&EPA
United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park NC 27711
EPA-450/3-82-010
April 1982
Air
Fugitive Emission
Sources of Organic
Compounds
Additional
Information on
Emissions, Emission
Reductions,
and Costs
-------
EPA-450/3-82-010
Fugitive Emission Sources
of Organic Compounds -
Additional Information on Emissions,
Emission Reductions, and Costs
Emission Standards and Engineering Division
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air, Noise and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
April 1982
-------
This report has been reviewed by the Emission Standards and Engineering Division of the Office of Air
Quality Planning and Standards, EPA, and approved for publication. Mention of trade names or commercial
products is not intended to constitute endorsement or recommendation for use. Copies of this report are
available through the Library Services Office (MD-35), U.S. Environmental Protection Agency, Research
Triangle Park, N.C. 27711, or from the National Technical Information Services, 5285 Port Royal Road,
Springfield, Virginia 22161.
-------
CONTENTS
CONTENTS.
TABLES. .
FIGURES .
Section 1 - INTRODUCTION AMP SUMMARY
1.1 PURPOSE
1.2 HIGHLIGHTS OF CONCLUSIONS
1.3 SUMMARY OF RESULTS
1.3.1 Emission Factors (Section 2) .
1.3.2 Model Units (Section 3). . . .
1.3.3 Emission Reductions (Section 4'
Section 2 - EMISSION FACTORS
2.1 TECHNICAL BASIS PRESENTED IN THE BID.
2.1.1 Petroleum Refinery Study . . .
2.1.2 Four Unit EPA Study
2.1.3 EPA 6-Unit Study /
2.1.4 DuPont Study
2.1.5 Exxon Cyclohexane Study. . . .
2.1.6 EPA 24-Unit Study
2.1.7 EPA's Position at Proposal . .
2.2 NEW INFORMATION
2.2.1 German Studies on Fuqitive Emissions ....
2.2.2 Union Carbide Study
2.2.3 Maintenance Study
2.2.4 Analysis Report
2.2.5 Analysis of Allied HOPE Unit Data
2.2.6 SCAOMD Study
2.2.7 Coke Oven By-Product Recovery Plant and
Gas Plant Studies
2.2.8 Revision of Emission Factors for Nonmethane
Hydrocarbons From Valves and Pump Seals in
SOCMI Processes
2.3 PUBLIC COMMENT,
2.4 EPA's CONCLUSION. . .
2.4.1 Approach . . .
2.4.2 Evaluation of
2.4.3 Conclusions. .
Fugitive Emissions Information
2.5 REFERENCES.
1-1
1-2
1-4
1-4
1-5
1-5
2-1
2-1
2-10
2-10
2-12
2-14
2-14
2-19
2-23
2-23
2-24
2-26
2-30
2-33
2-37
2-37
2-41
2-44
2-47
2-47
2-51
2-56
2-74
11
-------
Section 3 - MODEL UNITS
3.1 TECHNICAL BASIS IN THE BID 3-1
3.2 NEW INFORMATION 3-4
3.3 PUBLIC COMMENT 3-8
3.4 EPA's CONCLUSIONS 3-9
3.5 REFERENCES 3-13
Section 4 - EMISSIONS REDUCTIONS
4.1 VALVES 4-1
4.1.1 Technical Basis Presented in the BID 4-1
4.1.2 New Information 4-12
4.1.3 Public Comment 4-24
4.1.4 EPA's Conclusions 4-30
4.2 PUMPS 4-44
4.2.1 Technical Basis Presented in the BID 4-44
4.2.2 New Information 4-45
4.2.3 Public Comment 4-47
4.2.4 EPA's Conclusions 4-47
4.3 SAMPLING SYSTEMS, OPEN-ENDED LINES, COMPRESSORS,
SAFETY RELIEF VALVES 4-54
4.3.1 Technical Basis Presented in the BID 4-54
4.3.2 New Information 4-58
4.3.3 Public Comment 4-58
4.3.4 EPA's Conclusions 4-58
4.4 CONTROL DEVICE 4-60
4.4.1 Technical Basis Presented in the BID 4-60
4.4.2 New Information 4-63
4.4.3 Public Comments 4-63
4.4.4 EPA's Conclusions 4-64
4.5 REFERENCES 4-69
Section 5 COST ESTIMATES
5.1 VALVES 5-1
5.1.1 Technical Basis in the BID 5-1
5.1.2 New Information 5_3
5.1.3 Public Comments 5_3
5.1.4 EPA's Conclusions 5.5
iv
-------
5.2 PUMPS 5-6
5.2.1 Technical Basis in the Bin 5-6
5.?.2 New Information 5-12
5.2.3 Public Comments 5-12
5.2.4 EPA's Conclusions 5-13
5.3 SAFETY/RELIEF VALVES 5-18
5.3.1 Technical Basis in the BID 5-18
5.3.2 New Information 5-23
5.3.3 Public Comments 5-23
5.3.4 EPA's Conclusions 5-23
5.4 SAMPLING SYSTEMS 5-24
5.4.1 Technical Basis in the BID 5-24
5.4.2 Mew Information 5-24
5.4.3 Public Comments 5-24
5.4.4 EPA's Conclusion 5-24
5.5 OPEN-ENDED LIMES 5-28
5.5.1 Technical Basis in the BID 5-28
5.5.2 Mew Information 5-28
5.5.3 Public Comments 5-28
5.5.4 EPA's Conclusions 5-28
5.6 COMPRESSORS 5-28
5.6.1 Technical Basis in the BID 5-28
5.6.2 New Information 5-30
5.6.3 Public Comments 5-30
5.6.4 EPA's Conclusions 5-30
5.7 OTHERS 5-31
5.7.1 Technical Basis in the BID 5-31
5.7.2 New Information 5-34
5.7.3 Public Comments 5-34
5.7.4 EPA's Conclusions 5-35
5.8 REFERENCES 5-37
APPENDIX A METHODOLOGY FOR ECONOMIC ANALYSIS A-l
A.I ESTIMATION OF SOCMI PRODUCTION, SALES, AND
PRICE VALUES A-l
A. 2 REPLACEMENT INVESTMENT PROJECTIONS A-8
A.3 METHODOLOGY FOR COMPUTING COST OF CAPITAL TO
SYNTHETIC ORGANIC CHEMICAL MANUFACTURERS A-ll
-------
A. 4 METHODOLOGICAL CONSIDERATIONS: PRICE AND RATE OF
RETURN IMPACTS A-17
A.5 REFERENCES A-23
APPENDIX B - AGGREGATION OF MODEL UNIT IMPACTS B-l
VI
-------
LIST OF TABLES
Table Page
1-1 Emission Factors for Average SOCMI
Unit, kg/hr/snurce 1-4
1-2 Equipment Counts for Fugitive VOC Emission
Sources in SOCMI Model Units 1-6
1-3 Estimated Control Effectiveness for Leak
Detection and Repair Programs for Valves and
Pumps (decimal percent) 1-7
2-1 Sampled Process Units from Nine Refineries
During Refinery Study ?-2
2-2 Leak Frequency in Refineries by Process
Unit - Valves 2-4
2-3 Leak Frequency in Refineries by Process Unit -
Compressor Seals 2-5
2-4 Leak Frequency in Refineries by Process Unit -
Relief Valves' " 2-6
2-5 Leak Frequency in Refineries by Process
Unit Pump Seals 2-7
2-6 Leak Frequency in Refineries by Process Unit -
Flanges. " 2-8
2-7 Percent of Sources Leaking and Emission Factor?
for Fugitive Emission Sources in Petroleum
Refineries (95 Percent Confidence Intervals) 2-9
2-8 Organic Chemical Industry Emission Factors -
Four Unit Study 2-11
2-9 Frequency of Leaks from Fugitive Emission
Sources in Synthetic Organic Chemical Units
(Six Unit Study) . . . ~ 2-13
2-10 Frequency of Leaks from Fugitive Emission Sources
in Two PuPont Plants 2-15
''-11 Frequency of Leaks from Fugitive Emission Sources
in Exxon's Cyclohexane Unit 2-16
-------
2-1? Factors Considered in Selection of Types of
Process Units - 24 Unit Study 2-18
2-13 Summary of SOCMI Process Units Fugitive Emissions
(24 Unit Study) '. ^ 2-20
2-14 Comparison of Fugitive Emission Source Leak
Frequencies Available in the BID 2-21
2-15 Comparison of Emission Factors Available in the
BID, kg/hr 2-22
2-16 Leak Frequency in Union Carbide Study 2-25
2-17 Estimated Fugitive Emission Loss in the Union
Carbide Unit 2-27
2-18 Matrix of Sampling/Screening for All Units 2-29
2-19 Percent Leaking for Each Chemical Unit Type
as a Function of Source Type and Stream Service
in SOCMI 2-31
2-20 Emission Factors and Leak Frequencies Calculated
in the Analysis Report with 95 Percent
Confidence Intervals 2-34
2-21 leak Frequency by Source and Service - HOPE Unit 2-35
2-2? Leak Rates for Leakers by Source and Service -
HOPE Unit 2-36
2-23 Summary of Leek Frequencies by Source Type and
Stream Service in Two Refineries in SCAQMD --
All Process Units 2-38
2-24 Leak Frequency for Sources in Coke Oven
Byproduct Units 2-39
2-25 Emission Factors and Leak Frequencies for
Fittings in Gas Plants 2-40
2-26 Estimated Leak Rate to Screening Value
Models for Groups of Valves 2-43
2-27 Revised Emission Factor Estimates for Nonmethane
Hydrocarbons from Valves and Pump Seals in Ethylene,
Cumene, and Vinyl Acetate Units - kg/hr/source 2-45
VI 11
-------
2-?8 Emission Factors Used by Industry Commenters
to Estimate Emissions from SOCMI Units 2-48
2-29 Summary of Available Data on Fugitive VOC
Emission Sources - Emission Factor, kg/hr 2-49
2-30 Summary of Available Data on Fugitive VOC
Emission Sources - Leak Frequency 2-50
2-31 Summary of Aspects of Fugitive Emiscions
Studies 2-52
2-32 Development of Emission Factors for Leakina
and Non-Leaking Sources Based on Refinery
Emissions Data (kg/hr) 2-61
2-33 Development, of Average SOCMI Emission Factors 2-62
2-34 Comparison of Emission Factors for Illustrative
SOCMI Cases (Ethylene, Cumene, and Vinyl Acetate
Units), Average SOCMI Unit, and Petroleum Refineries,
kg/hr/source 2-64
2-35 Final Average SOCMI Unit Emission Factors 2-70
2-36 Comparison of Actual Emission Factors for Coke
Ovens and Gas Plants with Factors Estimated Using
the Leak/No Leak Procedure, kg/hr/source 2-71
2-37 Comparison of Emission Factors for "Average" SOCMI
Unit to Emission Factors Submitted by Industry,
kg/hr/source 2-73
3-1 SOCMI Valve Characterization 3-5
3-2 SOCMI Pump Seal Characterization 3-5
3-3 Fugitive Fmission Sources for Three Model Units 3-6
3-4 Summary of Selected Equipment Counts for Model
Units and Units in SOCMI 24-Unit Study 3-7
3-5 Equipment Counts for Fugitive VOC Emission
Sources in SOCMI Model Units 3-10
4-1 Summary of Maintenance Study Results from
the Union Oil Co. Refinery in Rodeo, California 4-4
-------
4-2 Summary of Maintenance Study Results from the
Shell Oil Company Refinery in Martinez, California. . . . 4-6
4-3 Summary of EPA Refinery Maintenance Study Results .... 4-7
4-4 Maintenance Effectiveness Unit D Ethylene Unit
Block Valves ". 4-9
4-5 Summary of Maintenance Study Results 4~14
4-6 Summary of Analysis Report Results 4-15
4-7 Summary of Results for the Allied HOPE Study 4-16
4-8 Leak Occurrence and Recurrence of Valves and
Open-Ended Lines Determined from Several
Inspections - SCAQMD Study 4-19
4-9 Inputs and Outputs for the Leak Detection
and Repair (LDAR) Model 4-25
4-10 Summary of Available Data on Valves 4-32
4-11 Summary of Valve Maintenance Test Results 4-37
4-12 Comparison of Overall LDRP Effectiveness for
Valves in Model SOCMI Units 4-39
4-13 LDRP Effectiveness for Valves Using the ABCD Model. . . . 4-40
4-14 LDRP Effectiveness for Valves Using the
Modified-ABDC Model 4-42
4-15 LDRP Effectiveness for Valves Using LDAR Model 4-43
4-16 30-Day Occurrence Rate Estimates for Pumps 4-46
4-17 Summary of Available Pump Data for SOCMI 4-49
4-18 Summary of Input Data for Calculation of
Emission Reductions Due to LDRPs for Light
Liquid Pumps 4-52
4-19 Comparison of LDRP Effectiveness for Light
Liquid Pumps 4-53
4-20 LDRP Effectiveness Using LDAR Model for Pumps
in Light Liquid Service 4-55
-------
4-21 Regulatory Alternatives for Some Fugitive
Emission Sources in SOCMI 4-56
4-2? Comparison of LDRP Effectiveness for Safety/Relief
Valves Based on ABCD and LDAR Models 4-61
5-1 Estimated Versus Actual Monitoring Times for
SOCMI Process Units in the 24 Unit Study 5-2
5-2 Summary of On-Line Repair Time Data (Six Unit
Maintenance Study) 5-4
5-3 _ Initial Leak Repair Labor-Hours Requirement
for Valves by Model Unit 5-7
5-4 Total Costs for Initial Leak Repair for
Valves by Model Unit 5-7
5-5 Annual Monitoring and Leak Repair Labor
Requirements (Monthly Leak Detection and Repair
Program for Valves) 5-8
5-6 Annual Monitoring and Leak Repair Costs for
Monthly Monitoring of Valves by Model Unit 5-9
5-7 Net Annual Monitoring and Repair Costs of
Leak Detection and Repair Programs for Valves
by Model Units " 5-10
5-8 Equipment Cost for Control of Emissions from a
Pump Seal (last quarter 1978 dollars) 5-16
5-9 Met Annualized Cost of Equipment for Control of Emissions
from Pump Seals (last quarter 1978 dollars) 5-17
5-10 Initial Leak Repair Labor-Hours Requirement for
Pump Seals by Model Unit 5-19
5-11 Total Costs for Initial Leak Repair for
Pump Seals by Model Unit 5-19
5-12 Annual Monitoring and Leak Repair Labor Requirements
(Monthly Leak Detection and Repair Program for
Pump Seals) 5-20
5-13 Annual Monitoring and Leak Repair Costs for Pump
Seals by Model Unit 5-21
XI
-------
5-14 Net Annual Monitoring and Repair Costs of Leak Detection
and Repair Programs for Pump Seals by Model Unit. . . . 5-22
5-15 Relief Valve Control Costs, Four Systems 5-25
5-16 Costs for Closed Loop Sampling Systems 5"27
5-17 Costs for an Open-Ended Line Cap 5~29
5-18 Equipment Cost for Control of Emissions from
Compressor Seals (last quarter 1978 dollars) 5-32
5-19 Net Annualized Cost of Equipment for Control
of Emissions from a Compressor Seal (last quarter
1978 dollars) 5-33
A-l U.S. Production and Sales of Synthetic3 Organic
Chemicals, 1978 A-2
A-2 Weights Used to Estimate Historical Production
Sales and Prices of Synthetic Organic Chemicals A-9
A-3 Projections of Replacement Investment A-ll
A-4 Yields of Rating Class for Cost of Debt Funds,
1979 (prime rate = 13,50%) A-15
A-5 Financial Data for 100 Chemical Firms A-18
B-l Example of Emissions Estimated for Model Unit B
in Absence of Standards B-2
B-2 Estimate of Emission Reductions Achievable for
Model SOCMI Units in Mg/Yr B-3
B-3 Summary of Aggregated Capital Cost Estimates
for Model SOCMI Units, 1978$ B-4
B-4 Annualized Cost Estimates for SOCMI Model
Units, 1978$ B-5
xn
-------
LIST OF FIGURES
Figure Page
2-1 Emission factor v. empirical leak frequency: gas
valves for petroleum refinery and SOCMI units 2-66
2-2 Emission factor v. empirical leak frequency: light
liquid valves for petroleum refinery and SOCMI units. . 2-67
2-3 Emission factor v. empirical leak frequency: light
liquid pumps for petroleum refinery and SOCMI units . . 2-68
3-1 Total number of pumps per process unit as a function of
the rated annual production capacity (million Ibs) . . 3-2
3-2 Total number of the rated annual production unit as a
function of the rated annual production capacity
million Ibs) 3-3
4-1 Schedule of the fugitive emissions study at the
Allied HOPE Unit 4-17
4-2 Schematic diagram of the LDAR model 4-21
4-3 Valve leak occurrence rates with 95 percent confidence
intervals for processes in SOCMI maintenance study. . . 4-35
4-4 Intervals between pump seal replacements, (cumulative
percentage), last six years 4-48
4-5 Pump leak occurrence rates with 95 percent confidence
intervals for processes in SOCMI maintenance study. . . 4-50
xi 11
-------
1. INTRODUCTION AND SUMMARY
1.1 PURPOSE
A number of equipment components used in chemical manufacturing
processes in the Synthetic Organic Chemical Manufacturing Industry (SOCMI)
are sources of fugitive emissions. These fugitive emission sources (pumps,
valves, flanges, compressors, sampling systems, open-ended lines, and
pressure relief valves) have a common feature -- a point of interface of
the process fluid with the atmosphere. These points of interface such as
seals, packings, and gaskets have a tendency to fail mechanically and
thereby leak process fluid. These leaks (if the process fluid is a volatile
organic compound) cause emissions of volatile organic compounds (VOC) to the
atmosphere. The nature of these emission sources is that leaks can occur at
any time and the majority of the emissions come from a few sources of
different types. These leaks can be reduced by using equipment which
prevents or reduces leakage, such as sealless pumps, or by detecting leaks
and repairing them.
Standards of performance for fugitive emission sources of VOC in the
Synthetic Organic Chemical Manufacturing Industry (SOCMI) were proposed on
January 5, 1981 (46 FR 1136). The proposed standards were supported by
technical information and analysis in the Background Information Document
(BID) (EPA-450/3-80-033a). The BID contains estimates of uncontrolled and
controlled fugitive emissions of VOC and costs for the emission reduction
techniques.
Since the BID was published, several relevant reports concerning
fugitive emissions have been completed. A Federal Register notice was
published on April 14, 1981 (46 FR 21789) to announce the availability of
and inviting comments on these reports. The Office of Air, Noise and
Radiation of the U.S. Environmental Protection Agency (EPA) reviewed the
new information in these reports and the comments received on the new
1-1
-------
information. Based upon a review of all information now available and in
the comments, EPA has reached conclusions on the data and methodology which
EPA believes to characterize SOCMI fugitive rates, effectiveness of control
techniques, and control costs.
These conclusions are being made available in this document as a basis
for public comment. Comments on this document will be used in making
decisions concerning fugitive emission sources. Even though much of the
technical information and methodologies are the same as those presented in
the BID, some changes and some additions have been made to the technical
information and methodologies. In some cases EPA has concluded that certain
data are inappropriate for use in decision making. Because these conclu-
sions are fundamental to decision making concerning fugitive emission
sources, they are being presented in advance for review before the final
decisions are made.
The information in this document is restricted to selection of
technical information, data, and calculation methods. It is not a general
discussion of fugitive emission sources and emission reduction techniques as
presented in the BID. This document contains specific technical details
applicable to some aspects of the more generalized BID discussions.
Comments are requested on these selections as they are presented in the four
major sections in the remainder of this document:
Emission Factors (Section 2)
Model Units (Section 3)
Emission Reductions (Section 4)
Costs (Section 5)
In each of these sections the technical basis presented in the BID is
briefly reviewed. Also, each section presents information developed since
the publication of the BID and relevant public comments received by EPA.
Each section concludes with EPA's conclusions and related rationale.
1.2 HIGHLIGHTS OF CONCLUSIONS
Although several minor changes to the analysis presented in the BID are
being considered by EPA, two important changes concern the emission factors
and the methodology used for calculating the effectiveness of a leak
1-2
-------
detection and repair program. Both of these changes would result from
taking full advantage of the latest information and data.
The SOCMI emission factors presented in the BID were derived from
petroleum refineries. For reasons only partially understandable, the SOCMI
fugitive emissions data showed a difference in the number of leaking and
non-leaking sources (leak frequency) when compared to the data derived from
petroleum refineries. Even though the SOCMI plants tested were not
necessarily representative of SOCMI as a whole, an average leak frequency
from these SOCMI plants may better approximate SOCMI fugitive emissions than
the average leak frequency from petroleum refining plants. Therefore, the
SOCMI emission factors have been adjusted to more accurately reflect the
number of leaking and non-leaking sources found in the process units tested.
The adjustment was made by dividing the petroleum refinery emission factors
into leaking and non-leaking factors. These two factors v/ere then applied
according to the number of leaking and non-leaking sources found in SOCMI.
Two of the emission factors were developed in slightly different ways.
The sampling connection emission factor is the same for SOCMI and petroleum
refineries. The factor is a quantitative estimate of the amount of
emissions due to purging the sample lines and will be the same in both types
of process units. The gas valve emission factor was developed using the
approach of applying leaking and non-leaking emission factors to SOCMI leak
frequencies. However, SOCMI gas valve leaking and non-leaking emission
factors were used instead of petroleum refinery leaking and non-leaking
emission factors.
The method presented in the BID (Chapter 4) for calculating the effec-
tiveness of leak detection and repair programs was the ARCD model. The ABCD
model adjusts the theoretical maximum control efficiency (A) for occurrence
and recurrence of leaks (B), time to repair (C), and repair to less than
10,000 ppm but not to zero emissions (D). The model, although easy to use,
has some disadvantages and does not fully represent phenomena observed for
a"Leaking" means screening at or above 10,000 ppm with a portable VOC
monitor. "Non-leaking" means screening below 10,000 ppm. "Non-leaking"
does not mean that no emissions occur.
1-3
-------
fugitive emissions. With new information available, EPA can use a Leak
Detection and Repair (LDAR) model which better represents observed fugitive
emission behavior. The LDAR model, developed in September, 1981, also is
more flexible in allowing calculations for different types of leak detection
and repair programs.
1.3 SUMMARY OF RESULTS
The major results of EPA's review of the technical basis presented in
the BID, new information which has become available, and public comments are
summarized here. Details and reasoning supporting the results may be found
in the corresponding sections.
1.3.1 Emission Factors (Section 2)
The emission factors which EPA plans to use in estimating fugitive
emissions of VOC from SOCMI are shown in Table 1-1. They were developed by
using (1) SOCMI leak frequencies (i.e., number of leaking and non-leaking
fugitive emission sources) and (2) emission factors determined in petroleum
refineries and SOCMI units for leaking and non-leaking fugitive emission
sources. The resulting emission factors are lower than the emission factors
presented in the BID for all equipment components except flanges and
sampling connections.
TABLE 1-1. EMISSION FACTORS FOR AVERAGE SOCMI UNIT, kg/hr/source
Source
Average SOCMI Emission Factor
(kg/hr/source)
Pumps - light liquid
- heavy liquid
Valves - gas
- light liquid
heavy liquid
Compressors
Safety/relief valves - gas
Flanges
Open-ended lines
Sampling connections
0.0494
0.0214
0.0056
0.0071
0.00023
0.228
0.104
0.00083
0.0017
0.0150
1-4
-------
1.3.2 Model Units (Section 3)
The model units which will be used by EPA to estimate fugitive
emissions of VOC in a process unit are shown in Table I-?.. The equipment
counts remain the same as those in the BID analysis. The valves associated
with open-ended lines have been incorporated in the valve counts leaving
only emissions from the open-end in the "open-ended lines" category.
1.3.3 Emission Reductions (Section 4)
EPA has concluded that the effectiveness of leak detection and repair
programs for valves and pumps is most accurately calculated by using the
LDAR model. The conclusions on the effectiveness of leak detection and
repair programs for valves and pumps have been used to modify the ABCD
estimates presented in the BID. The results obtained with the LDAR are
presented in Table 1-3. EPA's estimate of efficiency for the use of open-
ended lines with plugs or caps, closed purge sampling systems and dual seals
with barrier fluids and vent systems and rupture disks on safety relief
valves remain unchanged and are about 100 percent. Enclosed combustion
devices can be expected to achieve 98 percent reduction in VOC. Most other
types of control devices (new and existing) can achieve a reduction of
95 percent. Furthermore, a 90 percent control efficiency to flares will be
assumed for purposes of estimation.
1-5
-------
TABLE 1-2. EQUIPMENT COUNTS FOR FUGITIVE VOC EMISSION SOURCES IN SOCMI MODEL UNITS
I
CTi
Equipment Component
Pump Seals
Light Liquid Service
Single mechanical
Dual mechanical
Seal less
Heavy Liquid Service
Single mechanical
Packed
Valves
Vapor service
Light liquid service
Heavy liquid service
Safety/relief valves
Vapor service
Light liquid service
Heavy liquid service
Open-ended lines
Vapor service
Light liquid service
Heavy 1 iquid service
Compressor seals
Sampling connections
Flanges
Model Unit
A
5
3
0
5
2
90
84
84
11
1
1
9
47
48
1
26
600
BID Analysis
Model Unit
B
19
10
1
24
6
365
335
335
42
4
4
37
189
189
2
104
2400
Equipment Counts
Model Unit
C
60
31
1
73
20
1117
1037
1037
130
13
14
115
581
581
8
320
7400
Model Unit
A
5
3
0
5
2
99
131
132
llc
1
1
104d
1
26e
600
Revised Analysis
Model Unit
B
19
10
1
24
6
402
524
524
42C
4
4
415d
2
104e
2400
Model Unit
C
60
31
1
73
20
1232
1618
1618
130C
13
14
1277d
8
320e
7400
Equipment components in VOC service only.
52% of existing units are similar to Model Unit A.
33? of existing units are similar to Model Unit B.
15% of existing units are similar to Model Unit C.
Seventy-five percent of gas safety/relief valves are assumed to be controlled at baseline; therefore the emissions
estimates are based on the following counts: A,3; 8,11; C,33.
All open-ended lines are considered together with a single emission factor; 100* controlled at baseline.
eSeventy-five percent of sampling connections are assumed to be controlled at baseline, therefore, the emissions
estimates are based on the following counts: A,7; B,26; C,80.
-------
TABLE 1-3. ESTIMATED CONTROL EFFECTIVENESS FOR LEAK DETECTION AND
REPAIR PROGRAMS FOR VALVES AND PUMPS (decimal percent)
Monitoring Interval
Monthly
Monthly/Quarterly
Quarterly
Semiannual
Annual
Valves
Gas
0.73
0.65
0.64
0.50
0.24
Light Liquid
0.59
0.46
0.44
0.22
(0.19)
Pumps
0.608
-
0.325
(0.076)3
(0.800)9
Values in parentheses indicate negative efficiencies.
1.3.4 Cost Estimates (Section 5)
EPA's estimate for the cost to reduce fugitive emissions are presented
in Section 5. The cost basis is similar to the basis presented in the RID.
1-7
-------
2. EMISSION FACTORS
Estimates of emissions from fugitive emission sources; in a process unit
are made by applying fugitive emission factors for each source (pumps,
valves, compressors, etc.) to the number of sources in a process unit or
plant. (Model Units are discussed in Section 3). This section contains a
review of the data which were available when the Bin (EPA-450/3-80-033a) ,
was published, and an explanation of the basis for the emission factors in
the BID. New information which has been made available since proposal is
presented and pertinent comments received from the public are summarized.
Finally, this section presents EPA's conclusions concerning the new
information and the numerical values which should be assigned to emission
factors in SOCMI.
2.1 TECHNICAL BASIS PRESENTED IN THE Bin
Information available at the time of proposal concerning VOC emissions
from fugitive emission sources was presented in the Background Information
Document. The information consisted of the results of several studies of
fugitive emissions. Descriptions of the studies and their results were
presented in Appendix C.
o
2.1.1 Petroleum Refinery Study''
Data concerning the VOC emitter frequencies (an emitter was defined as
a source screening at or above POO ppmv) and emission factors for various
fugitive sources were obtained primarily at nine refineries. More complete
information for compressor and relief valve emissions was obtained by
sampling at four additional refineries. Refineries were selected to provide
a range of sizes and ages and all of the major petroleum refinery processing
units were studied. The type of process units and the number of each
studied in the first nine refineries are listed in Table 2-1.
In each refinery, sources in six to nine process units were selected
for study. The approximate number of sources selected for study and testing
in each refinery is listed below:
2-1
-------
TABLE 2-1. SAMPLED PROCESS UNITS FROM NINE REFINERIES
DURING REFINERY STUDY
Number of
Refinery process unit sampled units
Atmospheric distillation 7
Vacuum distillation 4
Thermal operations (coking) 2
Catalytic cracking 5
Catalytic reforming 6
Catalytic hydrocracking 2
Catalytic hydrorefining 2
Catalytic hydrotreating 7
Alkylation 6
Aromatics/isomerization 3
Lube oil manufacture 2
Asphalt manufacture 1
Fuel gas/light-ends processing 11
LPG 2
Sulfur recovery 1
Other 3
2-2
-------
Valves ?50-300
Flanges 100-750
Pump seals 100-125
Compressor seals 10-20
Drains 20-40
Relief Valves 20-40
There were normally 500-600 sources selected in each refinery.
The distribution of sources among the process units was determined
before the selection and testing of individual sources was begun. Indivi-
dual sources were selected from piping and instrumentation diagrams or
process flow diagrams before a refinery processing area was entered. Only
those preselected sources were screened. In this way, potential bias which
could have resulted from observation of individual sources was eliminated.
The screening of sources was accomplished with portable organic vapor
detectors. The principal device used in this study was the J.H. Bacharach
Instrument Co.* "TLV Sniffer" calibrated with hexane. The components were
tested on an individual basis and only those components with measured VOC
concentrations in excess of 200 ppinv (i.e. emitters) were considered for
further study. Leak frequencies based on 10,000 ppmv (TLV-hexane) could be
determined from the reported results as well.
Emitter frequencies were evaluated for each process type. In a later
analysis of the source data, leak frequencies were reported by source and
service for each process type. These leak frequencies are shown in
Tables 2-2 through 2-6.
A substantial number of sources were enclosed and both the methane and
nonmethane emission rates from the sources were determined. An important
result of this program was the development of emission factors and the
quantification of the relationship between the maximum observed screening
value (VOC concentration) and the measured nonmethane leak rate. Emission
factors, emitter frequency, and leak frequency information generated during
this study are given in Table 2-7. The values reported in the BID were from
s1
4
an interim report (EPA-600/2-79-044). The values in Table 2-7 are slightly
different since they came from the final report (EPA-600/2-80-075c).
*Mention of a company name does not represent endorsement by EPA.
2-3
-------
TABLE 2-2, LEAK FREQUENCY IN REFINERIES BY PROCESS UNIT - VALVES'
GAS SERVICE
Unit
Code
15
13
22
1
27
17
2,4,8
r-o
i. 33
23
32,34
35
18
5
11
36
Unit
Identification
Atmospheric Distillation
Fuel Gas/Light Ends
Processing
Catalytic Cracking
Catalytic Reforming
Alkylation
Vacuum Distil lation
Catalytic Hydrotreating/
Refining
Aromatics Extraction
Delayed Coking
Dewaxing, Treating
Sulfur Recovery
Hyd roc rack ing
Hydrogen Production
Hydrodealkylation
Other
Number
Screened
69
185
50
22
59
13
63
18
27
37
7
6
8
6
0
Percent
Leaking
8.7
12.4
16.0
27.3
23.7
0
6.3
5.6
0
16.2
0
16.6
12.5
16.1
-
95% Conf.
