&EPA
United States
Environmental
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park f<
EMB Report
January 1982
Air
Benzene Fugitive Leaks
Coke Oven
By-Product Plants
Leak Frequency And
Emission Factors
For Fittings
In Coke Oven
By-Product Plants
-------
DCN 82-222-018-01-26 EMB Report No. 81-BYC-12
FINAL REPORT
LEAK FREQUENCY AND EMISSION FACTORS
FOR FITTINGS IN
COKE OVEN BY-PRODUCT PLANTS
Prepared by:
D.P. Wiesenborn, J.I. Steinmetz, and G.E. Harris
RADIAN CORPORATION
8501 Mo-Pac Boulevard
Austin, Texas 78759
Prepared for:
Dan Bivins
U.S. Environmental Protection Agency
ESED/EMB (MD-13)
Research Triangle Park, North Carolina 27711
EPA Contract No. 68-02-3542
Work Assignment No. 1
ESED No. 74/4j
8 February 1982
-------
CONTENTS
Section
1 Introduction and Summary 1
2 Detailed Results 5
2.1 Screening Value Distributions 6
2.2 Benzene and Nonmethane Hydrocarbon Leak Rates 8
2.3 Comparison of Benzene in Leak and in Line 8
2.4 Emission Factors 8
2.5 Relationships Between Screening Values and Leak Rates 15
3 Estimation Methodology 35
3.1 Emission Factors 35
3.2 Service Factors 37
3.3 Relation of Total Leak Rate to Instrument Screening
Value 37
4 Quality Control/Quality Assurance 39
4.1 Calibration Checks on Screening Instruments 41
4.2 Repeated Screening of Fittings 41
4.3 Analysis of Blind Gas and Liquid Standards 43
4.4 Accuracy Checks 43
4.5 Interfering Compounds 43
Appendix A Statistical Procedures A-l
Appendix B Supplementary Figures and Tables B-l
Appendix C Detailed Process Descriptions C-l
-------
TABLES
Number Page
1-1 Comparison of Emission Factors (kg/day) for Sources in
Coke By-Product Plants and Refineries '4
2-1 Screening Result Frequency Distribution by Source Type
and Benzene Service for all Units 7
2-2 Benzene and Nonmethane Leak Rates (KG/Day) from Sampled
Sources 9
2-3 Comparison of Percent Benzene in Line to Percent Benzene
in Equipment Leaks for Coke Oven By-Product Plants 11
2-4a Emission Factors for Sources with > 10% Benzene in Line
According to Source 13
2-4b Emission Factors for all Sources According to
Source 14
4-1 Summary of Accuracy and Precision of Sampling and Analysis
Techniques 39
4-2 Variance Component Estimates for OVA Screening Measurements
on Sources 41
A-l Summary of the Number of Sources Screened, Found Emitting,
and Measured A-3
A-2 Prediction Equations for Nonmethane Vapor Leak Rates Based
on Instrument Screening Values A-5
A-3 Prediction Equations for Benzene Vapor Leak Rates Based
on Nonmethane Vapor Leak Rates A-7
B-l Screening Result Frequency Distribution by Source Type and
Benzene Service Wheeling-Pittsburgh. Steel B-18
B-2 Screening Result Frequency Distribution by Source Type and
Benzene Service Republic Steel B-19
11
-------
Tables Continued
Number Page
B-3 Screening Result Frequency Distribution by Source Type and
Benzene Service Bethlehem Steel B-20
B-4 Confidence Intervals for Mean and Individual Leak Rates
by OVA Screening Value - Valves B-21
B-5 Confidence Intervals for Mean and Individual Leak Rates
by OVA Screening Value - Exhausters B-22
B-6 Confidence Intervals for Mean and Individual Leak Rates
by OVA Screening Value - Pump Seals (without liquid leak) B-23
B-7 Confidence Intervals for Mean and Individual Leak Rates
by OVA Screening Value - Pump Seals (with liquid leak) .. B-24
B-8 Confidence Intervals for Mean and Individual Leak Rates
by TLV Screening Value - Valves B-25.
B-9 Confidence Intervals for Mean and Individual Leak Rates
for TLV Screening Value - Pump Seals (without liquid leak) B-26
B-10 Confidence Intervals for Mean and Individual Leak Rates
by TLV Screening Value - Pump Seals (with liquid leak) .. B-27
iii
-------
FIGURES
Number Page
2-1 Emission factor comparison - coke (sources with _> 10%
benzene service) and refinery (all sources) - valves 16
2-2 Emission factor comparison - coke (all sources) and
refinery (all sources) valves 17
2-3 Emission factor comparison - coke (sources with > 10%
benzene service) and refinery (all sources) - pump
seals 18
2-4 Emission factor comparison - coke (all sources) and
refinery (all sources) pump seals 19
2-5 Emission factor comparison - coke (exhausters) and
refinery (compessor seals) 20
2-6 Nonmethane leak rate to OVA screening valve relationship -
valves 21
2-7 Nonmethane leak rate to OVA screening value - exhausters 22
2-8 Nonmethane leak rate to OVA screening value relationship
pump seals (without liquid leakage) 23
2-9 Nonmethane leak rate to OVA screening value relationship
pump seals (with liquid leakage) 24
2-10 Nonmethane leak rate to TLV screening value relationship
valves 25
2-11 Nonmethane leak rate to TLV screening value relationship
pump seals (without liquid leakage) 26
2-12 Nonmethane leak rate to TLV screening value relationship
pump seals (with liquid leakage) 27
2-13 Benzene leak rate to OVA screening value relationship
valves 28
2-14 Benzene leak rate to OVA screening value relationship
exhausters 29
IV
-------
Figures Continued
Number Page
2-15 Benzene leak rate to OVA screening value relationship
pump seals (without liquid leakage) .30
2-16 Benzene leak rate to OVA screening value relationship
pump seals (with liquid leakage) 31
2-17 Benzene leak rate to TLV screening value relationship
valves 32
2-18 Benzene leak rate to TLV screening value relationship
pump seals (without liquid leakage) 33
2-19 Benzene leak rate to TLV screening value relationship
pump seals (with liquid leakage) 34
A-l Distribution of leak rates (kg/day) - valves A-8
A-2 Distribution of leak rates (kg/day) - pump seals A-9
A-3 Distribution of leak rates (kg/day) - exhausters A-10
B-la Percent benzene in line to percent benzene in total leak
comparison : B-2
B-lb Percent benzene in line to percent benzene in vapor leak
comparison B-3
B-2 Nonmethane leak rate to OVA screening value relationship
(log 10 scale) - valves B-4
B-3 Nonmethane leak rate to OVA screening value relationship
(log 10 scale) - exhausters B-5
B-4 Nonmethane leak rate to OVA screening value relationship
(log 10 scale) - pump seals (without liquid leakage) . B-6
B-5 Nonmethane leak rate to OVA screening value relationship
(log 10 scale) - pump seals (with liquid leakage) .... B-7
B-6 Nonmethane leak rate to TLV screening value relationship
(log 10 scale) - valves B-8
B-7 Nonmethane leak rate to TLV screening value relationship
(log 10 scale) - pump seals (without liquid leakage) . B-9
v
-------
Figures Continued
Number Page
B-8 Nonmethane leak rate to TLV screening value relationship
(log 10 scale) - pump seals (with liquid leakage) B-10
B-9 Benzene leak rate to OVA screening value relationship
(log 10 scale) - valves B-ll
B-10 Benzene leak rate to OVA screening value relationship
(log 10 scale) - exhausters B-12
B-ll Benzene leak rate to OVA screening value relationship
(log 10 scale) - pump seals (without liquid leakage) .... B-13
B-12 Benzene leak rate to OVA screening value relationship
(log 10 scale) - pump seals (with liquid leakage) B-14
B-13 Benzene leak rate to TLV screening value relationship
(log 10 scale) - valves B-15
B-14 Benzene leak rate to TLV screening value relationship
(log 10 scale) - pump seals (without liquid leakage) .... B-16
B-15 Benzene leak rate to TLV screening value relationship
(log 10 scale) - pump seals ( with liquid leakage) B-17
C-l Wash oil scrubber configuration at Bethlehem steel,
Bethlehem, Pennsylvania C-3
C-2 Light oil recovery unit at Bethlehem steel, Bethlehem,
Pennsylvania C-4
C-3 Light oil recovery unit, Republic steel C-7
C-4 Light oil recovery unit, Wheeling-Pittsburgh steel C-9
VI
-------
SECTION 1
INTRODUCTION AND SUMMARY
This report presents a statistical analysis of test data for fugitive
nonmethane hydrocarbon and benzene emissions from coke by-product plants.
