Environmental Protection Technology Series
EVALUATION OF STATIONARY SOURCE
PARTICULATE MEASUREMENT METHODS
Volume II. Oil-Fired Steam
Generators
ST
<
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Re&earch reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and application of
environmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the ENVIRONMENTAL PROTECTION
TECHNOLOGY series. This series describes research performed to develop and
demonstrate instrumentation, equipment, and methodology to repair or prevent
environmental degradation from point and non-point sources of pollution. This
work provides the new or improved technology required for the control and
treatment of pollution sources to meet environmental quality standards.
EPA RE VIEW NOTICE t
s
f
This report has been reviewed by the U.S. Environmental
Protection Agency, and approved for publication. Approval
does not signify that the contents necessarily reflect the
views and policy of the Agency, nor does mention of trade
names or commercial products constitute endorsement or
recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-77-026
February 1977
EVALUATION OF STATIONARY SOURCE PARTICULATE
MEASUREMENT METHODS
Volume II Oil-Fired Steam Generators
by
Edward T. Peters and Jeffrey W. Adams
Arthur D. Little, Inc.
Cambridge, Mass. 02140
Contract No. 68-02-0632
Project Officer
Kenneth Knapp
Emissions Measurement and Characterization Division
Environmental Sciences Research Laboratory
Research Triangle Park, NC 27711
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U. .S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
-------
ABSTRACT
An experimental study was conducted to determine the reliability of the
Method 5 procedure for providing particulate emission data from an oil-
fired steam generator. The study was concerned with determining first,
whether any "false" particulate, e.g., particulate formed as a result of
the collection process, was associated with Method 5 sampling and second,
the influence on collected particulate, if any, associated with parametric
variations to Method 5, including filter media, size and temperature, nozzle
type and sampling duration. Tests were conducted at a 350 MW source
utilizing low sulfur residual fuel. Stack gases contained approximately
200 ppm S02 and 1-2 ppm SOs. A first series of field tests included
simultaneous sample collection with two trains, one a Method 5 train and
the other modified to incorporate a flat in-stack filter; the latter was
employed as a control with the former being varied with respect to filter
media (glass fiber versus quartz), filter temperature (120°C versus 150°C)
and sampling duration (1 versus 2 hrs). Additional runs were performed
in which one or both trains were spiked with S02 at the nozzle to provide
sampled gas stream concentrations of -2000 ppm S02 as an approximation
of the combustion gases from a high sulfur fuel. By comparing particulate
and sulfate distributions between probe and filter catches for the two
trains, it was determined that there was no evidence for the formation of
"false particulate" for the conditions evaluated. Furthermore, it was
found that the in-stack train particulate catch (nozzle and filter)
was approximately 20 percent higher than for the Method 5 train; the
data for Method 5 train indicated the need for a blank correction of
-10 mg/m3. These observations are believed to be the result of an inability
to make a quantitative recovery of the probe catch. A second set of
twelve statistically designed, paired experiments was carried out with
Method 5 trains varied with respect to filter size and temperature and
nozzle type. An analysis of variance performed on the data indicated
that the nozzle type and filter temperature-size interaction can have a
statistically significant result, corresponding to a variance in result
of 7 mg/m3 out of a total average catch of 31 mg/m3. Most of the standard
error in the measurement, corresponding to 5 mg/m3, is associated with
the probe catch. Based upon these results, particulate sampling with an
in-stack collector appears to have an advantage over Method 5 sampling,
in that the need for recovering the probe catch is eliminated. Further-
more, for stack gas streams having SO^ concentrations higher than 2 ppm,
no increase to the particulate level {for the in-stack train) due to
condensation of $03 along the probe or on the external filter should
result, such as would occur in the Method 5 train if these components
were maintained at 120°C. Finally, an in-stack train could be much
simpler in terms of design, operation and sample recovery as compared
to the Method 5 procedure.
Arthur D Little Inc.
-------
TABLE OF CONTENTS
Page
LIST OF TABLES v
LIST OF FIGURES vii
ACKNOWLEDGEMENTS ix
I. INTRODUCTION 1
II. PROGRAM PLAN FOR OIL-FIRED BOILER EMISSION 5
A. Experimental Considerations 5
B. Sampling Apparatus 6
C. Source Description 8
III. EXPERIMENTAL PROCEDURE 11
A. Sampling Plan 11
B. Sample Collection 15
C. Sample Analysis 16
IV. EXPERIMENTAL RESULTS 19
V. DISCUSSION OF RESULTS 37
A. Set 1 - Generation of False Particulate 37
B. Set 2 - Method 5 Parameter Evaluation 46
C. Additional Field Tests 56
VI. CONCLUSIONS 59
m
Arthur D Little, Inc.
-------
LIST OF TABLES
Page
1 Stationary Source Participate Emission Standards 2
2 Experimental Design for Study of Filter Size, Filter 14
Temperature and Nozzle Type
3 Sampling Conditions for Set 1 Experimental Grid 20
4 Sulfur Oxides in Stack Gas for Set 1 Experiment Grid 21
5 Mass Distributions in Set 1 Experiment Grid Samples 22
6 Sulfate (504=) Distributions in Set 1 Experiment 23
Grid Samples
7 Sampling Conditions for Additional Set 1 Experiments 24
8 Sulfur Oxides in Stack Gas for Additional Set 1 25
Experiments
9 Mass Distributions in Additional Set 1 Experiment Samples 26
10 Sulfate ($04=) Distributions in Additional Set 1 27
Experiment Samples
11 Chemical Analysis of Collected Particulate 29
12 Sampling Conditions for Set 2 Experimental Grid 30
13 Mass Distributions in Set 2 Experimental Grid Samples 31
14 Sulfate ($04=) Distributions in Selected Set 2 Experiment 33
Grid Samples and Comparison to Set 1 Experiment Grid
Samples
15 Sampling Conditions for Additional Field Tests in Support 34
of Laboratory Evaluations
16 Mass Distributions for Additional Field Tests in Support 35
of Laboratory Studies
17 Comparison of A Train Versus B Train Particulate Mass and 39
Sulfate Distributions
18 Summary of Measured Particulate Masses in Set 2 47
Experimental Grid Samples
Arthur D Little Inc
-------
LIST OF TABLES (Cont'd)
Page
19 Analysis of Variance Table for Set 2 Experimental Grid 50
Samples: Total Participate
20 Analysis of Variance Table for Set 2 Experimental Grid 54
Samples: Filter Catch Particulate
21 Summary of the Analysis of Variance Results for the 55
Set 2 Experimental Grid Samples
vi
Arthur D Little In
-------
LIST OF FIGURES
Page
1 Schematic Sketch of Participate Sampling Trains 7
2 Sketch of Sampling Port Locations for Oil-Fired 9
Boiler
3 Comparison of Participate Catch A Train Vs B Train 41
4 Comparison of Sulfate Catch - B Train Versus A Train 42
5 Comparison of A Train and B Train Participate Catches 48
for Set 2 Experimental Grid (Numbers refer to Run No.)
6 Distribution of Particulate Between Probe and Filter 49
for Set 2 Experimental Grid
7 Schematic Presentation of the Significant Filter Size- 52
Temperature Interaction Determined by an Analysis of
Variance
vii
Arthur D Little Inc
-------
ACKNOWLEDGEMENTS
Acknowledgement is made to Philip L. Levins, Karl N. Werner and Clifford
Summers of Arthur D. Little, Inc., and to Arnold W. Doyle of Walden Research
for their participation in various aspects of the program. Appreciation
is extended to Kenneth Knapp, Fredric Jaye and John Davis of the EPA
for their suggestions and comments relative to experiment planning and
program review.
IX
Arthur D Little Inc
-------
I. INTRODUCTION
In response to the provisions of the Clean Air Act of 1970, the EPA has
identified stationary source categories that contribute significantly
to air pollution. Standards of performance applicable to new sources
within the listed categories have been promulgated for twelve source
categories and have been proposed for an additional twelve. These regu-
lations provide emission limits for various pollutant species and specify
the test methods that are to be utilized in determining compliance with
the performance standards.
Of the various source categories presently identified as contributing to
air pollution, fifteen are regulated with respect to particulate emissions.
Particulate is identified as "any material, other than uncombined water,
which exists in a finely divided form as a liquid or solid at standard
conditions (2Q°C, 760 mm Hg)," and is measured by EPA Method 5, "Deter-
mination of Particulate Emissions from Stationary Sources." A listing
of regulated industries, maximum permissible particulate emission levels
and appropriate test methods is given in Table 1.
During the three years since promulgation of the Method 5 test procedure,
a variety of comments has been made regarding its reliability and its
usability in measuring the particulate burden within source emissions.
The following types of questions are typical.
• Does Method 5 yield accurate values for the particulate emission
levels of the effected sources?
• Are there chemical interactions involving transformation from gas
phase to particulate taking place in the probe, on the filter or
anywhere else in the system that might influence the particulate
concentration?
• Does the moisture content, temperature or chemical composition
of the gases present influence the results or sampling procedure
in any way?
• Are the chemical compounds and/or elements present in the collected
particulates characteristic of this type of source? If characteris-
tic, to what degree; if not, have they been changed in any way by
the collection process?
• Do selected variations in the Method 5 testing procedure have an
influence on results?
•Do alternative sampling techniques or train components provide more
reliable emission values?
• Are there special sampling problems that are specific to just
one or to a few source categories?
Arthur D Little Inc.
-------
TABLE 1
STATIONARY SOURCE PARTICULATE EMISSION STANDARDS
Reference Subpart
Source
Standard for Participate Matter
1
1
1
2
2
2
2
2
2
3
3
3
4
5
5
1
1
Promulgated
D
E
F
I
J
L
M
N
0
Proposed
P
Q
R
Y
Z
AA
Methods
1
2
Fossil Fuel-Fired Steam
Incinerator
Portland Cement
Asphalt Concrete
Petroleum Refinery
Secondary Lead
Secondary Brass & Bronze
Iron and Steel (BOF)
Sewage Treatment
Primary Copper
Primary Zinc
Primary Brass
Coal Preparation
Ferroalloy Production
Steel (Electric Arc)
180 mg/106 cal
180 mg/m3
150 g/Ton of feed
90 mg/m
1 kg/1000 kg coke burnoff
3
50 mg/m
50 mg/m
3
50 mg/m
65 g/kg dry sludge
50 mg/rrr
50 mg/m
50 mg/m
3
40 mg/m
230 or 450 g/mw-hr.
12 mg/m
Sample and velocity traverses for stationary sources
Determination of stack gas velocity and volumetric flow
rate (Type S pilot tube)
1 3 Gas analysis for carbon dioxide, excess air and dry
molecular weight
1 5 Determination of particulate emissions
References
1 Federal Register 36 24875-24895 (December 23, 1971); 39 20790-20794
(June 14, 1974)
2 Federal Register 39 9307-9323 (March 8, 1974); 13776 (April 17, 1974)
3 Federal Register 39 37039-37049 (October 16, 1974)
4 Federal Register 39 37922-37924 (October 24, 1974)
5 Federal Register 39 37466-37468, 37470-37472 (October 21, 1974)
Arthur D Little Inc.