Interval
(3,18)
(8,18)
(7,29)
(11,50)
(14,37)
(0,25)
(2,15)
(1,35)
(0,13)
(6,32)
(0,41)
(0,64)
(0,53)
(0,64)
-
LIGHT LIQUID SERVICE
Number
Screened
63
246
59
85
151
2
116
24
29
159
0
32
4
24
1
Percent
Leaking
9.5
17.1
10.2
10.6
19.2
0
4.3
4.2
6.9
6.9
-
9.4
0
0
0
95% Conf.
Interval
(4,20)
(12,23)
(4,21)
(5,19)
(13,27)
(0,84)
(1,10)
(0,21)
(0,23)
(3,12)
-
(2,25)
(0,60)
(0,14)
(0,100)
HEAVY LIQUID SERVICE
Number
Screened
143
27
80
0-
4
42
57
2
30
93
3
15
15
0
11
Percent
Leaking
0
0
0
-
0
0
1.8
0
0
0
0
0
0
-
0
953! C.I.
(0,3)
(0,13)
(0,5)
-
(0,60)
(0,8)
(0,9)
(0,84)
(0,12)
(0,4)
(0,71)
(0,22)
(0,22)
-
(0,28)
Source: Reference 5.
Note: A leak is defined as a TLV screening value xl.0,000 ppmv, calibrated with hexane.
-------
TABLE 2-3. LEAK FREQUENCY IN REFINERIES BY PROCESS UNIT - COMPRESSOR SEALS'
ro
i
HYDROCARBON GAS
Unit
Code
15
13
22
1
27
17
2,4,8
33
23
32,34
35
18
5
11
36
Unit
Identification
Atmospheric Distillation
Fuel Gas/Light Ends
Processing
Catalytic Cracking
Catalytic Reforming
Alkylation
Vacuum Distil lation
Catalytic Hydrotreating/
Refining
Aromatics Extraction
Delayed Coking
Dewaxing, Treating
Sulfur Recovery
Hydrocracking
Hydrogen Production
Hydrodeal kylation
Other
Nimber
Screened
6
62
37
1
10
0
0
1
14
12
0
0
0
0
0
Percent
Leaking
66.7
37.1
54.1
0
70.0
-
-
100
100
25.0
-
-
-
-
SERVICE
95% Conf.
Interval
(22,96)
(25,50)
(37,71)
(0,100)
(35,93)
-
-
(0,100)
(77,100)
(5,57)
-
-
-
-
HYDROGEN GAS SERVICE
Number
Screened
0
0
0
41
0
0
26
2
0
1
0
9
0
2
2
Percent
Leaking
-
-
-
41.5
-
-
50.0
100
-
100
-
0
-
0
0
95* Conf.
Interval
-
-
-
(26,58)
-
-
(30,70)
(16,100)
-
(0,100)
-
(0,34)
-
(0,84)
(0,84)
Source: Reference 6.
Note: A leak is defined as a TLV screening value >10,000 ppmv, calibrated with hexane.
-------
TABLE 2-4. LEAK FREQUENCY IN REFINERIES BY PROCESS UNIT*
- RELIEF VALVES -
GAS SERVICE
Unit
Code
15
13
22
1
27
17
2,4,8
33
23
32,34
35
18
5
11
36
Unit
Identification
Atmospheric Distillation
Fuel Gas/Light Ends
Processing
Catalytic Cracking
Catalytic Reforming
Al kylation
Vacuum Distillation
Catalytic Hydrotreating/
Refining
Aromatics Extraction
Delayed Coking
Dewaxing, Treating
Sulfur Recovery
Hydrocracking
Hydrogen Production
Hydrodeal kylation
Other
Number
Screened
16
57
19
12
29
1
12
4
3
10
0
4
2
0
0
Percent
Leaking
0
3.5
5.3
8.3
13.8
0
16.7
0
0
0
-
0
0
-
-
95% Conf.
Interval
(0,21)
(0,12)
(0,26)
(0,38)
(4,32)
(0,100)
(2,48)
(0,60)
(0,71)
(0,31)
-
(0,60)
(0,84)
-
-
Source: Reference 7.
Note: A leak is defined as a TLV screening value >10,000 ppmv, calibrated
with hexane.
2-6
-------
TABLE 2-5 . LEAK FREQUENCY IN REFINERIES BY PROCESS UNIT - PUMP SEALS9
rv>
i
LIGHT LIQUID
Unit
Code
15
13
22
1
27
17
2,4,8
33
23
32,34
35
18
5
11
36
aSource
Note:
Unit
Identification
Atmospheric Distillation
Fuel Gas/Light Ends
Processing
Catalytic Cracking
Catalytic Reforming
Alkylation
Vacuum Distil lation
Catalytic Hydrotreating/
Refining
Aromatics Extraction
Delayed Coking
Dewaxing, Treating
Sulfur Recovery
Hydrocracking
Hydrogen Production
Hydrodeal kylation
Other
: Fcicrcncc 6.
leak is defined as
Number
Screened
51
127
34
33
70
0
40
38
18
32
0
22
0
5
0
a TLV
Percent
Leaking
19.6
25.2
26.5
30.3
41.4
-
7.5
15.8
11.1
21.9
-
13.6
-
40.0
screeni
SERVICE
95% Conf.
Interval
(10,33)
(18,34)
(13,44)
(16,49)
(30,54)
-
(2.20)
(6,31)
(1,35)
(9,40)
-
(3,35)
-
(5,85)
ng value
HEAVY
Number
Screened
94
26
43
0
0
25
20
3
19
33
0
18
6
0
5
^10,000 ppmv.
LIQUID
Percent
Leaking
9.6
3.8
0
-
-
0
0
0
0
3.0
-
0
0
-
0
SERVICE
95* Conf
Interval
(4,17)
(0,20)
(0,8)
-
-
(0,14)
(0,17)
(0,71)
(0,18)
(0,16)
-
(0,19)
(0,46)
-
(0,52)
calibrated with hexar
-------
TABLE 2-6. LEAK FREQUENCY IN REFINERIES BY PROCESS UNITC"
- FLANGES -
ALL SERVICES
Unit
Code
15
13
22
1
27
17
2,4,8
33
23
32,34
35
18
5
11
36
3
Unit
Identification
Atmospheric Distillation
Fuel Gas/Light Ends
Processing
Catalytic Cracking
Catalytic Reforming
Alkylation
Vacuum Distillation
Catalytic Hydrotreating/
Refining
Aromatics Extraction
Delayed Coking
Dewaxing, Treating
Sulfur Recovery
Hyd roc rack ing
Hydrogen Production
Hydrodealkylation
Other
Number
Screened
411
148
252
263
269
78
351
15
32
300
6
33
19
15
3
Percent
Leaking
0
1.4
0
2.3
0.4
0
0.9
0
0
0.3
0
3.0
0
0
0
95% Conf.
Interval
(0,1)
(0,5)
(0,1)
(1.5)
(0,2)
(0,5)
(0,3)
(0,22)
(0,11)
(0,2)
(0,46)
(0,16)
(0,18)
(0,22)
(0,71)
Source: Reference 9.
Note: A leak is defined as a TLV screening value >10,000 ppmv, calibrated
with hexane.
2-1
-------
TABLE 2-7. PERCENT OF SOURCES LEAKING AND EMISSION FACTORS FOR
FUGITIVE EMISSION SOURCES IN PETROLEUM REFINERIES
(95 Percent Confidence Intervals)3
1X3
I
Equipment Type
Valves
Gas
Light
Heavy
Pump seals
Light
Heavy
Compressor
Li
quid
Liquid
Li
Li
quid
quid
Seals
Pressure Reli
Valves
Gas
Light
Heavy
Flanges
Open-Ended
Li
Li
Li
ef
quid
quid
nes
Percent of Sources
Screening >200 ppmv
TLV-hexane
30 (25, 36)
37 (33, 43)
9 (4.2, 14)
66 (62, 70)
28 (21, 35)
79 (71, 88)
32 (21, 42)
5.7 (2.7, 8.6)
22. 4C
Percent of Sources
Screening >10,000 ppmv
TLV-Fexane
10 (6.5, 14)
11 (7.5, 14)
0.2 (0, 1)
24 (20, 26)
2 (0, 5.2)
36 (25, 44)
7 (1.8, 13)
0.5C
7.7C
0
0
0
0
0
0
0
0
0
0
0
0
Average Emjssi
Factor0
kg/hr/source
.0268
.0109
(0
(0
.00023(0
.114
.021
.636
.086
.16
.006
.009
(o.
(0.
(0.
(0.
.00025(0
.0023
(0
.014,
.008,
.00009
073, 0
0086,
30, 1.
032, 0
.00009
.0007,
on
0.050)
0.016)
,0.021)
-17)
0.050)
32)
.22)
,0.0011)
0.007)
Source: Reference 10.
Values presented in,the above table are from the finalized Petroleum Refining Study
(EPA-600/2-80-075c) and may differ slightly frora values presented in the BID which were based on
an intermediate draft report (EPA-600/2-79-044).
cNo confidence intervals reported.
-------
Another major conclusion drawn from the petroleum refinery study was
the correlation between fugitive emission rates and process variables. The
only equipment or process variable found to correlate with fugitive emission
rates was the volatility and/or phase of the process stream. This result
led to the separation of sources by service (gas/vapor, light liquid, heavy
liquid) within equipment categories. Other variables such as line
temperature and pressure indicated much lower degrees of correlation.
2.1.2 Four Unit EPA Study13
EPA-IERL (RTP), directed a study of fugitive emissions at four SOCMI
units. This study was designed according to the same plan as was used for
the Petroleum Refinery Study. Only four process units were surveyed,
whereas the Refinery Study was based on sampling more than 64 different
units. As seen in Table 2-8, this small number of units resulted in
measured emitter frequencies and emission factor estimates with large
confidence intervals for most source types. The process units tested were
monochlorobenzene (MCB), butadiene (BUT), ethylene oxide/glycol (E06), and
dimethyl terephthalate (DMT).
Due to the small number of plants/processes sampled and the
experimental design of this study, the results were not considered to be
technically sound and therefore conclusions about emissions from SOCMI could
not be drawn. Since valid conclusions could not be drawn concerning the
magnitude of fugitive emissions in the SOCMI, the results of the study were
not used in the development of standards for fugitive emissions control.
This study demonstrated the need for more intensive sampling and screening
which was undertaken by EPA.
2.1.3 EPA 6-Unit Study14
The objective of this test program was to gather data on the percentage
of sources which leak (as defined by a VOC concentration at the leak inter-
face of ^10,000 ppmv calibrated with methane). To achieve this objective,
an attempt was made to screen all potential leak sources (generally
excluding flanges) on an individual component basis with a portable organic
vapor analyzer. The test crews relied on plant personnel to identify
equipment handling organics. Normally, all pumps and compressor seals were
2-10
-------
TABLE 2-8. ORGANIC CHEMICAL INDUSTRY EMISSION FACTORS -
Four Unit Study3
ro
i
Source Type
Compressor
Compressor
Flange
Flange
Flange
Pump
Pump
Pump
Pump
Sample Valve
Valve
Valve
Valve
Valve
Process
BUT
EOG
DMT
EOG
MCR
BUT
DMT
EGO
MCB
DMT
BUT
DMT
EOG
MCB
Total
Mo.
18
4
63
91
36
23
73
72
25
14
194
63
90
37
Number .
Emitting
18
2
2
12
11
13
21
22
23
9
63
4
6
9
Percent
Emitting
100.0
50.0
3.17
13.2
30.5
56.5
28.8
30.6
92.0
64.3
32.5
6.3
6.7
24.3
95 Percent
Conf. Interval
(81 -
(6 -
(0.39
(7.0
(16 -
(34 -
(19 -
(20 -
(74 -
(35 -
(25 -
(1.8
(3.9
(11 -
100)
93)
- ID
- 2)
48)
11}
41)
43)
9Q)
87)
39)
- 16)
- 17)
39)
Emission Factor
Estimate 95 Percent Confidence
kg/hr Interval for Estimate
0.0564
0.0043
0.0068
0.00001
0.00108
0.0514
0.0029
0.00886
0.00266
0.0768
0.00306
0.00184
0.00002
0.00004
(0.0222, 0.114)
(neg.d, 0.764)
(neg., >10)
(neg., >10)
(0.00011, 0.491)
(0.0077, 0.271)
(0.00046, 0.0152)
(0.0013, 0.0550)
(0.00014, 0.0342)
(0.0073, 0.555)
(0.001, 0.00764)
(0.00002, 0.582)
(neg. , >10)
(neg. , >10)
Source: Reference 15.
Emitting source is defined as an OVA-1?8 measurement >200 ppm equivalent methane or a leak rate >0.000005 kg/hr;
the calibration standard used was the major component expected to be in the leak.
Analysis of samples was by gas chromatography using a flame ionization detector, calibrated with standard gas mixtures
made to correspond to the major constituents of the process lines tested.
"Neg. indicates the value of confidence limit was less than 4.5 x 10" kg/hr.
Upper confidence level is extremely large. The exact value has no physical meaning.
-------
examined, and the percentage of valves in VOC service which were screened
ranged from 33 to 85 percent. All tests were performed with a Century
Systems Corporation Organic Vapor Analyzer, Model 108, with the probe placed
as close to the source as possible. The results of this study are shown in
Table 2-9.
Six chemical process units were screened. Unit A is a chlorinated
methanes production facility in the Gulf Coast area which uses methanol as
feedstock material. The individual component testing was conducted during
September 1978. Unit B is a relatively small ethylene production facility
on the West Coast which uses an ethane/propane feedstock. Testing was
conducted during October 1978. Unit C is a chlorinated methanes production
facility in the Midwest. This plant also uses methanol as the basic organic
feedstock. During the years prior to screening, several pieces of equipment
had been replaced with equipment the company felt was more reliable. In
particular, the company installed certain types of valves which they found
not to leak "as much" as other valves. The individual component testing was
conducted during January 1979. Unit D is an ethylene production facility on
the Gulf Coast, using an ethane/propane feed. The facility is associated
with a major refinery, and testing was conducted during March 1979. Units E
and F are part of an intermediate size integrated petroleum refinery located
in the North Central United States. Testing was conducted during
November 1978. Unit E is an aromatics extraction unit that produces
benzene, toluene, and xylene by extraction from refined petroleum feed-
stocks. Unit E is a new unit and special attention was paid during the
design and startup to minimize equipment leaks. All valves were repacked
before startup (adding 2 to 3 times the original packing) and all pumps in
benzene service had double mechanical seals with a barrier fluid. Unit F
produces benzene by hydrodealkylation of toluene. Unit F was originally
designed to produce a different chemical and was redesigned to produce
benzene.
2.1.4 DuPont Study16
DuPont conducted a program of fugitive emission measurement from pumps
and valves at two of their plants. The process types of the 5 and 10 year
2-12
-------
TABLE 2-9. FREQUENCY OF LEAKS FROM FUGITIVE EMISSION SOURCES .
IN SYNTHETIC ORGANIC CHEMICAL UNITS (Six Unit Study)'
r\j
i
CO
Equipment type
Valves
Open-ended 1 ines
Pump seals
Compressor seals
Control valves
Pressure relief
valves
Flanges
Drains
Unit A
Chloromethanes
Number Percent with
of screening
sources values
tested >10,000 ppmv
600 1
52 2
47 15
t-
52 6
7 0
30 3
h
Unit B
Ethylene
Number
of
Percent with
screening
sources values
tested
2301
386
51
42d
128
h
u
b
b
> 10 ,000 ppmv
19
11
21
59
20
Unit C
Chloromethanes
Number
of
sources
tested
658
-b
39
3
25
-
_b
b
Percent with
screening
values
^10,000 ppmv
0.1
3
33
0
Number
of
Unit D
Ethylene
Percent with
screening
sources values
tested
862
90
63
17
25
-
b
39
>10,000 ppmv
14
13
33
6
44
10
BTX
Number
of
Unit E
Recovery
Percent with
screening
sources values
tested
715
33
33C
_b
53
-
b
b
> 10, 000 ppmv
1.1
0.0
3.0
4.0
Unit f
Toluene HDA
Number
of
sources
tested
427
28
30
b
44
-
-b
b
Percent with
screening
values
>1 0,000 ppmv
7.0
11.0
10.0
11.0
Source: Reference 17.
hNo Data.
"Pump seals in benzene service have double mechanical seals.
Compressors tested in this unit were reciprocating compressors found in the LD polyethylene production area.
Note: Screening conducted with an OVA-108 calibrated with methane.
-------
old plants were not revealed. The OVA-108 was used for screening (leak
identification) and for leak rate determination (analysis of collected leak
vapors). The leak rate was determined by taking Tedlar bags partially
filled with air and enclosing the leaking valve. The hydrocarbon concen-
tration in the bags was recorded as a function of time. Leak rates were
determined for a total of 6-8 valves. Visual estimates of the initial bag
volume were assumed to be ±5 percent. The screening data from the DuPont
study are shown in Table 2-10.
DuPont concluded from the data collected that no significant difference
in leak frequency exists between manual and automatic control valves.
Significant trends were observed with changes in product vapor pressure. It
also seemed that full open or closed valve seat positions resulted in lower
leak frequencies than intermediate positions.
I O
2.1.5 Exxon Cyclohexane Study
A fugitive emissions study was conducted by Exxon Chemical Company at
the cyclohexane unit at their Baytown plant. The total number of valves,
pump and compressor seals, and safety valves were determined. For all
sources, except valves, all of the fugitive emission sources were sampled.
For valves, a soap solution was used to determine leaking components. All
leaking valves were counted and identified as either small, medium or large
leaks. From the set of valves found to be leaking, specific valves were
selected for sampling so that each class of leaking valves was in approxi-
mately the same proportion as it occurred in the cyclohexane unit.
Heat resistant mylar bags or sheets were taped around the equipment to
be sampled to provide an enclosed volume. Clean metered air from the filter
apparatus was blown into the enclosed volume. The sampling train was
allowed to run until a steady state flow was obtained (usually about
15 minutes). A bomb sample was taken for laboratory analysis (mass
spectrometry). Table 2-11 presents the results of the Exxon study.
1 Q
2.1.6 EPA 24-Unit Study
EPA coordinated a study in 1980 to develop information about fugitive
emissions in the SOCMI. A total of 24 chemical process units were selected
2-14
-------
TABLE 2-10. FREQUENCY OF LEAKS3 FROM FUGITIVE EMISSION
SOURCES IN TWO DuPONT PLANTS
Equipment
type
Valves
Gas
Light
Heavy
Pumps
Light
Heavy
1 iquid
liquid
1 iquid
1 iquid
Mo. of
leakers
48
35
11
1
1
1
0
No. of
non-1 eakers
741
120
143
478
36
6
29
Percent
leakers
6.1
23.1
7.1
0.2
2.7
14.3
0
Leak defined as 10 ppm or greater. Screening conducted with an OVA-108
calibrated with hexane.
Source: Reference 20.
2-15
-------
TABLE 2-11. FREQUENCY OF LEAKS9 FROM FUGITIVE EMISSION SOURCES
IN EXXON'S CYCLOHEXANE UNIT0
Equipment
Source i
Valves
Gas
Light
Liquid
Safety
Valves
p
Pump Seals
Compressor
Seals
Total Screened and
n Unit Sampled
136
201
15
8
N/A
136
100
15
8
N/A
Percent Emission
Leaking factor(kg/hr)
32 0
15 0
87 0
83 0
100 0
017H .
.008d'e
.064
.255
.264
99.8% Confidence
Interval (kg/hr)
0
0
0
0
0
.008
.003
.013
.082
.068
- 0.035
- 0.007s
- 0.5
- 0.818
- 1.045
N/A - Not available.
Leak defined as 10,000 ppm or greater. Monitoring instrument and
calibration gas unknown.
Source: Reference 21.
Double mechanical seal pumps and compressors were found to have negligible
leaks.
Not clear whether these factors are kg VOC/leaking valve or kg VOC/valve.
0
The values presented are direct conversions from the values given in the
reference. There is an apparent typographical error.
2-16
-------
for this purpose. The process units were selected to represent a cross-
section of the population of the SOCMI. Several of the factors considered
during process unit selection included annual production volume, number of
producers, process conditions, corrosivity, volatility, toxicity, and value
of the final products. Table 2-1? shows some of the factors considered in
selection of process unit types. Several of the chemicals on the list are
either carcinogens or suspected carcinogens: acrylonitrile, ethylene
dichloride, formaldehyde, perchloroethylene and vinyl chloride.
The screening work began with the definition of the process unit
boundaries. All feed streams, reaction/separation facilities, and product
and by-product delivery lines were identified on process flow diagrams and
in the process unit. Process data, including stream composition, line
temperature, and line pressure, were obtained for all flow streams. Each
process stream to be screened was identified and process data were obtained
with the assistance of plant personnel, in most cases. Sources were
screened by a two-person team (one person handling the hydrocarbon detector
and one person recording data).
The Century Systems Models OVA-108 and OVA-128 hydrocarbon detectors
were used for screening. The HNU Systems, Inc., Model PI 101 Photoioniza-
tion Analyzer was also used to screen sources at the formaldehyde process
unit. The detector probe of the instrument was placed directly on those
areas of the sources where leakage would typically occur. For example, gate
valves were screened along the circumference of the annular area around the
valve stem where the stem exits the packing gland and at the packing
gland/valve bonnet interface. All process valves, pump seals, compressor
seals, agitator seals, relief valves, process drains, and open-ended lines
were screened. From five to twenty percent of all flanges were randomly
selected and screened. For the purpose of this program "flange" referred to
any pipe-to-pipe or tubing-to-tubing connection, excluding welded joints.
Each screening instrument was calibrated on a daily basis, at a
minimum. The model OVA-108 instruments, with logarithmic scales reading
from 1 n'pmv to 10,000 ppmv, were calibrated with high (8,000 ppmv) and low
2-17
-------
TABLE 2-12. FACTORS CONSIDERED IN SELECTION OF TYPES OF PROCESS UNITS'
- 24 Unit Study -
ro
i
co
Chemical Product
Acetaldehyde
Acetone
Acrylonitri le
Adipic acid
Cumene
Ethyl ene
Ethylene dichloride
Formaldehyde
Methyl ethyl ketone
Methyl methacrylate
Perchloroethylene
Phenol
1 ,1 ,1-Trichloroethane
Trichloroethylene
Vinyl acetate
Vinyl chloride monomer
Production
Q volume
103 Ib/yr (1977)
<1.7
2.2
1.6
1.5
2.6
25.4
11.0
6.0
0.5
0.7
0.6
2.3
0.6
0.3
1.6
6.0
Number of
producers
(1977)
5
13
4
5
14
31
12
16
5
3
8
13
3
5
6
12
Vapor pressure
of product,
mm Hg at 20°C
760+
160.5
84.9
<1.1
3.3
760+
60.2
760+
76.2
28.1
13.7
<1.7
95.6
57.0
84.2
760+
Threshold .
1 imit value
of product,
ppmv
100
1000
20(skin)
NAC
50(skin)
NAC
10
2
200
100
lOO(skin)
5(skin)
350
100
10
5
Cost of
product,
$/lb (1980)
0.29
0.30
0.38
0.53
0.27
0.24
0.14
0.07
0.35
0.50
0.23
0.36
0.31
0.27
0.34
0.22
aSource: Reference 22.
cThreshold limit value not available or not established.
-------
(500 ppmv) concentration methane-in-air standards to ensure accurate
operation at both ends of the instrument's range. The model OVA-128
instruments, with linear readouts ranging from 0 ppmv to 1,000 ppmv, were
also calibrated with high and low concentration standards. A pre-calibrated
dilution probe was used with the OVA-128 when calibrating with the
8,000 ppmv standard.
The HNU Photoionization instrument, used to screen the formaldehyde
process unit, was calibrated with isobutylene, which has an ionization
potential close to that of formaldehyde.
Results of the screening program at the 24 process units are summarized
in Table 2-13.
2.1.7 EPA's Position at Proposal
After considering the data available at proposal, EPA estimated
fugitive emissions of VOC from SOCMI by using the emission factors developed
in the study of fugitive emissions in petroleum refineries. The petroleum
refinery data are the most comprehensive and definitive fugitive emissions
data available and the study had been designed to produce emission factors
which would estimate uncontrolled fugitive emissions of VOC. The use of
petroleum refinery data to characterize emissions from SOCMI units was based
on the position that equipment handling similar substances would behave
similarly and therefore emissions from similar equipment in VOC service
should be similar. The available SOCMI fugitive emissions data also
supported this judgement. (Tables 2-14 and 2-15 show a comparison of the
results of the studies of fugitive emissions available at the time of
proposal.) Furthermore, the refinery fugitive emissions data were collected
before an awareness of the magnitude of fugitive emissions became wide-
spread, and, therefore, the petroleum refinery emission factors are
considered representative of fugitive emissions in the absence of a leak
detection and repair program. Thus EPA concluded that emissions from SOCMI
in the absence of a leak detection and repair program could be estimated
using emission factors developed for fugitive VOC emission sources in
petroleum refineries.
2-19
-------
TABLE 2-13. SUMMARY OF SOCMI PROCESS UNITS FUGITIVE EMISSIONS
(24 Unit Study)
i
r\o
o
Source Type
Flanges
Process Drains
Open-Ended Lines
Agitator Seals
Relief Valves
Valves
Pumps
Compressors
Otherb
Service
Gas
Light
Heavy
Gas
Light
Heavy
Gas
Light
Heavy
Gas
Light
Heavy
Gas
Light
Heavy
Gas
Light
Heavy
Light
Heavy
Gas
Gas
Light
Heavy
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
Liquid
(1)
Number
Screened
1,443
2,897
607
83
527
28
923
3,603
477
7
8
1
85
69
3
9,668
18,294
3,632
647
97
29
19
33
2
(2) (3)
% of Screened Sources
% Not with Screening Values
Screened >10,000 ppmv
4
2
2
23
1
0
17
10
21
46
11
66
72
40
66
17
12
9
4
40
9
9
19
33
.6
.6
.4
.1
.9
.0
.5
.4
.5
.1
.1
.7
.7
.5
.7
.5
.2
.9
.3
.5
.4
.5
.5
.3
4
1
0
2
3
7
5
3
1
14
0
0
3
2
0
11
6
0
8
2
6
21
6
0
.6
.2
.0
.4
.8
.1
.8
.9
.3
.3
.0
.0
.5
.9
.0
.4
.4
.4
.8
.1
.9
.0
.1
.0
(4)
9556 Confidence Interval
for Percentage of Sources
>1 0,000 ppmv
(3
(0
(0
(0
(2
(0
(4
(3
(0
(0
(0
(0
(0
(0
(0
.6,
.9,
.0,
.3,
.3,
.9,
.4,
.3,
-5,
.4,
.0,
.0,
.7,
.3,
-0,
(10.8
(6
(0
(6
(0
(0
(6
(0
(0
.1,
.2,
.6,
.3,
.9,
.0,
.7,
.0,
5.8)
1.8)
0.6)
8.4)
5.8)
23.5)
7.5)
4.6)
2.8)
57.9)
36.9)
97.5)
10.0)
10.1)
70.8)
, 12.1)
6.8)
0.7)
11.1)
7.3)
22.8)
45.6)
20.2)
84.2)
Source: Reference 23.
blncludes filters, vacuum breakers, expansion joints, rupture disks, sight glass seals, etc.
-------
TABLE 2-14.
COMPARISON OF FUGITIVE EMISSION SOURCE LEAK
FREQUENCIES AVAILABLE IN THE BID
Percent of SOCMI Sources
Having Screening Values
>10,000 ppmv, OVA-108,
Equipment Type Methane (six unit study)3
Valves (all) 11
Gas
Light Liquid
Heavy Liquid
Open-ended Lines (all) 10
Gas
Light Liquid
Heavy Liquid
Pumps (all) 17
Light Liquid
Heavy Liquid
Compressors (Gas) 10e
Pressure Pel ief
Valves (.all) 0
Gas
Light Liquid
Heavy Liquid
Flanges (all) 3
Gas
Light Liquid
Heavy Liquid
Process Drains (all ) N/A
Gas
Light Liquid
Heavy Liquid
Agitator Seals (all) N/A
Gas
Light Liquid
Heavy Liquid
Percent of SOCMI Sources
Having Screening Values
>10,000 ppmv, OVA-108,.
Methane (24 unit study) '
11.4
6.5
0.4
4
5.9
3.9
1.3
8.8
2.1
9.1
3
3.6
2.9
0.0
2
4.6
1.3
0.0
4
2.4
3.8
7.1
14.3
0.0
0.0
Percent of Petroleum
Refinery Sources Having
>Screening Values .
10,000 ppmv, TLV-Hexane
10
11
0.2
7.7
24
2
36
7
0.5
5
N/A
Source: Reference 24.
Source: Reference 25.
CValues changed only slightly from the initial report (Docket No. IV-A-11) to the Analysis Report
(EPA-600/2-81-111).
Source: Reference 26.
eValue does not include the reciprocating compressors screened in a LD polyethylene unit.
Note: N/A means not available.
2-21
-------
TABLE 2-15. COMPARISON OF EMISSION FACTORS AVAILABLE IN THE BID,
kg/hr
Equipment type
Valves
Gas service
Light liquid service
Heavy liquid service
Pump seals
Light liquid service
Heavy liquid service
Compressor seals (hydrocarbon
service)
Pressure relief valves (all)
Gas service
Light liquid service
Heavy liquid service
Flanges
Open-ended lines
-» [-1 /»
Petroleum refinery study ' Exxon study
0.0268 0.017d
0.0109 0.008d
0.00023
0.114 0.255
0.021
0.636 0.264
0.086 0.064
0.16
0.006
0.009
0.00025
0.0023
Source: Reference 27.
The emission factors presented in the BID were based on an interim
report (EPA-600/2-79-044); the values listed here represent the final
published emission factors (EPA-600/2-80-075c).
'Source: Reference 28.
Not clear whether these factors are kg VOC/hr/source or kg VOC/hr/leaking
source.
2-22
-------
2.2 NEW INFORMATION
Through the efforts of EPA and industry, several studies began on
fugitive emission sources in SOCMI units during the development of the
proposed standards. Since proposal, the results of these studies have been
finalized. In addition, the results of fugitive emissions studies by
industry in Germany and in a petrochemical unit in the United States became
available. The nature and the results of these studies are summarized
below.
?9
2.2.1 German Studies on Fugitive Emissions '
Four interrelated studies conducted in West Germany investigated
leakages from static and dynamic seals in chemical and petrochemical plants
and methods of leak prevention. The sealing elements were classified as
follows:
Static sealing elements
- flange connections
- manholes
- threaded connections
Dynamic sealing elements
- seals on oscillating machines (piston rod packing boxes
on gas compressors and piston pumps)
seals on rotating machines (slide ring seals on centrifugal
pumps, agitators and centrifugal compressors, labyrinth
packings on centrifugal compressors, seals for scaling
liquid on centrifugal compressors, packing boxes on
centrifugal pumps and agitators)
seals on fittings and control valves (stuffing box)
Various procedures were employed for measurement of leak rates. The
"pressure drop" method makes use of the measurement of the pressure drop of
a pressurized and completely sealed sealing element. In the "pressure
retention" method the pressure drop was compensated by injection of a
measured quantity of gas. In the "capsule method" the seals were encapsu-
lated with tape (for flanges) or globes (for valves). In the "spray method"
the seal was sprayed with soap solution or another foam agent. The leakage
was visible due to the formation of foam. The leak rate was estimated from
the rate of bubble formation.
2-23
-------
Qualitative statements were made by the studies about the dependence of
leakages from static seals (primarily flanges) on physical, chemical, and
design parameters, e.g. laminar capillary flow, pore diffusion in connection
with contact pressure, the dimensions and properties of sealing rings, the
overpressure of the fluid, etc. However, the studies pointed out that data
were insufficient to permit a quantitative statement on the effects of
changing or on more quantities. The number of parameters affecting dynamic
seals was said to be even greater. As with static seals, quantitative
relationships were determined for dynamic seals.