Test results have been previously presented in reports entitled, "Benzene
Fugitive Leaks; Coke Oven By-Product Plants; Emission Test Report" for the
following plants:
Plant . EMB Report No.
Bethlehem Steel Corporation, Bethlehem, PA. 80-BYC-9
Republic Steel Corporation, Gadsden, AL. 80-BYC-10
Wheeling-Pittsburgh Steel Corporation, Monessen, PA. 80-BYC-ll
This work was funded and administered by the Emission Measurement Branch
of the U.S. Environmental Protection Agency under Contract No. 68-02-3542. The
results of this study may be used in support of a National Emission Standard
for Hazardous Air Pollutants for benzene from coke oven by-product recovery
plants.
The purpose of this study can best be described by discussing the field
testing phase and the data analysis phase separately. Two objectives of the
field testing were as follows:
1) to count and screen all valves and pump seals and one-third of all
flanges, on process lines containing at least 4 weight percent
benzene; also to screen all exhauster seals and to determine the
percentage of benzene in each process line surveyed,
2) to measure the mass emission rate of benzene and of nonmethane
hydrocarbons at each leaking source identified during the screening.
-------
The objectives of the data analysis were as follows:
1) to compile leak frequency distributions for different benzene
service populations (all sources screened, all sources on lines
with at least 4 weight percent benzene, and all sources on lines
with at least 10 weight percent benzene),
2) to compare the percentage of benzene in the line to the estimated
percentage of benzene in the leak to determine if the benzene
concentration in the line is an adequate identifier of potentially
significant sources,
3) to prepare benzene and nonmethane hydrocarbon emission factors for
all sources and for sources on lines with at least 10 percent benzene,
4) to compare the coke oven by-product recovery emission factors with
emission factors for petroleum refineries.
The objectives of testing were met, and the results published in the
Emission Test Reports. The analysis objectives are met by results given in
Section 2 and briefly summarized in the following paragraphs. The methodology
used in estimating emission factors is discussed in Section 3, and the results
of the quality control practices are given in Section 4.
An examination of the population data indicates that no sources were
found in the 4 to 10 percent benzene service range. The contribution of
sources below 4 percent benzene service to total benzene fugitive emissions
was found to be quite small. These data indicate that the bulk of benzene
fugitive emissions can be attributed to sources on lines containing at least
10 percent benzene. Usually, only the light oil product lines contain 10
percent or more benzene.
The percentage of benzene in the line is generally a good indicator of
the percentage of benzene in the leak. The test results indicate, however,
that there are two exceptions. If benzene is, by far, one of the most volatile
components in the line, then there may be a higher percentage of benzene in
the leak than in the line. The second exception involves very small leaks,
where the benzene concentration in the sample gas may be only slightly higher
-------
than ambient. In these cases, the sampling and analytical precision is not
sufficient to resolve the benzene concentration accurately. This results in
a lack of correlation between sample and line benzene concentrations.
This program was not designed to produce an extensive data base from
which firm emission factors could be developed. A previous study of fugitive
emissions from petroleum refining, however, developed emission factors for
similar equipment types. Table 1-1 presents nonmethane hydrocarbon emission
factors for comparable sources in coke by-product plants and refineries.
The mean emission factors are reasonably close, especially for the important
valve category, and the confidence intervals for all categories show a
significant degree of overlap. Therefore, the use of refinery data to
characterize the coke by-products fugitive emissions is reasonable.
-------
TABLE 1-1. COMPARISON OF EMISSION FACTORS (KG/DAY) FOR SOURCES
IN COKE BY-PRODUCT PLANTS AND REFINERIES
Coke By-Product Planes
Source Type Emission Factor (Confidence Interval) Service
Valves 0.36 (0.03 - 3.3) > 10Z Benzene
Pump Seals 5.2 (1.3 - 18) > 101 Benzene
Exhausters 0.37 (0.006 - 10) All
(Compressors)
Refineries
Emission Factor (Confidence Interval) Service
0.26 (0.19 * 0.39) Light Liquid
2.7 (1.7 - 4.0) Light Liquid
1.2 (0.54 - 2.5) Hydrogen
Connections
0.007 (0.002 - 0.027)
All
-------
SECTION 2
DETAILED RESULTS
This section presents a detailed summary of all of the fugitive emission
data gathered at the:
Wheeling-Pittsburgh Steel Plant on November 24 to December
5, 1980,
Republic Steel Plant on December 8 to 12, 1980, and
Bethlehem Steel Plant on January 20 to 28, 1981.
Fugitive emissions testing was performed on fittings on process lines
containing at least 4 weight percent benzene. Benzene is concentrated in
the light oil recovery section, and therefore almost all of the testing was
performed in this area. All three plants have light oil recovery units that
operate by the absorption/stripping method of light oil recovery. At two
of three plants, the light oil is further fractionated. Light oil production
at the facilities ranges from 2,730 to 12,678 gallons per day and coke oven
gas production from 8.27 to 67.4 MMSCFD. A detailed description of each
process and the lines screened is included in Appendix^.
The fugitive emissions testing at each of these plants included both
"screening" and "bagging" procedures. Screening is a generic term covering
any quick portable method of detecting fugitive emissions. The initial
screening in this study was performed with a Century Systems Organic Vapor
Analyzer (OVA) Model 108. Bagging is a technique for measuring fugitive
emissions by enclosing the source in Mylarฎ and analyzing an equilibrium
flow of air through the enclosure. The screening and bagging procedures
are described in more detail in Section 4 of the individual test reports.
-------
2.1 SCREENING VALUE DISTRIBUTIONS
Screening value distributions are presented in Table 2-1 for all plants
combined and in Table B-l to B-3 for individual plants. These distributions
are reported by type of fitting and by the concentration of benzene in the
line. Three subcategories for the amount of benzene in the line were
considered:
All service (that is, all sources screened)
Sources on lines with at least 4 weight percent benzene
Sources on lines with at least 10 weight percent benzene
There were, however, very few sources found with benzene between 4 and 10
weight percent.