-------
• To what degree do process variables influence sampling procedures
or results?
To address these areas of concern, the EPA is sponsoring several research
investigations, each directed toward a specific stationary source category.
The present report relates to oil-fired steam generators. A study of coal-
fired boilers has also been conducted by Arthur D. Little, Inc., and is
reported separately. Other investigators have considered incinerators,
Portland cement plants, asphalt concrete plants and BOF steelmaking.
Mith respect to fossil-fuel steam generation, generally utilizing fuels
with a sulfur content of 0.5 to 3.0%, the sampling and analysis methods
developed to date have not adequately considered the chemical and physical
changes that can occur during the collection process which may lead to
the formation of what is expressed empirically as "false particulate".
Thus, the present Method 5 procedure may not accurately reflect the
particulate burden to the atmosphere as it would exist in a dispersed
plume.
False particulate, i.e., the matter that would not have formed in the
atmosphere at 20°C and 760 mm Hg, represents the increase in collected
mass due to the collection process itself. For example, particulate
that is collected near the start of a test, distributed along the probe
walls and on the filter, is continually exposed to the freshly sampled
stream; there is thus an opportunity for catalytic oxidation of S02 to
$03, physi- and chemisorption of organic vapors on particulate, hydra-
tion of compounds or compound conversion. In practice, all may occur to
some degree. The result is that various stationary source industries
could be found to be out of compliance due to appreciable but inappropriate
levels of "false particulate."
The current program has investigated the influence of selected variations
in sampling procedure and train configuration upon particulate levels.
In almost all cases, simultaneous sampling with two trains was carried out
to permit a direct comparison of results. Samples were analyzed to identify any
sampling conditions leading to false particulate as well as to determine the
difference in result for various collection procedures.
Sampling was conducted at a single point rather than by point-to-point
traverses. Thus, the comparative values of loading per unit volume of
aas reported herein do not reoresent emission values as would be determined
by a compliance test. They do, however, indicate the general range that
would be found by compliance testing.
Arthur D Little Inc.
-------
II. PROGRAM PLAN FOR OIL-FIRED BOILER EMISSIONS
A. Experimental Considerations
This program was conducted to develop the background and understanding
necessary to evaluate the accuracy and reliability of EPA Method 5 for
measuring particulate emissions from oil-fired steam generation plants.
The scope of work included the following elements:
Investigate the possibility of "false particulate" formation
within a Method 5 train.
Determine if minor modifications to the Method 5 train or
sampling procedure influences the amount or nature of the
particulate catch.
Evaluate alternate particulate collection procedures which
may have advantage over Method 5 in terms of accuracy, simplicity
of apparatus or procedure, etc.
To investigate experimentally these areas, two independent sets of tests
were conducted at an oil-fired electric utility. The design of individual
experiments was formulated on the basis of source and sampling variables
considered to have an influence on emission measurements. Parameters
that were considered included:
Source
Particulate - loading, chemical characteristics, particle size distribution
S02> S03 concentration
Stack temperature
Uniformity of conditions as a function of time
Sampling
Sampling train design
Filter size
Filter media
Nozzle configuration
Sampling time
System temperature
Sampling rate
Beyond the selection of a feasible location, little can be done regarding
source parameters. Once a site is chosen,experimental plans must be de-
veloped around typically existing conditions. Sampling parameters, on
the other hand, are readily adjusted and can influence the amount and
distribution (within the sampling train) of the collected particulate.
Sampling time depends largely upon particulate loading levels within the
sampled stream, but should not be so short as to be influenced by process
fluctuations; the range of 30 to 120 minutes is probably most suitable.
Arthur D Little Inc.
-------
Collection rate can be important in terms of the opportunity for stream
components to interact with participate that has already been collected.
The chemical reactivity between the gas stream and sampling train materials
represents an important category of sampling variables. This includes
filter media (silica, borosilicate glass, etc.) probe liner (glass, stain-
less), etc. Filter size can presumably influence the collection efficiency
of peirticulate as it reflects the face velocity of the approaching gas
stream. Nozzle configuration (90° or button hook) is considered as it
may influence the distribution of particulate within the system. Finally,
the temperature profile within the system may have a major influence
upon the preferential condensation of vapor or rates of reaction between
species of the sampled stream.
Previously mentioned factors were taken into account during preparation
of the sampling plan, and consultation with EPA personnel and the other
contractors resulted in mutually agreed upon sets of experiments. To
facilitate evaluation of these variables, simultaneous sampling was con-
ducted with two sampling trains as described in the next section.
B. Sampling Apparatus
Two Method 5 particulate sampling trains were employed during this study.
Both were modified versions of the Research Appliance Company "Stak-sampler."
The configuration of the sampling systems were varied to accommodate the
purpose of the various experiments. A schematic sketch of the two trains
is presented in Figure 1.
For the first set of experiments investigating "false particulate" forma-
tion, one sampling train was modified by the inclusion of a stainless
steel filter holder between the nozzle and probe assemblies. Both trains
included an external filter and a series of four impingers charged as
outlined in the Federal Register.* Samples collected from the various train
components were given the following codes:
AFS Instack filter, A train
AFE External filter, A train
APS Washings from nozzle and front half of instack filter
holder, A train
APE Washings from back half of instack filter holder, probe
and front half of external filter holder, A train
BF External filter, B train
BP Washings from nozzle, probe and front half of the
external filter holder, B train
*Federal Register, 36, 15704-15722 (August 17, 1971).
Arthur D Little, Incl
-------
APS
APE
'
K
I
BP
AFE
BF
Stack
Wall
A. FORMATION OF "FALSE PARTICIPATE" EXPERIMENTS
Back
Half
AF or BF
AP or BP
B. REPRODUCIBILITY EXPERIMENTS
FIGURE
SCHEMATIC SKETCH OF PARTICIPATE SAMPLING TRAINS
Arthur D Little Inc.
-------
The train configurations utilized during the second phase of this study,
during which the reproducibility of results as a function of train con-
figuration was investigated, were basic Method 5 arrangements. For these
tests, the nozzle design (90° versus button hook) and external filter
size (6.3 versus 10.0 cm) were varied according to a statistically designed
grid of experiments. Samples collected under this scheme were coded as
follows:
AP Washings of probe, nozzle and front half of external
filter holder, A train
AF External filter, A train
BP Washings of probe, nozzle and front half of external
filter holder, B train
BF External filter, B train
Additional equipment utilized during the evaluation included a S02/S03
sampling train, designed and constructed by ADL, as well as velocity
probes, an Orsat gas analyzer and proportional temperature controllers.
C. Source Description
Sampling was conducted at a New England power station equipped with two
oil-fired 350 MW boilers. The facility utilizes a low sulfur residual
oil with a magnesium oxide additive as a corrosion inhibitor. All sampling
was performed on Unit No. 2 which typically develops a load of 360 MW on
a fuel flow of 180,000 pounds per hour. Combustion is carried out with
about 30% excess air, employing 1 gallon of scale preventive per 3000
gallons of fuel. Sampling was conducted in the stack upriser ducting
after roughly 30 feet of straight run. A sketch of this duct is given
in Figure 2. Typical stack conditions are as follows:
Stack temperature 145 - 155°C
Velocity 24 - 26 m/sec
Stack Pressure 4.7 - 5.0 mm Hg
Duct Dimensions 3.66 m diameter
Stack Gas Volume 250 - 270 m3/sec
Port Size 10.2 cm nipple
Moisture 9 - 11%
02 3.5 - 5.0%
C02 9-12%
S02 200 - 250 ppm
SOs 1-3 ppm
Field work was performed in two stages; the first set of experiments was
conducted during April-May 1973 and the second set during June-July 1974.
Arthur D Little Incl
I
-------
UNIT #2
DUCT A
DUCT B
0.45m—i
W
15 m Apr ox. 3.7 mj- 0" D
8th Level
3.;
m D.
FIGURE 2 SKETCH OF SAMPLING PORT LOCATIONS FOR OIL-FIRED BOILER
Arthur D Little, Inc.
-------
III. EXPERIMENTAL PROCEDURE
A. Sampling Plan
To permit a comparison of results from several stationary sources, the
various contractors collaborated as much as possible with respect to
the design of experimental tests and commonality of equipment and sampling
procedures. Based upon discussions of our test plan with the EPA Project
Officer, the task of evaluating oil-fired boilers was separated into two
experimental sets of tests. These were designed to provide information
relevant to the formation of false particulate and to a parametric
study of Method 5, respectively.
1. Formation of False Particulate
To evaluate the possibility of false particulate formation, a series
of twelve experiments were proposed in which the research-type A train
was run as the control and the Method 5 B train was varied according
to the following experimental block.
Fl
2
Tl
T2
Tl
T2
h
X
X
X X
X X
L2
X
X
X X
X X
I
where
X
T
F
L
sample to be obtained
temperature of external filter, B train (T] = 120°C;
T£ = stack temperature or 175°C, whichever is lower)
filter media, B train (F] = MSA 1106 BH glass fiber;
F2 = Modified Quartz Fiber* - MQF)
level of loading (l_i = reasonable sample catch, i.e.,
100 mg; L£ twice sampling time of LI)
Pall flex "Tissuequartz" subjected to a strengthening treatment
developed by ADL under EPA Contract No. 68-02-0585.
11
Arthur D Little Inc.
-------
As indicated by the grid, three variables (effects) were selected as having
a possible influence on the formation of false participate, with two con-
ditions (levels) of each being evaluated. Data for the MQF filter type
was replicated to permit an evaluation of reproducibility.
The research-type A train was proposed to be the control, being equipped
with an instack filter to capture all particulate as it exists within
the stack, an external filter temperature maintained at 120°C and MQF
filter media for both the instack and external filter. The A train data
was intended to give an indication of the source variability.
The experiment design called for simultaneous sampling with both A and
B trains. Evaluation of the performance of the two trains was to be
carried out by comparing the levels and distributions of particulate
mass and sulfate content.
In addition to the twelve duplicate runs proposed in the grid, several
other experimental runs were proposed to study specifically the possibility
for S02 oxidation. These additional experiments and their purpose were
as follows:
1. Spike both trains with an order of magnitude enrichment in S02.
Is there an increase in S04= associated with any component of
the sampling train?
2. Run both trains as Method 5 trains.
a. Perform consecutive runs without cleaning probe in between.
Does a dirty probe show differences from a clean one relative
to sulfate/particulate ratio?
b. Spike one train with an order of magnitude enrichment in S02.
Are comparable results obtained between the two trains?
c. Use a soiled filter (containing a measured particulate loading)
from a prior run as a backup to a clean filter in one of the
trains.
Is there additional weight pickup signifying S02 reaction
with the filter catch?
Does the unsoiled backup filter exhibit a gain in weight,
implying an influence from the filter media?