The leak rates for the sources investigated were estimated to be about
1/10 of the values found in previous German literature. The studies
emphasized that valves showed many high leakage values; therefore, the
average leak rate increased. It was further reported that before any
directed maintenance of valves with high emission rates (i.e., screening
values), the leak rates were 200 percent of the literature values. However,
it was noted that regular maintenance of the valves reduced the average leak
rates to below 10 percent of the literature values. One of the studies
concluded that the most effective ways for avoiding excessive leak rates are
to train carefully the operating and maintenance personnel and to use
"maintenance-free" valves.
^n
2.2.2 Union Carbide StudyJU
Union Carbide conducted a study of fugitive emissions at a SOCMI unit.
The objective of the study was to find all leaking points in the process
unit and to quantify their leak rates. A secondary objective was to develop
a statistical fugitive emission sampling plan for future fugitive sampling
work.
The potential leaking points were screened with Century Systems Model
OVA-128 calibrated with hexane. Sources were screened at the surface. A
leak was defined as a source screened at 1000 ppm or greater. Table 2-16
presents a summary of the leak frequencies found in the Carbide study.
A total of 1,569 points were screened. Pipe joints, hand valves, and
valve bonnets accounted for 85 percent of the total sources. The overall
leak frequency was 6.7 percent. Pumps had the highest leak frequency
2-24
-------
TABLE 2-16. LEAK FREQUENCY IN UNION CARBIDE STUDY3
Safety/Relief Valves
Open-ended Lines
Gas
Liquid
Manholes
Valves - Total
Gas
Liquid
Hand
Motor
Pumps
Valve bonnets
Flanges
Threaded connections and
compressed ferrules
Total
Number Screened
26
127
19
108
9
484
120
364
442
42
18
279
507
112
1569
Percent Leaking
0
31
21
32
0
7.2
6.7
7.1
6.6
14
39
0.7
1.8
8.9
Average 6.7
Source: Reference 31.
Leak defined as 1,000 ppm as measured by OVA-128 calibrated with hexane.
2-25
-------
(39 percent). Since not all pumps were in operation during the screening
study, the overall leak frequency reported may be slightly biased on the low
side.
Leak rates were determined using different approaches depending on the
boiling point, the temperature of the process fluid, and the extent of
leakage as follows:
A: High leak rate, process fluid temperature low: The identified
leaking point was enclosed and sealed completely with a plastic bag. Hydro-
carbons saturated the enclosed bag and then condensed. The sampling time
started when a layer of liquid condensed on the bag surface and the liquid
started to run down the bag. After enough liquid accumulated, the bag was
punctured and the liquid was collected in a bottle for weight determination.
B: Medium leak rate: First, the leak point was enclosed in a plastic
bag. If no liquid condensation was observed in a few minutes, a tube was
connected to the bag to remove the leaking hydrocarbons. The other end of
the tube was connected to a condenser immersed in an ice bath. The majority
of the hydrocarbons condensed in the condenser and the weight of the
condensate was determined. The off-gas from the condenser was checked with
a FID GC and negligible hydrocarbons were found for the few leaking points
measured using this technique.
C: Low leak rate: After the leak point was enclosed in a plastic bag,
two tubes were connected to the bag and a known flow rate of compressed air
was introduced into the inlet tube. Gas samples were taken from the outlet
tube and the concentration of hydrocarbons was determined with a FID GC.
Table 2-17 shows the leak rates found in the study with the leak rates
broken down according to the extent of leakage. Table 2-17 shows that
25 percent of the top leakers accounted for 99 percent of the total
emissions.
??
2.2.3 Maintenance Study
A study of the effects of maintenance was performed concurrently with
the twenty-four unit screening study described previously. The maintenance
study was performed in six of the units studied in the twenty-four unit
study. The primary purpose of this study was the evaluation of maintenance
for fugitive VOC emissions control. The study yielded quantitative
2-26
-------
TABLE 2-17. ESTIMATED FUGITIVE EMISSION LOSS IN THE
UNION CARBIDE UNIT
Extent of Point leak rate, Estimated No.
Leakage kg/hr/source of leak points
Small leak 0.00002-0.00037
Wet surface 0.0019-0.0005
Dripping or 0.0188-0.3788
unbearable
odor
Continuous 0.9467-2.841
flow
Totals
50
30
20
5
105
Total leak
rate,
kg/hr/source
0.01
0.1704
3.788
9.47
13.44
Wt. % of total
Emissions
0.07
1.3
28.2
70.5
100
Source: Reference 33.
2-27
-------
estimates of the effects of maintenance on fugitive emissions and leak
occurrence and recurrence rates. As a secondary aspect of the study,
correlations of screening values with emission rates were developed. The
results of the study are discussed in detail in the section on emission
reductions.
The three process types studied were ethylene, cumene, and vinyl
acetate. Ethylene was chosen because typically these units are large and
widespread, operate with a wide range of process conditions, and handle very
volatile materials. Cumene was of interest because this type of unit
handles benzene, which has been listed as hazardous under Section 112 of the
Clean Air Act. Production of vinyl acetate from the reaction of ethylene
and acetic acid was chosen because some of the process streams are
corrosive. Two units of each process type were selected for the study. The
units were judged to be representative of the current level of control
existing in industry. One of the cumene units and one of the vinyl acetate
units employed leak detection and repair programs. And pumps in one of the
vinyl acetate units were equipped with dual seals and heavy liquid, positive
pressure barrier fluid systems.
In order to study the effects of maintenance on fugitive emissions,
valves and pump seals with various leak characteristics were to be studied
in detail. The total number of sources studied in all process units in each
category are given in Table 2-18. During the twenty-four unit screening
studies, leaking sources at the six units selected for the maintenance
studies had been tagged. The sources to be sampled were then selected at
random from these sources identified in the screening study.
Screening was done with a Century Systems Corporation OVA-108 and a
Bacharach Tl.V Sniffer. The valves were screened by traversing 360 degrees
around the stem seal and the seam where the packing gland merges with the
valve bonnet. The point of maximum concentration was identified. Pumps
were screened at the outer shaft seals by completely traversing 360 degrees
to locate the maximum concentration. Valve sampling was conducted by first
screening the valves with the OVA-108 and TLV Sniffer. The screening values
and time of day were recorded. Selected valves were then tented with Mylar
2-28
-------
TABLE 2-18. MATRIX OF SAMPLING/SCREENING FOR ALL UNITS'
I
IX)
Stream
Class
Gas
Light
Liquid
Initial Screening
Value (ppm)
< 1,000
1,000 to 9,999
10,000 to 49,999
>50,000
< 1,000
1,000 to 9,999
10,000 to 49,999
>50,000
Number to be
Control
M*
--
11
10
13
--
14
8
13
Valves
C**
84***
4
4
7
124***
5
2
6
studied i
Block
M*
--
14
12
14
--
19
17
18
n each fi
Valves
C**
161***
10
8
11
165***
10
8
9
ttinc[ type
Pump Seals
C**
--
--
--
--
87
11
9
6
Reference 34.
M* is the number of valves actually maintained.
C** is the number of valves (or pumps) in the control group.
*** "low leaking" valves for Unit 1 could not be matched to their original process information
since the identification tags were removed between the first and second visits. There were a
total of 106 "low leaking" valves in Unit 1.
-------
and duct tape and sampled. The tent was removed and the valve was
rescreened. Pump sampling was analogous to valve sampling.
35
2.2.4 Analysis Report
The results of the maintenance study were combined with the results of
the 24-unit study for more in-depth analysis. The analysis report presents
the findings of several data analysis tasks. Three of these tasks were the
analysis of leak frequency as a function of process parameters, the emission
factor development, and the analysis of leak frequency as a function of
equipment design. Other analysis tasks contained in the report, including
the impact of instrument response factors on leak frequency and the impact
on mass emissions due to leak occurrence and recurrence (see Section 4), do
not directly relate to emissions estimates and are not discussed here.
The process parameters that were examined for their effect on leak
frequency were process type, service, material in line, line pressure, line
temperature, and elevation of source. Data on four source types (valves,
pump seals, flanges, and open-ended lines) were used to examine the effects
of these parameters. As presented in Table 2-19, leak frequency for these
source types varied among the 15 process types as well as with service type.
And in almost every case examined, higher leak frequencies were associated
with higher line pressures. Line temperature was found to have no
consistent effect on leak frequency, while higher leak frequencies tended to
be associated with higher ambient temperatures. The effect of ambient
temperature, however, was not statistically significant in a majority of
cases. Finally, the elevation of the source, that is the height above
grade, had no consistent effect on leak frequency.
Emission factors were developed for gas valves, light liquid valves,
and light liquid pumps for the three process types studied in the
maintenance program. The sources included in the development of the
emission factors were all valves and pump seals screened in the seven
ethylene, cumene, and vinyl acetate process units or 51.2 percent of
all valves and pump seals screened in the twenty-four unit, screening
program. Because leak rate/screening value models that were determined
depend on the process type, emission factor estimation was limited to 'these
2-30
-------
TABLE 2-19.
PERCENT LEAKING FOR EACH CHEMICAL UNIT TYPE AS A FUNCTION OF SOURCE TYPE
AND STREAM SERVICE IN SOCMI3
,O
I
OJ
,b
bource/Chenuual (units)
Vj|_yes
Vinyl Arel.ite (1,3)
1 ihylei
l.umene
An-tinii
1 Un 1.-,
i In 1 i
t-i-lhyl
A' c l a ,
Hrl l-y i
A,l i pi,
I'hlor i
Ai ry lu
1.1,1-'
I'tnn^ Sc.l
V In/1
1 1 l.yli-
I'limi-tn-
A, i-l'.i
Mhylr
Vinyl
i<; (2.4, 11)
(5,6)
/riiiMii.l (12)
i.- ln.-l,,oride (21 . 29)
;,; MI nip Mnnnmer (2i\ 28)
i, l-.y.U- (22)
i II, vl 1 i-ti'iit (31,32)
iii-i.y.j.- l3j)
,';.-! .1.1. rv l/ii c 1 3»)
V- Id 1)5,64)
n.ilril Kllunes (60, 62)
n Hr 1 le (65.66)
I L- icn luroechaiia (o 1 )
U
Artt.itc- (1,3)
»,- (2,4,11)
(5,6)
,.-/|-'u-ii..l (12)
nu [HUilorlde (21,29)
Chloride Honomur (20.28)
GAS
Number Number
Screened Leaking
949 35
629<4 934
446 63
8 0
403 4
-12 30
41 1
207 19
17? 8
1<-0 0
95 0
48 0
396 9
LIGHT LIQUID
Percent
Leaking
3.7
14.8
14.1
0
1.0
7. J
2.4
9.2
4.5
0
0
0
2. 3
l-i.rm.ilduliyiJe (22)
M.-thyl
I'.thv 1 Ki't.me ( 3 1 , )2 1
.\, <-ta Idi-hydi, (i))
Mi-l liy 1
A,ll|>li
1 him 1
A, r y 1 1
1,1,1-
Hi.-lh.il rvl.UP (34)
Acid (35.64)
n.il I'd r.cli.incfl (60, 62)
mil r 1 le (65.66)
TrK-hlnr.icthani.- (hl'i
Nunber
Screened
2137
4176
799
1818
2256
1209
121
671
551
1058
17
I6:o
1494
373
89
76
25
86
58
65
8
31
32
45
60
61
10
Number
Leaking
3
969
84
6
IM
12
0
34
3
1
0
HI
28
4
(,
20
^
2
3
7
0
1
3
2
5
5
1
Percent
Leaking
0.4
23.2
10.5
0. 3
1.1
1.0
0
5. 1
0.5
0. 1
0
0.6
0.9
1.1
4.5
26.3
16.0
2.3
5.2
10.3
0
3.2
9.4
4.4
H. )
H. 2
1 0 . 0
HEAVY LIQUID
Number Number
Screened Leaking
124
1217
198
488
1478
12
95
5
15
3
36
30
B
0
13
0
0
0
0
0
0
0
0
0
---,-
0
2
Percent
Leak t ng
0
1. 1
0
0
0
0
0
0
0
0
0
0
25...
(Coin l
-------
TABLE 2-19. (CONTINUED)
IN3
1
Co
IX)
Source/Chemical (units) '
Kldngea
Vinyl Acetate (1,3)
Ethylene (2,4,11)
Cumene (5,6)
Acelime/1'ht.nol (12)
tlliylenu UKhluride (21,29)
Vinyl Ihlnnde Monomer (20,28)
KornMljL-hyde (22)
Mulhyl Klhyl Kecoue (11,32)
Acotaldehyde (33)
Methyl Muc.lidi.ryl.iic (J4)
A.ll|ite- Acid (35,64)
I liloi In.ited Ethanes (60, 62)
Acrylunitr 1 le (65.66)
Open tnded Lines
Vinyl Acetate (1,3)
Ftl.ylene (2,4,11)
Cumene (5,6)
Acetone/Phenol (12)
hlhylene Dichlorlde (21,29)
Vinyl Chloride Monomer (20,28)
Formaldehyde (22)
Helhyl Ethyl Ketoile (31, J2)
Acetaldehyde (33)
Methyl Muth.icryldte (34)
Adlplc Acid (35.64)
Chlorinated Ethanes (60.6Z)
AcrylonI tr 1 le (65,66)
I.I. l-Trl. hl.iri.etii.,nu (61)
Number
Screened
107
634
367
25
16
2
22
32
38
49
16
142
145
305
6
2
100
55
14
37
34
63
19
27
116
GAS
Number
Leaking
3
39
19
1
2
0
0
0
0
0
0
2
8
37
0
0
0
2
0
3
3
0
0
0
1
LIGHT LIQUID
Percent
Leaking
2.
6.
5.
4.
12.
1.
5.
12.
3.
8.
8.
0.
8
2
2
-
0
5
0
0
0
0
0
0
4
5
1
0
0
0
6
0
1
8
0
0
0
9
-
Number
Screened
173
407
468
82
163
47
8
76
144
247
2
461
382
318
214
15
518
475
340
36
186
158
335
1
412
486
111
Number
Leaking
0
25
9
0
1
0
1
0
0
0
0
0
0
8
41
2
8
16
18
0
19
8
1
0
6
12
2
Percent
Leaking
6.
1.
0.
12.
2
19
13
1
3
5
10,
5
0
1
2
1
0
1
6
0
6
0
5
0
0
0
0
0
0
.5
.2
.3
.5
.4
.3
0
.2
.1
.3
0
.5
.5
.8
Number
Screened
8
89
130
30
320
2
28
22
91
1
107
214-
4
38
HEAVY LIQUID
Number
Leaking
0
0
u
u
0
0
0
2
0
0
0
0
0
4
Percent
Leakl ng
0
L
I
(J
n
0
0
9.1
0
0
0
0
0
10.5
Source: Reference 36.
Unit numbers represent the unit number designation reported in the 24~Unit Study '(Reference
-------
process units. These emission factors are shown in Table 2-20. (Note:
These factors were recalculated at a later date.)
In addition to the above results, the pump and valve data were examined
in closer detail for design-oriented effects. Control valves had a higher
leak frequency than block valves. And for block valves, gate valves had the
highest leak frequency, while plug and ball valves demonstrated the lowest
leak frequencies. On-line pump seals had an overall leak frequency of
13.1 percent compared to 3.9 percent for off-line seals. Although no
difference in leak frequency was found for dual mechanical and single
mechanical pump seals, the barrier fluid composition was not known and could
not be accounted for in the analysis.
OQ
2.2.5 Analysis of Allied HOPE Unit Data
A 10-month study was performed by Allied and Kemron in Allied's newest
existing high density polyethylene unit (HOPE). The study consisted of six
screening and emissions measurement tests performed on valves and flanges
over a 9-month period. EPA contracted a review of the data from this study.
A Century Systems Flame lonization Analyzer (FIA), Model OVA-108
calibrated to 1000 ppm hexane was used in the screening studies. Calibra-
tion procedures differed from EPA's in that a high concentration standard
was not used and the calibration was not verified at the end of the day.
The leak frequencies determined in the six screening tests are shown in
Table 2-21.
A tenting method was used to bag samples. Suspected fugitive emission
points were enclosed in a Mylar tent. The source leak-rate was calculated
from the amount of total hydrocarbons collected in a specific time period.
The leak rates were determined only for sources screened at 10,000 ppm or
greater and should not be compared with average emission factors.
Table 2-22 shows the leak rates determined in the Allied study. The
emission factors determined during the study were "leaker" emission factors.
But some valves were maintained before being sampled. Therefore, the leaker
emission factor presented is based on valves that had, in part, been
subjected to a directed maintenance program.
2-33
-------
TABLE 2-20. EMISSION FACTORS AND LEAK FREQUENCIES CALCULATED.^ THE
ANALYSIS REPORT WITH 95 PERCENT CONFIDENCE INTERVALS5'13
Emission factor
kg/hr
Lear
frequency.
Gas Valves
Vinyl Acetate
Cumene
Ethylene
Light Liquid Valves
Vinyl Acetate
Cumene
Ethylene
Pump Seals (Light Liquid)
Vinyl Acetate
Cumene
Ethylene
0.0021 (0.0004, 0.01)
0.0052 (0.001, 0.02)
0.011 (0.004, 0.03)
0.0001 (0.00003, 0.001)
0.0025 (0.001, 0.01)
0.010 (0.003, 0.03)
0.0020 (0.00006, 0.06)
0.023 (0.0004, 1.2)
0.031 (0.003, 0.4)
3.7 (2, 5)
16 (13, 19)
15 (14, 16)
0.2 (0, 0.4)
12 (10, 13)
26 (24, 27)
1.7 (0, 4)
14 (1, 27)
30 (20, 39)
Source: Reference 39.
5These emission factors were
'Leak defined as 10,000 ppm or greater.
calibrated with methane.
ater recalculated.
Screening conducted with an OVA-108
2-34
-------
TABLE 2-21. LEAK FREQUENCY BY SOURCE AND SERVICE - HOPE UNIT0
Liquid
Source
Flanges
Valves
Statistic
No. of leakers
No. of sources
tested
Leak frequency,
%
LCL,^ %
UCL, %
No. of leakers
No. of sources
tested
Leak frequency,
%
LCL,^ %
UCL, %
Test
(0)
69
471
14
11
18
54
197
27
21
34
, 1
b
.6
.7
.1
.4
.7
.0
Test 2
(58)
72
474
15.2
12.2
18.7
47
206
22.8
17.6
29.2
Test 3
(110)
47
482
9.
7.
12.
22
208
10.
7.
15.
8
4
7
6
1
5
Service
Test 4
(148)
71
463
15.
12.
18.
34
206
16.
12.
22.
Test 5
3
3
9
5
1
2
(190)
31
469
6.
4.
9.
26
209
12.
8.
17.
6
7
2
4
6
6
Test 6
(250)
33
464
7.1
5.1
9.8
28
200
14.0
9.9
19.5
Gas Service
Flanges
Valves
No. of leakers
No. of sources
tested
Leak frequency,
%
LCL,C %
UCL, %
No. of leakers
No. of sources
tested
Leak frequency,
%
LCL,C %
UCL^d %
11
149
7
4
12
13
74
17
10
27
.4
.2
.7
.6
.6
.8
14
146
9.6
5.8
15.5
13
71
18.3
11.0
28.8
13
148
8.
5.
14.
6
74
8.
3.
16.
8
2
4
1
8
6
13
169
7.
4.
12.
6
77
7.
3.
16.
7
6
7
8
6
0
7
173
4.
2.
8.
2
76
2.
0.
9.
0
0
1
6
7
1
9
175
5.1
2.7
9.5
7
76
9.2
4.5
17.8
Source: Reference 40.
DThe day number in parenthesis indicates the number of days from the end of
Test 1 to the end of the specified test.
'Lower 95 percent confidence limit.
Upper 95 percent confidence limit.
2-35
-------
TABLE 2-22. LEAK RATES FOR LEAKERS BY SOURCE AND SERVICE - HOPE UNIT
a,b
Source
Flanges, mean
LCL,C I
UCL,C %
No. testedd
Valves, mean
LCL,C %
UCL,C %
No. tested
Liquid Service
kg/hr/source
0.018
0.011
0.029
70
0.021
0.014
0.034
56
Gas Service
kg/hr/souce
0.0059
0.0036
0.0100
20
0.0022
0.0012
0.0039
7
Source: Reference 41.
Results were determined by Kemron bagging tests on a sample of 153 leakers
(>10,000 ppmv). The bagging test data used in the analysis are in
Appendix E of the report (Reference 41). The bagging test data not used
(due to maintenance preceding the bagging test) are also in this appendix.
These numbers are leak rates for leaking sources in the presence of a leak
detection and repair program. The average magnitude of leak is decreased
if maintenance has previously (successfully or unsuccessfully) applied.
The lower (LCL) and upper (UCL) 95 percent confidence limits are determined
by using a pooled estimate of the variance for the four data sets.
The number of bagging tests by source and service.
2-36
-------
2.2.6 SCAQMD Study42
A study of fugitive emissions in two petroleum refineries in the South
Coast Air Quality Management District (SCAQMD) was undertaken by EPA to
investigate the effectiveness of fugitive emission control regulations. The
data were obtained by conducting leak detection surveys (screening) in
selected units and by examining the refinery records pertaining to leak
detection surveys conducted as requirements of the regulations. The effec-
tiveness of the regulations, as determined by this study, is discussed later
in the section on emissions reductions (Section 4).
A preliminary summary of the data collected in the two refineries is
presented in the draft report. All accessible valves (with the exception of
some heavy liquid valves), pumps, agitators, open-ended lines, drains, and
relief valves in eight process units were screened with a Century OVA-108
hydrocarbon detector to identify the sources of fugitive VOC emissions. No
flanges were screened. EPA Reference Method 21 was used for the screening
surveys. In addition to the maximum hydrocarbon detector reading, certain
information describing the source and its process conditions were recorded.
This included the type of process unit, type of source, type of service, and
the primary organic components in the line and their approximate
concentrations.
A summary of the leak frequencies found in the refineries is presented
in Table 2-23.
43 44
2.2.7 Coke Oven By-product Recovery Plant and Gas Plant Studies
Studies were conducted by EPA and industry groups to assess fugitive
emissions from sources associated with coke oven by-product recovery plants
and gas plants. The results of these studies are summarized in Tables 2-24
and 2-25.
Three coke oven by-product plants were tested by EPA for fugitive
benzene emissions. Source screening was conducted using an OVA-108 portable
hydrocarbon analyzer and was supplemented with measurements conducted using
the TLV analyzer for sources that were sampled for mass emissions. The
emissions data were categorized by source type and process stream benzene
content. For some emission sources, emission factors were generated for
2-37
-------
TABLE 2-23. SUMMARY OF LEAK FREQUENCIES BY SOURCE TYPE AND STREAM SERVICE
IN TWO REFINERIES IN SCAQMD ALL PROCESS UNITS9
I
CO
co
Source Type
Open-Ended Lines
Sealed Open-Ended
Lines
Open Process Drains
Relief Valves
Valves
Pumps
Compressors
Others
TOTAL
Stream Service
Gas
Light Liquid
Heavy Liquid
Unknown
Gas
Light Liquid
Heavy Liquid
Light Liquid
Heavy Liquid
Gas
Light Liquid
Heavy Liquid
Gas
Light Liquid
Heavy Liquid
Light Liquid
Heavy Liquid
Gas
Gas
Light Liquid
Gas
Light Liquid
Heavy Liquid
Unknown
Sources
Number
Screened
100
202
-
1
346
1019
-
87
-
7
12
-
2294
4761
-
116
-
9
1
11
2758
6208
-
1
Subject
Number
>10,000
8
23
-
0
8
41
-
2
-
1
0
-
108
274
-
20
-
4
0
0
129
360
-
0
to Rule
Percent
>10,000
8.0
11.4
-
0.0
2.3
4.0
-
2.3
-
14.3
0.0
-
4.7
5.8
-
17.2
-
44.4
0.0
0.0
4.7
5.8
-
0.0
Sources
Number
Screened
13
25
37
-
2
22
25
6
31
_
-
2
52
175
284
17
40
1
_
-
68
245
419
-
Exempt from Rule
Number
>10,000
1
0
0
-
0
0
0
1
0
_
-
0
3
3
0
0
0
1
_
-
5
4
0
-
Percent
>10,000
7.7
0.0
0.0
0
0.0
0.0
0.0
16.7
0.0
_
-
0.0
5.8
1.7
0.0
0.0
0.0
100.0
_
-
7.4
1.6
0.0
-
Source: Reference 45.
Note: Screening conducted with an OVA-108 calibrated with methane.
-------
TABLE 2-24. LEAK FREQUENCY FOR SOURCES IN COKE OVEN BYPRODUCT UNITSa
Source
Flanges
Threaded ,
Connections
Valves5
Pump Seals
Exhausters
Number of
Sources
66
59
135
20
34
Leak
Frequency
0
0
5.9
45
8.8
Emission Factor, kg/hr/source
(95% Confidence Interval)
c
c
0.015 (0.0013 - 0.14)
0.22 (0.054 - 0.75)
0.015 (0.0003 - 0.042)
aSource: Reference 46.
Sources were in service on process streams containing at least 10 percent
benzene.
d
cEmission factors were not determined for these sources.
These exhausters or compressors were mainly in hydrogen service.
2-39
-------
TABLE 2-25. EMISSION FACTORS AND LEAK FREQUENCIES FOR FITTINGS IN GAS PLANTS'
Source
Connections &
flanges
Open-ended lines
Pressure relief
devices
Valves
Pump seals
Compressor seals
Number
Screened
20,186
1,560
107
7,787
137
71
Percent Emitting
7.9
26.8
40.2
32.8
73.0
74.6
Leak
Frequency
3.1
11.9
17.5
16.4
29.7
52.8
Total Emission Factor,
kg/hr/source
(95% Confidence Interval)
0.0011 (0.0004, 0.002)
0.022 (0.008, 0.04)
0.19 (0.004, 4)
0.020 (0.008, 0.04)
0.063 (0.02, 0.2)
0.20 (0.03, 1.3)
Source: Reference 47.
Percent emitting means the percentage of sources scoring above zero using soap scoring or
instrument monitoring - API and EPA testing.
'Leak frequency means the percentage of sources screening at or above 10,000 ppm using instrument
monitoring - EPA testing only.
-------
nonmethane hydrocarbon emissions and benzene emissions. The leak
frequencies and nonmethane hydrocarbon emission factors determined in this
study are given in Table 2-24.
A total of six gas plants were screened in two studies: four by EPA
and two by the American Petroleum Institute (API). The results of the
combined studies are presented in Table 2-25. In both studies some of the
sources identified as emitting were selected for leak rate measurement.
There was a difference, however, in the screening technique used. The EPA
study used portable hydrocarbon detectors, supplemented in many cases by
soap bubble testing. The API study, on the other hand, relied primarily on
soap bubble testing, employing a portable hydrocarbon detector where soap
bubble testing was not possible. Using leak rate data collected during both
studies, leak rates were estimated for all the emitting sources in both data
sets to compute an emission factor for emitting sources alone. This factor
was then adjusted to account for non-emitting sources by weighting the leak
frequencies of emitting and non-emitting sources. These estimated total
emission factors are presented in Table 2-25.
2.2.8 Revision of Emission Factors for Nonmethane Hydrocarbons From Valves
48
and Pump Seals in SOCMI Processes
The purpose of this report was to update nonmethane hydrocarbon
emission factor estimates reported in the Analysis Report, based on further
review of the available data and recently developed methodology.
Recent review of the emissions data used to develop emission factors
for the vinyl acetate, cumene, and ethylene units indicated some differences
in the screening value to emission relationships for valves, depending on
whether maintenance had been performed on the valves during the screening
study. In addition, the statistical methodology for estimating emission
factors using screening data was refined. This methodology treats sources
with screening values measured at or above 100,000 ppmv, i.e. censored data,
separately from the sources with screening values which are not constrained.
This approach minimizes biases from the censored data and results in more
precise emission factor estimates.
2-41
-------
The available data for valves were analyzed for leak rate/screening
value relationships for each of the following groups:
1. All valves
2. Valves tested before maintenance was performed.
3. Valves tested after maintenance was performed.
4. Valves which were control sources (no maintenance performed).
5. Valves tested before maintenance and controls (2 and 4 combined).
The results of the analysis are summarized in Table 2-26. A significant
effect of on-line maintenance on the screening value/leak rate relationship
is seen. For valves in gas service the estimated leak rate (corresponding
to a given screening value) for a valve tested before maintenance was
performed was almost four times the estimate for a valve tested after
maintenance. For valves in light liquid service the estimated leak rate for
a valve tested before maintenance was performed is about two and one-half to
three times greater than the estimated leak rate for a valve tested after
maintenance.
The confidence interval for the modeled relationships indicated the
statistical significance of the difference in the leak rate/screening value
relationship for the before maintenance and after maintenance groups. The
study concluded that the valves tested after on-line maintenance was done
should not be used in developing models to estimate uncontrolled emission
factors. The valves which were tested before maintenance and the valves
used as control sources are appropriate for this purpose.
The OVA screening device used in the SOCMI emissions data collection
efforts is designed to measure VOC concentrations up to 10,000 ppm. A
dilution probe can be used to extend the scale to 100,000 ppm. A secondary
dilution probe which allows concentration as high as 1,000,000 ppm to be
measured was used in some cases in the maintenance study. But results
obtained using two dilution probes are of questionable value due to the
inaccuracies associated with measuring VOC concentrations of extremely small
sample streams.
In developing regression equations to estimate leak rates from
screening values in the maintenance study, those sources having screening
2-42
-------
TABLE 2-26. ESTIMATED LEAK RATE TO SCREENING VALUE MODELS FOR GROUPS OF VALVES'
Valves in Gas Service
Model Parameters
Estimate of Leak Rate, kg/hr, (95 Percent Confidence Interval)
for Different Screening Values
Correlation Standard
Coefficient Error
1,000 ppmv
10,000 ppmv
100,000 ppmv
Group 1 All Valves
(EPA-600/52-81-080)
Group 2 Before Maintenance
Group 3 After Maintenance
Group 4 Controls
Group 5 Before and Controls
301 0.350 0.79 0.70
82 0.705 0.81 0.68
129 0.309 0.76 0.73
93 0.211 0.81 0.67
175 0.427 0.81 0.67
0.73 0.0009 (0.00046, 0.0014) 0.0055 (0.0041, 0.0064) 0.033 (0.027, 0.041)
0.81 0.0018 (0.00091, 0.0036) 0.012 (0.0077, 0.019) 0.0077 (0.045, 0.13)
0.68 0.00046 (0.00036, 0.0091) 0.0036 (0.0027, 0.0045) 0.020 (0.014, 0.027)
0.73 0.00046 (0.00027, 0.0014) 0.0036 (0.0027, 0.0055) 0.025 (0.018, 0.036)
0.78 0.00091 (0.00046, 0.0018) 0.0072 (0.0055, 0.0095) 0.045 (0.032, 0.064)
-p.
oo
Valves in Light liquid Service
Group 1 All Valves
(EPA-600/52-81-080)
Group 2 Before Maintenance
Group 3 After Maintenance
Group 4 Controls
Group 5 Before and Controls
350 5.27 0.54 0.55
104 11.5 0.50 0.49
169 3.32 0.53 0.57
78 8.45 0.535 0.52
182 9.05 0.53 0.52
0.83 0.0023 (0.0018, 0.0032) 0.0077 (0.0064, 0.0095) 0.028 (0.020, 0.035)
0.0036 (0.0018, 0.0068) 0.011 (0.0077, 0.016) 0.036 (0.023, 0.055)
0.0014 (0.00091, 0.0018) 0.0041 (0.0032, 0.0055) 0.015 (0.010, 0.023)
0.021 (0.0073, 0.019)
0.021 (0.0091, 0.016)
0.78
0.79
0.94 0.0032 (0.0018, 0.0068)
0.86 0.0036 (0.0023, 0.0055)
0.040 (0.021, 0.073)
0.041 (0.028, 0.059)
Source: Reference 49.
qModel is of the form leak rate = a (10"5)(OVA Reading)*3
CNumber of test rate/screening valve pairs; the numbers do not add to the total because of missing descriptor codes for some of the data
pairs.