Sources with less than 4 weight percent benzene, other than exhausters,
were not intentionally screened. But at the Wheeling-Pittsburgh Steel and
Bethlehem Steel plants, it was not immediately known that the wash oil from
the light oil absorbers contained less than 4 weight percent benzene. Hence,
these wash oil lines were screened, even though subsequent analysis of
samples from these lines showed that the benzene concentration was less than
4 weight percent.
Exhauster seals were also tested, even though these are in the service
of coke oven gas with less than 4 weight percent benzene, because testing
in petroleum refineries indicated that this type of fitting can be a major
source of emissions. The exhausters are located on the coke oven gas line
upstream from the light oil recovery unit. The distribution of screening
values for exhausters is also presented in Table 2-1.
-------
TABLE 2-1. SCREENING RESULT FREQUENCY DISTRIBUTION BY SOURCE TYPE
AND BENZENE SERVICE FOR ALL UNITS
Benzene Screening
Service Value (PPMV)
All Service3 0 to 199
200 to 9,999
_> 10,000
Total Sources Screened
> < 10,000
Total Sources Screened
>_ 101 Benzene 0 to 199
200 to 9,999
> 10,000
Total Sources Screened
Flan
0
223
0
0
223
66
0
0
66
66
0
0
66
ges
I
100.0
0.0
0.0
100.0
100.0
0.0
0.0
100.0
100.0
0.0
0.0
100.0
Threaded
Fittings
0 X
70
0
0
70
59
0
0
59
59
0
0
59
100.0
0.0
0.0
100.0
100.0
0.0
0.0
100.0
100.0
0.0
0.0
100.0
Valves
0
226
13
8
247
117
10
8
135
117
10
8
135
*
91.
5.
3.
100.
86.
7.
5.
100.
86.
7.
5.
100.
5
3
2
0
7
4
9
0
7
4
9
0
Pump
Seals
1
18
5
9
32
6
5
9
20
6
5
9
20
Z_
56.3
15.6
28.1
100.0
30.0
25.0
45.0
100.0
30.0
25.0
45.0
100.0
Exhausters
a i
27 79.4
4 11.8
3 8.8
34 100.0
4 100.0
0 0.0
0 0.0
4 100.0
0
0
0
0 100.0
Total
0
564
22
20
606
252
15
17
284
248
15
17
280
%
93.1
3.6
3.3
100.0
88.7
5.3
6.0
100.0
88.6
5.4
6.1
100.0
& - Number of sources in each category
X Percent of total sources screened
a All service category includes all sources screened except exhausters, regardless of the percent benzene In the line.
No attempt was made 'to screen all sources with less than 6Z benzene, however, so these figures do not represent a
complete unit inventory.
-------
2.2 BENZENE AND NONMETHANE HYDROCARBON LEAK RATES
Table 2-2 summarizes the benzene and nonmethane hydrocarbon leak rates
in kilograms per day. All valves, pump seals, and exhausters that caused
an OVA reading greater than the ambient reading or that had a visible liquid
leak were sampled. Vapor phase leak rates were measured using the bagging
technique. Liquid leak rates were measured directly by timed collection in
a graduated cylinder, and a sample of the collected liquid was analyzed for
benzene. Each sampled source was screened immediately before sampling
with an OVA and with a J.W. Bacharach Instrument Company "TLV Sniffer."
These screening values are shown in Table 2-2 along with the weight percent
benzene in the line.
2.3 COMPARISON OF BENZENE IN LEAK AND IN LINE
Table 2-3 provides a comparison of the weight percent benzene in the
vapor, liquid, and total leak with the weight percent benzene in the line.
The weight percent benzene in the vapor sample is not directly comparable to
benzene in the line, because the sample is diluted with air. These values
for percent benzene are calculated as the ratio of benzene to nonmethane
hydrocarbon in the sample. This method is probably accurate unless the leak
is small. Those values of benzene in the leak that are much less than the
benzene in the line represent samples that had only slightly more benzene
and nonmethane hydrocarbon than the ambient air samples. The weight percent
benzene in the line is plotted against the weight percent benzene in the
total and vapor leak in Figure B-l (a & b).
2.4 EMISSION FACTORS
Benzene and nonmethane hydrocarbon emission factors were calculated
according to type of fitting. These factors are summarized in Table 2-4a
for sources with at least 10 percent benzene and in Table 2-4b for all sources
-------
TABLE 2-2. BENZENE AND NONMETHANE LEAK RATES (KG/DAY) FROM SAMPLED SOURCES
Plant"
Valves 1
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
Pump Seals 1
1
1
1
1
1
1
1
1
2
2
2
2
Source
I D^
18
23
32
73
121
122
123
124
125
129
40
84
87
91
103
104
108
114
115
116
120
121
124
129
139
141
98-1
98-0
117-0
131-0
139-1
139-ld
139-0
139-0(1
141-1
21-S
28-S
119-S
120-S
Before
Tenting
Value (ppmv)
1500
0
70000
2200
10000
100001
3600
100001
100000
10000
500
50
600
800
300
35
"iO
2oOO
350
350
29000
5000
65
110
140
30000
1000
45000
15000
6000
40000
5000
7000
1000
500
50000
100O01
60000
24000
Before
Tenting
TLV Screening
Value (ppmv)
400
No Data
8700
500
4200
10001
0
10001
600
800
2000
50
2600
1800
400
80
540
900
300
200
10001
2000
28
100
60
10001
4200
10001
10001
6500
7500
5400
4000
1100
700
8200
10001
8000
10001
Weight
Percent
In Line
39.00
71.50
71.50
71.50
71.50
71.50
71.50
71.50
71.50
71.50
NDC
ND
NAC
ND
NA
NA
63.00
63.00
63.00
63.00
63.00
63.00
63.00
63.00
63.00
63.00
85.00
85.00
77.00
77.00
39.00
39.00
39.00
39.00
0.97
71.50
71.50
71.50
71.50
Benzene Leak Rates
Vapor
0.0002
No Data
4.7389
0.0102
0.0028
0.4868
0.0172
1.4287
0.0147
0.0742
0.0000
0.0000
0.0014
0.0260
0.0015
0.0032
0.0000
0.0215
0.0031
0.0000
0.3511
0.0302
0.0000
0.0001
0.0006
0.2382
0.5953
1.6684
0.4086
1.1581
No Data
0.9945
No Data
0.6272
0.0095
3.1515
1.3240
4.8617
6.8238
Liquid
0.0000
3.2639
0.0000
0.0000
0.0000
0.4389
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
1.1425
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
1.3995
0.0000
1.7736
6.1584
6.1330
0.0000
0.0000
0.0000
0.0000
5.2077
13.1965
27.6367
Total
0.0002
No Data
4.7389
0.0102
0.0028
0.9257
0.0172
1.4287
0.0147
0.0742e
0.0000
0.0000
0.0014
0.0260
0.0015
0.0032
0.0000
0.0215
0.0031
0.0000
1.4936
0.0302
0.0000
0.0001
0.0006
0.2382
0.5953
3.0679
0.4086
2.9318
No Data
7.1275
No Data
0.6272
0.0095
3.1515
6.5317
18.0582
34.4604
Nonmethane Leak
Vapor
0.0015
No Data
5.1927
0.0162
0.0100
0.4929
0.0267
1.2395
0.0159
0.0701
0.0001
0.0000
0.0043
0.0339
0.0029
0.0050 f
No Data
0.0263
0.0071
0.0034
No Data
0.0386
0.0000
0.0005
0.0013
0.3488
0.7093
1.7137
0.4333
1.2764
1.5555
0.9069
1.0155
0.8073
0.0106
4.0297
1.4234
5.2887
7.9808
Liquid
0.0000
4.7578
0.0000
0.0000
0.0000
0.6029
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
2.0401
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
1.7860
0.0000
2.2610
15.3825
15.3172
0.0000
0.0000
0.0000
0.0000
7.8666
17.2277
36.0792
Rates
Total
0.0015
No Data
5.1927
0.0162
0.0100
1.0958
0.0267
1.2395
0.0159
0.0701
0.0001
0.0000
0.0043
0.0339
0.0029
0.0050
No Data
0.0263
0.0071
0.0034
No Data
0.0386
0.0000
0.0005
0.0013
0.3488
0.7093
3.4996
0.4333
3.5374
16.9380
16.2241
1.0155
0.8073
0.0106
4.0297
9.2900
22.5164
44.0599
(Continued)
-------
TABLE 2-2. Continued
Plane
Pump Seals 3
3
3
3
3
3
3
3
3
3
Exhausters 2
2
3
3
3
3
3
Source
IDb
26-1
27-0
66-1
68-1
98-0
98-0d
109-1.