To permit selective spiking of the sampled gas with S02» a 1/8 inch
diameter length of stainless steel tubing was tapped into the midsection
of the nozzle and silver soldered in place. The other end of the tubing
was connected through a rotometer to a tank of S02. During sampling,
proportionate amounts of S02 were bled into the sampled gas stream to
produce a 10 fold increase in S02 concentration thereby simulating the
worst case condition with respect to SO? content, such as might be ex-
perienced in the combustion of very high sulfur content fuels.
12
Arthur D Little Infl
-------
The experimental plan called for analysis of the collected samples for
evidence of false participate formation. If any unexplainable trends
were noted, additional well controlled laboratory experiments were
contemplated for studying possible mechanisms involved in false par-
ticulate formation.
2. Method 5 Parameter Evaluation
To permit a critical review of the applicability of Method 5 as written
for oil-fired steam generators, a statistically designed set of experiments
was developed to examine several procedural steps (factors), each at two
levels. After discussions with the other contractor and the EPA Project
Officer, it was mutually agreed to evaluate nozzle type, filter diameter
and filter temperature. Independent evaluations of these three factors
at two levels requires a total of 2^ or 8 combinations, as follows:
h
2
Ti
T2
Tl
T2
Nl
X
X
X
X
N2
X
X
X
X
where
X = samples to be collected
T = filter temperature (T] = 120°C; TZ = stack temperature)
F = filter diameter (F] = 6.3 cm [2-1/211]; F2 = 10 cm [4"])
N = nozzle type (N] = button hook; N£ = 90° bend)
The experimental design was based upon Simultaneous collection of two
samples, with each pair of simultaneous measurements comprising a "block."
For this purpose, identical Method 5 sampling trains were employed, varied
only with respect to nozzle type and filter size according to the experi-
ment plan. The A train sampling was carried out in Port SI (Figure 2). To
maximize the amount of information available from this series of tests,
it was decided to employ three replicates; separate test factors were
confounded with blocks within each replicate so that information would
be available to study both main effects and first order interaction
effects. The experimental design is given in Table 2.
13
Arthur D Little Inc.
-------
TABLE 2
EXPERIMENTAL DESIGN FOR STUDY OF
FILTER SIZE, FILTER TEMPERATURE AND
NOZZLE TYPE
Replicate
I
II
III
Block
1
2
3
4
5
6
7
8
9
10
11
12
Order
of Test
2
4
7
5
8
9
11
1
12
6
10
3
Filter
Size
6.3cm
10
6.3
10
10
10
6.3
10
6.3
10
10
6.3
A TRAIN
Temp
120°C
stack
stack
120
120
120
stack
stack
120
120
stack
stack
Nozzle
BH
90
BH
90
90
90
BH
BH
BH
BH
90
90
Filter
Size
6.3cm
10
6.3
10
6.3
6.3
10
6.3
10
6.3
6.3
10
B TRAIN
Temp
stack
120°C
120
stack
120
120
stack
stack
stack
stack
120
120
Nozzle
90
BH
90
BH
BH
90
90
90
BH
BH
90
90
14
Arthur D Little Inc
-------
The order of sample collection by blocks was randomized as a means of mini-
mizing the influence from any variation in source conditions. Furthermore,
it was assumed that Systems A and B were equivalent with respect to source
stream conditions as well as sampling and cleanup protocols.
The basis for this design is the analysis by variance technique, which assumes
that the observations are random variables with well defined distributional
properties. Utilizing the particulate catch data for all 24 experimental
runs, the variance in result can be independently determined for replicates,
blocks, each of the main effects, interaction effects and experimental
uncertainty (error). The value of the ratio of any of the above terms to
the experimental error permits one to test the hypothesis that the term
considered has no influence on weight collected. If the hypothesis is
rejected, one can conclude that different levels of the term lead to a
significant difference in the amount of collected particulate.
3. Laboratory Studies
In addition to the field experiments described above, several
laboratory studies were conducted to evaluate specific Method 5 criteria
and to provide guidance in improving sampling procedures for the second
set of field samples. The first of these independent studies included
a measurement of internal and external temperatures along the front half
of the sampling system for several stream conditions. The purpose of
this study was to determine the most reliable reference point for con-
trolling probe and filter box heaters to maintain a gas temperature of
120°C at the back of the probe and at the filter. A second study in-
volved an evaluation of the usability of several instack filter holders
and the physical and thermal properties of various filter media employed
in these holders. The results of these studies are presented in separate
reports.*
B. Sample Collection
The sampling plan followed throughout all runs called for the simultaneous
collection of samples with A and B trains from points of essentially
identical stream conditions; Ports AS1 and AS2 (Figure 2) were utilized,
with sampling points being approximately 90 cm within the ductwork and
15 cm apart. Except as previously mentioned, sampling was conducted
according to the Method 5 procedure, with the train configuration being
adapted to accommodate the desired experiment. An impinger train was
utilized in all runs. Besides permitting a measurement of the moisture
content of the stream, the impinger solutions were occasionally dried
down to yield the "back half" particulate level. Throughout the first
*
E.T. Peters and J.W. Adams, "The Measurement and Control of Gas Tem-
perature in a Method 5 Sampling Train," Special Report, EPA Contract
68-02-0632, in preparation.
E.T. Peters, J.W. Adams and A.L. Benson, "Properties of Several In-stack
Filters and Filter Holders for Source Sampling," Special Report, EPA
Contract 68-02-0632, in preparation.
15
Arthur D Little, Inc.
-------
set of experiments, the impingers were charged with 10% H202 in 0.1N NaOH
(prepared fresh before each run) to scrub out S02- The impinger solutions
were subsequently analyzed for sulfate, providing a measure of total SOX
for comparison to the S0£-S03 train results. For the second set of ex-
periments, the impingers were charged with distilled water.
Sampling duration was typically one to two hours for the first set of
runs and one hour for experiments in the second evaluation. Sampling was
isokinetic, except for those cases where particular train components being
utilized resulted in an excessive pressure drop across the system. For
these cases, both trains were operated at the same lower-than-isokinetic
rate.
Stack velocity, pressure and temperature measurements were made at fre-
quent intervals during the course of sampling. System temperatures were
initially monitored and controlled from convenient external reference
points. However, based on the results of a laboratory investigation of
exit gas temperature as a function of various reference control positions
(separately reported), an internal reference point was utilized for con-
trolling system temperatures during the second set of experiments.
Concurrent with the initial set of experiments, an S02-S03 sample was
withdrawn from port AE (Fig. 2). Separate values for SC>2 and $63
concentrations were obtained by EPA Method 8 and the barium chloranilate
procedure for analysis of sulfate. In addition, an Orsat sample
was obtained during each experiment to determine flue gas composition.
After sampling, the particulate catch was separated into several fractions,
as described previously. For experiments utilizing an instack filter,
the samples were identified by the code given in Figure 1A; for Method 5-
type runs where only external filters were used, sample coding is
described by Figure IB. In every case, the probe was scrubbed with a
nylon brush with several distilled water and acetone rinses. The
silica gel and impinger solutions were quantitatively recovered and
retained for a subsequent determination of moisture and additional
measurements such as sulfate and acidity. All samples were stored in
polyethylene bottles.
C. Sample Analysis
Gravimetric determinations of all filter and probe catches were performed
after evaporation to dryness and equilibration over a dessicant. To deter-
mine if relative humidity had an influence on the magnitude of the parti-
culate catch, weighings of initial samples were made after a 24-hour
dessication over Drierite and again after an additional 24-hour des-
sication over PZ®*' No differences were observed for the two dessicants,
and the more convenient Drierite was used for the remainder of the
runs.
16
Arthur D Little, Inc.
-------
To lessen the effect of any weighing error on final results, repetitive
weighings were performed on all samples,, If weighings agreed within 1 mg,
the average value was used. If differences of more than 1 mg were found,
additional determinations of weight were made by other analysts, with
the average taken as the true value.
For experiments conducted within the first set, sulfate and free acid
analyses were carried out on all probe, nozzle and filter residues. Initial
attempts to use a carbonate fusion led to high blank values and nonrepro-
ducible results. Finally, it was demonstrated that a 30-minute leach in
25 ml of hot water completely dissolved all sulfate detectable (as sulfur)
by x-ray fluorescence (XRF). By the use of known concentration standards,
XRF detection levels were found to be about 0.5 mg S04= under the condi-
tions of the filter leaching experiments. Leached solutions were analyzed
for sulfate by means of the barium chloram'late colorimetric procedure.
Total SOX data were obtained from impinger solutions by the chloram'late
procedure. Acidity was determined in the leachants by titration with
0.01N NAOH to a phenopthalein end point. No difficulty was encountered
in the end point determination for these samples even though past ex-
perience has shown that samples containing high Ni or V content could
result in interference. Occasionally, particulate samples were ana-
lyzed by x-ray fluorescence and emission spectrographic methods to
determine the presence and relative concentrations of cation species.
X-ray diffraction analyses were attempted in a few cases, but results
were inconclusive.
During the second set of experiments, residue samples were occasionally
analyzed by one or more of the above techniques to provide information
on the sulfate distribution within the sample fractions and to provide
qualitative descriptions of the elemental chemistry of various residue
samples.
17
Arthur D Little Inc.
-------
IV. EXPERIMENTAL RESULTS
A total of 30 paired particulate runs were performed at the oil-fired
boiler, 18 in response to evaluations of false particulate generation
and 12 with respect to the statistical sampling parameter grid. Within
the first set, 14 runs were conducted according to the mutually agreed
upon design (presented in Section III, A, I) involving selected variations
in sampling time, filter temperature and filter media; S02 and $63 analyses
were conducted concurrently. The sequence of runs with respect to the
experimental design is summarized by run no. as follows:
F = 1106 BH
F = MQF
T = 120°C
T = 150°C
T = 120°C
T = 150°C
L = 60 min
(1)
17
(3,6), 8t
lot
L = 120 min
(4), 11? 12
13
(2,5), 7
9
( ) - A Train data invalid due to leak in sampling system
* - A Train data incomplete due to loss of sample APE
t - Time was increased to 90 minutes to increase the
collected mass of sample.
Tables 3 through 6 present the sample collection conditions, S02 and
concentrations in the gas stream, particulate distributions and sulfate
distributions, respectively, for these runs. Four additional experimental
runs were performed to explore specific opportunities for S02 oxidation.
Comparable data for these experiments is presented in Tables 7 through
10.
During the initial runs, it was determined that isokinetic sampling was
not feasible due to the relatively high stack gas velocity and a substantial
pressure drop across the system equipped with an instack filter. As a
result, it was decided to sample equivalently with both systems at rates
below isokinetic. This deviation from the Method 5 procedure was
justified by the following points—results were interpreted on the
basis of a relative comparison of the particulate catch from the two
trains, very little influence on particulate loadings would result
from the predominately submicron size particulate in the gas stream,
and lower rates permitted longer sampling times thus tending to smooth
out short term source fluctuations.