-------
values higher than 1,000,000 ppm were treated as if the screening value was
100,000 ppm. This study made use of a recently developed statistical
methodology for estimating emission factors which treats those censored
sources separately from the uncensored screening values. This procedure
deletes the censored sources in the regression analysis, and develops an
estimate for censored sources using the deleted data. It also eliminates
the potential bias of using those censored sources in the regression
analysis.
Using the two changes in methodology described in the report, emission
factors for valves and pump seals in ethylene, cumene, and vinyl acetate
units were developed and are shown in Table 2-27.
2.3 PUBLIC COMMENT
Several commenters disagreed with using petroleum refinery emission
estimates for SOCMI emission estimates. The commenters made three major
points.
1. Commenters argued that both leak frequencies and mass emission
rates for SOCMI fugitive emission sources are lower than those corresponding
values for refinery sources. The commenters further felt that the lower
leak tendencies of SOCMI had been shown in SOCMI studies. Studies cited
50 51
included the twenty-four unit screening study, the maintenance study,
52 53
the analysis report, and the 4-unit EPA study, which were previously
discussed. The commenters felt, therefore, that SOCMI data should be used
instead of refinery data to estimate emissions.
Also presented as evidence of lower fugitive emissions in SOCMI were
results of studies performed by chemical companies in their own plants.
Testing in an acrylonitrile unit was reported to yield emission rates of
0.0002 to 0.011 kg VOC/hr for pumps and 0.0002 to 0.0087 kg VOC/hr for
valves. Emissions measured from a flange were 0.0003 kg VOC/hr. Leak
frequencies determined in the acrylonitrile plant were 5.9 percent for light
liquid pumps and 0.9 percent for light liquid valves. Another inplant
fugitive emissions testing project conducted in a chlorinated hydrocarbons
unit was reported to yield pump emission rates of 0.0068 to 0.0095 kq
CjC.
VOC/hr.
2-44
-------
TABLE 2-27. REVISED EMISSION FACTOR ESTIMATES FOR
NONMETHANE HYDROCARBONS FROM VALVES AND PUMP SEALS
IN ETHYLENE, CUMENE, AND VINYL ACETATE UNITS9
kg/hr/source
Source Type
Emission Factor
(95 Percent Confidence Interval)
Valves
Gas service
Ethylene
Cumene
Vinyl Acetate
Light Liquid Service
Ethylene
Cumene
Vinyl Acetate
Pump Seals
Light Liquid Service
Ethylene
Cumene
Vinyl Acetate
0.0086 (0.005, 0.012)
0.007 (0.003, 0.016)
0.0014 (0.0005, 0.004)
0.018 (0.01, 0.03)
0.006 (0.003, 0.013)
0.00023 (0.0001, 0.0006)
0.058 (0.01, 0.26)
0.018 (0.0014, 0.22)
0.002 (0.0002, 0.01)
Source: Reference 56.
2-45
-------
2. Industry commenters said that differences between SOCMI and
petroleum refining units could logically be expected to result in different
levels of emissions. A primary reason given for the expected differences in
emission factors was the dependence of emissions on the chemicals being
processed. The commenters said that process streams having different
compositions and, therefore, different chemical and physical properties will
probably produce different emission factors. One of the specific reasons
cited for differences in properties of refinery and SOCMI streams was higher
polarity as a result of substitution of heteroatoms (Cl, 0, N) for hydrogen.
This polarity was seen as causing larger dipole moments and stronger bonds
than those encountered in non-polar materials. Hydrogen bonding was also
suggested as a cause of increased polarity. The polarity was seen as
especially important because the more polar a substance is, the more
difficult it is to volatilize.
Several other reasons for expected differences in emissions between the
two industries were given. Many SOCMI materials were seen as more toxic and
hazardous than refinery products. Industry commenters said that the
toxicity of SOCMI chemicals often controls design and operating practices.
As a result, SOCMI units were seen as better controlled than refineries with
respect to fugitive emissions, and this level of control was expected to be
reflected in lower leak frequencies and emissions.
It was also pointed out that the chemical industry to a large extent is
characterized by smaller equipment and more batch processes that lend them-
selves more readily to improved fugitive emission control. Conversely,
refineries were characterized by much more strenuous conditions, larger
equipment, higher temperatures, and more outdoor continuous processes. This
difference was also seen as contributory to the expected differences in
emissions between the industries.
Finally, the materials produced in SOCMI were noted as of greater value
than those produced in refineries. This increased value was seen as
incentive for fugitive losses to be kept under better control in SOCMI than
in petroleum refineries.
2-46
-------
3. Industry commenters advocated the use of SOCMI emission factors
when SOCMI factors were available. In fact, one comment letter contained
emissions estimates that employed a weighted average of the SOCMI emission
factors presented in the Analysis Report previously discussed. The
average was weighted 40 percent for ethylene, 30 percent for cumene, and
30 percent for vinyl acetate. The commenters used emission factors for gas
valves, light liquid valves, and light liquid pumps from the Analysis
Report, and petroleum refinery emission factors for the remaining emission
sources. The factors used for the estimates are shown in Table 2-28.
2.4 EPA's CONCLUSIONS
2.4.1 Approach
In making a decision concerning what data to use in estimating fugitive
emissions of VOC from SOCMI, EPA identified three major criteria which would
form the basis for the decisions. The first criterion was relevance to
estimating fugitive emissions from SOCMI. That is, the information
considered had to be applicable to emissions of VOC from fugitive emission
sources. The second criterion was validity of the testing and analytical
methods used. Only those studies based on sound experimental design and
implemented with valid testing and analytical methods could be considered.
The third criterion was one of comparability to other work. The studies
considered had to be comparable to other studies to allow validation of
results by comparison with other studies. If a comparable basis for
comparison could not be found, it would be impossible to judge whether the
results of a particular study appeared reasonable in light of the results of
other studies.
Available data upon which emissions estimates could be based have been
presented previously in this section. Numerous studies of fugitive
emissions were evaluated for use in developing emission estimates for SOCMI.
As discussed earlier, six of these studies were considered prior to proposal
and nine additional studies have been considered as discussed in
Section 2.2. The results of these studies of fugitive emissions are
compared in Tables 2-29 and 2-30.
2-47
-------
TABLE 2-28. EMISSION FACTORS USED BY INDUSTRY COMMENTERS
TO ESTIMATE EMISSIONS FROM SOCMI UNITS8
Estimated Emission Factor
Equipment Component kg/hr/source
Pump Seals
Light Liquid 0.0020
Heavy Liquid 0.0200
In-Line Valves
Vapor 0.0065
Light Liquid 0.0045
Heavy Liquid 0.0003
Compressor Seals 0.441
Flanges 0.0003
aSource: Reference 58.
Includes: safety relief valves, open-ended valves & lines, and sampling.
connections.
2-48
-------
TABLE 2-29. SUMMARY OF AVAILABLE DATA ON FUGITIVE VOC EMISSION SOURCES -
EMISSION FACTOR, kg/hr/source
Emission Source
Valves
Gas
Light Liquid
All
Pumps
Light Liquid
Heavy Liquid
All
Compressors
Safety/Relief Vilves
Gas
All
Flanges
Open-ended Lines
Sampling Valves
Petroleum
Refinery
Study3
0.028
0.0109
0.114
0.021
0.636
0.16
0.086
0.00025
0.0023
0.015
Data Source
4-Unit Exxon . Coke Gas
SOCM1. Cyclohexane 6-Unit Maintenance Study Oven Plant,
Study Study Ethylene Cumene Vinyl Acetate Study Study
0.017 0.0086 0.007 0.0014
0.008 0.018 0.006 0.00023
0.037* 0.015 0.020
0.0045**
0.058 0.018 0.002
0.073* 0.255 0.22 0.063
0.022**
0«.035* 0.264 0.015 0.20
0.0143* 0.064 0.19
0.0051**
0.045* 0.0011
0.0014**
0.091* 0.022
0.040*
Value for leaking source only.
"Average value.
aSource: Reference 59.
Source: Reference 60.
Source: Reference 61.
Source: Reference 62.
Source: Reference 63.
Source: Reference 64.
-------
TABLE 2-30. SUMMARY OF AVAILABLE DATA ON FUGITIVE VOC EMISSION SOURCES -
LEAK FREQUENCY
i
en
O
Emission Source
Valves
Gas
Light Liquid
All
Pumps
Light Liquid
Heavy Liquid
All
Compressors
Safety/Relief Valves
Flanges
Open-ended Lines
Petroleum
Refining
Study3
10
11
24
2
36
7
0.5
7.7
4-Unit 6-Unit DuPont
SOCMI. Screening Study
Study Study
23.3
7.1
7 11 6.1
14.3
0
23 17
43
0
2 3
10
Exxon 24-Unit Union
Cyclohexane Screening Carbide
Study Study Studyy
32 11.4 6.7
15 6.5
8.2 7.2
8.8
2.1
83 39
9.1
3.2 0
2.1
3.9 31
Allied SCAQMD
HOPE. Study
Study"
17.6 4.7
27.4 5.6
24.7 5.1
15.0
0
11.6
50
4.8
1.8
8.5
3.5*
Coke Gas
Oven. Plant.
StudyJ Study
5.9 16.4
45 30
8.8 53
17.5
0 3.1
11.9
*Sealed open-ended lines.
Screening conducted with TLV Sniffer using hexane calibrant and 10,000 ppm leak definition. Reference 65.
Screening conducted with OVA-128 using the major constituent in the process stream as the calibrant and a 200 ppm equivalent methane leak
definition. Reference 66.
Screening conducted with OVA-108 using methane calibrant and 10,000 ppm leak definition. Reference 67.
Screening conducted with OVA-108 using hexane calibrant and 100 ppm leak definition. Number for ''all" includes heavy liquid valves.
0.2 percent of the heavy liquid valves were found leaking. Reference 68.
Leak definition was 10,000 ppm with an undefined instrument and calibrant. Soap screening used for valves. Reference 69.
Screening conducted primarily with OVA-108 and OVA-128 using methane calibrant and 10,000 ppm leak definition. Reference 70.
^Screening conducted with OVA-128 using hexane calibrant and 1,000 ppm leak definition. Reference 71.
Screening conducted with OVA-108 using hexane calibrant and 10,000 ppm leak definition. Reference 72.
Screening conducted with OVA-108 methane calibrant and 10,000 ppm leak definition. Reference 73.
JScreening conducted primarily with OVA-108 using hexane calibrant and 10,000 ppm leak definition. Reference 74.
1^
Screening conducted with OVA-108 using methane calibrant and 10,000 ppm leak definition. Supplemental data (reported in Table 2- ) was
provided using soap scoring. Reference 75.
-------
After examining the available fugitive emissions data, it is apparent
that no single, clear-cut set of emission factors from any one study can be
applied to SOCMI. There are no mass emissions data which can be considered
representative of emission factors for SOCMI as an industry. Furthermore,
close scrutiny of the leak frequency data available showed that many of the
studies were not comparable and could not be combined into one data set
because of differences in leak definitions, different monitoring instruments
calibrated with different gases, and different monitoring methods.
2.4.2 Evaluation of Fugitive Emissions Information
A summary of the data sets available on fugitive emissions is presented
in Table 2-31 with strengths and weaknesses of the data noted.
The most complete and comprehensive work on fugitive emissions
available prior to proposal was the study of emissions from petroleum
refineries. The data compiled in this study came from testing of over
64 units in 13 refineries. The units represented a broad cross-section of
sizes and ages. In addition to the determination of leak frequencies for a
number of fugitive emission sources, non-methane leak rate correlations were
developed which allowed average emission factors to be determined for these
sources.
The four-unit EPA study considered too small a number of sources and
units to be considered valid for emission factors by itself. A subsequent
study by EPA in six SOCMI process units gave some useful results in terms
of leak frequency only, with no mass emissions measurements made. These
data were also limited in that the service (gas, light liquid, heavy liquid)
of each source was not identified.
The four studies performed by industry which were available have
limitations with regard to their quantitative results. The leak frequency
78
for valves in the Exxon study of fugitive losses in a cyclohexane unit was
determined using a soap solution. And the mass emissions rate for valves
was based on selective sampling of valves from qualitative leak rate
categories (high, medium, and low leak). Further, the leak definition for
all sources was not specified. Only pumps and valves were screened for leak
79
frequency determination in the DuPont study. Although the monitoring
2-51
-------
TABLE 2-31. SUMMARY OF ASPECTS OF FUGITIVE EMISSIONS STUDIES
Study
Petroleum on
Refinery Study
4-Uni£,EPA
Study81
6-Unit
Screening
Study82
DuPonjU
Study83
Exxon
Cyclohexane
Study04
1.
2.
3.
4.
1.
2.
3.
4.
5.
1.
2.
3.
1.
2.
3.
4.
1.
2.
3.
4.
5.
6.
Remarks AID
Emission factor and emitter3 frequency
determined.
Emission factor and frequency (emitter
and leak) given by source and service.
Cross-section of refinery units.
"Uncontrolled" fugitive emissions data
collected.
Limited data base for test design
(4 units).
Leak definition of 200 ppm used for study.
Several sources considered for emission
factor and emitter frequency.
Leaking source emission factors determined.
Average emission factors estimated.
No emission factor determined.
Six sources considered for leak
frequency.
No delineation of sources by service
(gas, light liquid, heavy liquid).
No emission factors determined.
Only valves and pumps studied for
frequency of emissions.
Two older process units - limited
data base.
Leak definition unspecified.
Single unit study - limited base.
Four fugitive emission sources; valves
by service.
Emission factor and percent of leaks
determined.
Emissions rate for valves by "selected"
valve distribution.
Leak definition for 3 sources was 200 ppm.
Leaking valves determined by soap solution.
Section
2.1.1
2.1.2
2.1.3
4.1.1
2.1.4
2.1.5
2-52
-------
TABLE 2-31. (CONTINUED)
Study
24-Unit
Screening
Study
German Study
1.
2.
3.
1.
2.
Remarks
No emission factor determined.
Leak frequency by source and service.
Cross-section of SOCMI represented for
leak frequency.
Leak rates determined for sealing
mechanisms various means.
Qualitative results noted for seals (the
importance of maintenance in reducing
emissions from seals).
AIL) Section
2.1.6
2.2.1
UnionRCarbide 1.
Study07 2,
Leak definition of 1000 ppmv used.
Leak rates were categorized by quantity
of emission, not by source.
Only a single unit was surveyed.
2.2.2
6-Unit
Maintenance,
Analysis ggdog
Revisions '
Allied HOPE
StudyyJ
SCAQMD92
Maintenance
Study
(Prel i mi nary)
1.
i qn
9 3 \J
1.
2.
3.
1.
2.
3.
4.
Emission factor for three processes -
six units. (Subset of 24-unit results).
Emission factor for three fugitive
emission source typess.
Maintenance-oriented study.
Only valves and flanges studied.
Only leak emission factors determined.
Maintenance-oriented study - no mass
emissions studied.
Leak frequencies evaluated in two
refineries.
Five sources with pumps and valves by servi
Limited data base for compressors and
safety valves.
2.2.3
2.2.4
2.2.8
4.1.2
4.2.2
2.2.5
4.1.2
2.2.6
4.1.2
4.2.2
ce.
2-53
-------
TABLE 2-31. (CONTINUED)
Study
Coke Oven and
Gas Plagt QA
Studies '
1.
2.
3.
Remarks
Emission factors and leak frequencies
determined.
Limited number of units surveyed and
sources investigated.
Service not specified.
AID Section
2.2. /
4. Various methods used in source screening.
aAn emitter was defined as a source with a screening value >200 ppm.
Leak frequency was based on a leak definition of 10,000 ppm.
2-54
-------
instrument used (OVA-108) was of the same type used in the more recent EPA
studies, an action level of 10 ppm was used to determine leaking sources.
Likewise, the Union Carbide study was based on OVA measurements (OVA-128) at
a different action level (1000 ppmv). Although this study presented
detailed analysis of leak frequency as a function of many variables (valve
types, service, location, temperature, pressure, etc.), the applicability of
the results to SOCM1 in general is questionable since the study was
conducted in only a single process unit. The primary result of the series
of German studies was the effectiveness of good maintenance programs in
96
reducing fugitive VOC emissions on a qualitative basis.
The final study available before proposal was a screening study of
97
24 SOCMI process units conducted to determine leak frequencies for a
number of fugitive emission sources. The units were selected to represent a
cross-section of the SOCMI population. They were not randomly chosen.
Therefore, the units may or may not be representative of a true distribution
of SOCMI units. No mass emissions testing was conducted during this study.
Subsequent studies completed after proposal provided further insight
into the fugitive emissions from SOCMI. Non-methane leak rate functions
98
were developed in the SOCMI Maintenance report. In the SOCMI Analysis
go
Report, these functions were used with screening value distribution data
to establish emission factors. Only three process units and three emission
sources were studied, however, thus limiting the utility of the emissions
rate data in describing emissions from SOCMI in general. These factors were
later revised to account for some previous biasing due to maintenance.
The study conducted at an Allied HOPE unit was primarily focussed on
determining maintenance effects on valves and flanges. Although the leak
frequencies measured may be valid and applicable, the emissions rates deter-
mined in this report were represented as emissions from leaking sources
only. But some valves were maintained before they were sampled. The
emission factors, therefore, cannot be directly compared to those determined
for sources in SOCMI and petroleum refinery units because they represent
valves which had, in part, been subjected to a maintenance program.
2-55
-------
Several studies have been conducted to establish the effects of
maintenance on leak frequency and rate for valves in petroleum refineries.
Of these, the most recent was conducted by EPA to examine the effect of
fugitive emissions rules on emissions from refineries in the South Coast Air
Quality Management District (SCAQMD) in California. No mass emissions
data were collected in this study. But leak frequencies were measured for
several types of fugitive emissions sources. Only a small amount of data
was collected on compressors and safety/relief valves.
Although a large number of studies are available on fugitive emissions,
many of the studies had to be eliminated from rigorous consideration for the
purposes of estimating fugitive emissions of VOC from SOCMI. Incomplete
data is the primary limiting factor in most of the studies available. As
noted from the previous discussion, much of the leak frequency data could
not be incorporated in a quantitative analysis due to the leak definition
chosen or the lack of specificity in categorizing sources by service.
Similarly, some of the emission factor data could only be applied quali-
tatively since they represented only leaking sources or, as in the case of
the maintenance studies, included effects of attempted repair.
The two most complete studies of fugitive emissions of the studies
102
available are the twenty-four unit screening study and the petroleum
refinery study. The petroleum refinery study provides the most complete
set of emission factors for fugitive emission sources. The twenty-four unit
study, while it may not represent a true distribution of leaks in SOCMI,
does represent a cross-section of the industry and was more comprehensive
than other screening studies. These two studies were selected as the best
studies on which emission estimates could be based.
2.4.3 Conclusions
Summary of Conclusions -- In the BID, estimates of VOC emissions are based
on a comprehensive study of fugitive emission sources within petroleum
refineries. When the standards were proposed, EPA considered data for
fugitive emission sources within SOCMI. However, these data were not used
in making emission estimates because EPA did not gather the data for
2-56
-------
determining emission factors, and EPA did not consider them representative
of the industry.
In contrast, commenters requested that these data and other new data be
used in making emission estimates for this industry. EPA agrees that these
data could be used to estimate the percent of fugitive emission sources
which leak. EPA continues to believe, however, that data from petroleum
refineries are appropriate for estimating the quantity of VOC emissions from
sources which leak, except for valves in gas service. The percent of
fugitive emission sources which leak and the quantity of VOC emissions from
sources which leak are the primary factors which influence the quantity of
VOC emissions from fugitive emission sources.
Average Unit Concept -- Review of the available data on fugitive emissions
from SOCMI indicated considerable variability among the process units tested
(see Table 2-19). This variability is consistent with EPA's understanding
of the nature of the group of industries which make up the SOCMI industry
category. In an industry category composed of production facilities for
chemicals, no one type of process unit could be considered typical.
However, estimation of national impacts is dependent on the extrapolation of
emissions from a typical unit to the entire U.S. population of units in
SOCMI. Therefore, it is necessary to generate a single set of emission
factors which can be considered as an "industry average". The situation is
similar to the one found in the petroleum refining industry. Just as in the
SOCMI units tested, wide variability was seen in the emissions from
different units in the refineries tested (see Tables 2-2 through 2-6). EPA
uses a single set of emission factors to represent an "average" refinery
even though in reality an "average" refinery may not exist. In the same
manner EPA decided to generate a set of emission factors which represent an
"average" SOCMI unit.
Development of Emission Factors -- As indicated earlier (p.2-56), EPA
determined that the best studies on which emission estimates for SOCMI
104
emission sources could be based were the Petroleum Refinery Study and the
105
Twenty-four Unit Study. Also indicated earlier, EPA considered these
data sets to show differences between the SOCMI data and the petroleum
2-57
-------
refinery data. The assessment of differences and similarities between the
data sets was not clearcut. There were some apparent differences, but they
could not be explained conclusively. The differences may be due to factors
mentioned by the commenters. It is impossible to tell because there are so
many variables. It seemed illogical that on the average, identical equip-
ment handling similar organic compounds would behave differently. However,
EPA determined that the differences, as indicated by the data, were evident.
Because of the differences, EPA decided that an adjustment of the emission
factors used in the BID is warranted.
Reviewing the available studies of fugitive emissions from SOCMI units,
no studies were found that resulted in a full set of emission factors
applicable to SOCMI in general. Furthermore, no study had been designed to
establish a single set of emission factors for SOCMI fugitive emission
sources. Therefore, the results of more than one study would be needed to
estimate fugitive emissions from SOCMI in general. Several approaches were
considered for estimating emissions from SOCMI, in addition to the approach
in the Background Information Document , including the following:
Approach 1 -
Approach 2 -
Approach 3 -
Approach 4 -
Approach 5 -
applying a ratio of leak frequencies (SOCMI to refinery) to
the refinery factors to obtain SOCMI factors;
applying SOCMI leak frequencies to a correlation of
emission factors versus leak frequency generated from
SOCMI and refinery data;
presenting a range of emission factors based on available
SOCMI factors and factors generated from ratios of leak
frequencies applied to refinery factors;
determination of leaking source and non-leaking source
emission factors from the refinery data set (as often as
technically reasonable) and applying these factors to SOCMI
leak frequencies to yield SOCMI factors for an average unit;
using a weighted average of the available SOCMI factors as
the average industry emission factor for those three sources
(pump seals, gas valves, and light liquid valves) and
refinery factors for the remaining sources.
2-58
-------
After considering these alternative approaches, EPA concluded that the
best method of arriving at a complete set of emission factors was by using
108
leak frequencies determined in the 24-unit study to weight the emission
109
factors determined in the petroleum refinery studies (Approach 4). The
approach of averaging emission factors from three types of SOCMI units
suggested by commenters (Approach 5) was not chosen because the average of
those three unit types is not representative of an average for SOCMI. Also,
the level of^existing control in the units tested was undetermined.
Furthermore, emission factors from only three types of emission sources
resulted from this approach. Approach 1 was not chosen because the ratios
constructed in this manner included some with zeros in the denominator.
Approach 2 was discarded because the correlation was not strong enough to
make inferences. The third approach was discarded because ranges would be
too unwieldy when carried through aggregation for the decision making
process.
The approach chosen makes use of the two most comprehensive sets of
fugitive emissions data available at this time. "Leaking source" emission
factors and "non-leaking source" emission factors can be estimated from the
complete petroleum refinery studies and leak frequency data (including
leaking/non-leaking source distributions) representing a cross-section of
SOCMI was available from the 24-unit screening study. Based primarily on
these data sets, emission factors for an average SOCMI unit were computed by
weighting the petroleum refinery emission factors. The manner in which this
weighting was accomplished is described in the following paragraphs.
When a group of sources leak (screening value >_10,000 ppm), they leak
on the average at a certain mass rate. Likewise, those sources not leaking
(screening value <10,000 ppm) have a certain average mass emission rate
associated with them. Overall emission factors for emission sources, then,
are a combination of two components: emissions due to leaking sources and
emissions due to non-leaking sources.
Following this approach, emission factors for leaking sources (LEF) and
emission factors for sources not leaking (NLEF) were determined for fugitive
VOC emission sources using data from the petroleum refinery studies. The
leak/no leak factors were computed according to the following equations:
2-59
-------
LEF = OEF x PCM and NLEF - OEF x (100 - PCM)
PCL (100 - PCL)
where: LEF = emission factor for leaking sources
NLEF = emission factor for sources not leaking
OEF = overall emission factor
PCM = percent of mass emissions due to leaking sources
PCL = percent of sources found leaking
The development of the LEF and NLEF factors is presented in Table 2-32.
The emission factors determined for SOCMI using leak/no leak emission
factors from petroleum refineries in combination with leak frequencies
determined in SOCMI111 are shown in Table 2-33.
The emission factor development for two of the fugitive emission
sources deserves specific comment. The emission factor for sampling
connections is based on the amount of sampling purge reported for every
112
1,000 barrels of refinery throughput and the average count of sampling
113
connections per 1,000 barrels of refinery throughput reported. The ratio
of these values results in an emission factor of 0.0150 kg/hr/source. The
emission factor for open-ended lines, represents valve seal leakage only.
The emissions attributable to the valve, such as from around the stem and
packing, are normal valve leakage and are accounted for in the valve
114
emission factor. In the BID, the emission factor for open-ended valves
included contributions to emissions from the valve and the leakage through
the seat. This combination was confusing and was therefore changed as
explained in Chapter 3.
These "average" SOCMI emission factors were then compared to the
emission factors determined for light liquid pumps, light liquid valves, and
115
gas valves in seven actual SOCMI units. The comparison is shown in
Table 2-34.
It can be seen from the comparison of light liquid pump emission
factors that the average SOCMI factor is lower than the factors for
refineries and ethylene units and higher than the factors for cumene and
vinyl acetate. The heavy liquid pump emission factors are the same for the
average SOCMI unit and refineries.
2-bO
-------
TABLE 2-32.
DEVELOPMENT OF EMISSION FACTORS FOR LEAKING AND NON-LEAKING
SOURCES BASED ON REFINERY FUGITIVE EMISSIONS DATA
(kg/hr/source)
ro
Equipment
Valves
Pump Seals
Compressor
Seals
Pressure
Relief Valves
Flanges
Open-Ended
Lines
Service
Gas
LLd
HL6
LL
HL
Gas
Gas
All
All
Emission
Factor
0.0268
0.0109
0.00023
0.114
0.021
0.636
0.16
0.00025
0.0023
Percent of
Sources
>10,000 ppm
10
11
0.2
24
2
36
7
0.5
7.7
Percent of
Emissions
from Sources
>10,000 ppm
98
86
0.04
92
37
91
74
75
40
Leaker
(_>10,000 ppm) .
Emission Factor
0.2626
0.0852
0.00023f
0.437
0.3885
1.608
1.691
0.0375
0.01195
Non-Leaker
(<10,000 ppm)
Emission Factor
0.0006
0.00171
0.00023
0.0120
0.0135
0.0894
0.0447
0.00006
0.00150
From Appendix B of the finalized refinery assessment report (EPA-600/2-80-075c). Reference 116.
Emission factor times the ratio of percent of mass emissions: percent of sources
screening >10,000 ppm.
Emission factor times the ratio of percent of mass emissions: percent of sources
screening <10,000 ppm.
LL - light liquid service.
eHL - heavy liquid service.
Leaking emission factor assumed equal to non-leaking emission factor since the computed leaking
emission factor (0.00005 kg/hr/source) was less than non-leaking emission factor.
-------
TABLE 2-33. DEVELOPMENT OF AVERAGE SOCMI EMISSION FACTORS
Pump Seals: Light Liquid
Leak
Non-Leak
Total
Heavy Liquid
Leak
Non-Leak
Total
Valves: Gas
Leak
Non-Leak
Total
Light Liquid
Leak
Non-Leak
Total
Heavy Liquid
Leak
Non-Leak
Total
Compressor Seals: Gas
Leak
Non-Leak
Total
Pressure Relief Valves: Gas
Leak
Non-Leak
Total
# Sources
Per 1000
88
912
21
979
114
886
65
935
4
996
91
909
36
964
2
Emission
Factor
(kg/hr)
0.437
0.012
0.3885
0.0135
0.2626
0.0006
0.0852
0.0017
0.00023
0.00023
1.608
0.0894
1.691
0.047
Emissions
per 1000
(kg/hr)
38.5
10.9
49.4
8.20
13.20
21.40
29.9
0.5
30.4
5.5
1.6
7.1
0.00
0.23
0.23
147
81
228
61.0
43.0
104.0
Overall
Emission
Factor
(kg/hr)
0.0494
0.0214
0.0304
0.0071
0.00023
0.228
0.104
2-G2
-------
TABLE 2-33. (CONTINUED)
Open-Ended Lines: All
Leak
Non-Leak
Total
Flanqes: All
Leak
Non-Leak
Total
# Sources
Per 1000
39
961
21
979
2
Emission
Factor
(kg/hr)
0.01195
0.0015
0.0375
0.00006
Emissions
per 1000
(kg/hr)
0.47
1.44
1.91
0.79
0.04
0.83
Overall
Emission
Factor
(kg/hr)
0.0017
0.00083
From SOCMI analysis report (Reference 117): fraction of leaking sources
for the 24 unit screening study (Reference 118).
"From petroleum refinery emission factor development for fugitive VOC
emission sources (Reference 119).
2-63
-------
TABLE 2-34. COMPARISON OF EMISSION FACTORS FOR ILLUSTRATIVE SOCMI CASES
(ETHYLENE, CUMENE, AND VINYL ACETATE UNITS), AVERAGE SOCMI UNIT,
AND PETROLEUM REFINERIES,
kg/hr/source
Source
Pumps - 1 ight 1 iquid
- heavy liquid
Valves - gas
- light liquid
- heavy liquid
Compressors
Safety/relief valves - gas
Flanges
Open-ended lines
Sampling connections
Average
SOCMI
Unit
0.0494
0.0214
0.0304
0.0071
0.00023
0.228
0.104
0.00083
0.0017
0.0150
. . Vinyl .
Refinery3 Ethyl ene Cumene Acetate
0.114 0.058* 0.018* 0.002*
0.021
0.0268 0.0086* 0.007* 0.0014*
0.0109 0.018* 0.006* 0.00023*
0.00023
0.636
0.16
0.00025
0.0023
0.0150
Source: Reference 120.
^Source: Reference 121.
^Emission factors were determined in the SOCMI Analysis report and updated
to account for data biasing (see Section 2.2.8).
2-64
-------
The average SOCMI unit emission factor for light liquid valves is lov;er
than those for refineries and ethylene units. It is about the same as the
light liquid valve emission factor for cumene units, and it is higher than
the emission factor for vinyl acetate units. Emission factors for the heavy
liquid valves in refineries and average SOCMI units are the same.
The average SOCMI unit emission factor for compressors is lower than
the emission factor for refineries, as is the average SOCMI emission factor
for safety relief valves. The average SOCMI emission factor for flanges,
however, is higher than the one for refineries. The factors for open-ended
lines in the average SOCMI unit and in refineries are almost the same. The
petroleum refinery emission factor for sampling connections was not modified
for application to SOCMI because there is no reason to believe that the
amount of sample purge would be different.
The average SOCMI emission factors for light liquid valves and light
liquid pumps are verified by comparison with the emission factors determined
for the three types of SOCMI units. These average SOCMI factors fall within
the range of the emission factors determined in the SOCMI units. However,
the average SOCMI emission factor for gas valves does not compare favorably.