109-Id
333-0
334-1
2
It
18
19
2ฐ4
20d
23
Before
Tenting
OVA Screening
Value (ppmv)
140
25
30
15
20000
1SOOO
6000
15000
3200
75000
2000
500
75000
100001
40000
65000
15000
Before
Tenting
TLV Screening
Value (ppmv)
700
35
35
14
10001
7600
8000
8000
10001
10001
1300
1100
10001
10001
10001
10001
10001
Weight
Percent
Benzene
In LineC
ND
NO
ND
ND
63.00
63.00
63.00
63.00
63.00
63.00
3.10
3.10
2.10
2.10
2.10
2.10
2.10
Benzene Leak Rates
Vapor
0.0257
0.0013
0. 0081
0.0017
0.8473
0.6937
0.8701
0.7878
3.0992
2.2132
0.0207
0.0082
2.1791
1.6867
1.8928
0.2317
0.0071
Liquid
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
4.4133
96.2911
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Total
0.0257
0.0013
0.0081
0.0017
0.8473
0.6937
0.8701
0.7878
7.5125
98.5043
0.0207
0.0082
2.1791
1.6867
1.8928
0.2317
0.0071
Nomnethane Leak
Vapor
0.0354
0.0028
0.0111
0.0026
Ho Data
1.0233
No Data
1.2266
3.1692
1.8661
0.1248
0.0598
5.3717
4.1286
No Data
0.5790
0.0303
Liquid
1.2716
0.0000
0.4187
0.1821
0.0000
0.0000
0.0000
0.0000
6.8424
149.289
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Rates
Total
1.3070
0.0028
0.4298
0.1848
No Data
1.0233
No Data
1.2266
10.0115
151.155
0.1248
0.0598
5.3717
4.1286
No Data
0.5790
0.0303
Plant codes are as follows:
1) Wheeling-Pittsburgh Steel
2) Republic Steel
3) Bethlehem Steel
1 denotes inboard seal and 0 denotes outboard seal of a pump with two seals. S denotes a single seal pump.
NA= Stream was not sampled and there was not sufficient data to make an estimate.
N0= No benzene detected (Benzene < 1 weight percent).
These sources were sampled twice because problems occurred during Initial sampling (leaky ambient air bag, THC not operating).
The benzene analyses and the nonmethane hydrocarbon analyses were performed on two different Instruments. The vapor sample for
these two sources was approximately 100Z benzene, and normal experimental error between the two analyses resulted In the anomalous
results of the benzene leak rates being larger than the nonmethane leak rates.
No data = No dnta was collected because vnpor leak stopped before fitcino could be sanmled, THC was not operating or sample bag leaked.
-------
TABLE 2-3. COMPARISON OF PERCENT BENZENE IN LINE TO PERCENT BENZENE
IN EQUIPMENT LEAKS FOR COKE OVEN BY-PRODUCT PLANTS
Equipment
Type
Block
Valves
Control
Valves
On-Llne
Pump Seals
Source
Plant" IDb
1 16
2 23
73
121
122
123
124
125
129
3 40
84
87
91
103
104
108
114
116
120
124
129
139
141
2 32
3 115
121
1 98-1
98-0
117-0
131-0
139-1
139-I8
139-0
139-0
141-1
Before
Tenting
Screening
Value (ppmv^
1500
0
2200
10000
100001
3600
100001
100000
10000
500
50
600
800
300
35
160
2000
350
29000
65
170
140
30000
70000
350
5000
1000
45000
15000
6000
40000
5000
7000
1000
500
Nonme thane
Leak Rate
(kg/Day)
0.00147
No Data
0.01622
0.00996
1.09578
0.02665
1.23947
0.01590
0.07013
0.00011
0.00000
0.00432
0.03388
0.00286
0.00503
No Data
0.02634
0.00336
No Data
0.00000
0.00050
. 0.00129
0.34882
5.19267
0.00714
0.03857
0.70926
3.49964
0.43330
3.53743
16.93801
16.22410
1.01553
0.80732
0.01057
Weight
Percent
Benzene
In I.lnec
39.00 (M)
71.50 (M)
71.50 (E)
71.50 (M)
71.50 (M)
71.50 (M)
71.50 (H)
71.50 (M)
71.50 (M)
(NO) (M)
(ND> (M)
(NA)
(ND) (M)
(NA)
(NA)
63.00 (M)
63.00 (H)
63.00 (M)
63.00 (H)
63.00 (M)
63.00 (M)
63.00 (M)
63.00 (M)
71.50 (M)
63.00 (M)
63.00 (M)
85.00 (M)
85.00 (M)
77.00 (H)
77.00 (M)
39.00 (H)
39.00 (M)
39.00 (H)
39.00 (M)
0.97 (M)
Weight
Percent
Benzene
In Total
Leakd
15.83
63.15
28.48
84.48
64.38,
115.26'
92.21,
105.78
0.00
32.61
76.77
53.57
62.74
81.62
0.00
20.00
46.83
68.30
91.26
43.65
78.30
83.93
87.66
94.30
82.88
43.93
77.69
90.02
Weight
Percent
Benzene
In Vapor
Leak
-------
TABLE 2-3. Continued
Equipment Source
Type Plant3 lDb
On-Line
Pump Seals 2 21-S
1L9-S
120-S
3 26-1
27-0
66-1
68-1
98-0
98-0B
109-1
109- Is
333-0
334-1
Ofฃ-Llne
Pump Seals 2 28-S
Exhausters 2 2
4
3 18
19
20
208
23
Before
Tenting
Screening
Value (ppmv)
50000
60000
24000
140
25
30
15
20000
15000
6000
15000
3200
75000
100001
2000
500
75000
100001
40000
65000
15000
Nonmcthane
Leak Rate
(ke/Dav)
4.02973
22.51639
44.05992
1.30700
0.00282
0.42978
0.18475
No Data
1.02333
No Data
1.22656
10.01154
151.155
9.29001
0.12483
0.05975
5.37174
4.12858
No Data
0.57896
0.03030
WelRht
Percent
Benzene
In LlneC
71.50 (M)
71.50
-------
TABLE 2-4a. EMISSION FACTORS FOR SOURCES WITH >; 107. BENZENE
IN LINE ACCORDING TO SOURCE
Source Type
Valves
Pump Seals
1 Sources were
Number Number
Screened1 Emitting1
135 21
20 -15
NONM ETHANE
Number Emission Factor
Liquid Estimate 95Z Confidence Number
Emitters5 (kg/day/source) Interval Emitting2
3 0.36 (0.03,3.3) 20
8 5.2 (1.3,18) 15
BENZENE
Number Emission Factor
Liquid Estimate 95Z Confidence
Emitters1 (kg/day/source) Interval
3 0.21 (0.02,1.7)
8 4.0 (L.I, 13)
I t
t_*J Eml11ing Sources were those releasing emissions detectable by che screening method (excluding those sources with measured
leak race of zero).