19
Arthur D Little Inc
-------
TABLE 3
SAMPLING CONDITIONS FOR SET 1 EXPERIMENTAL GRID
Run
Number
1A
B
2A
B
3A
B
4A
B
5A
B
6A
B
7A
B
8A
B
9A
B
10A
B
11A
B
12A
B
ISA
B
17A
B
Fi 1 ter
Temp
(°C)
120
120
110
120
120
120
120
120
145
145
12D
120
150
150
Filter
Type*
1106
MQF
MQF
1106 ,
MQF
MQF
MQF
MQF
MQF
MQF
1106
1106
1106
1106
Stack
Date Temp
(°C)
4-16-73 150
150
4-24-73 145
145
4-25-73 135
130
5-01-73 145
145
5-02-73 145
145
5-08-73 145
145
5-10-73 145
5-17-73 140
Run
Time
(min)
60
60
120
120
60
60
120
120
120
120
60
60
135
135
90
90
120
120
90
90
109
113
108
109
125
114
60
60
Moisture
Content
(X)
4.1
12.3
4.5
11.8
4.1
11.9
7.5
12.0
6.5
11.0
9.8
11.1
11.5
11.7
11.5
11.8
12.0
12.2
12.6
12.7
12.1
12.2
12.3
12.4
11.3
11.9
12.6
12.8
Gas Volume
Collected
(m3)
0.885
0.909
1.693
1.811
1.187
1.309
2.461
2.498
1.651
1.820
0.846
0.911
1.929
2.033
1.338
1.453
1.568
1.661
1.306
1.352
1.626
1.750
1.550
1.652
1.781
1.746
0.890
0.925
* 1106 - MSA Type 1106 BH Glass Fiber Filter
MQF - Pallflex "Tissuequartz" Filter with a strengthening
treatment applied by Arthur D. Little, Inc.
20
Arthur D Little, Inc.
-------
TABLE 4
SULFUR OXIDES IN STACK GAS FOR SET 1 EXPERIMENT GRID*
Particulate Trains
Run
Number
1
2
3
4
5
6
7
8
9
10
11
12
13
17
Average
(All Runs)
Average
(Runs 7-17)
so2
(mg
A-Train
NAt
NA
NA
NA
NA
NA
870
750
900
990
910
860
820
830
865
(Impinger)
S04=/m3)
B-Train
1050
1030
1000
630
670
870
900
870
890
860
900
920
850
800
880
875
SO Train
/\
so2
(mg S04=/m3) (mg
965
1300
1220
1050
875
875
560
560
615
530
570
590
580
—
790
570
so3
S04=/m3)
5.9
5.7
2.1
1.4
8.3
7.9
2.9
0.6
1.3
0.9
3.7
5.6
1.9
1.7
3.6
2.3
* 4 mg S04~W 1 ppm S02 or S03
t not analyzed
21
Arthur D Little Inc
-------
TABLE 5
MASS DISTRIBUTIONS IN SET 1 EXPERIMENT GRID SAMPLES
Run No.
APS
1
2
3
4
5
r\5 p.
PO D
7 15
8 11
9 14
10 8
11 tt
12 8
13 4
jr 178
£ Avg (7-17)
ET Avg A/B
Parti culate Mass (mg/m )
AFS APE AFE A* A-TOTAL BP
30
15
25
12
10
27
60 5 2 75 82 44
48 6 1 59 66 39
63 2 3 77 82 51
53 3 - 61 64 41
20 5 - - - 15
69 1 0 77 78 37
17 2 1 21 24 9
25 1 5 33 39 18
58 62
BF
26
26
22
24
17
17
25
23
30
23
27
41
23
24
B/A*
B-TOTAL
56
41
47
36
27
34
69 0.92
62 -1.05
81 f.05
64 1 .05
42
78 1.01
32 1.52
42 1.27
61 1.05
1.12
B/A
0.84
0.94
0.99
1.00
--
1.00
1.33
1.08
0.98
1.03
t A* = APS -i- AFS ttSample APS was lost.
-------
TABLE 6
SULFATE (S04=) DISTRIBUTIONS IN SET 1 EXPERIMENT GRID SAMPLES
Run
Number
Sulfate Mass -- mg/nT
B/A* B/A
APS AFS APE AFE A*t A-Total BP BF B-Total
21
0 26 28 15 11 26 1.00 0.93
8
15
0 19 21 13 10 23 1.21 1.09
29
0 33 33 11 10 21 0.64 0.64
10
19 1 0 21 22 15 10 25 1.19 1.14
12
23 1 0 26 27 20 13 33 1.27 1.22
13
0 1 11 12
10 13 1.18 1.08
17
1 1 11 13 8
15 23 2.09 1.77
Average
(7-17)
21 22
23 1.10 1.05
Average
(A/B)
1.23 1.12
= APS+AFS
23
Arthur D Little Inc
-------
TABLE 7
SAMPLING CONDITIONS FOR ADDITIONAL SET
Run
Number
A.
15 A
B
B.
14 Bl
B2
14'B1
B2
16 Bl
B2
Fil
tev Filter
Temp
(°
SO
1
C)
2
10
Both
1.
Type*
Date
Stack
Temp
(°C)
1 EXPERIMENTS
Run
Time
(min)
Moisture
Content
(
%
)
Gas Volume
Collected
3
(m )
Enrichment (^2000 ppm S02)
1106
Trains Run As
5-15-73
140
129
129
1
9
0
.1
.0
1.83
1.97
EPA Method 5 Trains
Influence of Soiled Probe
120
120
1
1
2.
1
1
3.
20
20
25
20
(Probe was not
MQF
1106
MQF
1106
Influence of SO
(B2 spiked for
MQF
MQF
Influence of SO
(B2 contained a
cleaned between
5-10-73
5-10-73
2 Enrichment
£2000 ppm S02 i
5-17-73
? on Soiled Fil
soiled filter
trains were spiked for ~2000
113 Bl
B2
1
1
20
20
MQF
MQF
5-18-73
Runs 14 and 14')
145
145
n sampled
140
ter
behind the
ppm S02 in
130
60
60
90
90
gas)
120
120
regular
sampled
120
120
1
1
1
1
1
1
filter
gas).
1
1
1
1
2
0
9
9
9
.6
.7
.6
.7
.1
.8
.4
.7
1.17
1.11
1.32
1.37
1.75
1.78
1.72
1.80
* 1106 - MSA Type 1106 BH Glass Fiber Filter
MQF - Pallflex "Tissuequartz" Filter with a strengthening
Treatment Applied by Arthur D. Little, Inc.
24
Arthur D Little Inc
-------
TABLE 8
SULFUR OXIDES IN STACK GAS FOR ADDITIONAL SET 1 EXPERIMENTS*
Run
Number
Particulate Trains
S0? (Impinger)
(rag S04 /m3)
A-Train
B-Train
SO Train
/\
SO,
so.
(mg S04=/m3) (mg S04=/m3)
14+14'
560
1.9
15
8120
5780
2.9
16
18
730 3930 — 2.7
8530 6160 — 2.8
* 4 mg S04 ft* 1 ppm S02 or S03
25
Arthur D Little Inc
-------
TABLE 9
MASS DISTRIBUTIONS IN ADDITIONAL SET 1 EXPERIMENT SAMPLES
Run No. Particulate Mass - mq/m3
APS AFS APE AFE A*t A-Total BP
15 10 35 2 0 45 47 30
14-1
_2
14'-1 23
-2 28
16-1 21
16-2 18
18-1 18
18-2 14
BF
33
15
21
26
33
17
13
12
12
B-Total
63
—
--
64"
82tt
38
31
30
26
B/A"
B/A
1.40 1.34
B1/B2 = 0.78
B1/B2 = 1.22
B1/B2 = 1.15
rA* = APS + AFS
ftTotal for Runs 14 and 14'
26
Arthur D Little; Inc
-------
TABLE 10
ro
Run
Number
15
14+14 '-1
14+14'-2
16-1
16-2
18-1
18-2
SULFATE (S04 ) DISTRIBUTIONS IN ADDITIONAL
Sulfate Mass — mg/m
APS AFS APE AFE A*+ A-Total BP
43147 12 12
10
12
8
6
4
5
SET 1 EXPERIMENT S
BF B-Total
2 14
--
—
9 17
7 13
5 9
IS1"1" 20
fA* = APS + AFS
"""includes sulfate on soiled filter (Run 141)
B/A"
B/A
2.00 1.17
B1/B2 = 1.30
B1/B2 = 0.45
c
-------
The data presented for the A train, Runs 1 through 6, were influenced
by a leak between the instack filter and probe assemblies, allowing
the collected sample to be diluted by ambient air. Thus, the particulate
catches for these runs were biased to the low side. Progressive steps
were taken in the field to correct this problem, with improved results
for Runs 4 and 5 and even greater improvement for Run 6. An approximate
measure of the ambient air dilution can be estimated from the observed
moisture content of the stack gases; on this basis, the dilution factor
was about 200% for Runs 1-3, 70% for Runs 4 and 5 and 10% for Run 6.
Because of the magnitude of correction necessary, no correlation was
attempted between the A train and B train for these runs.
Concurrent with the initial 18 particulate samples, Orsat and SOx samples
were obtained from Port AE (see Figure 2). The SOX sample was withdrawn
through a specially designed Method 8 train, selectively condensing $03
prior to scrubbing out S02 in a 10% ^2 solution. Condenser washings
and impinger solutions were segregated and retained for varium chloranilate
colorimetric analysis. Stack gas compositions determined by Orsat analysis
of a gas sample withdrawn from Port AE were found to agree with the re-
sults obtained from samples routinely collected by station personnel at
a point just above the economizer. Typical values were 12-14% CO? and
1-3% 02.
A few additional analyses were conducted on the particulate on occasion
to provide some understanding of the distribution and concentration of
cation species. Optical emission spectrographic and x-ray fluorescence
analysis results for several samples is provided in Table 11.
Twelve paired samples were obtained during the second evaluation; data
regarding sampling collection conditions and measured values of particulate
distributions are presented in Table 12 and 13. Experimentally, no dif-
ficulties were encountered during this latter investigation, excepting
the necessity to conduct two runs at approximately 70% of the isokinetic
condition. Both trains were operated equivalently.