The average SOCMI emission factor for gas valves is higher than the one
determined for gas valves in petroleum refineries and falls outside the
range of the emission factors determined in the SOCMI units tested.
Comparisons of the emission factors for light liquid pumps and valves
and gas valves from petroleum refineries and the three types of SOCMI units
are shown in Figures 2-1, 2-2, and 2-3. As Figure 2-1 shows, the confidence
intervals for the SOCMI gas valve emission factors exhibit almost no overlap
with the petroleum refinery emission factor confidence interval. And the
confidence intervals for the SOCMI gas valve emission factors are narrower
than the interval for the petroleum refinery emission factor. This
comparative analysis of the statistical measures of confidence indicates
that the gas valve emission factors for SOCMI are different from the gas
valve emission factor for petroleum refineries. The smaller confidence
intervals also suggest that, the SOCMI gas valve numbers are better
estimators of SOCMI gas valve emissions.
2-65
-------
i
01
01
(D
U
o
I/)
en
CD
O
5- 4-
O C
+J O
O 0
fO
LJ_ ^S
un
C. CT»
O '
0.050
0.040
0.030
0.020
0.010
15
Vinyl
Acetate
5 10
Leak Frequency (10,000 ppm),
* f 4 T
Petroleum Cumene Ethylene
Refinery
Figure 2-1. Emission factor v. empirical leak frequency:
gas valves for petroleum refinery and SOCMI units.
Sources: Reference 122, Reference 123.
-------
01
o
ro
i
01
o
c
o>
-a
o o
4_J (_)
fO ^S
u_ LO
CTl
0.030
0.020
0.010
olf
Vinyl
Acetate
o
Q
Cumene P<
'etroleum 15
Refinery
20
30
Ethylene
Leak Frequency (10,000 ppm), %
Figure 2-2. Emission factor v. empirical leak frequency:
light liquid valves for petroleum refinery and SOCMI units.
Sources: Reference 124, Reference 125.
-------
0.25
0.20
i
CTi
CO
o>
u
S-
3
o
co
en
S-
o
o
01
o
O)
o
c
o
o
o
to
co
p
UJ
LO
cn
0.15
0.10
0.05
10
Cumene
_ j__,j
Petroleum Ethylene
Refinery
30
Factor 2-3.
Leak Frequency (10,000 ppm), %
Emission factor v. empirical leak frequency:
light liquid pumps fo~r petroleum refinery & SOCMI units.
Sources: Reference 126, Reference 127.
-------
The same comparative analysis for light liquid valves and light liquid
pumps (Figures 2-2 and 2-3) does not result in the same conclusions.
Substantial overlapping of confidence intervals is seen for petroleum
refinery and SOCMI light liquid valves and pumps. This overlap indicates
that the emission factors are similar in value within the confidence of the
estimates. The SOCMI pump emission factors have very broad confidence
intervals, a fact which indicates a small amount of certainty for the value
of the emission factor. The petroleum refinery light liquid pump emission
factor has a smaller confidence interval and, therefore, is considered a
better estimator for light liquid pump emissions.
Based on these considerations, EPA concluded that the gas valve
emissions data developed in the SOCMI units are a better basis for
estimating gas valve emissions in SOCMI than the data from petroleum
refineries. The data further indicate that the petroleum refinery emission
factor data for light liquid valves and pumps are the better of the two sets
of data for estimating emissions from light liquid valves and pumps.
Therefore, EPA has applied the methodology used to develop SOCMI
emission factors from leaking and non-leaking emission factors to the SOCMI
gas valve data. The resulting average SOCMI gas valve emission factor is a
better estimate of SOCMI gas valve emissions. The final slate of average
SOCMI emission factors is shown in Table 2-35.
Assuming these leak/no leak emission factors are applicable to fugitive
emission sources in other industries, the overall emission factor for any
source of fugitive VOC emissions in any industry can be determined if the
leak frequency for the fugitive emission source is known. As shown in the
examples in Table 2-36, the emission factors estimated with this technique
for gas plants and coke ovens compare favorably with the actual emission
factors measured for sources associated with coke ovens and gas plants.
As discussed before, industry commenters also presented SOCMI emission
128 129
factors based on the leak rates reported in the SOCMI analysis report
(see Section 2.3). An apparent decimal point error was corrected in the
emission factor for light liquid pump seals. Also, using the methodology
presented by the commenters and the revised emission factors determined for
2-69
-------
TABLE 2-35. FINAL AVERAGE SOCMI UNIT EMISSION FACTORS
Equipment
Component
"Average" SOCMI Factors
kg/hr/source
Pump Seals
Light Liquid
Heavy liquid
Valvesd
Gas
Light liquid
Heavy liquid
Compressor Seals
Safety relief valves - gas
Flanges
Open-ended lines
Sampling connections
0.0494
0.0214
0.0056
0.0071
0.00023
0.228
0.104
0.00083
0.0017
0.0150
2-70
-------
TABLE 2-36. COMPARISON OF ACTUAL EMISSION FACTORS FOR COKE OVENS AND GAS PLANTS
WITH FACTORS ESTIMATED USING THE LEAK/NO LEAK PROCEDURE, kg/hr/source
IX)
i
Coke Ovens3
Source
Pump seals - overall
Valves - overall
(95%
0
0
Actual
Confidence
.22
.015
(0.054 -
(0.0014
Interval )
0.75)
- 0.14)
Estimated
0.20C
0.0066C - 0.01
(95?
0.
$ 0.
Gas Plants
Actual
Confidence
063
020
(0.02 -
(0.008
Interval )
0.2)
- 0.04)
Estimated
0.14C
0.015C - 0.044
d
Compressor seals -
gas
Flanges
Open-ended lines
Pressure relief devices
0.015 (0.0003 - 0.042]
0.0166 0.020 (0.03 - 1.3)
0.00006 0.0011 (0.0004, 0.002)
0.022 (0.008, 0.04)
0.19 (0.004, 4)
0.89U
0.0012
0.0027
0.33
Source: Reference 130.
Source: Reference 131.
Light liquid service.
Gas service.
eHydrogen service.
-------
SOCMI sources (see Section 2.2.8), another set of emission factors was
developed. These are compared in Table 2-37 with those emission factors
determined using the leak/no leak emission factors in combination with leak
frequencies.
Since industry only presented new factors for light liquid pump seals
and vapor and light liquid valves, only those'factors are compared here.
The methodology employed in generating the industry emission factors
involved a weighted average of the factors determined in-the SOCMI studies.
And, as such, they do not consider the leak rates that were well established
during the refinery studies. This procedure resulted in lower estimated
emission factors by industry for light liquid pump seals and vapor valves.
The industry estimate of emission factor for light liquid valves is higher
than the "average" SOCMI factor since it was heavily influenced by the
factor determined in ethylene units.
2-72
-------
TABLE 2-37. COMPARISON OF EMISSION FACTORS FOR "AVERAGE" SOCMI UNIT
TO EMISSION FACTORS SUBMITTED BY INDUSTRY,9 kg/hr/source
Equipment
Component
Pump Seals
Light liquid
Heavy liquid
Valvesd
Gas
Light liquid
Heavy liquid
Compressor Seals
Flanges
Industry Emi
In Comment
0.02C
0.02
0.0065
0.0045
0.0003
0.441
0.0003
ssion Factors
Revised
0.035
0.02
0.0085
0.0109
0.0003
0.441
0.0003
"Average" SOCMI Factors
0.0494
0.0214
0.0056
0.0071
0.00023
0.228
0.00083
a n 100
"In Comment" refers to the values used by industry; "Revised" presents new
estimates of emissions factors using industry's methodology and the updated
emission factors for SOCMI.
CApparent decimal point error. The number was reported as
0.0045 kg/hr/source. According to the calculation methodology reported, it
should have been 0.045 kg/hr/source.
Estimate presented by industry considered a single value for valves
("valves" included in-line valves, safety relief valves, open-ended valves
& lines, & sampling connections).
2-73
-------
2.5 REFERENCES
1. U.S. Environmental Protection Agency. Background Information for
Proposed Standards for VOC Fugitive Emissions in Synthetic Organic
Chemicals'Manufacturing Industry. Research Triangle Park, N.C.
Publication No. EPA-450/3-80-033a. November 198(f.
2. Wetherold, R.G., L.P. Provost, and C.D. Smith. (Radian Corporation.)
Assessment of Atmospheric Emissions from Petroleum Refining,
Appendix B: Detailed Results. (Prepared for U.S. Environmental
Protection Agency.) Research Triangle Park, N.C. Publication
No. EPA-600/2-80-075c. April 1980.
3. Wetherold, R., and L. Provost. (Radian Corporation). The Assessment
of Environmental Emissions from Oil Refining. (Prepared for U.S.
Environmental Protection Agency.) Interim Report. EPA
No. 600/2-79-044. Research Triangle Park, N.C. February 1979.
4. Reference 2.
5. Analysis of data contained in Reference 2.
6. Reference b.
7. Reference 5.
8. Reference 5.
9. Reference 5.
10. Reference 5.
11. Reference 2.
12. Reference 3.
13. Memo from Hustvedt, K. C., EPA:CPB, to Durham, J. F., EPA:CPB.
December 2, 1980. 170 p. MRC SOCMI Fugitive Testing.
14. U.S. Environmental Protection Agency. Air Pollution Emission Test at
Dow Chemical Company. Research Triangle Park, N.C. EMB Report
No. 78-OCM-12C.
U.S. Environmental Protection Agency. Air Pollution Emission Test at
Union Carbide Corporation. Research Triangle Park, N.C. EMB Report
No. 78-OCM-12A.
2-74
-------
U.S. Environmental Protection Agency. Air Pollution Emission Test a.
Phillips Petroleum Co. Research Triangle Park, N.C. EMB Report
No. 78-OCM-12E.
U.S. Environmental Protection Agency. Air Pollution Emission Test at
Refinery E. Research Triangle Park, N.C. EMB Report No. 78-OCM-12F.
Reference 11.
15. Reference 13.
16. Meeting Report. Honerkamp, P.., Radian Corporation, to Hustvedt, K.C.,
EPA-.CPB, and distribution list. June 12, 1979. 14p. Minutes of
meeting between EPA and DuPont representatives about fugitive emission
sampling.
17. Reference 14.
18. Letter and attachment from Cox, J.B., Exxon Chemical Company, to
Weber B., EPA:CPB. February ?1, 1978. 4p. Copy of letter about
cyclohexane unit fugitive loss data to Hydroscience.
Letter and attachment from Cox, J.R., Exxon Chemical Company, to
Walsh, R.T., EPA:CPB. March 21, 1979. 4p. Information about
cyclohexane unit.
19. Blacksmith, J.R., et al. (Radian Corporation.) Problem Oriented
Report: Frequency of Leak Occurrence for Fittings in Synthetic Organic
Chemical Plant Process Units. (Prepared for U.S. Environmental
Protection Agency.) Research Triangle Park, N.C. EPA Contract
No. 600/2-81-003. September 1980.
20. Reference 16.
21. Reference 18.
22. Reference 19.
23. Reference 19.
24. Reference 14.
25. Reference 19.
26. Reference 2.
27. Reference 2.
28. Reference 18.
2-75
-------
29. Schwanecke, R., "Air Pollution Resulting from Leakage from Chemical
Facilities." Luftverun-reinigung, 1970, pp.9-15. Translated for the
U.S. Environmental Protection Agency by SCITRAN. Santa Barbara,
California.
Kremer, H. "Leakages from Static and Dynamic Seals in Chemical and
Petrochemical Plants." 4th Meeting OG EW/DGMK. Salzburg.
October 1976. Translated for the U.S. Environmental Protection Agency
by SCITRAN, Santa Barbara, California.
Bierl, Alois, et al. "Leakage Rates of Sealing Elements." Chem. Ing.
Tech. 49_ (No.2) 1977, pp.89-95. Translated for the U.S. Environmental
Protection Agency by SCITRAN. Santa Barbara, California.
Schwanecke, R. "Air Polltuion Through Small Leakages for Equipment of
the Chemical Industry and Ways for Their Prevention." Translated for
the U.S. Environmental Protection Agency by SCITRAN. Santa Barbara,
Cal iform'a.
30. Lee, Kun - Chich, et al. (Union Carbide Corporation. South
Charleston, West Virginia.) "A Fugitive Emission Study in a
Petrochemical Manufacturing Unit." Presented at the 73rd Annual
Meeting of the Air Pollution Control Association in Montreal, Quebec.
June 22-27, 1980.
31. Reference 30.
32. Langley, G.J. and R.G. Wetherold. (Radian Corporation.) Evaluation of
Maintenance for Fugitive VOC Emissions Control. (Prepared for the
U.S. Environmental Protection Agencv.) Research Triangle Park, N.C.
Publication No. EPA-600/52-81-080. "May 1981.
33. Reference 30.
34. Reference 32.
35. Langley, G.J., et al. (Radian Corporation.) Analysis of SOCMI VOC
Fugitive Emissions Data. (Prepared for the U.S. Environmental
Protection Agency.) Research Triangle Park, N.C. Publication
No. EPA-600/2-81-111. June 1981.
36. Reference 35.
37. Reference 19.
38. Harvey, "'Cynthia M. and A. Carl Nelson. (PEDCo Environmental Inc.)
Fugitive Emission Data - High Density Polyethylene Process Unit.
(Prepared for the U.S. Environmental Protection Agency.) Research
Triangle Park, N.C. Publication No. EPA-600/2-81-109. June 1981.
VOC
2-76
-------
39. Reference 35.
40. Reference 38.
41. Reference 38.
42. Honerkamp, R.L. and M.L. Schwendeman. (Radian Corporation.)
Evaluation of the Maintenance Effect on Fugitive Emissions from
Refineries in the South Coast Air Quality Management District. Draft
Final Report. (Prepared for the U.S. Environmental Protection Agency.)
Research Triangle Park, N.C. December 1981.
43. Wiesenborn, D. P., et al. (Radian Corporation.) Lea'k Frequency and
Emission Factors for Fittings in Coke Oven By-Product Plants.
(Prepared for U. S. Environmental Protection Agency.) Draft Final
Report. Research Triangle Park, N. C. February 1982.
44. DuBose, D. A., et al. (Radian Corporation.) Emission Factors and Leak
Frequencies for Fittings in Gas Plants. (Prepared for U. S. Environ-
mental Protection Agency.) Draft Final Report. Research Triangle
Park, N. C. September 1981.
45. Reference 42.
46. Reference 43.
47. Reference 44.
48. Langley, G.J. and L.P. Provost. (Radian Corporation.) Revision of
Emission Factors for Nonmethane Hydrocarbons from Valves and Pump Seals
in SOCMI Processes. Technical Note. (Prepared for the U.S. Environ-
mental Protection Agency.) Research Triangle Park, N.C. November 1981
49. Reference 48.
50. Reference 19.
51. Reference 32.
52. Reference 35.
53. Reference 13.
54. Letter with enclosures from Schroy, Jerry M., Monsanto Company, to
Central Docket Section, Docket No. A-79-32. March 26, 1981. Technical
Comments on Proposed VOC Emission Regulations.
2-77
-------
55. Letter from Samel son, R.J., PPG Industries, Inc. to Central Docket
Section, Docket No. A-79-32. March 19, 1981. Comments on proposed
regulation of VOC fugitive emissions within the Synthetic Organic
Chemical Manufacturing Industry.
56. Reference 48.
57. Letter with enclosures from Matey, Janet S., Chemical Manufacturers
Association to Central Docket Section, Docket No. A-79-32. August 7,
1981. Comments on SOCMI Reports.
58. Reference 57.
59. Reference 2.
60. Reference 13.
61. Reference 18.
62. Reference 48.
63. Reference 43.
64. Reference 44.
65. Reference 2.
66. Reference 13.
67. Reference 14.
68. Reference 16.
69. Reference 18.
70. Reference 35.
71. Reference 30.
72. Reference 38.
73. Reference 42.
74. Reference 43.
75. Reference 44.
76. Reference 2.
2-78
-------
77. Reference 14.
78. Reference 8.
79. Reference 15.
80. Reference 2.
81. Reference 13.
82. Reference 14.
83. Reference 16.
84. Reference 18.
85. Reference 19.
86. Reference 29.
87. Reference 30.
88. Reference 32.
89. Reference 35.
90. Reference 48.
91. Reference 38.
92. Reference 42.
93. Reference 43.
94. Reference 44.
95. Reference 30.
96. Reference 29.
97. Reference 19.
98. Reference 32.
99. Reference 35.
100. Reference 38.
101. Reference 42.
2-79
-------
102. Reference 19.
103. Reference 2.
104. Reference 2.
105. Reference 19.
106. Reference 1.
107. Reference 57.
108. Reference 19.
109. Reference 2.
110. Reference 2.
111. Reference 19.
112. U. S. Environmental Protection Agency. Compilation of Air Pollutant
Emission Factors. Research Triangle Park, N. C. AP-42. February
1980.
113. Powell, D., et al. (PES, Inc.) Development of Petroleum Refinery Plot
Plans. (Prepared for U. S. Environmental Protection Agency.) Research
Triangle Park, N. C. EPA Publication No. EPA-450/3-78-025". June 1978.
114. Reference,1.
115. Reference 48.
116. Reference 2.
117. Reference 35.
118. Reference 19.
119. Reference 2.
120. Reference 2.
121. Reference 48.
122. Reference 2.
123. Reference 48.
124. Reference 2.
2-60
-------
125. Reference 48.
126. Reference 2.
127. Reference 48.
128. Reference 57.
129. Reference 35.
130. Reference 43.
131. Reference 44.
132. Reference 57.
2-81
-------
3. MODEL UNITS
In order to estimate industry wide fugitive emissions of VOC, control
costs of emission reduction techniques and environmental impacts for SOCMI
units, three model units were developed in the BID. These three model units
represented the range of emission source populations that may exist in SOCMI
process units. The model units are, therefore, the starting point for
projections of industry wide emissions, costs, and environmental impacts
associated with various control options. This section presents the basis
for the development of the model units in the BID, new information and
public comments regarding the model units, and the decisions made by EPA in
view of the new information and the comments.
3.1 TECHNICAL BASIS IN THE BID
The model units presented in the BID were used as the basis for the
analysis of impacts of the standards. The model units were based primarily
on process complexity because fugitive VOC emissions generally are related
to the number of equipment components in a process unit, and are not related
to equipment size or process unit capacity. The model units used in
characterizing fugitive VOC emissions from SOCMI were developed in the study
of SOCMI performed for EPA by IT Enviroscience (formerly Hydroscience).
Fifty-one process units were surveyed in the Hydroscience study. Based on
information from this study and as illustrated in Figures 3-1 and 3-2, there
is a general lack of correlation between the number of equipment components
(pumps and valves) and the rated production capacity of the process unit.
The results of the Hydroscience study and equipment counts provided by
engineering design and construction firms are the basis of the model units.
In addition, the equipment counts in new units were expected to be
comparable to the equipment counts in existing units.
In examining available information on SOCMI units, the most complete
information available on equipment counts was found to be the total number
3-1
-------
Log Y - .82 + .27 Log X (Slop*)
r2 - 0.07 (Correlation Coefficient)
1OOO
50O
~ \QQ
c
a
~4
o.
u
o.
a
Q.
O
10
. UARGE.
PL.AUT
ME.DIUK
MOO
' 15 %
4 MO
5.1
DEI.
PUAWT 3i "T.
SMALl.
PLAMT
i
MOC
52.'
>£.!_
.
M
^-"^
t
9 1
_
»
^
""^
1
^^
^v
1
^f
t
^*
*~'
[
I
1
-
V
1
. '
10
SO IOO
Plant Capacity M Ibs (X)
500
IOOO
Figure 3-1. Total number of pumps per process unit as a
function of the rated annual production capacity
(million Ibs)/
3-2
-------
Log Y - 2.13 « 0.31 Log X (Slop*)
r -0.10 (Correlation Coefficient)
HJVJtAJ
aooo
6OOO
4OOO
2000
>
"" 1OOO
u
g 400
a,
h bco
0.
« 4OO
10
*
«
^~*~
-- "
. .
.
.
1
^rf**
^^^^
"
***~~~
.-
-r~
i .
.
a
.^^
-
1
i
.
»
IOO _ IOOO
Plant Capacity M Ibs (X)
Figure 3-2.
Total number of valves per process unit as a
function of the rated annual production capacity
(million Ibs) .3
33
-------
of pumps in each process unit. To quantify the model units across the range
of possible units in SOCMI, the numbers of pumps for the model units were
chosen as 15, 60, and 185. The number of valves in each unit was determined
by using the valve-to-pump ratio (25:1) computed as the overall average from
n
the Hydroscience report. Flange estimates were made using the flange-to-
valve ratio (1.6:1) determined from a study of 8 SOCMI process units
performed by Pullman Kellogg.
The differentiation of valves and pumps by service and types was based
on the average numbers of these pieces of equipment in each type of service
in the data base formed by the Hydroscience report and information from
construction firms. These average numbers were determined from continuous
and batch operations of varying capacity processes. The typical industry
distributions for valves and pump seals thus generated are presented in
Tables 3-1 and 3-2. In order to simplify the analysis, EPA assumed that on
the average for SOCMI, pump seals are split 50/50 between light liquid and
heavy liquid service, that all packed pumps are used in heavy liquid
service, and that all pumps with dual seals are used in light liquid
service. EPA also assumed that valves in liquid service were split 50/50
between light and heavy liquid service.
The estimated number of sampling connections in each model unit was
based on the equipment count data which showed that 25 percent of the
open-ended valves were used for sampling. The total number of compressor
seals for each model unit was selected from the Hydroscience and construc-
tion firm data and was considered representative of the industry.
The resulting model units for SOCMI are presented in Table 3-3. These
units spanned the ranges of equipment counts per process unit in the
industry. The model units thus derived and presented in the BID based on
equipment counts incorporated information concerning the degree of control
currently practiced by industry.
3.2 NEW INFORMATION
Some new information was acquired since the publication of the BID as a
result of the 24-unit screening study (previously described in the section
on emission factors, Section 2). As seen in Table 3-4, the equipment counts
3-4
-------
TABLE 3-1. SOCMI VALVE CHARACTERIZATION3'b
Type of
Valve
Safety-relief
Open-ended (sample
vent, drain)
In-line (process,
control )
All valves
Percent of
Total Valves
3.4
27.6
69.0
100.0
Percent in
Liquid Service
17.0
91.0
65.0
71.0
Percent
Vapor (Gas)
83.0
9.0
35.0
29.0
in
Service
Includes only valves in VOC service.
Reference 8.
Check valves are excluded.
TABLE 3-2. SOCMI PUMP SEAL CHARACTERIZATION9'b
Pump Seals Percent in Use
Mechanical ~~~~~
Single 71.9
Double 16.7
Packed 10.6
None (sealless) 0.8
Total 100.0
alncludes only pump seals in VOC service.
Reference 9.
3-5
-------
TABLE 3-3. FUGITIVE EMISSION SOURCES FOR THREE MODEL UNITS
Number of comoonents in model um't^
Equipment component3
Pump seals
Light liquid service
Single mechanical
Dual mechanical
Sealless
Hea-vy liquid service
Single mechanical
Packed
In-line valves
Vapor service
Light liquid service
Heavy liquid service
Safety/ relief valves
Vapor service
Light liquid service
Heavy liquid service
Open-ended valves and lines'*
Vapor service
Li.ght liquid service
Heavy liquid service
Compressor seals
Sampling connections
Flanges
Cooling towers
Model unit
A
5
3
0
5
2
90
84
84
11
1
1
9
47
48
1
26
600
-_e
Model unit
B
19
10
1
24
6
365
335
335
42
4
4
37
189
189
2
104
2400
e
Model unit
C
60
31
1
73
20
1117
1037
1037
130
13
14
115
581
581
3
320
7400
__e
bEquipment components in VOC service only.
52% of existing units are similar to Model Unit A.
33% of existing units are similar to Model Unit B.
C15% of existing units are similar to Model Unit C.
.Sample, drain, purge valves and the associated open end.
Based on 25% of open-ended valves.
.Data not available.
Reference 10.
3-6
-------
TABLE 3-4. SUMMARY OF SELECTED EQUIPMENT COUNTS FOR MODEL UNITS
AND UNITS IN SOCMI 24-UMIT STUDY '
CO
I
In-line valves
Unit
Number
A
B
C
1
2
3
4
5
6
11
12
20
21
22
28
29
31
32
33
34
35
60
61
62
64
65
66
Safety/relief
Process Type Valves
Model unit
Hodel unit
Model unit
Vinyl acetate
Ethylene
Vinyl acetate
Ethylene
Curaene
Cunene
Ethylene
Acetone/phenol
Ethylene dlchloride
Vinyl chloride monomer
Formaldehyde
Ethylene dlchloride
Vinyl chloride monomer
Methyl ethyl ketone
Methyl ethyl ketone
Acetaldehyde
Methylmethacrylate
Adipic acid
Trichloroethylene/
perchloroethylene
1,1.1 - Trlchloroethane
Ethylene dlchloride
Adipic add
Acrylonltrile
Acrylonitrile
11
42
130
31
47
31
101
10
18
65
29
30
IB
3
10
4
1
2
15
21
5
5
-
-
-
-
-
Gas
90
365
1117
377
2563
872
2425
123
422
2396
8
390
93
48
1C8
420
82
169
236
220
48
52
-
29
61
292
221
Light
Liquid
84
335
1037
788
1412
1570
2295
354
573
1118
2075
916
751
126
474
1806
348
389
610
1179
17
1800
430
-
-
723
1093
Heavy
Liquid
84
335
1037
67
1311
61
54
71
177
59
530
-
-
-
-
-
-
-
-
-
1232
12
-
-
342
114
-
Open-ended lines
Gas
9
37
115
37
56
142
79
2
4
208
2
43
4
16
27
lib
12
27
51
73
4
21
-
22
29
77
68
Light
Liquid
47
189
581
49
39
305
59
6
10
144
582
230
44
37
136
470
107
97
168
351
1
483
130
-
-
202
370
Heavy
Liquid
48
189
581
8
93
15
3
-
1
-
Ill
-
-
-
-
-
-
-
-
-
200
4
-
-
122
48
-
Pumps
LightTTeavy
Liquid Liquid
8
29
91
44
22
49
35
10
15
27
90
45
10
8
22
51
15
16
32
49
-
60
10
-
-
24
42
7
30
93
5
16
-
-
1
2
-
36
-
-
-
-
-
-
-
-
-
60
-
-
-
33
10
-
Compressors
1
2
8
4
8
4
3
-
2
7
-
-
-
-
-
-
-
1
-
-
-
-
-
-
1
2
-
Flange;
600
2400
7200
1940
5340
3820
15980
875
2990
1220
2280
1020
980
200
280
2840
820
1140
080
1425
1175
2760
410
45
770
1310
1935
Reference 11.
Reference 12.
-------
for the 24 process units in the screening study generally fall within the
range of equipment counts described by the model units presented at
proposal. For example, safety relief valves ranged from none to 101 per
process unit, light liquid pumps ranged from none to 90 and compressors
ranged from none to 8. Ethylene units are the exception with a high number
of valves and flanges.
The degree of control existing in SOCMI for compressors was indicated
to be higher than assumed in the BID. It was found from the 24 unit study
that the seal areas of 60 percent of the compressors were shrouded and
13
vented either back to the process or to a flare header. This high degree
of control resulted in a low leak frequency measured for compressors. At
this time, EPA is gathering more information concerning the existing level
of control for pumps in SOCMI. Details about the pumps and barrier fluid
systems in the SOCMI units tested are being requested.
3.3 PUBLIC COMMENT
Few comments were received from the public concerning the model units
presented at proposal. The comments received from industry centered around
four major points.
1. The emission factor for open-ended lines in the BID included two
components: one for leakage through the valve seat ("the open end") and one
for leakage around the valve stem (or valve leakage). This distribution of
emissions was found to be confusing to several commenters.
2. Specific comments were also received on the degree of control
practiced in the industry. Commenters stated that open-ended lines in
process units built today use plugs, caps, and blinds. And 75 percent of
14
the sampling connections used today are of the* closed purge variety.
3. One commenter presented the premise that there is a relationship
between fugitive emissions and plant size or production rate. For example,
smaller capacity units leak less on a mass basis than larger capacity units.
4. Another commenter stated that the model units should not be
presented as small, medium, and large units. Rather, the model units should
be classified as low-leak, high-leak, and ethylene type units as was done in
the SOCMI analysis report.
-------
3.4 FPA's CONCLUSIONS
After reviewing the data bases and comments received on the model
units, EPA decided to retain the model units presented in the BID for
purposes of analysis. But, given some of the comments and new information
concerning the degree of control presently practiced by the industry, some
clarifications and adjustments have been made. A comparison of the old and
new model units is presented in Table 3-5.
First, to avoid the confusion concerning the emission factor for open-
ended lines, the emission factor was changed to represent only the emissions
through the valve seat ("the open end"). The valve counts were then
adjusted to include those valves that were previously included as part of
the open-ended line. Also, since only one emission factor for open-ended
lines in all VOC services (gas, light liquid, heavy liquid) was used, a
single equipment count was used for all open-ended lines instead of
individual counts for gas, light liquid, and heavy liquid services. In
addition, nearly all open-ended lines have been assumed to be currently
controlled since industry had stated that capping and plugging open-ended
lines is standard industry practice.
Second, the number of uncontrolled sampling connections represented in
the model units was reduced by 75 percent. EPA decided to make this change
because industry commented that 75 percent of the sampling connections in
the existing process units currently use closed purge sampling mechanisms.
Third, the number of compressor seals was not adjusted to reflect the
higher degree of control indicated by the 24-unit screening study. Instead,
the emission factor for compressors accounted for the level of control. The
degree of control for compressors was indicated by the low leak frequency
determined for compressors in the 24-unit study. The leak frequency
reported for the 24-unit study was based on all compressors, controlled as
well as uncontrolled. And the emission factor generated for compressors
consisted of the emissions due to all leaking compressors as well as the
emissions due to all non-leaking compressors (including controlled compres-
sors). Therefore, the fugitive emissions due to compressors must be
estimated based on all compressors, not just uncontrolled compressors.
3-9
-------
TABLE 3-5. EQUIPMENT COUNTS FOR FUGITIVE VOC EMISSION SOURCES IN SOCMI MODEL UNITS
CO
1
Equipment Component3
Pump Seals
Liyht Liquid Service
Single mechanical
Dual mechanical
Seal less
Heavy Liauid Service
Single mechanical
Packed
Valves
Vapor service
Light liquid service
Heavy liquid service
Safety/relief valves
Vapor service
Light liquid service
Heavy liquid service
Open-ended lines
Vapor service
Light, liquid service
Heavy 1 iquid service
Compressor seals
Sampling connections
Flanges
Model Unit
A
5
3
0
5
2
90
84
84
11
1
1
9
47
48
1
26
600
BID Analysis
Model Unit
B
19
10
1
24
6
365
335
335
42
4
4
37
189
189
2
104
2400
Equipment Counts
Model Unit
C
60
31
1
73
20
1117
1037
1037
130
13
14
115
581
581
8
320
7400
Model Unit
A
5
3
0
5
2
99
131
132
11C
1
1
104d
1
26e
600
Revised Analysis
Model Unit
B
19
10
1
24
6
402
524
524
42C
4
4
415d
2
104e
2400
Model Unit
C
60
31
1
73
20
1232
1618
1618
130C
13
14
1277d
8
320e
7400
Equipment components in VOC service only.
52* of existing units are similar to Model Unit A.
33% of existing units are similar to Model Unit B.
15% of existing units are similar to Model Unit C.
Seventy-five percent of gas safety/relief valves are assumed to be controlled at baseline; therefore the emissions
estimates are based on the following counts: A,3; B,ll; C,33.