1 Liquid emitting sources were chose releasing liquid leakage detectable by visual inspection.
-------
TABLE 2-4b. EMISSION FACTORS FOR ALL SOURCES SCREENED ACCORDING TO SOURCE
NONMETHANE BENZENE
Number Emission Factor Number Emission Factor
Number Number Liquid Estimate 95% Confidence Number Liquid Estimate 95Z Confidence
bounce Type Screened1 Emitting2 Emitters3 (kg/day/source) Interval Emitting2 Emitters' (kg/day/source) Interval
Valves
Pump Seals
Exhausters
247
32
34
29
20
7
3
11
0
0.19
6.3
0.37
(0.02,1.4)
(1.3,28)
(0.006,10)
27
20
7
3
8
0
0.11
2.6
0.087
(0.01,0.71)
(0.56,11)
(0.002,1.4)
1 Sources were screened using instrument screening and inspection for visible leakage.
* Emitting sources were those releasing emissions detectable by the screening method (excluding those sources with measured
leak rate of zero).
' Liquid emitting sources were those releasing liquid leakage detectable by visual inspection.
-------
screened. In Figures 2-1 to 2-5, these factors are compared with emission factorc
developed during the refinery program and published in "Assessment of Atmospheric
Emissions from Petroleum Refining," EPA Report No. 600/2-80-075 (Vol. a-e),
Radian Corporation, Austin, Texas, July 1980.
2.5 RELATIONSHIPS BETWEEN SCREENING VALUES AND LEAK RATES
Empirical relationships between instrument screening value and total
(vapor plus liquid) leak rate were developed from the data for valves,
exhausters and pump seals. The relationships for total nonmethane leak
rate are presented graphically on arithmetic scales in Figures 2-6 through
2-12 for both OVA and TLV Sniffer instrument screening values. The same
relationships are presented on logarithmic scales in Appendix B. Similar graphs
for total benzene leak rate are given in Figures 2-13 through 2-19 and in
Appendix B. Each figure also gives the parameters of the fitted equation used
to develop the plot. Ninety-five percent confidence intervals for the predicted
mean leak rate bound each curve. Sources screening at greater than or equal
to 100,000 ppmv for the OVA instrument or greater than 10,000 ppmv for the
TLV Sniffer were not included in developing the graphs Cthese values are the
instrument scale maximum). Insufficient data were available to develop rela-
tionships for flanges and threaded fittings.
The relationship of total leak rate to instrument screening value are
presented in tabular form in Appendix B. These tables also include indivi-
dual leak rate confidence intervals. The mean leak rate confidence intervals,
which are presented in Figures 2-6 through 2-19, apply to the estimated
mean leak rate. There is 95 percent confidence that the actual mean leak
rate for a particular screening value falls between the mean leak rate
confidence limits. However, if a single source is to be screened and
measured, the measured leak rate should fall within the broader individual
leak rate confidence limits with 95 percent confidence.
15
-------
cr-
E 10
M
I
S
s 1
0
N
p 0.1
A
C
T
0
R 0.01
k
G
j 0.001
a
.
-t
i i
1
I
_i
t_
i
-*
_
.
V
NONMI
:THANE .BEI^
1 1
ZENE
HYDROCARBON
f f\17 T* T\ A m A
COKE DATA
T
t 1 I-T-I
i-Lj it
i i
u
i i
1 1 1
GAS LIGHT HEAVY HYDROGEN
LIQUID LIQUID
REFINERY DATA" ' '
(Nonmethane Hydrocarbon)
Bracketed intervals are 95% confidence intervals.
Figure 2-1. Emission factor comparison
and refinery (all sources)
- coke (sources with
Valves.
10% benzene service)
-------
E
M
I
S
s
I
0
N
F
A
C
T
0
R
k
8
10
o.i
0.01
0.001
! I
NONMETHANE BENZENE
HYDROCARBON
Till
GAS LIGHT HEAVY HYDROGEN
LIQUID LIQUID
(Nonmethane Hydrocarbon)
Bracketed intervals are 95% confidence intervals.
Figure '2-2. Emission factor comparison
and refinery (all sources)
- coke (all sources)
Valves.
-------
ฃ 1UU
M
I
S
S
I
o 10
N
F
A
C
T i
0
R
k
a
t 0.1
d
a
V
_ i
| ,
j j
1 1
1 1 f i
^i
. T ,
i i i i
1 1
1
NONMETHANE BENZENE LIGHT HEAVY ]
HYDROCARBON LIQUID LIQUID j
i
(Nonmethane Hydrocarbon)
Bracketed intervals are 95% confidence intervals,.
Figure 2-3. Emission factor comparison - coke (sources with
and refinery (all sources) pump seals.
10% benzene service)
-------
C J-vJU
H
I-
S
S
I
o 10
N
F
A
C
T 1
0
R
k
8
/ 0.1
d
0
V
i 1
i i
H
i i
II H
f
i
-
,
1 t
I 1
J
-
NONMETHANE BENZENE LIGHT HEAVY
HYDROCARBON LIQUID LIQUID
POKF DATA . - T?T?TT'TMrir?Y HATA . . .
!-- - V/WIN.C* ur\ i rv - ' ExiLr i.LNHit\I 1J A Lc\ .......
(Nonmethane Hydrocarbon)
Bracketed intervals are 95% confidence intervals.
Figure 2-4. Emission factor comparison - coke (all sources) and
refinery (all sources) pump seals.
-------
to
O
M
I
S 10
c
o
I
N i
F
A
C 0.1
T
0
R
0.01
k
a
d o.ooi
a
V
i ป
it
.
.
i
m
_
i_
t
ป
1
I
-f
i_
~ป II
f
I
-t
.
\
NONMETHANE BENZENE HYDROCARBON HYDROGEN
HYDROCARBON
I
1 rr\T/i7 nATA . REFINERY DATA
! (Exhausters) (Nonmethane Hydrocarbon- j
j Compressor Seals)
i i
!
Bracketed intervals are 95% confidence intervals,
Figure 2-5. Emission factor comparison - coke (exhausters) and
refinery (compressor seals).-
-------
edlcted Mean Line
9-
8-
7-
Total Leak Rate- 3.36(10"5) x (Screening Value)0'96
Least Squares Equation Used to Develop Chart
in(Total Leak Rate)- -11.40 + 0.96 ln(Screenlng Value)
Correlation Coefflcent-
Standard Error- 1.57 in(Leak Rate)
Number of Data Pairs- 18
Scale Bias Correction Factor- 3.00
-Mean Estimate
-95 % Confidence Interval
for Mean /'
L
E
A
K
R
A
T
E
K
Q
/
P
A
V
6-
B-
4-
3-
2-
26088
40080 60883
OVA SCREENING VALUE CPPMV)
80008
100000
Figure 2-6. Nonmethane leak rate to OVA screening valve relationship - valves.