The first set of experiments (Runs 1-18) had been carried out with the probe
liner temperature at the back of the probe maintained at 120°C. In the
interim between the first and second sets of experiments, a separate
laboratory evaluation was made of gas temperature versus external system
temperature at several positions along a Method 5 sampling train as a
function of variations in stream condition. It was determined that
for sampling conditions similar to those employed in the field the ex-
ternally measured temperatures at the back of the probe could be 25°C
or more colder than the gas temperature at this point. As a result,
the second set of experiments utilized an internal reference thermocouple
at the back of the probe. The output of this thermocouple was fed to a
proportional controller for the probe heater, set to maintain a gas tem-
perature of either 120°C or stack temperature (150°C). The filter box
temperature was maintained at 120+10°C. Because of the difference in
28
Arthur D Little Inc
-------
TABLE 11
CHEMICAL ANALYSIS OF COLLECTED PARTICULATE
Optical Emission
Pnnf*ont^at *i A n
l/UilLeMtral lUil
> 10%
3-30
1-10
0.3-3
0.1-1
Spectroscopy
5AFS SAFE
-
Mg
V
Ni
_ _
(Qualitative)
_.Mp. F_
OMI Ir L.C.O
5BF 9APS 9APE
_
- - -
Mg
Ni ,V Fe Fe
Mg
9BP
Mg
V
Fe
Ni ,Si
Al.Zn
10BP
Mg
V
Fe
Ni.Si
Al.Zn
B. X-ray Fluorescence (Qualitative)1
niiiuuri u
Strong
Medium
Weak
Trace
5AFS
s,v
K,Ni,Fe
Pb
_
ottrirLto
5AFE 5BF
s,v
Ni,K,Fe
Pb
Pb,S
Mg cannot be analyzed by the apparatus employed
29
Arthur D Little Inc
-------
TABLE 12
SAMPLING CONDITIONS FOR SET 2 EXPERIMENTAL GRID
>
rthur D Li
Run
Ntnber
39 A
B
40 A
B
41 A
B
42 A
B
43 A
B
OJ
0 44 A
B
45 A
B
46 A
B
47 A
3
48 A
3
49 A
3
50 A
3
Experiment Description
Filter Filter Nozzle
Size Temp Config. Date
10
6.3
6.3
6.3
6.3
10
10
10
10
10
•10
6.3
6.3
6.3
10
6.3
10
6.3
10
6.3
6.3
10
6.3
10
Stack
Stack
120
Stack •
Stack
120
Stack
120
120
Stack
120
Stack
Stack
120
120
120
120
120
Stack
120
Stack
Stack
120
Stack
BH 7-10-74
90°
BH - 7-10-74
90°
90° 7-10-74
90°
90- 7-11-74
BH
90" 7-11-74
BH
BH 7-11-74
BH
BH 7-12-74
90"
90° 7-12-74
BH
BH 7-12-74
90°
90° 7-15-74
90°
BH 7-15-74
90°
BH 7-15-74
BH
TJ
i,
•8 a
3$
in t—
150
150
150
150
150
150
150
150
150
150
150
150
*t>
E
P'c'
§5
C£
50
50
32
32
60
60
60
60
60
60
60
60
56
56
45
45
40
40
60
60
60
60
60
60
««
TJ
I. 4->
= C
*J (U
t/» 4->
6 O
£ 0
14.9
11.3
9.6
13.2
10.8
11.4
10.5
10.3
9.7
10.0
8.8
9.3
10.7
11.1
11.3
9.4
11.7
10.1
11.6
11.0
10.9
11.6
11.0
10.7
U
.*£ tu
*-> E
"" =^
-«-
ro U '
«- 0
OJ
«t >
24.37
24.13
24.07
24.07
24.13
24.13
24.16
24.16
24.11
24.11
24.07
.24.07
25.00
25.00
25.02
25.02
25.03
25.03
24.15
24.12
24.20
24.21
24.21
24.21
i"
— 'T3
3 OJ O
•—•*-» O i-
•— O
I/I i — U
ra 0 c
003
1.470
1.245
.937
.797
1.081
1.850
1.528
1.772
1.535
1.778
1.735
1.641
• 1.597
1.459
1.155
1.331
1.076
1.187
1.815
1.144
1.757
1.578
1.726
1.783
<*>
^
3 0) Gl
QUO
>• «l 0)
I/I— S-
HJ 0 0
00
0. Wl
E •-,
<0 E
•" >,
OJ •»*
C71-»-
(O O
(. 0
QJ r—
> <1J
«I >
25.65
23.64
24.17
23.99
16.81
16.91
24.57
24.22
24.42
25.15
24.14 .
2-3.49 ;
24.47
25.41
24.91
25.00
25.64
24.64
16.95
18.13
24.42
24.83
24.90
24.73
0 W
^ *^
c o>
.tt t.
c o>
•25
105.2
98.0
100.4
99.7
69.7
70.1
101.7
104.4
101.3
104.3
100.3
97.6
97.9
101.6
99.5
99.9
102.4
98.4
70.2
75.1
100.9
102.6
102.8
102.2
-------
TABLE 13
MASS DISTRIBUTIONS IN SET 2 EXPERIMENTAL GRID
Run No.
39
40
41
42
43
44
45
46
47
48
49
50
SAMPLES
Participate Mass (mg/m^)
AP
14
12
13
9
9
7
9
29
11
9
8
10
AF
18
23
20
19
15
18
18
25
26
22
14
16
ATotal
32
35
33
28
24
25
27
54
37
31
22
26
BP
29
22
10
9
9
13
8
10
11
23
8
9
BF
18
24
23
17
12
17
18
20
24
16
16
18
BTotal
47
46
33
26
21
30
26
30
35
39
24
27
31
Arthur D Little Inc
-------
gas temperature at the back of the probe for the two sets of experiments
(estimated to be 150-160°C for the first set versus 120°C for the second
set),, several samples from the second set were analyzed to obtain sulfate
distributions for comparison to the values obtained in the first set.
In this way, it was hoped to determine if a higher gas temperature in
the probe resulted in less condensation of $03 or conversely provided
improved kinetics for S02 conversion to sulfate. The results are given
in Table 14.
Finally, five pairs of samples were collected in which one train was
operated by the Method 5 procedure and the other train employed either
an in-stack particulate collector or an external flat quartz filter.
These runs were performed as an extension of a laboratory study con-
cerned with evaluating the usability of various instack collectors and
filter media. The results of these experiments are described in a
separate report. However, so that all test data for this oil-fired
utility boiler are together in one report, the sampling conditions and
particulate distributions obtained for these five experiments are in-
cluded in Tables 15 and 16, respectively.
32
Arthur D Little Inc.
-------
TABLE 14
SULFATE (S04=) DISTRIBUTIONS IN SELECTED SET 2 EXPERIMENT GRID
SAMPLES
A. Set 2 Data
Filter Temp
AND COMPARISON TO SET 1
(Buttonhook
Run No.
nozzle, 2-1/2
EXPERIMENT GRID
" filter)
Mass of Sulfate (mg/m3)
SAMPLES
Percent
Sulfate/
Total Particulate
120°C
150°C
40A
46B
50A
44B
45A
49A
P
T
2
2
2
2
2
F
7
9
12
12
10
11
P
4
18
18
17
25
24
F
21
43
71
71
54
80
B. Comparison of Average Results
Probe Temp* Filter Temp Tests Mass of Sulfate (mq/m3) Percent Sulfate/
Set 1
-150°C 120°C 6
150°C 4
Set 2
120°C 120°C 3
150°C 3
*
Gas temperature at the back of the probe
Total Particulate
£
11
10
2
2
£
10
11
9
11
£
33
36
13
22
£
37
43
47
68
33
Arthur D Little Inc
-------
TABLE 15
SAMPLING CONDITIONS FOR ADDITIONAL FIELD TESTS IN SUPPORT OF LABORATORY EVALUATIONS
Run
Number
33A
B
34A
to
* B
35A
B
37A
B
38A
B
Experiment Description Date
In-stack thimble 6-20-74
10 cm external filter
MSA-1106-BH
In-stack thimble 6-21-74
10 cm external filter
MSA-1106-BH
In-stack thimble 6-27-74
10 cm external filter
MSA-1106-BH
10 cm external filter 6-28-74
10 cm external filter
MSA-1106-BH
10 cm external filter 6-28-74
10 cm external filter
<_> O -— ^ •— ~ CU O
o ^ E ojeucu fQf? E T3 T3 OO -r-
4->CU > 3 3 O> O> >> +•>
to E--~> $-•(-> i— -MO) r- 4J-M -r- E CD
_ii dJ • i_ >• a)
+->O) 3 OO >(U rOOE fOOO >O) io>
001— C£ SO ct=> OC_>3 t300-<«=C
155 60 11.9 25.23 1.780 1.902 25.61 101.5
60 12.1 25.23 1.804 1.956 25.73 102.0
155 60 9.1 25.25 1.785 1.847 25.20 99.8
60 9.6 25.25 1.793 1.884 25.05 99.2
150 60 10.0 24.77 1.850 2.019 26.57 107.2
60 10.7 24.77 1.795 1.977 25.36 102.4
150 60 11.3 24.96 1.768 1.964 25.74 103.1
60 11.5 24.83 1.773 1.992 25.46 102.5
150 60 11.2 25.64 1.821 2.015 26.39 102.9
^. MSA-1106-BH 60 11.5 25.64 1.830 2.028 25.90 101.0
c
-\
ii_r^ £
r- Svenka Flakt Jabrikan (Sweden) Thimble distributed in the United States
. by Carborundum Environmental Systems, Inc., Knoxville, Tenn
-------
co
en
-i
D
TABLE 16
MASS DISTRIBUTIONS FOR ADDITIONAL FIELD TESTS
Part ic ul
Run No.
APS AFS APE
IN SUPPORT OF
LABORATORY STUDIES
ate Mass Distributions
A*
AFE TOTAL
A
TOTAL
Carborundum In-stack Thimble
33 2 20 4
34 2 15 8
35 18 12 1
Method 5 (Quartz Fiber Fi
AP
37 9
38 6
1 22
2 17
4 30
Her)
AF
14
13
27
27
35
23
19
(mg/m3)
BP BF
Method 5
14 18
12 15
8 13
Method 5
9 17
4 15
B
TOTAL
32
27
21
(Glass F
26
19
B/A* B/A
32 1.45 1.18
27 1.58 1.00
21 0.70 0.60
1.13
1.00
A* = APS + AFS
-------
V. DISCUSSION OF RESULTS
A. Set 1 - Generation of False Particulate
1. Experimental Grid (Runs 1-13. 17)
Simultaneous sampling was carried out with a Method 5 train (Train B)
and a comparative research train (Train A) employing an in-stack filter.
The results of concurrent measurements of S02-S03 concentrations in the
stack gas compared to S02 measurements based upon analysis of impinger
scrubber solutions from the particulate trains yielded average S02 values
of -160 and -220 ppm, respectively. The difference in result of approxi-
mately 30 percent cannot be explained. Separate laboratory evaluations
of the collection efficiency of 10% H^2 yielded greater than 90 percent
recoveries for conditions simulating the field tests. The calculated
concentration of S02 for stiochiometric combustion of 0.5 percent sulfur
fuel corresponds to 165 ppm (or 130 ppm at 30 percent excess air). Be-
cause of these uncertainties, it is necessary to say that typical SOX
levels for this source are within 20 percent of 200 ppm, with S03 com-
prising about 1 percent of the total.
Train B particulate, designated B-total (or B), included the nozzle and
probe washings residue and the filter catch. For Train A, the parti-
culate catch was categorized in two ways. The catch designated A* in-
cluded particulate collected in the nozzle and by the in-stack filter.
The catch designated A-total (or A) also included the weight of the dried-
down residue from the probe washings and the catch on the external filter,
both ostensibly due to condensation of vapor species and penetration of
any particulate through the in-stack collector. Train A was used as
the control to indicate source variations and was therefore always run
according to the same procedure. Three collection times were utilized,
yielding the following results:
•3
Sampling Time Tests Catch-mg/m Ratio
(mi")A* A A*/A
60 1 33 39 0.85
90 2 60 65 0.92
120 4 66 70 0.94
It is observed that 4-6mg of particulate are collected behind the in-
stack filter, corresponding to less than 10 percent of the total catch.