All open-ended lines are considered together with a single emission factor; 100% controlled at baseline.
eSeventy-five percent of sampling connections are assumed to be controlled at baseline, therefore, the emissions
estimates are based on the following counts: A,7; B.26; C,80.
-------
However, the compressors that are already controlled need not be considered
for the purpose of the cost analysis. That is, since 60 percent of all
compressors are controlled, the cost of control due to regulation is
applicable only to 40 percent of the compressors in the model unit. This
gives 0.4, 0.8, and 3.2 compressors for cost analysis in model units A, B,
and C respectively rather than 1,2, and 8 compressors for the environmental
analysis.
Next, the number of uncontrolled safety relief valves was reduced to
25 percent. The degree of control found in the 24-unit screening study was
higher than that reflected in the model units. Some units had nearly all
safety relief valves tied into flares. Based on this information, it is
expected that at least half of the safety relief valves will be controlled.
Therefore, the assumption was made that 75 percent of safety relief valves
in new, modified, or reconstructed SOCMI units will be controlled.
The model units were not changed to reflect a relationship between
fugitive emissions and production rate. There are no data relating fugitive
emissions to throughput or production rate. Instead, data collected in the
refinery and SOCMI studies indicate that process fluid vapor pressure is the
primary factor influencing the fugitive emissions rate. Furthermore, data
indicate " that there is no relationship between number of fugitive emission
sources and throughput. Therefore, EPA judged that fugitive emissions
varied with the number of components in a process unit (i.e., complexity)
and not with the production rate of the process unit.
And finally, based on the results of the SOCMI and petroleum refining
screening and maintenance studies, leak frequency might have been used as a
supplementary basis for considering model units as suggested by one
commenter. However, several complications make it impracticable to do so.
For example, there is no information available on the breakdown of the
number of process units in each of the three leak frequency categories. It
is also impracticable to do a cost analysis for model units that are based
on leak frequencies. Without any of this 'information, industry wide impacts
cannot be estimated. In addition, it is not practicable to categorize a
given process type as low, high, or ethylene type. For instance, one methyl
3-11
-------
ethyl ketone (MEK) unit in the 24 unit study was a low leak unit. Another
MEK unit fell in the high leak category. Some units may have low leak
frequency for one type of fugitive emission source (e.g. valves) and a high
leak frequency for another type of emission source (e.g. pumps). These leak
frequencies may also change over time. Therefore, EPA concluded that leak
frequencies should not be the basis for the model units. However, EPA plans
to evaluate leak frequencies on a fugitive emission source basis in
determining whether alternative control strategies should be selected for
fugitive emission sources with low leak frequencies.
3-12
-------
3.5 REFERENCES
1
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
Erikson, D.G. and V. Kalcevic. (IT Enviroscience.) Organic Chemical
Manufacturing Volume 3: Storage, Fugitive, and Secondary Sources.
Report 2. (Prepared for U.S. Environmental Protection Aaency.)
Research Triangle Park, N.C. EPA Publication No. EPA-450/3-80-025.
December 1980.
Reference 1, p. B-4.
Reference 1, p. B-5.
Reference 1, p. 11-13.
Pullman Kellogg. Equipment Component Analysis for Identification of
Potential Fugitive Emission Sources. (Prepared for U.S. Environmental
Protection Agency.) Held on file at Emission Standards and Engineering
Division. Research Triangle Park, N.C. June 1978.
Reference
Reference
Reference
Reference
Reference
Reference
1,
1,
1,
1,
1,
1,
P-
P-
P-
P-
P-
P-
I
I
I
I
1-13.
1-13.
1-10.
1-10.
11-12.
I
1-12.
Blacksmith, J.R., G.E. Harris, and G.L. Langley. (Radian Corporation.)
Frequency of Leak Occurrence for Fittings in Synthetic Organic Chemical
Plant Process Units. (Prepared for U.S. Environmental Protection
Agency.) Washington, D.C. EPA Publication No. EPA-600/2-81-003.
September 1980.
Reference 12.
Memo from Dimmick, F., FPA:SDB, to Wyatt, S., EPA:SDB. September 4,
1980. lOp. Notes of meeting with Texas Chemical Council on July 18,
1980.
15. Reference 1, p.B-4 and B-5.
3-13
-------
4. EMISSION REDUCTIONS
Estimates of the effectiveness of control techniques for fugitive
emissions of VOC from SOCMI sources were presented in the BID (EPA-450/3-80-
033a). Some new information has become available concerning the effec-
tiveness of the control techniques presented in the BID. Some of this
information allows more accurate estimation of emission reductions which can
be achieved through the use of these control techniques. This section
contains a discussion of emission reductions achievable in light of the
basis presented in the BID and that new information. The discussion is
arranged by fugitive emission source: valves, pumps, sampling systems,
safety relief valves, open-ended lines, compressors, and control devices.
For each of these fugitive emission sources the basis presented in the BID
is reviewed, new information is presented, and relevant public comments are
summarized. Finally, EPA's view of the effectiveness of controls and how
that effectiveness should be estimated is presented.
4.1 VALVES
4.1.1 Technical Basis Presented in the BID
The BID (EPA-450/3-80-033a) included a discussion of fugitive emissions
from valves and the control techniques for reduction of those emissions.
Two general control techniques were considered in the BID: a leak detection
and repair program and leakless equipment.
Leakless equipment for valves, such as diaphragm and bellows seal
valves, eliminate the seals which allow fugitive emissions and, thus, their
control effectiveness is virtually 100 percent. However, as noted in the
BID, the applicability of these types of valves is limited. Because these
leakless types of equipment are limited in their applicability, the regula-
tory alternatives developed in the BID only incorporated leak detection and
repair programs. The leak detection methods considered were individual
4-1
-------
component surveys (with soap and instruments), area surveys, and fixed point
monitors.
The leak detection method selected for incorporation in the regulatory
alternatives was individual component surveys using a portable VOC instru-
ment. The individual component survey was selected because it was more
reliable in detecting leaks than fixed area monitors or walk-through
surveys. The portable VOC monitor was selected over the other semi-quanti-
tative method of leak detection, application of soap solutions because the
temperature of the fugitive emission source, the physical configuration, and
the relative movement of parts often interfere with bubble formation.
Repair procedures for valves were also presented in the BID. The basic
repair procedure for leaking valves is the tightening of the packing gland.
Depending on site specific factors, it may be possible to repair valves by
injection of a sealing fluid into the source. In some cases it would be
necessary to replace the packing or to replace the valve.
Data from five studies on the effects of maintenance on fugitive
emissions from valves were presented in the BID. These studies are briefly
described below.
2
Union Oil Maintenance Study -- The Union Oil valve maintenance study
consisted of performing undirected maintenance on valves selected from 12
different process units. Maintenance procedures in this study consisted of
adjusting the packing gland while the valve was in service. Undirected
maintenance consists of performing valve repairs without simultaneous
measurement of the VOC concentration detected. Directed maintenance
involves simultaneous measurement during the repair procedure. With
directed maintenance repair procedures are continued until the VOC
concentration detected drops to a specified level or further reduction in
the emission level is not possible. Also, maintenance may be curtailed if
increasing VOC concentrations result.
The Union Oil data were obtained with a Century Systems Corporation
Organic Vapor Analyzer, OVA-108, calibrated with methane. All measurements
were taken at a distance of 1 cm from the seal. Correlations developed in a
3
study of six refineries in the Bay Area were used to convert the data from
4-2
-------
OVA readings taken at one centimeter to readings equivalent to those
measured with a Bacharach TLV (calibrated to hexane) at the leak interface
(TLV-0). Two sets of results were provided; the first includes all repaired
valves with before maintenance screening values greater than or equal to
5,300 ppmv (OVA-108), and the second includes valves with before maintenance
screening values below 5,300 ppmv (OVA-108).a Estimates of emissions were
made using correlations of emissions with screening values developed by EPA
4
in the Petroleum Refinery Study.
The results of the Union Oil data are shown in Table 4-1. The results
of this study indicated that maintenance on valves with initial screening
values above 10,000 ppmv (OVA-108) is much more effective than maintenance
on valves leaking at lower rates. In fact, in this study 124 out of 133
valves originally screening over 5300 ppm had lower screening values after
maintenance while only 13 out of ?1 valves originally screening under
5300 ppm had lower screening values after maintenance.
Shell Maintenance Study -- The Shell maintenance program consisted of two
parts. First, valve repairs were performed on 171 leaking valves. In the
second part of the program, 162 of these valves were rechecked and
additional maintenance was performed. Maintenance consisted of adjusting
the packing gland while the valve was in service. The second part of the
program was conducted approximately one month after the initial maintenance
period. The information reported by Shell did not indicate whether the
maintenance procedures were directed or undirected.
VOC emissions were measured using the OVA-108 calibrated with methane
and readings were obtained one centimeter from the source. These data were
aA screening value of 5,300 ppmv, obtained with an OVA at 1cm from the leak
interface, is equivalent to a screening value of 10,000 ppmv measured
directly at the leak interface by a Bacharach Instrument Co. "TLV Sniffer"
calibrated with hexane. The OVA-lcm readings were converted to equivalent
TLV-Ocm readings from the following reasons:
(1) EPA correlations which estimate leak rates from screening values were
developed from TLV-Ocm data.
(2) Additional maintenance study data existed in the TVL-Ocm format.
(3) Method 21 specifies Ocm screening procedures.
4-3
-------
TABLE 4-1. SUMMARY OF MAINTENANCE STUDY RESULTS FROM THE
UNION OIL CO. REFINERY IN RODEO, CALIFORNIA
All valves
with initial
screening values
>5300 ppmva
Number of repairs attempted
Estimated emissions before maintenance, kg/hr
Estimated emissions after maintenance, kg/hr
Number of successful repairs (<5300 ppmv after maintenance)
Number of valves with decreased emissions
Number of valves with increased emissions
Percent reduction in emissions
Percent successful repairs
Percent of valves with decreased emissions
Percent of valves with increased emissions
133
9.72
4.69
67
124
9
51.8
50.4
90.2
6.8
All valves
with initial
screening values
>5300 ppmv
21
0.323
0.422
--
13
8
-30.5
--
61.9
38.1
aThe value 5300 ppmv, taken with the OVA-108 at 1 cm., generally corresponds to a value of
10,000 ppmv taken with a "TLV Sniffer" at 0 cm.
Source: Reference 6.
-------
transformed to TLV-0 cm (calibrated to hexane) values as were the Union
data. And, the same methods of data analysis used in the Union Oil study
were applied to the Shell data. Emission rates were generated by using
correlations of emission rates and screening values generated by EPA in the
Petroleum Refinery Study.
The results of the Shell maintenance study are given in Table 4-2. The
results show that all of the valves with screening values above 5300 ppmv
had lower screening values after maintenance in March. The results for
successive maintenance attempts in April show that 151 out of 152 valves had
decreased screening values after maintenance.
o
EPA Refinery Maintenance Study -- Repair data were collected on valves
located in four refineries. The effects of both directed and undirected
maintenance were evaluated. Maintenance consisted of routine operations,
such as tightening the packing gland or adding grease. Other data,
including valve size and type and the processes' fluid characteristics, were
obtained. Screening data were obtained with the Bacharach Instrument
Company's "TLV Sniffer" calibrated to hexane and readings were taken as
close to the source as possible.
Unlike the Shell and Union studies, emission rates were not based on
the screening value correlations. Rather, each valve was sampled to
determine emission rates before and after maintenance. These values were
used to evaluate emissions reduction.
The results of this study are given in Table 4-3. Of interest here is
a comparison of the emissions reduction for directed and undirected mainte-
nance. The results indicate that directed maintenance is more effective in
reducing emissions than is undirected maintenance, particularly for valves
with lower initial leak rates. While for most of the valves tested, the
emission reductions achieved were greater than 80 percent, the results
showed an increase in total emissions of 32.6 percent for valves with
initial screening values less than 10,000 ppmv which were subjected to
undirected maintenance. However, this increase is due to a large increase
in the emission rate of only one valve.
4-5
-------
TABLE 4-2. SUMMARY OF MAINTENANCE STUDY RESULTS FROM THE SHELL OIL COMPANY
REFINERY IN MARTINEZ, CALIFORNIA
March maintenance
Number of repairs attempted
Estimated emissions before maintenance, kg/hrb
Estimated emissions after maintenance, kg/hrb
Number of successful repairs (<5300 ppmv after
maintenance)
Number of valves with decreased emissions
Number of valves with Increased emissions
Percent reduction 1n emissions
Percent successful repairs
Percent of valves with decreased emissions
Percent of valves with increased emissions
All repaired valves
with initial screening
values >5300 ppmv
161
11.08
2.66
105
161
0
76.0
65.2
100.0
0.0
All repaired valves
with initial screening
values <5300 ppmv
11
0.159
0.0
--
11
0
100.0
--
100.0
0.0
April maintenance
All repaired valves with
initial (March) screening
values 25300 ppmv
152C
2.95
0.421
45
151
1
85.7
83.3
99.3
0.7
All repaired valves with
initial (March) screening
5300 (note nine valves
from Initial data set not rechecked In April).
Initial value of 10 of these valves was <1500 ppmv-TLV at 0 cm.
Source: Reference 9.
-------
TABLE 4-3. .SUMMARY OF EPA REFINERY MAINTENANCE STUDY RESULTS
Repaired valves with initial Repaired valves with initial!
screening values >10,000 ppmv screening values <10,000 ppmv
Number of valves repaired
Measured emissions before maintenance
kg/hr
Measured emissions after maintenance
kg/hr
Number of successful repairs
(<10,000 ppmv after maintenance)
Number of valves with decreased
emissions
Number of valves with increased
emissions
Percent reduction in emissions
Percent successful repairs
Percent of valves with decreased
emissions
Percent of valves with increased
emissions
Directed
Maintenance
9
0.107
0.0139
8
9
0
87.0
88.9
100.0
0.0
Undirected Directed
Maintenance Maintenance
23 10
1.809 0.0332
0.318 0.0049
13
21 6
2 4
82.4 85.2
56.5
91.3 60.0
8.7 40.0
Undirected
Maintenance
16
0.120
0.159
-
15
1
-32.6
-
93.8
6.3
Source: reference
10.
-------
Unit D (Ethylene Unit) Maintenance Study -- Maintenance was performed by
Unit D personnel. VOC concentration measurements were made using the
OVA-108 calibrated with methane, and readings were obtained as close as
possible to the source. The results of this study are shown in Table 4-4.
Directed and undirected maintenance procedures were used. The results show
that follow up directed maintenance on valves not repaired by undirected
maintenance results in more repairs being successfully completed than when
undirected maintenance is used alone.
1?
Chevron Refinery Study " -- The Chevron study included inspection of 33,000
valves. A Century OVA-108 hydrocarbon analyzer was used and the VOC concen-
tration was measured at 1 cm from the source. Approximately 4 percent of
the screened valves were found leaking. Maintenance consisted of adjustment
of the packing. Following this maintenance procedure, 93 percent of the
leakers were repaired. Approximately 5 percent of the leakers required
repacking. Less than 1 percent required a more elaborate repair procedure
such as injecting a sealant type material. Similarly, less than 1 percent
of the leakers required replacement.
Summary of Studies -- The following conclusions were drawn from the five
maintenance studies presented in the BID:
1. A reduction in emissions may be obtained by performing maintenance
on valves with screening values above 10,000 ppm (measured at the
source).
2. The reduction in emissions due to maintenance of valves with
screening values below 10,000 ppm is not as dramatic and may
result in increased emissions if undirected maintenance is used.
3. Directed maintenance is preferable to undirected maintenance for
valve repair.
Based on the results of these five studies and other factors, four
regulatory alternatives were constructed which incorporated leak detection
and repair programs. The regulatory alternatives constructed for valves
presented in the BID were as follows:
4-8
-------
TABLE 4-4. MAINTENANCE EFFECTIVENESS UNIT D ETHYLENE UNIT BLOCK VALVES
UNDIRECTED MAINTENANCE
1. Total number subjected to repair attempts 37
?.. Successful repairs (VOC <10,000 ppm) 2?
"In Repaired 59%
Followup
DIRECTED MAINTENANCE
3. Number of valves unrepaired by undirected
maintenance subjected to directed maintenance 14
4. Number repaired by followup directed maintenance 5
% of unsuccessful repaired by
directed maintenance 36%
UNDIRECTED AND DIRECTED MAINTENANCE
5. Total number repaired based on undirected
maintenance and follow-up directed maintenance 71
% Repaired 73%
Source: Reference 13.
4-9
-------
1. Regulatory Alternative I: No requirements
2. Regulatory Alternative II: Quarterly monitoring and repair of all
gas valves and annual monitoring and repair of light liquid
valves.
3. Regulatory Alternative III: Monthly monitoring and repair of all
valves.
4. Regulatory Alternative IV: Same as regulatory alternative III.
The emission reductions from valves due to leak detection and repair
programs were calculated using the ABCD model presented in Chapter 4 of the
BID in the following manner.
Reduction efficiency =AxBxCxD
where:
A = Theoretical maximum control efficiency = fraction of total mass
emissions for the source with VOC concentrations greater than the
action level.
B = Leak occurrence and recurrence correction factor = correction
factor to account for sources which start to leak between
inspections (occurrence); for sources which are found to be
leaking, are repaired and start to leak again before the next
inspection (recurrence); and for known leaks which are not
repaired.
C = Non-instantaneous repair correction factor = correction factor to
account for emissions which occur between detection of a leak and
subsequent repairs; that is, repair is not instantaneous.
D = Imperfect repair correction factor = correction factor to account
for the fact that some sources which are repaired are not reduced
to zero emission levels.
The correction factors can, in turn, be determined from the following
expressions:
n
(1) B = 1 -
m
4-10
-------
(2) C - 365 - t
365
(3) D = 1 - f
F
where:
n = Average number of leaks occurring and recurring over the
monitoring interval (including known leaks which were not
repai red).
N = Total number of sources at or above the action level.
t = Average time before repairs are made.
f = Average emissions for repaired sources.
F = Average emission factor for all sources at or above the action
level.
The inputs to the model were selected based on a combination of the
conclusions from the five maintenance studies and engineering judgement.
The following are the inputs and the rationale behind their selection.
A factor: The theoretical maximum control efficiency depends on the action
level. An action level of 10,000 ppm was chosen for two main reasons:
(1) the monitoring instrument is designed for measuring a maximum concentre
tion of 10,000 ppm without the use of a dilution probe and (2) based on
refinery data, a large proportion of mass emissions is from valves at
10,000 ppm or greater (i.e., 98 percent for light liquid valves and
84 percent for gas valves).
D factor: Since quantitative data were not available, the leak occurrence
and recurrence correction factor was chosen based on engineering judgement.
It was estimated that 10 percent of initially leaking sources will occur,
recur, and remain leaking between monthly monitoring intervals (i.e.,
n - 0.1 N). Therefore, the average number of occurring, recurring, and
remaining leaks between monitoring intervals was n = 0.05N; so
B = 1 - 0.05N = 0.95 for monthly monitoring.
4-11
-------
C factor: The regulatory alternatives in the BID considered a 15-day time
limit for the repair of leaking valves. Some valves were expected to be
repaired immediately, while some were expected to take as long as 15 days.
On the average, it was expected that leaks would be repaired in
7.5 days (i.e., t = 7.5), so C = 365-7.5 = 0.98.
365
D factor: All sources which were repaired were assumed to be reduced to
a 1000 ppm concentration level. Results from the petroleum refinery study
of fugitive emissions showed that the average emission factor at
1000 ppm(f) was estimated to be 0.001 kg/hr (gas valves) and 0.004 kg/hr
(light liquid valves). From the same study the average emission factor
at ^10,000 ppm(F) was 0.21 kg/hr (gas valves) and 0.07 kg/hr (light
liquid valves). Therefore, D = 1 - jf was computed to be 0.99 for gas valves
and 0.94 for light liquid valves.
Based on the above inputs to the model, the control efficiency
(AxBxCxD) of a monthly leak detection repair proposed (Regulatory Alterna-
tives III and IV) for valves were calculated to be 0.90 for gas valves and
0.74 for light liquid valves. The control efficiency for Regulatory Alter-
native II (quarterly monitoring for gas valves and annual monitoring for
light liquid valves) was calculated to be 0.86 for gas valves and 0.62 for
light liquid valves.
4.1.2 New Information
Results from three new studies became available after proposal. Also,
an improved model was developed for calculating emissions and emission
reductions for fugitive emission sources operating under a leak detection
and repair program. This information developed since proposal is summarized
below.
14
Maintenance Study -- Effectiveness of maintenance on reducing fugitive
emissions from valves in six process units was evaluated in this study. The
methodology employed to generate the technical results of the study is
presented in the section on Emission Factors. The results of the study
relevant to emissions reduction include the determination of occurrence rate
4-12
-------
recurrence rate, overall emission reduction, and percent successful repair.
These results are summarized in Table 4-5.
The successful repair rate determined in this study was 29 percent.
However, the emission reduction achieved by reducing the screening values of
29 percent of the valves from ^10,000 ppm to <10,000 ppm was 71.3 percent.
These figures indicate that attempting maintenance, even if it is
unsuccessful in terms of screening value, reduces emissions significantly.
Analysis Report As indicated in the Emission Factor Section, the
results of the Maintenance Study and the twenty-four unit study were
combined for more in-depth analysis. Estimates of emission reductions due
to LDRPs from successful and unsuccessful repair and mass emissions from
valves screening >_10,000 ppm were made. These estimates are shown in
Table 4-6.
An Analysis of Allied HOPE Study Data -- The details of this study are
described in the emission factors section. This study was performed over a
period of 10 months in Al "Ned's newest existing high density polyethylene
unit (HOPE). The fugitive emissions data collected by Allied was analyzed
for EPA by PEDCo. The study consisted of six screening and emissions
measurement tests performed on valves and flanges. The relevant portions of
this study are summarized in Table 4-7. The results of this study showed a
high rate of successful repair (over 80 percent). The numbers also
indicated a 30-day recurrence rate of 7.8 to 11.1 percent which was higher
than the 30-day occurrence rate of 3.6 to 7.4 percent.
There are some additional considerations that must be made when
examining these results. The schedule for the series of six tests is shown
in Figure 4-1. As can be noted in Figure 4-1, there are some inconsisten-
cies in the repair effectiveness measures that were determined. For
example, valves were not rescreened immediately after maintenance in Test 1;
therefore, immediate leak recurrence is included in the Test 1 estimate of
repair effectiveness. And rescreening for Test 5 followed the maintenance
by a month. This means that the repair effectiveness for Test 5 includes
both immediate recurrence and long-term recurrence. As seen in Figure 4-1,
rescreening in the remaining tests closely followed maintenance efforts,
4-13
-------
TABLE 4-5. SUMMARY OF MAINTENANCE STUDY RESULTS
30-day
Occurrence rate,
(% of non-leaking
valves)
Ethylene Units
2.0 (All valves)
0.9 (G valves)
4.1 (LL valves)
Cumene Units
1.9 (All valves)
2.8 (G valves)
0.6 (LL valves)
Vinyl Acetate Units
0.3 (All valves)
0.7 (G valves)
0.2 (LL valves)
All Units
1.3 (All valves)
1.0 (G valves)
2.4 (LL valves)
-p-
i
30-day recurrence
rate, (% of repaired
valves)
Early failures, (% of
repaired valves)
17.2 (All valves)
14.3 (All valves)
Overall emission
reduction, (% of
before maintenance
emissions) b b
Successful repair, % b b
b 71.
84.
42.
b 29
3 (All valves)
5 (G valves)
0 (LL valves)
(All valves)
Leak recurrence within two weeks of repair.
Not determined for individual units.
°Recurrence within five days.
Source: Reference 19.
-------
TABLE 4-6. SUMMARY OF ANALYSIS REPORT RESULTS
Emission reduction
from successful
repair, (% of before
maintenance emissions'
Ethylene Units
Cumene Units Vinyl Acetate Units All Units
97.7 (AIT valves)
Emission reduction
from unsuccessful
repair, (% of before
maintenance emissions)
62.6 (All valves;
Mass emissions from
valves screening
>10,000 ppm, %
94 (G valves)
89 (LL valves)
94 (G valves)
80 (LL valves;
90 (G valves)
25 (LL valves'
Not determined for individual units.
Source: Reference 20.
-------
TABLE 4-7. SUMMARY OF RESULTS FOR THE ALLIED HOPE STUDY
30-day occurrence 3.6 - 7.4 (All valves)
rate, % 2.6 - 5.6 (G valves)
4.1 - 8.2 (LL valves)
30-day recurrence 7.8 - 11.1 (All valves)
rate, I 4.3 - 6.2 (G valves)
9.1 - 12.9 (LL valves)
Successful repair,9 % 83.3 (all valves)
100.0 (G valves)
78.7 (LL valves)
Results are based on leaking valves which were successfully repaired for
Test 2.
Source: Reference 21.
4-16
-------
I
I*
~vl
Days from Start
Test Number
Sampling
Rescreening
Time increments for
maintenance effectiveness
Maintenance
Screening Start
Duration of Test
Estimated Increments for
occurrence and recurrence
Actual increments for recurrence
Actual increments for occurrence
Date
40
80
2
120
160
200
5
240
280
PT.,1
N79 D79 J80 F80 M80 A80 M80 J80 J80 A80
Figure 4-1. Schedule of the fugitive emissions study at the Allied HOPE Unit.
Source: Reference 22.
-------
thus providing good indications of the repair effectiveness of on-line
maintenance.
Also, the time increments from which the occurrence and recurrence
rates were determined were based on the end of the sampling periods, which
were not always similar to the actual time increments for occurrence or
recurrence. While this simplification appears to have had little effect on
the occurrence rate estimates, the recurrence rate estimates, especially for
the last few inspections, appear to be based on time increments substan-
tially larger than the actual time to recurrence measurements. This error
is compounded by the fact that some of the recurrence was included in the
maintenance effectiveness estimates, as previously discussed.
SCAQMD Study23 -- The SCAQMD study (described in the emission factor
section) had several objectives, one of which was the investigation of the
effects on fugitive emissions of fugitive emission control regulations.
Screening data were gathered from eight process units in two refineries
complying with SCAQMD Rule 466.1.
To determine the effects of the regulations on occurrence and
recurrence rates, the results of previous screening inspections by the
refinery were obtained. The screening effort for the study comprised an
additional inspection. The results of the screening study (the additional
inspection) were presented in Table 2-16. Table 4-8 summarizes the results
of these inspections for valves. When a valve was found to be leaking
(screening value ^10,000 ppm), it was repaired (screening value <10,000 ppm)
within two working days. Then, the source was rescreened in three months,
and if the screening value was found to >10,000 it was repaired within two
days. The valve was screened again in one month. If it was found leaking
at this time, it was repaired and screened again in two weeks. If it was
found leaking again, it was repaired in one week. The process was continued
until the valve was screened and repaired every day or taken out of service.
It can be seen that there were significant recurrence rates over the three
month interval.
Leak Detection and Repair (LDAR) model -- A mathematical model using
recursive equations was developed by Radian Corporation. The leak detection
4-18
-------
TABLE 4-8. LEAK OCCURRENCE AND RECURRENCE OF VALVES AND OPEN-ENDED LINES
DETERMINED FROM SEVERAL INSPECTIONS - SCAQMD STUDY
UD
Process
Month
of
Unit Screening
Alkylation2
I somax
FCC6
Crude 1)1 s-
t Illation
Of f-Uas
Plant7
Plat former
Varuum Dis-
til lat Ion
2/793
7/793
9/80*
2/813>5
?/793
9/793
10/80^
2/813'3
3/793
8/793
9/80^
2/81J'5
n/794
& 9/80^
3/813-
11/79^
9/804
3/813'
10/804
3/813.
Number and percent of repaired valves found >10,000 at inspection aftes
Percent
Not Screened
10.3
1.1
5.2
2.3
5.0
4.3
9.0
4.8
4 5
5.8
6.4
9.0
-
5 29.4
_
-
5 20.3
_
5 36.0
Number
Screened
1768
2043
1931
1700
1818
1860
2322
1983
1925
1949
1983
1749
2648
722
988
233
171
396
37
242
Number 1
>10,000
257
111
159
153
48
41
84
47
79
24
232
115
4
24
19
19
6
23
0
0
Percent
210,000
14.5
5.4
8.2
9.0
2.6
2.2
3.6
2.4
4.1
1 .2
11.7
6.6
1.5
3.3
2.3
8.2
3.5
5.9
0.0
0.0
3 Months
Number
34
16
9
0
0
2
6
0
15
4
20
0
0
1
_
2
-
-
0
1 Month 2 Weeks
Percent Number Percent Number Percent
13.2 1 2.9 0 0.0
14.4 8 50.0 3 37.5
5.7 0 0.0 -
0.0
0.0
4.9 2 100.0 0 0.0
7.1 0 0.0 -
0.0
19.0 2 13.3 1 50.0
16.0 4 100.0 0 0.0
8.6 4 20.0 0 0.0
0.0
_
_
_
_ _ _ _
_
_
_ _ _ _
1 Ueek
Number Percent
1 33.3
-
-
-
_
-
0 0.0
_
_
-
_
-
-
_ _
-
-
_
The number of valves with screening values MO,000 pprov reflect the combined effect of occurrence and recurrence.
-Ki'corJs were obtained for the alkylatlon separator subunlt only.
Screening performed with Century OVA 108 calibrated with methane.
''Screening performed with Century OVA 108 calibrated with hexane.
Contractor's inspection.
^Refinery records do not distinguish between FCC Gas Recovery and FCC Gas Reactor.
Refinery records do not distinguish between crude distillation and off-gas plant.
8Does not Include any fuel gas valves or curde lines; many rule sources missed.
Source: Reference 24.
-------
and repair model (LDAR) approximates the behavior of fugitive emission
sources more closely than the ABCD model presented in the BID. A computer
25
program with variable inputs was prepared for this model. The ability to
use variable inputs makes the model flexible for use in many situations.
The model can be used to evaluate programs requiring leak detection and
repair of leaking sources at regular intervals (1 month, 3 months, 6 months,
9 months, or one year). The model also includes an option to evaluate a
program requiring quarterly inspection of all valves, attempted repair of
leaking valves, reinspection of repaired valves monthly until they are
determined not to be leaking for two successive months, and repair of
leaking valves (including those that could not be repaired within 15 days)
during a process turnaround.3 In addition, the model allows a variable
input for repair effectiveness, process unit turnaround frequency, leak
occurrence and leak frequency. The model can also incorporate the
uncertainty of the inputs and calculate approximate confidence intervals.
The details of the model are described in Reference No.26. A brief
description of the model is given below.
Figure 4-2 is a schematic diagram of the model (for quarterly screening
with monthly follow up on repaired sources). The model is based on the
following assumptions:
1. All sources at any given time are assumed to be in one of four
categories:
a) Non-leaking sources (sources screening < action level),
b) Leaking sources (sources screening > action level),
The model is simplified from the proposed valve standard in one area. In
the proposed standard, a source that is repaired must be screened and
found to be non-leaking for two successive months before the source is
considered to be in the non-leaking source population. The model imple-
ments this rule for sources repaired during the quarterly inspection or
in the second month of the quarter. But the model only requires that
sources maintained in the third month of the quarter are non-leaking
during the next quarterly screening. In most cases, the fraction of
sources affected should be small, and thus, the impact of this simplifica-
tion was not considered important enough to further complicate the model.
4-20
-------
tf*»ft.*m mot
l*Mlr*^
ItW jtttfc
Carlf R*c«*r«Mc»
Jt*J Kith U»k
Bccurrvac* Duri*| Mo*i
L»*k >«curr«ac« I
Jwrlng Month
the «nJ of tha
Source: Reference 27.
Figure 4-2. Schematic Diagram of the LDAR Model.
-------
c) Leaking sources which cannot be repaired on-line and are awaiting
a shutdown for repair, and
d) Repaired sources with early leak recurrence.
