2
S
T
A
N
D L
A
R
D
CO n
' R U
VD K.
SO E
S
I
D _±
A
L
b
e
0 e
e
»
• . e
o o
o
C3 .4*.
0
O
k • - • -
^
€
. 0 .
© «
o
LEGEND:
©One
Observation
£i. Two
Observations
O
O
0
O
8
-3
1 2 3 4 56 7 8 9 10 11 12 13 14 15 16
ORDER OF PRESENTATION
Figure 15. Graph: Standard Residual versus Order of Presentation
-------
Appendix E
The wire feed rate is a unitless dimension on the metallizer unit.
The rate of feed is dependent on the wire feed rate setting and the
amount of material that is being feed. Time trials at various settings
with Walcomonoy #10 wire were used to graph a wire feed rate. This
graph is shown in Figure 16.
Using the developed equation for the converted mass of the aerosol:
Y = -0.81 + 0.01 (WF) + 0.048 (V)
a relation between increased metallizer wire feed rate and the percentage
of converted mass can be found. Assuming the open voltage to be 31 volts
3
and a ductwork volumetric flow of 5.6 m /min., the predicted percent
conversion for the various wire feed rates is given by the following
expression:
J rmwrainT. . 1-0-81+0.048(31) + 0.01 (WF)] 5.6
/o conversion = •* 1 . Q^ .—•=—-—*—"——— x ±00
mass input @ wire feed rate, WF
Table 37. Relating wire feed rate and percent of mass converted
Wire Feed
Rate
30
40
50
60
Percent
Conversion
14.8%
11.6%
9.6%
8.3%
3-100
-------
gm
mln
90
80
70
60
50
40
30
20
10
30
40
50
60
WF
Figure 16. Wire Feed Rate Calibration
3-101
-------
TECHNICAL REPORT DATA
(Phase read luanictions on the reverse before complctin/:)
1. NtPORT NO.
2.
4. TITLL AND SUOTITt.fc
Generation and Simulation of Metallic Particulate
Air Pollutants by Electric Arc Spraying
3, RECIPIENT'S ACCESSION NO.
B."REPO"RT DATE
September 1977
6. PERFORMING ORGANIZATION CODE
7. AUTHOHIS)
B.Linsky, R.Hedden, M.Naylor, and F.Dimmick
8. PERFORMING ORGANIZATION REPORT NO.
9, PERFORMING ORGANIZATION NAME AND ADDRESS
University of West Virginia
Morgantown, West Virginia 26506
10. PROGRAM ELEMENT NO.
1AB012; ROAP 21ADM-025
11. CONTRACT/GRANT NO.
Grant R801858
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
Grant Final; 7/73-1/75
COVERED
14. SPONSORING AGENCY CODE
EPA/600/13
15. SUPPLEMENTARY NOTES IERL-RTP project officer for this report is Dennis C. Drehmel,
Mail Drop 61, 919/541-2925.
16. ABSTRACT The report gives results of efforts to provide a generated output with an
appropriate mass and concentration of fresh, dry, fine metal oxide particles for bench
or pilot scale fine particulate collection research and development work. The work
involved two electric arc aerosol generators: one using a single consumable electrode
of welding wire; the other, two cornsumable wire electrodes of a commercially avail-
able electric arc metallizer. The generated aerosols were exhausted into a duct sys-
tem and sampled using membrane filters. The single electrode generator produced
0.67 g/cu m of 0.1 micrometer diameter iron oxide particles as sampled by an Ander-
sen Stack Sampler. The mass emission rate with an average of 1.95 g/min varied
within a + or - 12% range. The double electrode generator produced submicron parti-
cles (measure by scanning electron microscopy). Mass volumetric concentrations
ranged from 0.7 to 2.0 g/cu m for zinc oxide aerosols. The mass emission rate aver-
aged 8.6 g/min for the zinc oxide aerosols and 7.7 g/min for the iron oxide aerosols.
(The zinc oxide wasZnO andlhe Iroh oxide,TFe3O4.TThe double electrode generator
was further tested and validated for reproducibilities of total mass volumetric concen-
tration and basic particle diameter distributions. Variables of operation were inves-
tigated to determine their effect on the mass volumetric concentration of the "aerosol.
7.
KEY WORDS AND DOCUMENT ANALYSIS
a.
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Air Pollution
Simulation
Aerosols
Metal Powder
lectric Arcs
Metallizing
Zinc Oxides
Iron Oxides
Air Pollution Control
Particulate
Electric Arc Spraying
Metallic Oxide Particles
Particle Collection
13B
14B
07D
11F
20C
13H,11C
07B
3. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (ThisReport)
Unclassified
21. NO. OF PAGES
334
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
3-102
------- |