This amount is not considered significant. The increase in particulate
catch with time is attributed to random variations in source emission
levels rather than false particulate formation on the in-stack collector.
37
Arthur DLittk Inc.
-------
Total particulate and sulfate values obtained during concurrent sampling
with A and B trains are compared on the basis of B/A* and B/A-total ratios
as well as on the combined average results for the seven valid paired runs.
These comparative values have been presented earlier in Tables 5 and 6;
the average results are repeated here for convenience.
o
Particulate - mg/m (Average of 7 runs)
A*_ A-Total B-Total B/A* B/A
Total Particulate 58 62 61 1.05 0.98
Sulfate 21 22 23 1.10 1.05
Sulfate/Total Part. 0.38 0.37 0.40
These average results indicate a near equivalence between the A and B
trains; furthermore, about 95% of the total particulate and sulfate is
collected in the in-stack section of the A train.
A detailed summary of the distribution of particulate and sulfate as well
as sulfate/total particulate ratios is given in Table 17 as a function of
collection point within the sampling trains. The average distributions
for particulate and for sulfate are almost identical for the two trains,
with equivalent scatter in the particulate and sulfate data as indicated
by similar values for the coefficient of variation (c.v.). The average
ratio of sulfate/total particulate is 0.38 for A train experiments com-
pared to 0.40 for B train experiments—a remarkably close agreement.
An especially uniform trend in the sulfate/particulate ratio exists for
the nozzle, probe and filter samples for both trains. No data are re-
ported in this table for the AFE sample fraction due to its very small
amount (<1 mg), which could result in large sulfate/particulate variations
on a percentage basis; on the average, this fraction accounts for only
about 3 percent of the total particulate catch (and 1 percent of the
total sulfate).
Identical sampling procedures and conditions were employed for collecting
all the A train data. The spread in results shown in Table 17, therefore,
reflects the level of variation to be expected in repetitive sampling.
The fact that values for the coefficient of variation are approximately
the same for A and B trains implies that there is no major influence
on the B train data as a result of differences in test conditions accord-
ing to the experimental grid. Likewise, the rather constant sulfate/total
particulate ratio of ~0.40 for all sample collection points within both
the A and B trains precludes the generation of false sulfate particulate
as a result of sampling configuration.
The total particulate and total sulfate collected by the A* and A-total
sections of the A train are compared to B train values in Figures 3 and 4,
respectively. Excellent correlation to a straight line fit is obtained
for B versus A* particulate values, with a coefficient of 0.98. Regression
38
Arthur D Little Inc
-------
TABLE 17
COMPARISON OF A TRAIN VERSUS B TRAIN PARTICULATE MASS AND SULFATE DISTRIBUTIONS
RUN NO.
A.
APS
AFS
PARTI CULATE
TOTAL PARTICULA
7
8
9
10
12
13
17
Average
c.v.(%)
18
17
17
12
11
16
20
16
20
74
72
78
83
88
72
65
76
10
APE
JE(%)
6
9
2
5
1
8
4
5
59
AFE
2
2
3
0
0
4
11
3
120
A* A-TOTAL
92
89
95
95
99
88
85
92
5
BP
64
63
63
64
44 '
23
42
1 52
31
BF B- TOTAL
36
37
37
36
56
77
58
48
23
TOTAL S04= (%)
7
8
9
10
12
13
17
Average
c.v.(%)
18
19
13
9
10
13
19
14
29
77
74
86
88
86
79
72
80
8
5
7
1
2
3
3
5
4
55
0
0
0
1
1
5
4
2
131
95
93
99
97
96
92
91
95
3
58
58
63
62
61
22
35
51
31
42
42
37
38
39
78
65
49
33
39
Arthur D Little Inc
-------
RUN NO. APS AFS APE
C. S0£
PARTICULATE
TABLE 17 (Cont'd)
AFE A* A-TOTAL BP
7
8
9
10
12
13
17
Average
c.v.(%)
33
36
30
24
33
39
32
32
15
35
31
46
36
33
54
38
39
21
29
26
19
17
9
21
43
23
47
BF B-TOTAL
35
32
43
34
34
52
33
38
19
33
31
42
34
34
50
34
37
18
33
33
33
37
54
30
44
38
22
43
41
33
41
31
41
59
41
22
37
36
33
38
42
43
53
40
16
40
Arthur DLittk Inc.
-------
100
80
60
40
co
20
B/A
y = 12.6 + 0.7B*
SE = 4.6, c.v. > 7.5*
Cor.Coef. « 0.973
A—A Runs 35-37
(Chapter V,C.)
J I
20
40
60
•0
100
A-Traln Particulatc (mg/m )
FIGURE 3 Cftmparison of Particulate Catch A Train Versus B Train
41
Arthur D Little Inc
-------
40
~ 30
n
E
5 »
(0
i
CO
10
— O
y = 16.0 + 0.35x
S£ « 5.8, c.v. « 24.5*
Cor.Coef. =0.481
0
- B/A
y = 14.5 + 0.40x
SE = 5.6; c.v. = 24.0%
Cor.Coef. = 0.303
10
20
30
40
A-Train Sulfate (mg/m )
FIGURE 4 Comparison of Sulfate Catch - B Train Versus A Train
42
Arthur D Little Inc
-------
analysis yields a slope of 0.80, an intercept of 15 mg and a coefficient
of variation corresponding to 6 percent. A similar comparison with A-
total values yields comparable results, with a displacement in intercept
of about 3 mg. These results imply a background (noise) level of ~15 mg
for the B train together with an improvement in collection efficiency for
the A train of about 20 percent in comparison to the B train. These two
factors result in approximately equal catches for the two trains when
loadings are in the range of 60-80 mg/m3. The comparative sulfate data
presented in Figure 4 provide similar results. A much larger scatter in
the data reduces the confidence in making a linear fit. The results,
however, indicate a B train noise level of perhaps as much as 10 mg.
The indication of substantial blank value for the B train suggests a
difference in A and B train collection (or sample recovery) characteris-
tics independent of the experimental grid variations in collection time
and external filter media and temperature. A possible explanation is the
difficulty in obtaining a quantitative recovery from the probe. Inde-
pendent laboratory studies conducted at ADL have shown that successive
washings of a probe liner previously employed for several Method 5 stack
tests each continue to provide indicated catches of 5 to 15 mg even
after four clean-ups by the Method 5 procedure. IR analysis of these
blank residues indicates the presence of considerable sulfate and x-ray
fluorescence analysis shows, in addition to large amounts of S, the
presence of Ca, K and Si, all typical components of flyash. Thus, the
empirical evidence is that glass probes employed for stack sampling (of
fossil-fuel boilers) become degraded through some unknown mechanism such
that successive probe clean-ups by the EPA procedure yield catches of
5 to 15 mg, much of which is sulfate. This leads to an additional 5 to
15 mg being collected during a Method 5 test that would not be present
for a new, non-degraded probe.
The experimental data also indicate that the in-stack section of the A
train collects approximately 20 percent more particulate than the B train,
based upon the observed least square slope of 0.80. This difference does
not appear to be due to "false particulate" generation, for the result
is the same irrespective of B train filter temperature, media or sampling
time. A reasonable explanation is incomplete recovery and measurement
of the actual particulate burden to the probe. Opportunities are avail-
able for particulate loss during probe washing, recovery from sample
bottles and solution dry down to a residue. It is not unreasonable to
suggest that incomplete sample recovery from the probe results in as
much as a 20 percent reduction in total measured particulate.
The elemental chemistry of the various samples from Trains A and B is
rather uniform, as shown by the analytical data presented in Table 11.
The optical emission spectroscopy (OES) was carried out on residue from
a hot water extraction combined with ultrasonic agitation of the filter
samples arid directly on the residues from the probe washings. The XRF
analyses were performed directly on the filters. All of the detected
elements are expected for this type of source. Mg and Al are present
as solids in the scaling additive, and Ni, V and Fe are the major
43
Arthur D Little Inc.
-------
metallic impurities in the fuel; sulfur, of course, is the major non-
metallic impurity in the oil. Most of the sulfur in the collected par-
ticulate is present as MgSO/j. The small amount of Pb observed by XRF is
associated with the MQF filter, where Pb is present as the major impurity;
the Pb is not extracted by a hot water leach and therefore is not observed
in the OES analyses. Based upon calibration standards, the trace of S
observed on the external filter of Train A (sample 5AFE) corresponds to
less than 0.5 mg S04= for the entire filter. The close agreement between
Train A and B samples and between filter and probe samples tends to
discount the presence of any significant amount of inorganic "false
particulate."
In summary, for the particular source tested, there has been no indica-
tion cf any "false particulate" being formed as a result of variation in
filter media, external filter temperature or level of loading (sampling
time). However, it can be argued that this facility discharges a "clean"
emission stream (nominally 200 ppm S02, 1-2 ppm $03 and -0.04 grams/m3)
and is not representative of oil-fired utility boilers in general. To
address this point, several additional experiments were conducted to
simulate high sulfur fuel combustion, thereby improving the opportunity
for forming "false particulate." These experiments are described in
the following section.
2. Additional Set 1 Experiments (Runs 14, 15, 16 and 18)
Several additional experiments were carried out under conditions
which were expected to maximize the opportunity for false particulate
formation in various parts of a sampling train. The results for these
experiments have been given in Tables 9 and 10. The results are inter-
preted as follows:
a. Run 15 ($02 enrichment)
Both trains were spiked with 502, resulting in a sampled gas
stream S02 concentration of -2000 ppm in the A train and -1500 ppm in
the B train. The A train utilized MQF filters, whereas an MSA 1106
filter was utilized in the B train. The data in Tables 9 and 10 show
that the B train has collected about 30 percent more particulate and
sulfate, respectively, than the A train. However, the sulfate to total
particulate ratio is constant at 23 percent for both trains compared
to a typical (average) ratio of -40 percent. The filter media thus
does not appear to have an influence on the amount of sulfate collected.
It is concluded that there was no conversion of SOz to sulfate, i.e.,
no "false particulate" was formed.
b. Both Trains Run as Method 5 Trains
1. Runs 14 and 14* (two consecutive runs without cleaning
probe in between)
The two trains were run in identical fashion except the Bl
train utlized MQF filters and the B2 train an 1106 filter. The primary
44
Arthur D Little Inc.
-------
purpose of this run was to obtain a supply of soiled filters for utiliza-
tion in other experiments. Nevertheless, the combined particulate catch
for Runs 14 and 14' yielded identical distributions for probe (35%) and
filter (65%) catches for both trains. The probe catch of both A and B
trains corresponded to 43 percent sulfate, in good agreement to the
typical value of 38 percent observed for the experimental grid. These
results tend to preclude the formation of "false particulate" by inter-
action of the sampled gas stream and particulate already collected along
the probe. Also, there is apparently no difference in result with either
a MQF or an 1106 filter.