2. There are only three distinct leak rates for all sources. A
leaking source leak rate, a non-leaking source leak rate and a leak rate for
unrepairable and early recurring sources. That is, repaired sources with
early leak recurrence are assumed to have the same leak rate as sources
which cannot be repaired on-line.
3. A non-leak has the same leak rate as a repaired leak.
4. The model does not evaluate gradual changes in leak rates over
time but assumes that all sources in a given category have the same average
leak rate.
5. Effect of time is accounted for as follows: a) all monthly
repairs occur at the end of the month; b) the effect associated with the
time interval during which repairs occur are negligible.
6. Early recurrences occur essentially instantaneously; i.e., the
time from repair to early recurrence is negligible.
7. If a leak stays repaired for a week, leak occurrence rate equals
leak recurrence rate.
8. Unsuccessfully repaired sources fall into the unrepaired category
instantaneously.
9. Leaks occur (other than unsuccessful maintenance and early
recurrences) at a linear rate with time during a given period.
10. When a turnaround occurs, it occurs essentially instantaneously at
the end of a quarter, before the beginning of the next monitoring period.
11. All unrepaired sources are repaired at the turnaround. It should
be noted that repairs at turnaround are not counted as additional repair
attempts in computing the fraction of sources with attempted maintenance
because time for complete repair is included in the time estimate for the
first attempt, based on 75 percent repaired in 10 minutes and 25 percent
repaired in 4 hours.
I?. The fractional reduction in emissions reported in the program is
determined by comparing emission levels expected under the leak detection
4-22
-------
and repair program to the initial level of emissiors. The initial level is
defined by data collected in the field at operating SOCMI process units.
The significance of the last assumption may not be readily obvious. In
the absence of a leak detection and repair program (i.e. uncontrolled) the
fraction of valves leaking would increase with time, reaching a maximum at
the unit turnaround. If this were the case, the model would compare these
"uncontrolled" emissions to those occurring under the leak detection and
repair program. However, for valves which are subject to leak detection and
repair programs and for other fugitive emission sources which require
frequent maintenance, the fraction of sources leaking do not accumulate with
time. Instead, periodic maintenance has the effect of causing the fraction
of sources leaking and their emissions to increase and decrease in a
cyclical manner.
The major implication of this behavior is that when field testing a
population of fugitive emission sources which are subjected to periodic
maintenance, it is impossible to tell where in the cycle the testing
occurred. The field data represent a "snapshot" in time of fugitive
emissions. In spite of this obvious shortcoming, the model uses the level
of control found in the field as the point of reference against which the
effects of a leak detection and repair program are compared. Unless the
point of reference is at an average point in the cycle, the LDAR model and
other models will underpredict emissions and emission reductions due to leak
detection and repair. Furthermore, if the population of fugitive emission
sources is near the minimum in the cycle, the measured occurrence rate can
exceed the measured leak frequency, and the models may even predict negative
emission reductions.
There are two possible ways in which negative results may arise. One
possibility is that the units tested actually had programs in place which
were more stringent than the leak detection and repair program being
evaluated during testing. To perform the testing, the usual program must be
abandoned and strict adherence to the program being evaluated must be
maintained. If the original program was more effective than the program
being evaluated, negative emission reductions would result.
4-23
-------
Another possibility is that the program abandoned for testing of the
program being evaluated is less effective, but the baseline level measured
was measured at a time in the maintenance cycle of the original program when
emissions were low. If the maintenance cycle is irregular, it is difficult
to judge just where the unit is in the cycle and when emissions can be
expected to be low. Maintenance performed just prior to testing, then,
could cause negative emission reductions to result even if the original
program was not very effective. Maintenance just prior to testing occurred
JQ
at least once during the 6-Unit Study.
Thus, one of the most important parameters involved in calculating
emission reductions is one in which there is a large amount of uncertainty.
It is possible to make two sets of comparisons: one to the uncontrolled
level and one to the assumed baseline level. The uncontrolled level would
be represented by the situation in which leaks are allowed to accumulate in
the absence of a leak detection and repair program until the unit turn-
around. The baseline level would be the same as the initial level used for
comparison in the model (i.e. the level measured in the field). The actual
emission reduction would probably be between the two calculated values.
Consideration is being given to making such calculations and whether they
would provide useful comparisons. At any rate, the present comparisons
should be considered conservative estimates of the effects of leak detection
and repair programs.
The input parameters required for the LDAR model and the outputs which
the model calculates are listed in Table 4-9.
4.1.3 Public Comments
Many of the comments received challenged EPA's estimates of emission
reductions achievable by leak detection and repair programs for valves
challenged the methodology used to arrive at the estimates. Four major
comments were made in this area.
1. Some commenters objected to the use of engineering judgement to
arrive at an estimate. Others disagreed with specific judgements.
2. Some of the comments also expressed concern that the emission
reduction calculations were done before the SOCMI data were available and
that these data would impact the calculations. More specifically, the
4-24
-------
TABLF 4-9. INPUTS AND OUTPUTS FOR THE LEAK
DETECTION AND REPAIR (LDAR) MODEL
Inputs
1. Emission factor - the initial emission factor for all sources in units
of mass/time/source. Note: This input is a key value because it forms the
baseline level to which all emissions are compared for calculation of
emission reductions. See assumption #12.
2. Occurrence rate - the fraction of sources operating properly at the
beginning of the monitoring interval that become leakers during a monitoring
interval.
3. Initial leak frequency - the fraction of sources leaking initially (from
leak occurrence).
4. Fractional emission reductions from unsuccessful repair - emission
reductions for valves for which maintenance did not reduce the screening
values to below the action level.
5. Fractional emission reductions from successful repair - emission
reductions for valves for which maintenance reduced the screening values to
below the action level.
6. Fraction of sources that are leaking and for which attempts at repair
have failed - the fraction of sources screening above the action level for
which maintenance has failed to decrease the screening value to below the
action level.
7. Fraction of repaired sources that experience early failure fraction of
sources which screened above the action level, were repaired to screening
values below the action level, and which were screened above the action
level within five days.
8. Turnaround frequency the length of time between plant shutdowns.
Outputs
1. Estimated emission factors by turnaround - the average emissions per
source for the period between plant shutdowns in units of mass per time with
an approximate 90 percent confidence interval.
2. Fractional reduction in mass emissions by turnaround - the average
fractional reduction in emissions for the period between plant shutdowns
relative to initial emissions with an approximate 90 percent confidence
i nterval.
4-25
-------
TABLE 4-9. (CONTINUED)
3. Total fraction of sources screened per year - the fraction of sources
screened in each year of a five year period.
4. Total fraction of sources with attempted repair during a year - the
fraction of sources for which repair is attempted for each year of a five
year period. (Note: does not include subsequent maintenance of sources at
turnaround.)
5. Fraction of sources screened per month - the fraction of the sources
which are screened during each month of a five year period.
6. Fraction of sources with attempted repair per month - the fraction of
sources for which maintenance is attempted in each month of a five year
period. (Note: does not include subsequent maintenance of sources at a
turnaround.)
7. Estimated emission factor for the monitoring interval - average
emissions per source for the monitoring interval in units of mass per time.
8. Fractional reduction in mass emissions between monitoring "intervals -
the average fractional reduction in emissions from the monitoring interval
relative to initial emissions with an approximate 90 percent confidence
interval.
9. Fractional distribution of leakers due to occurrence - fraction of
sources screening below the action level initially which screen above the
action level at the end of the monitoring period.
10. Fractional distribution of unrepaired sources - the fraction of sources
screening above the action level for which maintenance failed to reduce the
screening value below the action level at the end of the monitoring period.
11. Fractional distribution of sources experiencing early failures -
fraction of sources screening above the action level which were repaired to
screening values below the action level but screened above the action level
within five days of repair.
12. Fractional distribution of non-leaking sources - fraction of sources
screening below the action level at the end of the monitoring period.
4-26
-------
results of the maintenance study were cited in support of the argument that
maintenance effectiveness would be lower than assumed in the BID
calculations because on-line valve repair efficiency found in the study was
lower than assumed.
3. Of the factors used in the ABCD model for estimating emission
reductions achievable, the B factor was the most frequently challenged.
Disagreement was expressed with the complex, non-linear occurrence/
recurrence relationship assumed in the development of emissions reductions
calculations. The leak occurrence/recurrence rates developed in the SOCMI
29
studies were cited in support of this argument. Commenters suggested
assuming a linear leak occurrence rate with a proportional recurrence rate
or because of the small amount of recurrence data, using only an occurrence
rate.
4. Industry commentnrs submitted a calculation method which they felt
more appropriately predicted emission reductions achievable with leak
detection and repair programs for valves. Specifically, a mathematical
model was recommended to calculate the number of leaks which occur, recur,
and remain between monitoring intervals to replace the value of n presented
in the BID. The industry commenters presented a set of equations (along
with their derivations) for calculating this value. This value, n',
computed by industry's suggested set of equations is not the same value as
used in the BID, which was a function of the fraction of leaks found
initially. An example calculation using results (modified in some cases)
determined in the maintenance study was also presented.
The following quantities were defined:
P = occurrence rate -: 100
o
PD = recurrence rate : 100
K
N - percent of sources found leaking initially
Two case? were developed for comparison with the calculations presented in
the BID. The first case assumes 100 percent maintenance effectiveness which
was also assumed in the BID calculations. The second case takes into
account maintenance efficiency based on data from the maintenance study.
4-27
-------
Case 1: (assumes 100 percent maintenance effectiveness)
Period Percent Leakers
tj PQ(100-N) + PRN
t0 .P [100-N-P (100-N)] + PjN+P (100-N)]
c 0 0 K 0
t3 PO[IOO-N-PO(IOO-N)-PO[IOO-N-PO(IOO-N)]] +
PD[N+P (100-N) + PjlOO-N-P (100-N)]]
K 0 0 0
In summary, the rates were calculated as follows:
Period Percent Leakers
'o N
tk(k>o) PO[IOO-N]SK(PO) + PR[IOO-(IOO-N)SK(PO)] =
100[PoSK(Po) + PR[1-SK(P0)]] + (PR-P0)SK(P0)N
where
k-i (k-i)i(-ijV
SK(PO) = x °
* ° 1=0 (K-i-l)li!
Case 2: (takes into account maintenance effectiveness less than 100%)
The following definition is necessary for this case:
P = (maintenance effectiveness) = 100
This case was considered since the maintenance study data show less than
100 percent maintenance effectiveness.
Period % Leakers
tK(K>o) PO(IOO-N)SK(PO) + (i-Pm)[ioo-(ioo-N)sK(Po)] +
PmPR[100-(100-N)SK(P0)] =
100[PoSK(Po) + (l-Pm , PmPR)[l-SK(Po)]] + (l-Pm + PmPR-P0)SK(P0)N
The industry model and the LDAR model are comparable in the general
approach to assessing the impacts of leak detection and repair programs on
the number of leaks found during any particular monitoring period. They are
4-28
-------
both based on recursive equations that apply occurrence rate, recurrence
rate, and maintenance effectiveness to the population of sources of
interest. There are fundamental differences in the assumptions made during
the development of the models and in definitions of terms that impact the
final form of the equations and the method in which the available data must
be applied to the equations. These differences also affect the number of
leaks which the models calculate.
A major difference between the models is the handling of leak
recurrence. The industry model applies the recurrence rate to all sources
that have leaked at some time between turnaround periods. Thus, the
occurrence rate is applied only to those sources that have never leaked.
Furthermore, the recurrence rate does not include early recurrences (those
happening within five days of repair). Early recurrences are considered to
be maintenance failure and are accumulated as unrepairable sources. The
industry model also does not subject those sources that were successfully
repaired to leak occurrence. Basically, when a source has been found to be
an unrepairable source, a recurrence, or a fixed source, it remains in that
category until the next turnaround.
Although the model suggested by industry and the LDAR model are similar
in their general approach, the industry model incorporates some simplifying
assumptions that add to the differences described above. The LDAR model is
more complex and better represents the expected performance of leak
detection and repair programs required by the standards. Where the industry
model provides only the number of leaks determined at each monitoring
internal, the LDAR model provides the number of sources occurring,
recurring, maintained, and those that cannot be repaired. Furthermore, the
LDAR model accounts for different emission factors for the different
classification of source (unrepairable source, recurring source, occurring
source, successfully maintained source) and it can simulate a variety of
LDRPs including hybrid monitoring schemes such as quarterly monitoring with
monthly follow-up of recently maintained sources. As currently presented,
the industry model does not afford this flexibility. The results of this
industry model are compared to the LDAR results later in this section.
4-29
-------
4.1.4 EPA's Conclusions
Estimating emissions of VOC from fugitive emission sources involves the
use of data and engineering judgement. Engineering judgement is used to
select the modeling approach and to select the data to use in the model. It
is sometimes necessary to select a numerical value which is not based on
completely representative data. In this situation the numerical values must
be reasonably consistent with the available information or understanding of
fugitive emission sources. The analysis presented in the BID necessarily
used engineering judgement to select modeling approaches and numerical
values. Even though the Agency uses the best modeling approaches and
numerical values and improves these as new information becomes available,
engineering judgement remains an important part of the estimating process
(and appropriately so) no matter which method for estimation is chosen.
To determine how emission reduction estimates for valves should be
calculated, EPA had to make two choices: the computational method to be
used and the data to use in the computations. The choices made and the
reasons for the selections are discussed in the next two subsections. In
the last two subsections the results of the analysis are presented and
compared to the results from other methods.
Selection of Model for Emissions Reduction Calculations -- Two mathematical
models are available for use in estimation of emissions reduction for leak
detection and repair programs:
(1) The ABCD reduction efficiency model and
(2) the LDAR model for evaluation of the impact of a leak detection
and repair program.
The ABCD model can be used in two different ways:
(a) Using engineering judgement for estimation of n (as used at
proposal) and
(b) Using the mathematical model presented by the industry for the
determination of n'.
The ABCD model, especially when combined with the mathematical computa-
tion of n presented by the industry, is a straightforward calculation
4-30
-------
procedure. However, it does not incorporate all the latest information
which has been developed concerning fugitive emissions. The addition of the
industry recommendation for estimation of n' is an improvement in that the B
m
factor is no longer based on engineering judgement. However it is a
simplified model and does not necessarily improve the accuracy of the
emission reduction estimation procedure. Care must be taken in choosing
inputs to the industry model that will be consistent with the assumptions
made during its development. Also, the ABCD model does not make provisions
for turnarounds. Furthermore, it does not provide for several phenomena
which were noted in the field. For example, it does not consider early leak
recurrence as a separate phenomenon.
The LDAR model provides these features and also provides the flexi-
bility for evaluating programs such as quarterly monitoring with monthly
followup for leakers. It provides additional information, such as the
number of valves screened and maintained during each period. Furthermore,
it incorporates the phenomenon of early recurrence. The LDAR, therefore, in
spite of its relative complexity, has some advantages over the ABCD model
and was selected for use in estimating the effectiveness of leak detection
and repair programs. A comparison of these results with those of the ABCD
model are presented later in this section.
Input data selection -- The second choice to be made was the data to be used
in the estimating model. The results of all available fugitive emissions
studies were compiled for consideration in making decisions on the data to
select for calculations of emission reductions. Table 4-10 presents a
summary of the available data on valves. Input parameters that represented
the most reasonable estimates for SOCMI units were selected from the
available data base. A discussion of the selected inputs and the reasons
for their selection is presented below.
1. Emission factors: The estimated emission factors selected are
0.0056 kg/hr for gas valves and 0.0071 kg/hr for light liquid valves. These
emission factors were estimated using the leak rates for leaking and
non-leaking sources. The details of the estimating procedure are discussed
in the emission factors section.
4-31
-------
TABLE 4-10. SUMMARY OF AVAILABLE DATA ON VALVES,
I
CO
ro
Study
PR/BID
24-Unit
6-Unit
Maintenance
6-Unit Leak
Frequency
Cyclohexane6
Chevron Refinery
DuPont3
Allied HOPE
Occurrence Initial Overall Emission Emission Successful Early Reference
Rate, '.'- Leak Emission Reduction Reduction Repair, % Failures,
(30 days) Frequency, % Reduction, % From From (Recurrence), X
Unsuccessful Successful
Repair, % Repair, %
10 (G) - - -
12 (LL)
8.2 (All Valves) - ...
11.4 (G)
6.5 (LL)
1.3 (All Valves) - 71.3 (All 62.6 (AH 97.7 (All 29 (All Valves) 14 All
1.0 (G) All Valves) Valves) Valves) (17.2) Valves
2.4 (LL) Units 84.5 (G)
42.0 (LL)
1 (Unit A) - - -
19 (Unit B)
0.2 (Unit C)
14 (Unit D)
1.1 (Unit E)
17 (Unit F)
32 (G)
15 (LL)
4 (All Valves) - - 93 (All Valves)
6.1 (All) .....
23.1 (G)
7.1 (LL)
3.6-7.4 (All) 24.7 (All)b - - - 83.3 (All)c (7.8-11.1) (All)
2.6-5.6 (G) 17.6 (G)D. 100.0 (G)Cr (4.3-6.2) (G)
4.1-8.2 (LL) 27.4 (LL)D 78.7 (LL)C (9.1-12.9) (LL)
30
31
32
33
34
35
36
37
-------
TABLE 4-10. (CONTINUED)
Study Occurrence
Pate, f
(30 days)
Union Oil
Maintenance
Shell 01 ld
Maintenance
EPA Refinery
Ha intenance
Phillips
Ethylene
EPA 4-Unitf
German Studie?
Initial Overall Emission Emission
Leak Emission Reduction P.eduction
Frequency, % Reduction, " From From
Unsuccessful Successful
Repair, * Repair, v
51.8
76.0 (Test 1)
85.7 (Test 2)
87.0 (Directed
Maintenance)
82.4 (Undirected
Maintenance)
-
32.5
6.3
6.7
24.3
9W
Successful Early
Repair, - Failures,
(Recurrence), 1-
50.4
65.2 (Test 1)
83.3 (Test 2)
88.9 (Directed
M;> intensnce)
56.5 (Undirected
Maintenance)
63
-
-
Referencp
38
39
40
41
42
43
SCAQMD
96 (All Valves)
100 (All Valves)
44
Leak defined as 10 ppm or greater. "All valves" include heavy liquid valves.
Test 1 results only.
GDirected maintenance only.
dLeak defined as 5300 ppm, OVA - 108 at 1 cm.
elpaks detected by using soap bubbles.
Leak defined as 200 ppm.
*
95T, confidence intervals.
-------
2. Occurrence rate: Occurrence rate estimates were available from
two studies. First, the maintenance study had occurrence rate estimates
45
developed from tests in three SOCMI processes. Estimates were presented
for each type of process (vinyl acetate, cumene, and ethylene) and by
service (gas, light liquid). An overall estimate for all units was also
developed. Second, the Allied study presented occurrence rate estimates for
46
a high density polyethylene unit. Due to some inconsistencies previously
noted in this study and due to the broader range of processes covered by the
maintenance study, occurrence rates generated in that study were considered
to be the best available estimates of occurrence rates for valves. Because
the confidence intervals showed substantial overlap, (See Figure 4-3) the
overall 30-day occurrence rate of 1.3 percent was selected.
3. Leak frequency: Leak frequency data were available from several
studies (see Emission Factors Section). These studies are the Petroleum
47 48 49
Refinery study, EPA 24 unit study, EPA 6 unit study, Exxon cyclohexane
study,50 DuPont study,51 and Allied HOPE study.52 Some of these studies had
leak definitions different from the proposed action level (10,000 ppm). For
instance, valve leaks in the cyclohexane study were discovered by using soap
bubbles. The DuPont study used a leak definition of 10 ppm. The results of
these studies are, therefore, not directly comparable to the rest of the
data base. Of the remaining studies the 24 unit study covers the broadest
range of processes and was, therefore, considered to be the best available
estimate for initial leak frequency for SOCMI. This estimate is
11.4 percent for gas valves and 6.5 percent for light liquid valves.
However, there are uncertainties which remain in these values as they are
used in the model. These uncertainties were discussed previously in the
description of the LDAR model.
4. Emissions reduction from an unsuccessful repair: The results from
the maintenance study showed that emissions were reduced by about 63 percent
even if the screening value could not be brought down to below 10,000 ppm
(TO
(i.e., the maintenance was unsuccessful). Since the maintenance study was
the only available source of information on emissions reduction from
unsuccessful repair, 63 percent was used as the input value.
4-34
-------
to
CJ
o
c:
O)
-a
c
o
o
4 -.
OJ
CJ
Ol
Q-
LT)
cn
c
(1)
o
s_
CD
Q.
O)
2 -
Ol
(J
c
HI
o
-------
5. Emissions reductions from a successful repair: The analysis
report was the only available source of information presented in this
manner. The value used for emissions reduction from successful repair is
55 '
about 98 percent. The 90 percent figure from the German study was the
effect on the overall emission factor due to directed repair of leaking
components. Therefore, it could not be used as this input parameter.
6. Fraction of sources that are leaking and for which attempts at
repair have failed: A summary of all available data on maintenance effec-
tiveness (in terms of percent successful repair) is shown in Table 4-11.
The overall average successful repair rate is seen to be close to
80 percent. It should be noted that these data are for attempts at quick
on-line repair. The successful repair input value should be based on all
repairs that can resonably be completed within 15 days. It is also expected
that as maintenance crews gain experience in valve repair, their effective-
ness will improve as evidenced by the work reported by Chevron in which a
57
repair rate of 93 percent was achieved and the work performed in the
ro
SCAQMD in which 100 percent were repaired. Also, newer units are expected
to have a higher incidence of successful repair resulting from the proper
selection and maintenance of valves to reduce fugitive emissions by
equipment design and operation. Therefore, 90 percent successful repair is
used for the analysis, i.e., fraction of sources that are leaking and for
which attempts at repair have failed = 0.10.
7. Fraction of repaired sources that experience early failure: The
maintenance study showed that about 14 percent of all repaired sources
59
started to leak within 5 days of repair. The only other recurrence rate
data is from the Allied HOPE study. However, that study does not provide
information for early failures. Therefore, the maintenance study data were
used for input to the model.
8. Turnaround frequency: A two-year turnaround is assumed to be a
reasonable estimate for an industry average and is, therefore, used in the
calculations. Some units may go longer between turnarounds; some may
turnaround more frequently. A two year frequency has been assumed for this
analysis.
4-36
-------
TABLE 4-11. SUMMARY OF VALVE MAINTENANCE TEST RESULTS
Maintenance Test
Union
Shell
March 1979
April 1979
EPA-4 Refineries
Refinery ( Ethyl ene Unit
EPA-6 unit maintenance
Allied #la
Allied #2
Allied #3a
Allied #4a
Allied #5a
Allied #6a
Chevron
SCAQMD
TOTALS
OVERALL AVERAGE
SUCCESSFUL REPAIRS
Number of Valve
Repairs Attempted
133
161
54
32
) 46
97
33
60
30
40
28
33
1320
382
2449
Number of
Successful Repairs
67
105
45
21
29
28
14
50
18
17
16
15
1228
382
2035
Percent
Repaired
50
65
83
66
63
29
42
83
60
42
57
45
93b
100
83
Actual repair effectiveness may be higher than shown. In some cases the
valves were not rechecked until 30 days after maintenance.
Simple on-line repairs. Others fixed by more elaborate methods.
4-37
-------
Computation Results -- The inputs discussed above were used in the LDAR
model to evaluate the effectiveness in reducing fugitive emissions of LDRPs
of varying monitoring intervals for control of fugitive emissions from
valves in gas and light liquid VOC service. Where possible, the same inputs
were also used in the ABCD model, as presented in the BID and as modified to
incorporate industry's model for occurrence/recurrence rate estimates. But
where engineering judgement was needed in lieu of data, such as for the
B-factor in the ABCD model, the original estimate used in the BID was
retained. By using equivalent input parameters in the different models,
comparisons of the models themselves can be made. A summary of the LDRP
effectiveness estimates generated in this manner is presented in Table 4-12.
These values represent the overall program effectiveness for gas and light
liquid valves for the model units presented in a previous section.
The highest emissions reductions for LDRPs were estimated with the ABCD
model presented in the BID. But these values were based on assumed occur-
C 1
rence/recurrence correction factors (B-factors). These B-factors were
larger than those derived using the ABCD model with industry's recommended
changes, indicating that emissions from occurrence and recurrence and
unrepaired leaks are more significant than originally estimated in the BID.
Still lower LDRP effectiveness was computed using the LDAR model. The
modified-ABCD model and the LDAR model compare closely over the range of
monitoring intervals from annual to monthly when implemented with equivalent
inputs. They indicate the same trend of increasing LDRP effectiveness with
increasing monitoring frequency. In contrast to the ABCD model, the
modified-ABCD and LDAR models indicate essentially no emission reduction for
the LDRPs with annual monitoring plans.
The ABCD model was executed as it was presented in the BID (See
Table 4-13). The theoretical maximum control efficiency (A-factor) and the
fractional emission reduction attributed to repair of sources originally
screening above the action level (D-factor) were both based on the petroleum
refining study and are the same values used in developing emission factors.
But the values for A and D presented here differ by a small amount from the
values presented in the BID. This discrepancy is due to the fact that an
4-38
-------
TABLE 4-12. COMPARISON OF OVERALL LDRP EFFECTIVENESS FOR
VALVES IN MODEL SOCMI UNITS
Monitoring
Interval
Service
ABCD
Model
Modified-ABCDb
LDARC
Monthly
Monthly/Quarterly
Quarterly
Semiannual
Annual
Gas
Light Liquid
Gas
Light Liquid
Gas
Light Liquid
Gas
Light Liquid
Gas
Light Liquid
0.93
0.90
0.88
0.85
0.78
0.76
0.78
0.62
0.69
0.46
0.56
0.24
0.30
(0.21)e
0.73
0.59
0.65
0.46
0.64
0.44
0.50
0.22
0.24
(0.19)6
These are overall effectiveness values for LDRPs for gas and light liquid
valves in the SOCMI model units presented in a previous section.
Modified to incorporate industry's model for nm.
CLDAR model presented in Reference 62.
Quarterly monitoring with monthly follow-up of repaired sources.
eNumbers in parentheses indicate an estimated negative control efficiency.
Negative numbers are generated when the occurrence rate for the monitoring
interval exceeds the initial leak frequency. Negative results are subject
to interpretation and may not be meaningful.
4-39
-------
TABLE 4-13. LDRP EFFECTIVENESS FOR VALVES USING THE ABCD MODEL
Monitoring
Interval
Monthly
Quarterly
Annual
Valve
Service
Gas
Light Liquid
Gas
Light liquid
Gas
Light Liquid
A
0.998
0.980
0.998
0.980
0.998
0.980
Ba
0.95
0.95
0.90
0.90
0.80
0.80
C
0.98
0.98
0.98
0.98
0.98
0.98
D
0.999
0.983
0.999
0.983
0.999
0.983
ABCD
0.928
0.896
0.879
0.849
0.781
0.755
aThese are the same values used in the BID that were based on engineering
judgement.
4-40
-------
interim report of this study was used in developing the SOCMI BID; the new
numbers are based on the final report. Engineering judgement had been used
to assign B-factors in the SOCMI BID. Therefore, the B-factors used here
were not revised for this estimate.
Input parameters for the modified-ABCD model were chosen to be as
nearly equivalent as possible to those used in executing the LDAR model. In
executing the modified-ABCD model, the recurrence rate was assumed to be
equal to the occurrence rate as in the LDAR model. Early recurrence, as
considered in the LDAR model, was incorporated in repair effectiveness in
the modified-ABCD model. Since early recurrence was taken as 14 percent and
repair effectiveness was 90 percent in implementing the LDAR model, the
maintenance effectiveness in industry's model was 90 - 14 = 76 percent.
Assuming a two year turnaround cycle, the average number (or fraction) of
leaks occurring, recurring, and remaining over a monitoring interval was
calculated for gas and light liquid valves, considering various monitoring
intervals.
The corresponding B-factors and LDRP effectiveness values which
resulted are presented in Table 4-14. The program effectiveness was
computed using the same A, C, and D-factors previously discussed and used in
the ABCD model. The B-factor incorporated the value for n' calculated
according to the algorithm submitted by industry using inputs equivalent to
those used in executing the LDAR model.
The effectiveness of leak detection and repair programs with various
monitoring intervals was calculated with the LDAR model. The overall
effectiveness and emission reduction achievable under each plan is presented
in Table 4-15. As shown, the highest degree of emission reduction is
obtained from monthly monitoring with an overall effectiveness of about
63 percent. The quarterly monitoring plan resulted in a lower effectiveness
(about 50 percent), and small differences were seen between the quarterly
program and the program incorporating quarterly monitoring with monthly
follow-up of repaired sources. The least effective LDRP was an annual
monitoring plan that indicated no net benefit.
4-41
-------
TABLE 4-14. LDRP EFFECTIVENESS FOR VALVES USING THE MODIFIED-ABCD MODEL
Monitoring
Interval
Monthly
Quarterly
Semiannual
Annual
Valve
Service
Gas
Light
Gas
Light
Gas
Light
Gas
Light
Liquid
Liquid
Liquid
Liquid
0
0
0
0
0
0
0
0
A
.998
.980
.998
.980
.998
.980
.998
.980
0
0
0
0
0
0
0
(0
B
.802
.652
.709
.489
.573
.251
.304
.222)
C
0.98
0.98
0.98
0.98
0.98
0.98
0.98
0.98
0
0
0
0
0
0
0
0
D
.999
.983
.999
.983
.999
.983
.999
.983
ABCD
0.
0.
0.
0.
0.
0.
0.
(0.
784
616
693
462
560
237
297
210)
Numbers in parentheses indicate negative control efficiency. Negative
numbers are generated when the occurrence rate for the monitoring interval
exceeds the initial leak frequency and may not be meaningful.
4-42
-------
TABLE 4-15. LDRP EFFECTIVENESS FOR VALVES USING LDAR MODEL
Monitoring
Interval
Monthly
Quarterly/
Monthly
Quarterly
Semiannual
Annual
Valve LDRP
Service Effectiveness
Gas
Light Liquid
Gas
Light Liquid
Gas
Light Liquid
Gas
Light Liquid
Gas
Light Liquid
0.73
0.59
0.65
0.46
0.64
0.44
0.50
0.22
0.24
(0.19)a
Emission
Model
Unit A
3.6
4.7
3.2
3.7
3.1
3.5
2.4
1.8
1.1
(1.5)
Reductions, Mg/yr
Model Model
Unit B Unit C
14.5
18.8
12.8
14.7
12.6
14.1
9.8
7.2
4.6
(6.0)
44.3
59.1
39.3
45.3
38.6
43.6
30.2
22.1
14.2
(18.5)
Numbers in parentheses indicate negative numbers. Negative numbers may be
generated when the occurrence rate for the monitoring interval exceeds the
initial leak frequency. Negative results are subject to interpretation and
may not be meaningful.
4-43
-------
4.2 PUMPS
4.2.1 Technical Basis Presented in the BID
As discussed in the BID (EPA-450/3-80-033a)63, fugitive VOC leaks from
pumps used in SOCMI may occur between the shaft and sealing mechanism
separating the process from the surrounding environment. The resulting
fugitive emissions can be reduced by equipment designed to prevent leakage
and by implementing leak detection and repair programs. Both of these
techniques were evaluated in the BID and were incorporated in the four
regulatory alternatives for light liquid service pumps.
The equipment alternatives for pumps described in the BID include
sealless pumps, dual mechanical seals, and seal area enclosures connected to
effective control devices. Sealless pumps have no junction between the
shaft and pump casing and, therefore, do not leak under normal operating
circumstances.