-------
55-
N>
NJ
50-
4S-
40-
35-
30-
25-
20-
16-
10-
,0.64
Equation tor Predicted Mean Line
Total Leak Rate- 0.00197 x (Screening Value)''
Least Squares Equation Used to Develop Chart
Jtn(Totol Leak Rate)- -7.28 + 0.64 ln(Screenlng Value)
Correlation Coefficient- 0.68
Standard Error- 1.76 ฃn(Leak Rate)
Number of Data Pairs- 5
Scale Bias Correction Factor- 2.86
Mean Estimate
95% Confidence Interval
for Mean x
'
20000
30000
OVA SCREENING VALUE CPPMVJ
100000
Figure 2-7. Nonmethane leak rate to OVA screening value - exhausters.
-------
20-d
18-
16-
ป-*-
L
E
A 12-
K
R
A
T 10-
E
K
G
A
Y
6-
4-
Z-
.0.73
Equation for Predicted Mean Line
Total Leak Rate- 0.00214 x (Screening Value)'
Leaat Squares Equation Used to Develop Chart
In (Total Leak Rate)- -6.73 + 0.73 ln(Screenlng Value)
Correlation Coefficient- 0.92
Standard Error- 1.11 in(Leak Rate)
Number of Data Pairs- 31
Scale Bias Correction Factor- 1.79
Mean Estimate
95% Confidence Interval
for Mean
20000
teaaa eeeea
OVA SCREENING VALUE CPPMV)
80883
I 00800
Figure 2-8. Nonmethane leak rate to OVA screening value relationship - pump seals
(without liquid leakage).
-------
275-
250-
225-
200-
L
E
A
K 1 75-
R !
A
T 150-
E ;
K '.
G 125-
y
D
A
Y iaa-
Equation for Predicted Mean Line
Total Leak Rate- 0.0292 x (Screening Value) '
Least Squares Equation Used to Develop Chart
Zn(Total Leak Rate)- -4.12 + 0.73 ln(Screenlng Value)
Correlation Coefficient- 0.92
Standard Error- 1.11 ln(Lcak Rate)
Number of Data Pairs- 31
Scale Bias Correction Factor- 1.79
,'
-*
tป
ป'
jf*1^
.'
s"
..^
,-'
7S-
58-
25-
Mean Estimate
95 % Confidence Interva^-
for Mean *^'
2B000
eaa0a
OVA SCREENING VALUE CPPMV>
80800
180000
Figure 2-9. Nonmethane leak rate to OVA screening value relationship
(with liquid leakage).
pump seals
-------
NJ
Lr.
5.0-
4.5-
4.0-
3.5-
L
E
A
K 3.0-
R
A
T
E 2.E-
K
G
/
D 2.0-
A
Y
I .5-
I .0-
0.5-
0.0-
Equatlon for Predicted Mean Line
Total Leak Race- 1.66(10"*) x (Screening Value)0'89
Least Squares Equation Uaed to Develop Chart
tn(Total Leak Rate)- -10.52 -I- 0.89 ฃn(Scrcenlng Value)
Correlation Coefficient- 0.53
Standard Error- 2.06 ln(Leak Rate)
Number of Data Palra- 17
Scale Bias Correction Factor- 6.15
Mean Estimate
95 % Confidence Interval
for Mean .
2000
4000 6000
TL.V SCREENING VALUE CPPMV5
8000
10000
Figure 2-10. Nonmethane leak rate to TLV screening value relationship - valves.
-------
8-
N3
en
7-
6-
L
A
K
R
A
T
E
K
G
/
D
A
Y
S-
3-
Z-
0-
Equatlon for Predicted Heap Line
Total Leak Race- 4.84(10~4) x (Screening Value)0'93
Least Squares Equation Used to Develop Chart
Kn(Total Leak Rate)- -8.14 + 0.93 ฃn(Screenlng Value)
Correlation Coefficient" 0.93
Stanadord Error- 1.04 ln(Leak Rate)
Number of Data Palre- 22
Scale Bias Correction Factor- 1.66
-Mean Estimate
-95 % Confidence Interval
for Mean
2000
4000 6000
TLV SCREENING VALUE CPPMV3
6000
10000
Figure 2-11.
Nonmethane leak rate to TLV screening value relationship - pump seals
(without liquid leakage).
-------
I20H
I teH
80H
L
A
K
R
A
T
E
K
6
/
P
A
V
70H
B0H
S0H
40H
30H
I0H
a-t
,0.93
Equation for Predicted Mean Line
Total Leak Race- 0.00636 x (Screening Value)"
Least Squares Equation Used to Develop Chart
tn(Total Leak Rate)- -5.29 + 0.93 ฃn(Screenlng Value)'
Correlation Coefficient- 0.93
Standard Error- l.OA ฃn(Leak Rate)
Number of Data Palra- 22
Scale Bias Correction Factor- 1.66
-Mean Estimate
-95 % Confidence Interval
for Mean .-'
2000
4000 6000
TLV SCREENING VALUE CPPHV3
8000
10000
Figure 2-12.
Nonmethane leak rate to TLV screening value relationship
pump seals (with liquid leakage).
-------
18-
CO
16-
L
E
A
K
R
A
T
E
K
G
/
D
A
Y
12-
18-
6-
6-
Equation for Predicted Mean Line
Total Leak Rate- 1.32(10"5) x (Screening Value) 'ฐ
Least Squares Equation Used to Develop Chart
tn(Total Leak Rate)- -12.64 + 1.08 fcn(Screenlng Value)
Correlation Coefficient- 0.80
Standard Error- 1.79 In(Leak Rate)
Number of Data Pairs- 17
Scale Blaa Correction Factor- A.07
Mean Estimate
95 % Confidence Interval
for Mean
j
4-
2-
20000
40880 60000
OVA SCREENING VALUE CPPMV3
88888
180888
Figure 2-13.
Benzene leak rate to OVA screening value relationship
- valves.
-------
38-
33-
30-
27-
24
L
E
A
K 21
R
A
T 18
E
K
G 15
/
O
A
8-
6-
3-
8-
0 99
Equation for Predicted Mean Line
Total Leak Rate- 3.32(10 ) x (Screening Value)'
Least Squares Equation Used to Develop Chart
MTotnl Leak Rate)- -11.56 + 0.99 tn(Screenlng Value)
Correlation Coefficient- 0.77
Standard Error- 1.87 fcn(Leak Rate)
Number of Data Palra- 6
Scale Bias Correction Factor- 3.48
/
Mean Estimate
95 % Confidence Interval
for Mean
/
S
s
'
40000 60000
OVA SCREENING VALUE CPPMVJ
60000
100000
Figure 2-14. Benzene leak rate to OVA screening value relationship - exhausters.
-------
30.0-
OJ
o
L
E
A
K
R
A
T
e
K
G
/
0
A
y
27.5-
25.0-
22.5-
20.0-
17.6-
IB.0-
12.5-
10.8-
7.5-
5.8-
2.5-
0.0-
Equation for Predicted Mean Line
Total Leak Rate- 6.08(10"*) x (Screening Value)0'86
Least Squares Equation Used to Develop Chart
ฃn(Total Leak Rate)- -8.01 + 0.86 ln(Screenlng Value)
Correlation'Coefficient- 0.92
Stanadard Error- 1.13 ฃn(Leak Rate)
Number of Data Pairs- 27
Scale Bias Correction Factor- 1.83
Mean Estimate
95 % Confidence Interval
for Mean
2O003
40000 60080
OVA SCREENING VALUE CPPMVJ
80000
|00000
Figure 2-15.