2. Run 16 (One train (B2) was enriched with SO? (-1000 ppm
S02 added to the sampled stream) and the other
train (Bl) was run in the normal manner, with
-200 ppm S02 in the stack stream.)
The ratio of sulfate to the total particulate in the probe
washings (BP) was slightly higher for the train sampling the enriched
stream (46%) than for the normal -200 ppm $03 stream (38%); this dif-
ference (corresponding to 3 mg sulfate) is not considered to be signi-
ficant. The ratio of sulfate to the total particulate in the filter
catch (BF) was the same for both trains (52%). It is of interest to
note that 20% more particulate was collected by the normal train than by
the train with S02 enrichment. It is concluded that there was no con-
version of S02 to particulate sulfate in a stream containing ~1000 ppm S02.
3. Run 18 (Each train utilized a three-deep filter sandwich,
with the second filter facing downstream. Train
Bl utilized clean filters, whereas Train B2
utilized a filter sandwich consisting of a clean
filter, a soiled filter (from Run 14'Bl) and a
clean filter. The soiled filter was reversed to
contact the third filter, with the two being
weighed as a unit. S02 enrichment was employed
to provide gas streams containing -2000 and
-1500 ppm S02 for the two trains, respectively.)
The soiled filter showed no weight gain, precluding S02 con-
version to particulate on the filter. The two trains yielded comparable
results in terms of particulate distributions. Likewise, the clean
backup filter in Train Bl did not change in weight, reinforcing the con-
clusion that there is no S02~filter media interaction even with very
high S02 challenges.
None of these additional tests gave any indication of "false particulate"
formation, even under the extreme conditions of very high S02 concentrations,
45
Arthur D Little Inc.
-------
B. Set 2 - Method 5 Parameter Evaluation
A total of 24 experimental conditions were evaluated in order to deter-
mine the influence of varying filter size (6.3 and 10 cm), temperature
(250° cind stack) and nozzle type (buttonhook and 90°). Three replications
of the 23 = 8 unique combinations were tested according to an experi-
mental pattern based on statistical design principles. The observed
weight gains of filter and probe catches, normalized to a volume of 1 nr,
were determined for each test conditions, providing the basis for this
analysis.
Three Independent measurements were obtained for the probe catch. One
of the initial weights for Run 43B appeared to be in error and was deleted.
Differences among the three measurements were negligible (in a statistical
sense), and these values were averaged for this analysis. As a foreign
particle was noted in Run 50A, an overall average probe weight gain was
computed from the other 23 observations and inserted as an estimate.
The observed values, to be described, are summarized in Table 18 and
are presented graphically in Figure 5. The experimental design of the
test grid was given earlier in Table 2.
The particulate distribution between the probe and filter is presented
in Figure 6. The data show that, on the average, 40 percent of the catch
is in the probe. This compares to 52 percent found for the Set 1 experi-
mental grid runs.
The normalized values of particulate catch given in Table 18 were analyzed
according to a widely used statistical technique known as the "Analysis
of Variance," The basic assumption implicit in this technique is that
the observations are random variables, with well-defined distributional
properties, that can be expressed as an additive function of parameters
associated with possible sources of variation under investigation. If
appropriate assumptions are satisfied, it is possible to test hypotheses
and estimate effects concerning the parameters of interest. For these
data, the Analysis of Variance computations are summarized in Table 19.
The average (mean) value of the total particulate catch is 31.40 mg/m^,
with ei standard error of 4.67 and a coefficient of variation corresponding
to 15 percent. For example, under the null hypothesis that there is no dif-
ference among the three replicates, the value given for Mean Square (74.75
in this case) is an independent estimate of experimental error mean square.
Thus, computed F-values, corresponding to the ratio (Replicate Mean Square/
Error Mean Square) can be used to test the hypothesis that the replicates
have no effect on weight collected, as the theoretical behavior of this
ratio is known if this hypothesis is correct. For these data, this
46
Arthur D Little Inc
-------
TABLE 18
SUMMARY OF MEASURED PARTICULATE MASSES
IN SET 2 EXPERIMENTAL GRID SAMPLES
Block Run No. Probe Catch Filter Catch Total Catch
(mg/m3)(mg/m3) (mg/m3)
Train A Train B Train A Train B Train A Train B
1 40 11.75 21.70 23.15 23.67 34.90 45.37
2 42 9.00 9.06 19.29 17.19 28.29 26.25
3 45 8.48 8.04 17.89 18.44 26.37 26.48
4 43 8-57 8.66 14.93 11.90 23.50 20.56
5 46 28.68 9.99 24.62 20.16 53.30 30.15
6 47 10.63 11.37 25.86 23.51 36.49 34.88
7 49 7.98 7.71 13.60 15.71 21.58 23.42
8 39 14.35 28.75 17.62 18.10 31.97 46.85
9 50 10-02 8.94 16.30 18.30 26.32 27.24
10 44 6.95 13.04 17.72 17.42 24.67 30.46
11 48 8-47 22.83 21.84 16.08 30.31 38.91
12 41 12.80 9.52 19.91 23.09 32.71 32.61
47
Arthur D Little Inc
-------
50
40
30
£
CD
O
to
«£ 20
-------
40
30
e
>E
u
4->
O
(U
.0
o
s_
Q_
20
10
• A Train
O B Train
45P-55F
10 20 30
Filter Catch (mg/m )
40
FIGURE 6 Distribution of Participate Between Probe and
Filter for Set 2 Experimental Grid
49
Arthur D Little Inc
-------
TABLE 19
ANALYSIS OF VARIANCE TABLE FOR SET 2
EXPERIMENTAL GRID
Degrees of
Source of Variation Freedom
Replicates
Blocks with reps
Filter Size
Temperature
Nozzle Type
Size x Temp Interaction
Size x Type Interaction
Temp x Type Interaction
Experimental Error
2
9
1
1
1
1
1
1
6
SAMPLES: TOTAL PARTI CULATE
Sum of
Squares
149.49
894.88
0.21
3.68
181.04
150.43
5.56
25.07
131.06
Mean
Square
74.75
99.43
0.21
3.68
181.04
150.43
5.56
25.07
21.84
F- Value
3.42
4.55
0.01
0.17
8.30
6.88
0.25
1.14
« .
50
Arthur D Little Inc.
-------
hypothesis cannot be rejected and we conclude that there is no measurable
difference among the three replicates. This, of course, is a desirable
result since the reps were randomly determined, and should not contribute
significantly to variations in the data.
Performing similar analyses on the other sources of variation, it was
determined that effects due to blocks, nozzle type, and size by tempera-
ture interaction were statistically significant.
The significant difference noted among blocks is not surprising since
blocks were constructed for the purpose of explaining and estimating these
differences. The effect due to changing nozzle type can be estimated from
the first two reps only (blocks 1-8), as the contribution in the third
rep cannot be disassociated with the block effect; i.e., the same nozzle
type was tested in both cells of each block for blocks 9-12. Therefore,
the difference in collected particulate for the two nozzle configurations
is estimated as follows:
Nozzle Type Effect = (Average of observations in Blocks
1-8 with smooth nozzle)
= (Average of observations in Blocks
l-8w1th BH nozzle)
= 35.3 - 28.5
=6.8 mg/m
If the observed difference (6.8 mg/m^ in this case) has a "very unlikely
chance" of occurrence when, in fact, there should be no difference, the
hypothesis is rejected. For this analysis, the difference of 6.8 (or
larger) can be expected to occur with probability less than .05, thus
leading to rejection of the hypothesis.
In terms of estimation, the error mean square can be used to compute a
95% confidence interval about the "true" difference between nozzle con-
figurations; namely,
Nozzle Type Effect = 6.8 + 1.96 J-^^- x 2
= 6.8 + 4.6
3 3
Thus, there is 95% confidence that the interval 2.2 mg/m to 11.4 mg/m
includes the true increase in particulate that the 90° nozzle actually
yields over that of the button hook nozzle.
The size x temperature interaction, also declared significant, is illus-
trated in Figure 7. Again, valid comparisons can be made only from data
in blocks 1-8 as this interaction effect is also confounded with blocks
in the third replicate. With reference to Figure 7, it is observed that
a change in temperature level has a substantially different effect on
51
Arthur D Little Inc
-------
TEMPERATURE
FILTER
SIZE
2-1/2"
4"
Stack 250C
35.0
26.1
31.6
34.9
en
E
38 h
34
u
4->
30
OJ
O
-M
I.
26
24
22
2-1/2"
250C
Stack
4"
FIGURE 7 Schematic Presentation of the Significant Filter Size-
Temperature Interaction Determined by an Analysis of
Variance
52
Arthur D Little Inc.
-------
collected particulate when using a 2-1/2" filter compared to a 4" filter.
This mean difference and the 95% confidence interval about this mean
difference is 6.1 + 4.6.
In summary, the analysis of combined filter and probe weight gains has
indicated a measurable difference attributable to nozzle type, as well
as to the combined filter size-temperature interaction. In an attempt
to determine the relative importance of the probe catch and filter catch
with respect to the above observations, an Analysis of Variance was also
conducted for the filter catch alone. These data are summarized in
Table 20. Based upon this analysis, a significant difference is found for
the filter size, the nozzle type and the filter size-nozzle type inter-
action.
The results of this analysis are summarized in Table 21. It is observed
that the largest source of random variation is associated with the probe
catch, with a coefficient of variation corresponding to almost 40 percent.
This compares to about 5 percent for the filter catch and yields a
value of 15 percent coefficient of variation for the total particulate.
This suggests that it would be highly desirable to avoid particulate col-
lection in the probe as could be accomplished by the use of an in-stack
collector. Because of the substantially lower standard error associated
with the filter catch data, a more sensitive test can be made for statis-
tical significance among the various effects tested than by using the
probe (or total) catch values. Thus, an analysis of variance for the
probe data shows three significant effects—filter size, nozzle type
and a filter size nozzle type interaction. However, the resulting un-
certainty to an emission measurement of 1 to 3 mg out of 30, i.e., 10
percent or less, is acceptable. On the other hand, in utilizing values
for the total catch, only two effects are found to be significant—the
nozzle type and the filter size temperature interaction. In light of
the large confidence interval, however, these result in an uncertainty
of anywhere from 5 to. 35 percent.
Based upon an analysis of variance, it is concluded that there is sub-
stantial experimental scatter associated with the probe catch. Effects
found to have significant differences in particulate catch for the levels
tested and to also result in a substantial bias to the measured emission
levels include the type of nozzle and filter size-filter temperature
interaction.
As described earlier, a separate laboratory study of temperature distri-
butions and gradients in a Method 5 sampling train indicated that there
can be substantial differences between external control temperature and
gas temperature at the back of the probe. From this work, it is esti-
mated that the actual gas temperature at the back of the probe for the
Set 1 experiments was 150-160°C (300-320°F) rather than 120°C as inferred
from the external reference temperature. As a result of this difference
between Set 1 and Set 2 Method 5 experiments, several additional experi-
ments were conducted on selected samples from the Set 2 grid. A comparison
53
Arthur D Little Inc
-------
TABLE 20
ANALYSIS OF VARIANCE TABLE FOR SET 2 EXPERIMENTAL
GRID SAMPLES:
FILTER CATCH PARTICULATE
Degrees of Sum of
Source of Variation Freedom
Replicates
Blocks v/ithin reps
Filter Size
Temperature
Nozzle Type
Size Vs Temp Interaction
Size Vs Type Interaction
Temp Vs Type Interaction
Experimental Error
2
9
1
1
1
1
1
1
6
Squares
10.50
243.68
24.21
0.69
7.42
2.31
7.16
0.12
3.92
Mean
Square
5.25
27.08
24.21
0.69
7.42
2.31
7.16
0.12
0.65
F- Value
8.08
41.66
37.25
1.06
11.42
3.35
11.00
0.18
„
54
Arthur D Little, Inc.
-------
TABLE 21
SUMMARY OF THE ANALYSIS OF VARIANCE RESULTS
FOR THE SET 2 EXPERIMENTAL GRID SAMPLES
A. Summary of Data
Probe Filter Total
Mean Participate Catch (mg/m3) 12.4 19.0 31.4
Standard Error (mg/m3) 4.85 0.81 4.67
Coefficient of Variance (%) 39.0 4.2 15.0
B. Analysis of Variance Results
Significant Effects Mean Pifference 95% Confidence Internal
(mg/m3) (mg/mj)
Probe catch 0 +4.8
Filter catch -
• Filter Size: 2.5 1.7-3.3
• Nozzle Type: 1.4 0.6-2.2
• Size vs Type Interaction 1.3 0.5-2.1
Total catch -
• Nozzle Type 6.8 2.2-11.4
• Size vs Temp Interaction 6.1 1.7-10.9
55
Arthur DLittklnc
-------
of sulfate contents in the probe and filter catches and a comparison to
previous (Set 1) data on the basis of gas temperature at the back of the
probe and filter temperature was given in Table 14. With respect to
these data, there is little difference in sulfate content or distribution
as a function of filter temperature. However, there is an indication
of a higher level of sulfate collected in the probe for the higher
probe gas temperature. Also the proportion of sulfate in the particulate
is lower in the probe catch (and higher in the filter catch) for a gas
temperature 1n the probe of 120°C. The significance of these observations
is not considered very important due to the relatively small amounts of
sulfate mass involved.
C. Additional Field Tests
Several additional field tests were conducted at the same oil-fired elec-
tric utility representing an extension of laboratory evaluations of in-
stack particulate collectors. The sampling conditions and observed
particulate distributions were presented earlier in Tables 15 and 16.
Paired experiments were conducted in which one train (the B train) was
run according to the Method 5 procedure and the A train either utilized
an in-stack particulate collector or a quartz fiber external filter.
The in-stack collector was the Svenska Flakt Jabriken (Sweden) thimble
distributed in the United States by Carborundum Environmental Systems,
Inc., SKnoxville, Tenn. For our experiments, the thimble was slightly
modified by increasing the pipe thread connections from 1/4 inch to
3/8 inch. The thimble was mounted in line with the probe axis, and was
fitted with a buttonhook nozzle. The advantage of such a thimble over
the flat filter arrangement utilized in Set 1 experiments is a substan-
tially increased collection surface permitting more particulate to be
collected with no significant increase in system pressure drop.
The experimental results for these comparative runs given in Table 16
indicates that the majority of the A train particulate is, in fact,
collected in the thimble. An exception is noted for Run 35 in which
more than half of the particulate was found in the nozzle; it is probable
that the majority of this nozzle mass is extraneous,resulting from the
nozzle scraping the side of the port during removal of the sampling
train. Although some particulate is collected behind the thimble in
these runs, the masses are relatively small and are not especially
significant. Average distributions for the B train catch indicate
about 40 percent in the probe and 60 percent on the filter. The com-
parative B/A ratios are not especially sensitive due to the very low
levels of particulate mass collected.
These data are compared to the Set 1 experimental results in Figure 3,
in which the Run 35-APS sample was assigned a value of 2 mg/m3. The
agreement between the sets of data is very good, providing additional
support to the conclusions drawn from the Set 1 experiment grid samples.
56
Arthur D Little Inc.
-------
Runs 37 and 38 were both carried out by the Method 5 procedure, with the
A train utilizing a high purity quartz filter and the B train, a con-
ventional MSA 1106 BH filter. The quartz filter has the advantage of
a very low chemical background (other than silica) and is therefore
useful for in situ chemical analysis by such techniques as x-ray fluores-
cence analysis. The quartz filter was developed by Arthur D. Little,
Inc., under EPA Contract No. 68-02-0585; the samples used here were from
a pilot scale run carried out by Balston, Ltd. Runs 37 and 38 yield
almost identical results for the two trains, indicating an equivalence
between the quartz and glass fiber filters for use in Method 5.
57
Arthur D Little Inc
-------
VI. CONCLUSIONS
Based upon the experimental studies described herein of particulate col-
lection from an oil-fired steam boiler by Method 5 and in-stack collection
procedures, the following conclusions are reached:
1. The Method 5 procedure does not lead to the formation of
false particulate as a result of S02 oxidation or conversion
to sulfate or by other mechanisms for the particular source
conditions tested. Selective experiments involving enrich-
ment in S02 by as much as an order of magnitude to levels
of about 2000 ppm (representing combustion gases from fuel
oil containing about 3% sulfur) likewise failed to indicate
the formation of "false particulate" by Method 5 collection.
2. There is no evidence for chemical interactions involving
transformations from the gas phase to particulate at any
location within the sampling train that would have an in-
fluence on mass loading.
3. Based upon a set of experiments in which a research train
with an in-stack collector was used as the control, selected
variations to the Method 5 train resulted in no difference
to the expecte'd collection levels; variations included filter
media (glass fiber vs quartz), filter temperature (120°C
vs 150°C) and collection time (1 hr vs 2 hrs), where under-
lined conditions represent the Method 5 condition.
4. Based upon a straight line fit through the comparative
Method 5 versus in-stack collector particulate loadings
(normalized to 1 m^), the Method 5 train was observed
to collect about 20 percent less particulate than the in-
stack procedure. Also, a blank correction of 10-15 mg/m3
was indicated for the Method 5 train. It is thought that
these differences are due to the difficulties of obtaining
a quantitative recovery of the probe catch.
5. Particulate collected by the Method 5 train typically con-
tained 40 percent sulfate, irrespective of filter tempera-
ture for the range 120-150°C. With a gas temperature at
the back of the probe of approximately 150°C, both filter and
probe catches contained r40. percent sulfate, but for a tem-
perature of 120°C, the distribution of sulfate was ~17 percent
in the probe catch and ~57 percent in the filter catch. The
total particulate distributions between the probe and filter
were about equal at the higher gas temperature as compared
to a 37-63 percent distribution, respectively, for the lower
gas temperature. In summary, less of the particulate (and
sulfate) is caught in the probe for a gas temperature of 120°C
59
Arthur D Little; Inc
-------
compared to approximately equal probe-filter catches for a
gas temperature of ~150°C. Because of the substantially
lower standard error for the filter catch compared to the
probe catch (i.e., ~1 mg/m^ versus -5 mg/m^, respectively,
as given in Table 21), there is an advantage in maintaining
the gas temperature at the back of the probe at ~120°C.
6. The type of nozzle (buttonhook versus 90 degree bend) and a
combined filter size-temperature interaction (6.3 cm versus
10 cm dia and 120°C versus 150°C) are found to have a sta-
tistically significant result on the level of the particulate
catch, with differences of 6 to 7 mg/m3 with an experimental
uncertainty of +5 mg/m^.
7,, For Method 5 testing, the probe catch exhibits a standard
error of 5 mg/m^ and a coefficient of variation of 40 percent,
compared to 1 mg/m^ and 5 percent, respectively, for the
filter catch.
The above conclusions are made wifch respect to the particular stationary
source studied and are not necessarily valid for all oil-fired steam
generators. For example, combustion of a high sulfur fuel could yield
stack gases with a $03 content of 5 ppm or higher. If the probe and
filter temperatures were maintained at 150°C, the SO-} would remain in
the vapor state, and none should be collected. At 120°C, however,
amounts corresponding to 10 mg/m3 or more would be collected as parti-
culate,, Such a situation could alter some of the conclusions, such
as No. 5 given above.
In terms of these observations, there appears to be an advantage in
utilizing an in-stack train for particulate sampling. In this way,
sample recovery from the probe is precluded, thus, eliminating the
largest source of error in the Method 5 procedure. Also, any con-
densation of S03 in the probe or on the external filter would not be
included in the particulate catch. Finally, the design,operation
and sample .recovery aspects of an in-stack particulate train could
be greiitly simplified with respect to the present Method 5 procedure.
60
Arthur D Little Inc.
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-60Q/2-77-026
2.
3. RECIPIENT'S ACCESSION>NO.
4. TITLE AND SUBTITLE
EVALUATION OF STATIONARY SOURCE PARTICULATE
MEASUREMENT METHODS
Volume II Oil-Fired Steam Generators
5. REPORT DATE
February 1Q77
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
Edward T. Peters and Jeffrey W. Adams
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Arthur D. Little, Inc.
Acorn Park
Cambridge, Mass. 02140
10. PROGRAM ELEMENT NO.
1AD712 (1AA01Q)
11. CONTRACT/GRANT NO.
68-02-0632
12. SPONSORING AGENCY NAME AND ADDRESS
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Interim 10/73 - 2/76
14. SPONSORING AGENCY CODE
EPA-ORD
15. SUPPLEMENTARY NOTES
Volume I was issued as EPA-650/2-75-051a, June 1975
16. ABSTRACT
An experimental study was conducted to determine the reliability of the Method 5 pro-
cedure for providing particulate emission data from an oil-fired steam generator. The
study was concerned with determining whether any "false" particulate resulted from the
collection process of from the collected particulate. Variations to Method 5 tested
included filter media, size and temperature, nozzle type and sampling duration. Tests
were conducted at a 350 MW source utilizing low sulfur residual fuel.
Two series of field tests were conducted. In both cases simultaneous samples were
collected with two trains. In all cases one train was the standard Method 5 train.
The other train was either a modified Method 5 train with an in-stack filter or a
second standard Method 5 train.
Runs were preformed in which one or both trains were spiked with S02 at the nozzle to
provide sampled gas stream concentrations of ^2000 ppm S02 approximating a high sulfur
fuel. By comparing particulate and sulfate distributions between probe and filter
catches for the two trains, no evidence for the formation of "false particulate" for
the condition evaluated were found.
The in-stack filter method used has several advantages over EPA Method 5 including the
elimination of recovering the probe catch.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
* Air pollution
* Particles
* Collecting methods
* Evaluation
* Field tests
* Electric power plants
* Oil burners
13B
14B
10B
13A
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (ThisReport)
UNCLASSIFIED
21. NO. OF PAGES
66
20. SECURITY CLASS (This page)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
61
------- |