Benzene leak rate to OVA screening value relationship - pump seals
(without liquid leakage).
-------
250-
225-
200-
176-
L
E
A iee
K
R
A
T 125-
E
K
G
/ 108-
P
A
V
76-
58-
2S-
0-
,0.86
Equation for Predicted Mean Line
Total Leak Rate- 0.00479 x (Screening Value)''
Least Squares Equation Used to Develop Chart
tn(Total Leak Rate)- -5.95 + 0.86 tn(Screenlng Value)'
Correlation Coefficient- 0.92
Standard Error- 1.13 ฃn(Leak Rate)
Number of Data Pairs- 27
Scale Bias Correction Factor- 1.83
Mean Estimate
95% Confidence Interval
for Mean
-r
60008
OVA SCREENING VALUE CPPMVJ
2B000
80000
100000
Figure 2-16.
Benzene leak rate of OVA screening value relationship - pump seals
(with liquid leakage).
-------
7-
6-
L ;
E 5-
A :
R :
A
T 4-
E :
K
G :
/
t> 3-
A :
Y
z-
Equation for Predicted Mean Line
Total Leak Rate- 1.50(10"5) x (Screening Value)1'20
Least Squares Equation Uaed to Develop Chart
tn(Total Leak Rate)- -12.86 + 1.20 fcnfScreening Value)
Correlation Coefficient- 0.66
Standard Error- 2.04 tn(Leak Rate)
Number of Data Pairs- 15
Scale Bias Correction Factor- 5.78
Mean Estimate
95 % Confidence Interval
for Mean
I "-
zeea
4000 eeaa
TLV SCREENING VALUE CPPMV)
8000
10000
Figure 2-17. Benzene leak rate to TLV screening value relationship - valves.
-------
7-
6-
S-
Equatlon tor Predicted Mean Line
Total Leak Rate- 7.94(10~5) x (Screening Volue)1'13
Least Squares Equation Used to Develop Chart
tn(Total Leak Race)- -9.93 + 1.13 gn(Scrcenlng Value)
Correlation Coefficient- 0.95
Standard Error- 1.0} ฃn(Leak Rate)
Number of Data Fairs- 17
Scale Bias Correction Factor- 1.63
-Mean Estimate
-95 % Confidence Interval /'
for Mean s"*
U)
OJ
aaaa
eaea
TLV SCREENING VALUE CPPMV3
aeaa
teaea
Figure 2-18.
Benzene leak rate to TLV screening value relationship -
pump seals (without liquid leakage).
-------
OJ
70-
60-
50-
L
A
K 40-
R
A
T
e
K 3B-
G
/
D
A
Y
20-
10-
0-
Equatlon for Predicted Mean Line
Total Leak Rate- 5.53(10~4) v. (Screening Value)1'13
Least Squares Equation Used to Develop Chart
fai(Total Leak Rate)- -7.99 + 1.13 ln(Screening Value)
Correlation Coefficient- 0.95
Stanadard Error- 1.03 ฃn(Leak Rate)
Number of Data Fairs- 17
Scale Bias Correction Factor- 1.63
-Mean Estimate
-95 % Confidence Interval
for Mean
X
X
2000
ea00
teaea
TLV SCREENING VALUE CPPMVi
Figure 2-1?.
Benzene leak rate to TLV screening value relationship -
pump seals (with liquid leakage).
-------
SECTION 3
ESTIMATION METHODOLOGY
This section summarizes the methods used to compute the emission
factors and prediction equations presented in Section 2. The mathematical
details and background are discussed in Appendix A.
3.1 EMISSION FACTORS
In developing emission factors, sources were divided into three groups:
sources identified as not emitting, emitting sources which were not measured,1
and emitting sources that were measured. The emissions were detected with
instrument screening and/or visual inspection. The data from measured sources
was used to develop an empirical relationship between the screening valve
and the nonmethane vapor leak rate. This relationship was then used to pre-
dict the nonmethane vapor leak rate for the unmeasured valves, pump seals
and exhausters. Separate relationships were developed for the different
source types.
After nonmethane vapor leak rates were predicted for the unmeasured sources,
as described above, the benzene vapor leak rate was predicted using the
relationship between nonmethane vapor leak rate and benzene vapor leak rate.
For the several cases in which only the benzene vapor leak rate was measured,
the nonmethane vapor leak rate was predicted using the relationship between
benzene leak rate and nonmethane leak rate. One relationship was found to
1 Almost all emitting sources were measured; however, the analysis of a few
samples was incomplete or erroneous and emission rates for these sources
were estimated.
35
-------
adequately describe the emissions from all source types, but a second rela-
tionship was required for sources in coke oven gas service (see Table A-3).
Emission factors for total (vapor plus liquid) nonmethane hydrocarbon
and benzene emissions were estimated from the combined measured and predicted
leak rates. Because of the high degree of skewness in the distribution of
total leak rates, a lognormal distribution was used to model the distribution
of leaking sources. An evaluation of the transformed data based on tests of
normality indicates that the data for most sources appeared to approximate
a normal distribution.
The emission factor computed from emitting sources was adjusted to
account for the non-emitting sources. This adjustment was made as a weighted
average with the weights being the frequency of emitting and non-emitting
sources. An "emitting" source is defined here as any source with a screening
value greater than zero.
As described earlier, fitted regression equations were developed to
estimate nonmethane vapor leak rate from instrument screening value for
unmeasured valves, pump seals, and exhausters. Regression relationships
between nonmethane vapor leak rate and instrument screening value were
developed from measured sources. The equations were fitted to the data on
a logarithmic scale. No relationship between instrument screening value and
nonmethane vapor leak rate could be developed for flanges and threaded
fittings due to insufficient data. Sources with instrument screening values
of zero or measured leak rates of zero were treated as having zero emissions.
If the nonmethane vapor leak rate was not measured, but the benzene
vapor leak rate was measured, then the nonmethane rate was estimated from
the benzene rate using regression equations. Conversely, the benzene rate
was estimated from the nonmethane rate, if measurements were available for the
nonmethane but not the benzene. The regression equations were fitted on a
logarithmic scale using measured source data using two product groupings.
36
-------
Total (vapor plus liquid) emission factors were estimated from the
combined measured (vapor and liquid) and estimated (vapor) leak rates. The
estimated mean total leak rate for emitting sources was first computed (leak
rates being averaged on a logarithmic scale). The average was converted to
the arithmetic scale by exponentiation and multiplication by a transformation
bias correction factor appropriate for the lognormal distribution. The esti-
mated mean total leak rate for emitters was combined with the zero leak rate
for nonemitters by a weighted average using the proportion of emitters and
nonemitters respectively as weights. This final weighted average is the
estimated emission factor per source.
3.2 SERVICE FACTORS
Four of the pump seals tested (sources 119-S and 120-S at Republic Steel
and 333-0 and 334-1 at Bethlehem Steel) were on loading pumps that only op-
erated intermittently, for an average of about one hour per day. These pump
seals were tested while in operation, and the measured leak rates presented
in Tables 2-2 and 2-3 are representative of the higher emissions during
loading. Screening of the pumps during their idle periods indicated .that
the vapor leak rate was comparable to that under operating conditions. There
was, however, significant liquid leakage during pump operation that stopped
completely when it was idle. In computing the average daily emission factors
for Table 2-4, the measured vapor leak rate was assumed to be constant, but
the measured liquid leak rate was multiplied by a service factor of 0.042 (one
hour of operation per 24 hour day) to account for the reduced emissions during
the idle period.
3.3 RELATION OF TOTAL LEAK RATE TO INSTRUMENT SCREENING VALUE
The presented relationships between total (vapor plus liquid) leak rate
and instrument screening value (.Figures 2-6 through 2-19, Figures and Tables
in Appendix B) were developed using regressing techniques on a logarithmic
scale.
37
-------
To investigate the relationship between instrument screening value and
leak, rate, a representative screening value was first chosen. Rescreening
values for both the OVA and TLV Sniffer were taken before and after each
source was sampled. In several cases the before and after leak measurement
screening value was found to be significantly different. Since the effect
(if any) of sampling a source immediately before taking an instrument screening
reading is unknown, the screening reading before leak measurement was chosen
as the representative instrument screening value and was used in the model
development.
The logarithm of the total (vapor plus liquid) leak rate (nonmethane
and benzene) was regressed on the logarithm of the instrument screening value
(OVA and TLV Sniffer) for valves and exhausters a relationship between
TLV instrument screening value and total leak rate could not be developed
for exhausters due to insufficient data.
In order to model the relationship between total leak rate and instrument
screening value for pump seals, an additional independent variable (indicating
whether or not the source released liquid leakage) was included in the
above-described regression equation. No relationship between instrument
screening value and total leak rate could be developed for flanges and thread-
ed fittings due to insufficient data.
Details of the model and the above described techniques are included
in Appendix A.
38
-------
SECTION 4
QUALITY CONTROL/QUALITY ASSURANCE
Quality control procedures were implemented to insure accurate, consistent,
and unbiased techniques during the testing. These procedures included:
calibration checks on screening instruments
repeated screening of fittings
analysis of blind gas and liquid standards
accuracy checks on leak rate measurement
analysis for interfering compounds.
The results for each of these are summarized in Table 4.1. The results are
described more fully in Section 5 of each of the individual test reports.
An analysis of quality control data indicates that some error is introduced
through screening, sampling, and analysis procedures. However, the cumulative
error from.sampling and analysis procedures is not the limiting factor with
respect to the confidence intervals of emission factors. The variance component
due to variations in sampling and analysis procedures is small compared to the
variance component associated with variation between individual sources and
day-to-day variation within one source. Based on this, it can be said that
the precision of the sampling and analytical techniques is adequate to support
the emission factors and other analyses.
39
-------
TABLE 4-1. SUMMARY OF ACCURACY AND PRECISION OF
SAMPLING AND ANALYSIS TECHNIQUES
Accuracy
Repeatability
Percent of Variation
in Data Attributable
to Measurement
Screening
More than 50 percent of calibration checks were
within + 30 percent of the standard; however,
approximately 20 percent of the differences
found were greater than + 100 percent of the
known concentration. Consistent negative drifts
were noted at the high level for both OVA devices.
Screening value for a given source with an average
screening value of x can be expected to vary from
x/7.4 to 7.4x within a short time period.
About 11 percent of the variability in the screen-
ing values from the selected sources with multiple
readings can be attributed to the screening devices.
More than 86 percent of the variation is attribu-
able to differences between the sources.
Accuracy
Vapor Analysis
Liquid Analysis
Sampling and
Analysis
Interfering Compounds
Sampling/Analysis
The average percent difference from standards for
the two analytical systems used is -1.1 percent,
indicating no significant bias in the analysis.
The average percent difference from liquid
standards for the analytical system used is 1.6
percent, indicating no significant bias in the
analysis.
Average recovery for the benzene and hexane
standards were 109 and 93 percent, respectively,
indicating slight biases in the total sampling/
analysis system.
No compounds were found with the same retention
time in the gas chromatograph as benzene.
40
-------
4.1 CALIBRATION CHECKS ON SCREENING INSTRUMENTS
The OVA and TLV instruments were calibrated in the morning each day
before they were used. The instruments were first tested on gas standards
to check the calibration drift. These checks indicate that some significant
drift did occur; however, other studies have shown that such drift often
occurs after a prolonged shut-down. Thus, the over-night drift does not
necessarily indicate that the previous day's screening results were inaccurate.
4.2 REPEATED SCREENING OF FITTINGS
Two valves and one pump seal were screened once in the morning and once
in the afternoon for five days to determine the reproducibility of screening
results. The variation in screening values was analyzed statistically to
identify the most significant sources of variation. Table 4-2 summarizes
the results of this variance analysis. About 87 percent of the variation
is attributed to differences between sources, 11 percent to the screening
devices and procedures, and 2.5 percent to the day-to-day variation in the
source itself.
The variation from replicate screenings (.variation in the screening value
due to such factors as screening instruments and procedures), can be used to
define the repeatability of the screening measurement. The. variance component
for replicate readings is 0.189 log (ppmv)2, and this determines that the
standard deviation between replicate readings of one source is 0.435 log
(.ppmv). This standard deviation means that a source with/'a mean screening
value of x may have screening values ranging from x/7.4 to 7.4x with. 95 percent
confidence. For example, screening values from a source which, has an average
screening value of 8000 ppmv could he between 1081 ppmv and 59_, 200 ppmv, 95
percent of the time.
41
-------
TABLE 4-2. VARIANCE COMPONENT ESTIMATES FOR OVA
SCREENING MEASUREMENTS ON SOURCES
Log 10 (Screening Value)
Source of
Variation
Total Variation
Variation Between
Degrees of
Freedom
29
2
Variance
Component
Estimate
Logio(ppmv)2
1.778
1.544
Percent Of
Total Variation
In Screening
Values
100
86.9
Individual Sources
Day-to-Day Variation
(within a source)
Variation From
Replicate Screenings
12
15
0.045
0.189
2.5
10.6
42
-------
4.3 ANALYSIS OF BLIND GAS AND LIQUID STANDARDS
Gas and liquid standards were prepared and submitted to the analysts,
without divulging the composition, to evaluate the quality of the analytical
data. Gaseous hexane standards were used to evaluate the precision of the
data for nonmethane hydrocarbons as determined by a Total Hydrocarbon Analyzer.
Similarly, gaseous and liquid benzene standards were used to check the preci-
sion of the benzene -analysis of gas and liquid samples by gas chromatograph.
The results of the analyses were an average of -1.1 percent different from the
standards for vapor analysis, and 1.6 percent different from the standards
for liquid analysis.
4.4 ACCURACY CHECKS
Accuracy checks were used to evaluate the overall accuracy of the sampling
and analysis techniques. The checks basically involve inducing a known
flow rate of a concentrated calibration gas, and then measuring this simulated
leak rate by the same techniques used to measure the leak rates from fittings.
Comparison of the measured leak rate with the known leak rate indicates that
the average recovery of the benzene and hexane standards were 109 and 93
percent, respectively. These values indicate an acceptable accuracy level
for the overall sampling and analysis effort.
4.5 INTERFERING COMPOUNDS
Liquid samples were selected from each plant for analysis by gas chromato-
graph/mass spectrometer (GC/MS). GC/MS would detect any compounds with the
same retention time on gas chromatograph as benzene, that might interfere with
the benzene analysis. No such compounds were detected in the samples from
coke by-products plants.
43
------- |