Dual mechanical seals are commonly applied to centrifugal pumps in
back-to-back or tandem arrangements. A barrier fluid is flushed between the
seals to provide protection against the process fluid leaking to the
environment because any process fluid that may be leaked into the barrier
fluid would be emitted from barrier fluid degassing reservoirs. Sometimes
VOC fugitive emission control by dual seals requires venting of the
, degassing reservoir to a control device. Although the efficiency of such a
system would be dependent upon the efficiency of the control device and the
frequency of seal failure, a control effectiveness of 100 percent was used
in the BID for dual seal/non-VOC barrier fluid systems. In instances when
dual seals might not be applied, enclosing the seal area and venting to a
control device was assumed to achieve nearly 100 percent effectiveness in
eliminating fugitive VOC losses. These equipment alternatives were taken as
the most stringent control option for pumps and represented Regulatory
Alternative IV.
Leak detection and repair programs for pumps were also evaluated and
were incorporated in the regulatory alternatives presented in the BID.
Under Regulatory Alternative II, light liquid pumps would be subject to an
annual LDRP, resulting in a 63 percent emission reduction. The monthly LDRP
4-44
-------
examined as Regulatory Alternative III resulted in an estimated 75 percent
control effectiveness. These estimates were made using the same ABCD model
discussed previously for valves. The A-factor for pumps was derived from
figures for pumps in light liquid service similar to Figure 4-1 of the BID,
"cumulative distribution of total emissions by screening values - valves -
gas/vapor streams." The leak frequency, N, of 23 percent was taken from
Table 4-2 of the BID. The B and C factors were the same as for valves. The
D factor was computed in the same manner as for valves.
4.2.2 New Information
Three reports have presented additional data on fugitive emissions from
light liquid pumps: Evaluation of Maintenance for Fugitive VOC Emissions
Control (6-unit Maintenance Study) ,64 Analysis of SOCMI VOC Fugitive
Emissions Data (SOCMI Analysis report), and Evaluation _of the Maintenance
Effect 0£ Fugitive Emissions from Refineries jh^ the South Coast Air
Management District. Two of these reports were based on studies conducted
in SOCMI units. The SOCMI 6-unit maintenance study presented leak
occurrence rates for the three SOCMI process types studied (see Table 4-16).
A 30-day leak occurrence rate of 5.5 percent was found for all light liquid
pumps in six process units. Recurrence rates and measures of the effective-
ness of maintenance were not developed in this study because none of the
pumps studied were maintained.
The SOCMI Analysis report presented several observations concerning
fugitive emissions from pumps. On-line pump seals had an overall leak
frequency of 13.1 percent compared to 4.9 percent for off-line pump seals.
No difference in leak frequency was found between double mechanical pump
seals and single mechanical pump seals. Data on the barrier fluid systems
used were not accounted for in this analysis, however. It is possible that
VOC barrier fluids were used in which case VOC would not be emitted from the
process fluid to the environment but VOC from the barrier fluid could be.
A preliminary evaluation of the effects of maintenance on fugitive
emissions from pumps in refineries was conducted by EPA's IERL in the South
Coast Air Quality Management District in California. Screening values were
examined as a function of pump seal age and were shown to have a slight
4-45
-------
TABLE 4-16. 30-DAY OCCURRENCE RATE ESTIMATES FOR PUMPS
Unit Type
Cumene
Ethyl ene
Vinyl Acetate
All
Number
Pumps With
Screening Values
Less Than
1,000 ppmv
12
31
39
82
of Sources Followed
Pumps With
Screening Values
Between 1,000
and 10,000 ppmv
3
2
2
7
30-Day
Occurrence
Rate
Estimate
5.8
18.4
2.8
5.5
95 Percent
Confidence
Interval
(0.7, 20)
(2.8, 42)
(0.8, 6.2)
(2.2, 10)
*A leak from a source was defined as having occurred if it initially
screened <10,000 ppmv and at some later date screened >_10,000 ppmv.
Source: Reference 67.
4-46
-------
positive correlation. As can be seen from Figure 4-4, almost 90 percent of
pump seals are replaced within two years and almost all seals were replaced
cp
within two to three years.
4.2.3 Public Comment
Many of industry's comments concerning fugitive emission reductions
achievable with purnps were similar to ones made for EPA's estimates of
emission reductions for valves. The assumed leak occurrence relationship
used in evaluating leak detection and repair programs using the ABCD model
was questioned by industry with respect to pumps as well as valves.
Industry commenters recommended modification of the ABCD model to
incorporate a calculated n' value as previously discussed.
4.2.4 EPA's Conclusions
After evaluating the available information concerning fugitive
emissions from pumps, EPA found no information which would alter the
estimate of the control efficiency expected to be achieved by dual
seal/barrier fluid systems. The system, when vented to an efficient control
device should achieve virtually 100 percent control, depending on the
efficiency of the control device. The option of enclosing the seal area and
venting to a control device is also expected to achieve a similar level of
control.
The control efficiency of leak detection and repair programs for pumps
was reevaluated in light of the Maintenance Study results and the LDAR
model. The LDAR model represents a refinement over the ABCD model and the
modified ABCD model. And, the data necessary for executing the model are
available. EPA saw the chance for making improvements in their estimates of
leak detection and repair program effectiveness and, therefore, chose to use
the LDAR model.
Table 4-17 shows the available data from which inputs to a leak
detection and repair program model could be selected. As shown in
Figure 4-5, the confidence intervals for all leak occurrence rates
determined in the Maintenance Study overlap. Therefore, the single
occurrence rate for all pumps in the study was chosen to represent the
typical SOCMI case. This occurrence rate was then adjusted to account for
4-47
-------
CUMULATIVE PERCENTAGE
100
00
60
I
-fe
CO
20
4-
#
#
#
+
<£
£
*
^. *****
^ *****
£ *****
/ *****
+ *****
^ ***** *****
£ ***** *****
£ ***** *****
1
<
= 4
1
M
M 0
0 S
. ,
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
H
-
a
M
0
s
,
*****
*****
*****
*****
*****
*****
**.***
*****
*****
*****
*****
*****
*****
*****
a
-
i
2
M
0
S
.
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
1
,
0
-
1
.
b
Y
R
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
1
*
5
-
2
Y
R
S
*
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
2
-
3
Y
R
S
*
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
3
-
14
Y
R
S
.
*****
*****
*****
*****
*****
*****
*****
*****
***** .
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
H
-
5
Y
R
S
,
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
5
,.
6
Y
R
S
,
TIME BETWEEN PUMP SEAL REPLACEMENTS
Figure 4-4'. Intervals Between Pump Seal Replacements,
(Cummulative Percentage), Last Six Years.
Source: Reference 69.
-------
TABLE 4-17. SUMMARY OF AVAILABLE PUMP DATA FOR SOCMI
Data Source
BID*
24-Unit Screening
Maintenance Study*
6-Unit Frequency*
Exxon Cyclohexane
(Summerfield)*
German Studies
EPA 4 Units
DuPont*
SCAQMD
Estimate*
Emission Factor,
kg/hr
0.12 (LL)
0.02 (HL)
VA:0.002 (LL)
CU:0.018 (LL)
ET:0.058 (LL)
0.255(LL)
0.0028*
90% < 0.004
0.006 - 0.063
0.029*
0.073 (LEF)#
0.022 (AEF)#
0.011**
0.049 (LL)
0.021 (HL)
Occurrence Rate, % Initial Leak
(30-day, LL only) Frequency, *
23 (LL)
2 (HL)
8.8 (LL)
2.1 (HL)
VA: 2.8 (LL) VA: 6.7 (LL)0
CU: 5.8 (LL) CU: 16 (LL)
ET: 18.4 (LL) ET: 33 (LL)#
All Units: 5.5 All Units: 14.5
15 (Unit A)
21 (Unit B)
3 (Unit C)
33 (Unit D)
3 (Unit E)
10 (Unit F)
17 (All)
83 (LL)*
23#
5.9**
14.3 (LL)
0 (HL)
15 (LL)
0 (HL)
Comments
*Based on Petroleum Refinery Studies
*VA = vinyl acetate; CU = cumene;
ET = ethylene
^Adjusted for shrouded pumps and pumps
in methane service.
*Considered all pumps.
*Leak definition unclear.
*Based on laboratory studies, field
studies, and interpreted results.
^Average of 20 values.
*Results adjusted in Docket No.
£Leak definition was 200 ppm OVA-128;
LEF = emission factor for leaking
specifically related.
*Emission factor and leak frequency not
source; AEF = average emission factor.
*Leak definition unclear.
*Developed in section on emission
factors using leak and non-leak emission
factors from Petroleum Refinery Studies
and leak frequencies from 24-unit
screening study.
Reference
70
71
72
73
74
75
76
77
78
79
-------
(O
s-
OJ
U
c
O)
a
P
M-
C
O
O
c
(U
-------
seal replacement which normally occurs in SOCMI units. The adjustment was
made based on a total seal life of two years. Using a two year average seal
life would mean that an average of 4.2 percent of the pump seals were
replaced every month. To adjust for the fact that the set of pump seals
replaced would not necessarily correspond exactly to those replaced in a
leak detection and repair program, only half of the 4.2 percent was included
in the adjustment. Then, the 30-day occurrence rate input to the LDAR model
became 5.5 - 1/2 (4.2) = 3.4 percent.
The leak frequency determined for light liquid pumps in the 24 unit
study (8.8 percent) was chosen as an input to the model since it was based
on the most comprehensive data. This leak frequency is also consistent with
the emissiorrfactor which was chosen from the same data set.
Because leaking pump seals are usually replaced with new seals, all
repairs were assumed to be successful. The percent emissions reduction for
successful repair of pumps is estimated based on the reduction of emissions
from the average leaking source emission rate to the non-leaking source
emission rate, resulting in 97.2 percent emissions reduction for replaced
seals. It should be noted that this estimate is probably conservative
because repair would likely decrease the emission rate to a value lower than
the average non-leaking source emission rate. The actual emission reduction
will, therefore, probably be greater.
Summary of Input Values -- The parameters selected to be used as inputs to
the emission reduction calculations for light liquid pumps are summarized in
Table 4-18.
Computation Results -- Using these inputs, the overall effectiveness of leak
detection and repair programs for light liquid pumps was evaluated for
several monitoring intervals using two of the models discussed previously
for valves. The results of these calculations are shown in Table 4-19. As
was indicated for valves, the highest effectiveness was estimated using the
ABCD model presented in the BID and the lowest values were given by the LDAR
model. The higher values computed using the ABCD model were primarily the
01
result of the B-factor assumed for the various monitoring plans. The
B-factors used in executing the ABCD model are the same as those presented
4-51
-------
TABLE 4-18. SUMMARY OF INPUT DATA FOR CALCULATION OF EMISSION
REDUCTIONS DUE TO LDRPs FOR LIGHT LIQUID PUMPS
Occurrence Rate (30-day) 3.4%
Emission Factor 0.049 kg/hr/source
Leak Frequency 8.8%
Emission Reduction for Successful Repair 97.2%
Successful Repair Rate 100%
4-52
-------
TABI F. 4-19. COMPARISON OF LDRP EFFECTIVENESS FOR LIGHT LIQUID PUMPS
Monitoring
Interval
Monthly
Quarterly
Semiannual
Annual
aThe B-factor used in the BIP
Model
ABCDa Modified-ABCD
0.892 0.759
0.845 0.443
0.075
0.751 (0.347)
analysis was based on enr
LDAR
0.608
0.325
(0.076)
(0.800)
jirif.'pring judgement
and was, therefore, retained in this analysis. The other factors were
computed based on the best available data, as presented in Table 4-13.
Note: Numbers in parentheses indicate negative control efficiency.
Negative numbers will be generated when the occurrence rate over
the monitoring interval exceeds the initial leak frequency and may
not be meaningful.
4-53
-------
in the BID. Since they had been based on engineering judgement and not
computed, there was no reason to update the values with new judgements.
Because a method had been devised for estimating the other factors (A, C,
and D), however, they were updated using the inputs established for the LDAR
model. Also, by using the industry model for occurrence/recurrence rate
estimates, new B-factors were computed that resulted in lower effectiveness
when the modified ABCD model was executed. The discrepancy between the
estimates for the ABCD and modified-ABCD models became larger with
decreasing monitoring frequency. And still lower program effectiveness was
op
estimated using the LDAR model.
Both the modified-ABCD and LDAR models predicted negative program
effectiveness values for long monitoring intervals. This anomaly is caused
by the comparison of emissions resulting when a leak detection and repair
program is in place to emissions indicated by the initial fraction of
sources leaking. As explained earlier, there is much uncertainty in fixing
this baseline for purposes of comparison. As further explained, if the
measured occurrence rate for the given monitoring interval exceeds the
initial leak frequency, the models may predict negative emission reductions.
It should be noted that both of these models did indicate decreasing
effectiveness with increasing monitoring interval. The impact of a pump
LDRP on emission reductions in model units is present in Table 4-20 for LDAR
estimates.
4.3 SAMPLING SYSTEMS, OPEN-ENDED LINES, COMPRESSORS, SAFETY RELIEF VALVES
4.3.1 Technical Basis Presented in the BID
Various control techniques were examined for sampling systems, open-
ended lines, compressors and safety/relief valves in the BID. These control
techniques which were incorporated in the regulatory alternatives are shown
in Table 4-21. The control techniques considered included leak detection
and repair programs (LDRPs), equipment, and combinations of equipment and
LDRPs. The type of control evaluated for each source depended upon the leak
characteristics of the source.
As explained in the BID, emissions from sampling systems primarily
result from purging of liquid trapped in sample lines. These emissions will
4-54
-------
TABLE 4-20. LDRP EFFECTIVENESS USING LDAR MODEL FOR PUMPS IN
LIGHT LIQUID SERVICE
Monitoring
Interval
Monthly
Quarterly
Semiannual
Annual
Note: Numbers in par
LDRP
Effectiveness
0.61
0.33
(0.076)
(0.80)
entheses indicate
Emission
Model
Unit A
2.1
1.1
(0.3)
(2.8)
estimated nee
Reductions,
Model
Unit B
7.6
4.1
(1.0)
(10.0)
Mg/yr
Model
Unit C
23.9
12.8
(3.0)
(31.5)
jative control
efficiency. Negative numbers will be generated when the occurrence
rate over the monitoring interval exceeds the initial leak
frequency. These values are subject to interpretation and may not
be meaningful.
4-55
-------
TABLE 4-21. REGULATORY ALTERNATIVES FOR SOME FUGITIVE EMISSION SOURCES IN SOCMI
Regulatory Alternative
II
Source
Sampling Systems
Safety/ relief
valves-gas
"f" Open-ended lines -
S 9as
Open-ended lines -
light liquid
Open-ended lines -
heavy liquid
Compressors
Monitoring
Interval
(% Efficiency)
None
(0)
Quarterly
(59)
Quarterly
(86)
Annually
(62)
None
(0)
Quarterly
(72)
Equipment
Specification
(% Efficiency)
None
(0)
None
(0)
Caps, etc.
(100)
Caps, etc.
(100)
Caps, etc.
(100)
None
(0)
III
Monitoring
Interval
(% Efficiency)
None
(0)
Monthly
(62)
Monthly
(90)
Monthly
(74)
None
(0)
Monthly
(76)
Equipment
Specification
(% Efficiency)
None
(0)
None
(0)
Caps, etc.
(100)
Caps, etc.
(100)
Caps, etc.
(100)
None
(0)
IV
Monitoring
Interval
(% Efficiency)
None
(0)
None
(0)
None
(0)
None
(0)
None
(0)
None
(0)
Equipment
Specification
(% Efficiency)
Closed loop sampling
(100)
Upstream Rupture
Disks (100)
Caps, etc.
(100)
Caps, etc.
(100)
Caps, etc.
(100)
Seal area/degassing
vents to control
device (100)
Regulatory Alternative I was uncontrolled for all sources.
Source: Reference 83.
-------
also depend upon the frequency of sampling. By using closed loop, or closed
purge, sampling systems where the purge VOC is returned to the process,
purge emissions can essentially be eliminated. A control efficiency of
almost 100 percent, therefore, was assumed to be attainable.
Open-ended lines are primarily drain, purge, sample, and vent valves.
The emissions through the seat of these valves were assumed to be eliminated
by installing plugs, caps, blinds, etc. Plugs, caps, etc, were, therefore,
chosen as the equipment requirement for open-ended lines. In the BID,
emissions estimates for open-ended lines included emissions through the seat
as well as emissions around the stem of the associated valve. Therefore, a
combination of equipment and LDRP was considered. The fugitive emissions
around the stem of the valve would be controllable under the LDRPs assumed
for in-line valves. These programs and their effectiveness in reducing
emissions were discussed in detail in the section on emission reductions for
valves.
Fugitive emissions from compressors can result from seal failure. As
part of the regulatory analysis, two LDRPs were considered for compressors:
quarterly and monthly monitoring. The efficiencies of these programs were
estimated by the ABCD model to be 72 percent and 76 percent, respectively.
Regulatory Alternative IV for compressors consisted of a mechanical seal
with non-VOC barrier fluid system. An efficiency of 100 percent was used
for such a system because the emissions from the degassing reservoir were
sent to a control device.
The three control techniques evaluated in the regulatory alternatives
for safety/relief valves in gas service were LDRPs of quarterly and monthly
monitoring intervals and control equipment. Emissions from safety/relief
valves result from failure of seating surfaces, improper reseating after a
release, and "simmering" due to processes operating too near the set
pressures of the valves. Reduction of fugitive emissions from safety/relief
valves requires removal of the leaking valve for repair or replacement.
Even after effecting repairs, elimination of leaking after another over-
pressure is not ensured. The effectiveness of LDRPs for safety/relief
valves was estimated to be approximately 60 percent using the ABCD model
4-57
-------
previously described. Equipment considered in making the selection included
rupture disks, soft-seated (0-ring) relief valves and piping of relief valve
exhaust to a flare or other combustion device.
4.3.2 New Information
Since proposal, additional information has become available only for
open-ended lines. Information on the effectiveness of repair of open-ended
lines was gathered in the SCAQMD in a study of fugitive emissions in
84
refineries subject to Rule 466.1 of that district in California. Of the
open-ended lines screened, 86 percent were sealed by means of plugs, caps,
blinds, or second valves. These sealed open-ended lines leaked at a
frequency of 4.0 percent, while those not sealed had 9.0 percent leak
frequency. Repair of leaking sources was attempted during this study.
The emission reduction achieved was 97 percent.
4.3.3 Public Comment
Industry commenters questioned the emission reductions estimated for
two of the four fugitive emissions sources discussed in this section:
safety relief valves and sampling systems.
1. Commenters disagreed with the assumption that the installation of a
rupture disk beneath a relief valve eliminates leakage. They said that
considerable leakage can occur at the gaskets between the disk holder and
the mating of the valve.
2. The emission reductions estimated for sampling systems were also
questioned. One commenter noted that there are typically four valves
associated with sampling systems and that emissions from these valves easily
exceed the emissions estimated for sampling systems. Another pointed out
that a 100 percent control efficiency is not realistic, since VOC would be
retained and lost from coupling points around the sample container.
4.3.4 EPA's Conclusions
EPA reviewed the information available concerning emission reductions
achievable by the control techniques for sampling systems, safety relief
valves, open-ended lines, and compressors. Rupture disks can be used to
eliminate leakage from safety relief valves. When they fail, they are
replaced, so that fugitive emissions will not occur. The emissions reduc-
tions being considered are emissions through the valve seat, not emissions
4-58
-------
from flanges. Moreover, EPA test results show fugitive emissions from
flanges to be low. Therefore, the estimate of 100 percent control
efficiency is appropriate. Similarly, emissions from capped open ends would
approximate those of flanges and are anticipated to be low, especially in
comparison to the emissions expected from the valve seat and the open end.
EPA realizes that a small amount of VOC may be lost when sampling
connections are broken on a closed purge sampling system. This amount of
VOC is so small compared to the volume of sample purge, however, that the
overall control efficiency approaches 100 percent. Fugitive emissions from
valves associated with the sampling system would be controlled by the LDRP
for valves and should not affect the emission reductions achieved by
eliminating the purge stream. Thus, the estimates in the BID are considered
appropriate. Furthermore, any sampling system closed purge or otherwise
will have valves and when these are properly controlled by a leak detection
and repair program, their emissions will be minimized.
In addition to reevaluating the control efficiencies estimated for
equipment, EPA considered the emission reduction estimates for LDRPs for
safety/relief valves and compressors. The applicability of leak detection
and repair programs to compressor seals would be of limited use considering
the effectiveness of equipment techniques. If a compressor is found
leaking, the repair procedure would be the installation of control
equipment. Because compressors are not generally spared, repair would be
delayed until the next turnaround, thereby reducing the effectiveness of a
leak detection and repair program to essentially zero.
Leak detection and repair programs, however, may be a viable control
alternative for safety/relief valves. Two methods of estimating LDRP effec-
tiveness were considered. First, the ABCD model with values presented in
the BID could be applied. These estimates were based on assumptions that
may not be representative of the actual situation, considering the
comparison of model results for valves and pumps. The results of this model
are given in Table 4-21. The second approach would be based on the LDAR
model. The LDAR model is a better indicator of program effectiveness than
the ABCD model presented in the BID. In order to evaluate the effectiveness
4-59
-------
of LDRPs in reducing emissions from any source using the LDAR model, the
following data would be needed:
1) emission factors,
2) leak frequencies,
3) leak occurrence rates,
4) leak recurrence rates,
5) rate of successful repair,
6) emissions reductions due to successful repair, and
7) emissions reductions due to unsuccessful repair.
Of these data, only leak frequencies and estimated emission factors are
available for safety/relief valves. Because all the data necessary for
evaluating LDRPs for safety/relief valves are not available, the LDAR model
cannot be applied directly using data on safety/relief valves alone. In
lieu of these data, LDAR results for gas valves were compared to ABCD
results for gas valves in developing a revised estimate of LDRP effective-
ness for safety/relief valves in gas service. The LDRP effectivness for
safety/relief valves was estimated using the effectiveness for gas service
valves based on the LDAR model multiplied by the ratio of the effectiveness
for safety/relief valves based on the ABCD model to the effectiveness for
gas service valves based on the ABCD model. This is in essence a rejudgment
of the B-factor in the ABCD model based on the results of the LDAR model for
valves in gas service. A comparison of these estimates is presented in
Table 4-22.
4.4 CONTROL DEVICE
4.4.1 Technical Basis Presented in the BID
As explained in the BID, several control devices were considered for
o c
control of fugitive VOC emissions. These devices included thermal
incinerators, process heaters, and vapor recovery systems, such as carbon
adsorbers. As reported in the BID, through proper design and operation,
carbon adsorption systems reportedly achieve 95-99 percent control
efficiency. Condensation systems can also achieve VOC capture of
oy
90 percent or better. Carbon adsorption systems generally achieve
constant concentrations of VOC in the exhaust stream, based on the specifics
4-60
-------
TABLE 4-22. COMPARISON OF LDRP EFFECTIVENESS FOR SAFETY/RELIEF VALVES
BASED ON ABCD AND LDAR MODELS
Monitoring
Interval
Quarterly
Monthly
ABCD3
0.59
0.62
Model Basis
Estimated
0.44
0.50
LDARb
Effectiveness presented in Table 4-21 and in the BID.
LDAR effectiveness for gas valves multiplied by the ratio of ABCD
effectiveness for safety/relief valves to ABCD effectiveness for gas
valves.
4-61
-------
of the adsorption bed design and gas stream conditions. Thus, the
efficiencies will depend upon the inlet concentration of VOC in the gas
stream, as well as the system design. But, considering the VOC exhaust
streams from fugitive emission sources (except safety relief valves) most
vapor, recovery systems in a process unit can achieve greater than 95 percent
control.
Achievable VOC destruction efficiency by thermal incineration was
evaluated using EPA, industry, and Los Angeles County incinerator test data.
Data from a laboratory incinerator were also compared to actual field test
results. Based on an analysis of these data, it was concluded that new
incinerators could achieve 98 percent control efficiency or 20 ppmv by
OQ
compound exit concentration, whichever is less stringent.
The use of flares was also considered in the BID. The efficiency of
flares in eliminating VOC is not well established and there is no approved
EPA reference method for establishing flare performance. Efficiencies
reported in the literature ranged up to 99 percent. But, based on theore-
tical calculations, an estimate of only 60 percent efficiency had been made
for a flare used on a vent stream in an ethylbenzene/styrene plant (EPA-450/
89
3-79-035a). The estimates of destruction rates were based on the "After-
Qf)
burner Systems Study" by Shell Development Company (EPA-R2-72-062) and
represented a generalized correlation for hydrocarbons combusted at 1410°F.
Design requirements were then considered to ensure that control devices
would achieve a desired VOC destruction. The temperature and residence time
specified for enclosed combustion devices in the proposed standards were
based on data analyzed in an EPA memo ("Thermal Incinerators and Flares")
91
dated August 22, 1980. The data base contained in this memo included
Union Carbide laboratory studies, EPA and industry field tests, and
147 tests on incinerators in Los Angeles county. These data indicate that
greater than 98 percent efficiency is attainable by all new incinerators
operating at 1500°F (816°C) and 0.75 seconds residence time. The memo
concludes that 98 percent efficiency, or less than 20 ppmv, is achievable in
many situations at less than 1600°F (871°C) and 0.75 seconds residence time.
4-62
-------
While thermal Incinerators are proven control devices for destruction
of VOC emissions, they are not the only enclosed combustion devices that
could be used. In fact, boilers and process heaters are expected to be used
for eliminating the small VOC streams covered by the standards. Other
systems, such as catalytic incinerators, and vapor recovery systems are also
applicable to control of these streams.
A control efficiency of at least 95 percent was chosen as the design
requirement because it is a reasonable control efficiency which is
achievable for vapor recovery systems such as carbon adsorption or
condensation units used for fugitive emission sources. This control
efficiency can be achieved by boiler furnaces, incinerators, process
92
heaters, and carbon adsorption units.
4.4.2 New Information
A study was undertaken for EPA by Battelle Memorial Laboratories in
93
order to develop a methodology for measuring flare performance. This
program was carried out in conjunction with John Zink. Although the program
was to develop flare test methods and was not specifically designed to
provide flare performance data, the data gathered during this development
program indicate that properly designed and operated flares may attain from
94 to 99+ percent local burnout efficiency. Both the single air-augmented
flare and the flare consisting of three multi-stage burners attained local
destruction efficiencies of greater than 99 percent under normal operating
conditions. When the single flare was operated without supplemental air
(nonsmokeless operation), the local destruction efficiencies measured were
greater than 94 percent, with an average of about 96 percent. As was the
94
case in the Siegel flare study, the efficiencies measured were local
burnout efficiencies, not overall flare efficiencies.
4.4.3 Public Comments
A number of comments were received after proposal concerning the use of
flares and the performance level achievable by them. Chemical producers,
with a long history of flare use, stated that flares are efficient control
mechanisms capable of achieving 95-99 percent VOC destruction efficiency.
They cited as their primary support a 1980 study of flare conversions in
4-63
-------
95
refinery flares by K.D. Siegel. Based on this efficiency estimate, they
said that flares should be considered equivalent control devices to the
other accepted techniques (thermal incineration, vapor recovery).
Efficiencies greater than 95 percent were cited in the literature for newer
flaring techniques, such as multistage flares and flares with air or steam
premixing. Multistage flares result in improved efficiency since variable
flowrates are handled in stages, keeping close to design conditions.
Commenters also expressed concern that time and temperature requirements
precluded the use of catalytic incinerators.
4.4.4 EPA's Conclusions
To date, four studies of flare performance that have been conducted
include a 1972 study of ethylene destruction by flares conducted by DuPont,
a 1980 study of flare conversions by K.D. Siegel, a recent flare study by
Union Carbide, and the John Zink flare study recently conducted for EPA by
Battelle Memorial Laboratories. Each of the studies is discussed below with
respect to the discussion on control devices in the BID.
DuPont study. In 1972, DuPont conducted a series of tests on a
bench-scale flare to develop sampling procedures applicable to large flare
systems. Ethylene was used as the waste gas stream; helium was used as a
tracer gas so that air dilutions in the combusted gas stream could be
determined. The flare efficiency was assumed to be equal to the ethylene
destruction efficiency.
Flame lengths measured during test runs without steam addition
corresponded to those estimated by currently accepted empirical relation-
97
ships such as presented in the API reference RP-521. The measured flame
lengths demonstrated no apparent dependence upon gas velocity, indicating
that flame length was bouyancy-dependent. With the addition of steam, flame
length was reduced and its characteristics changed from long, reddish, and
smokey to short and yellow. This is an indication of improved combustion
due to better oxidation and probably higher flame temperature (these
temperatures were not measured).
The quantitative utility of these data is limited for the following
reasons. (1) The analysis used did not consider total hydrocarbons in
4-64
-------
determining flare efficiency; it considered only ethylene destruction.
Although all ethylene fed was destroyed by the flare, one series of tests
indicated that higher molecular weight hydrocarbon tars were produced.
Taking these compounds into account, the flare efficiency based on total
hydrocarbon destruction would be lower than the 100 percent reported.
(2) All data collected during the study were in the turbulent flow region.
This situation represents the design case for a flare system, i.e. emergency
venting. Flares designed for emergency venting and operated for fugitive
emission control might be represented better by a lower flow region.
(3) Finally, the results of the tests on steam addition are inconclusive in
terms of flare efficiency. The limited data do not allow for quantitative
analysis of steam addition on efficiency.
98
Siegel flare study. An extensive study of an industrial flare was
presented by K.D. Siegel in February 1980. The study considered the
conversion of flare gas with varying gas composition, steam addition rate,
and crosswind conditions. Approximately 1300 data points were collected
during 12 test runs made on the flare. But this number of data points is
misleading. The data describe point conditions for contours around the
flame region for various elevations. No evaluation of overall flare
efficiency is presented. All of the data collected were for flow conditions
in the turbulent region (although most of the data was collected in the
lower end of this region). As previously mentioned, lower flow regions may
be more representative of fugitive emissions venting from a flare designed
for emergency situations. One of the most significant limitations of
Siegel's data is the flare design used; his design allowed a degree of
air-fuel premixing prior to the flare tip exit. This is atypical of the
majority of the flare designs used on industrial applications in the United
States.
99
In a previous review of Siegel's paper EPA pointed out limitations in
the gas compositions and carbon balancing used. The high concentration of
hydrogen (about 50 percent), low concentration of hydrocarbon, and
relatively high BTU content (over 1000 Btu/scf) of the flare gases used by
Siegel are not commonly found in the chemical industry. These
4-65
-------
characteristics also promote high combustion efficiency. EPA further noted
that a carbon mass balance could not be made.
Based on these points, the quantitative results of Siegel's flare study
are of questionable utility and applicability to the flares used for
fugitive VOC streams. But there are some useful results of this study.
Siegel does examine the relationship between the measured visible flame
length and existing flame length correlations (applicable to turbulent flow
conditions). His results also indicate enhancement of flare efficiency with
increasing steam addition.
Union Carbide Study. Researchers at Union Carbide Corporation
presented the initial results of a study of flare destruction efficiency.
This study was focused on obtaining a better understanding of the fundamen-
tals of hydrocarbon destruction in a flare; establishing a quantitative
flare efficiency was not the primary intent of the study. Two bench-scale
flares (1/8 inch diameter and 2-inch diameter) were used to burn a propane
fuel gas. Helium was used as the tracer gas so that air dilutions of the
flared gas could be determined. For all cases, no steam-enhancement was
used and the resulting flame was sooty.
Both continuous and integrated grab sampling was used and propane
destruction efficiencies determined. The local burnout efficiencies
reported for the 2-inch flare ranged from 95.8 percent to greater than
99.9 percent. These values must be considered in a qualitative sense only
since some sampling problems were encountered during the study and some
samples were taken from eddy flames that were still burning.
John Zink flare study. EPA issued a contract to Battelle Memorial
Laboratories to develop measuring techniques for use on flare systems. The
Battelle study was just recently conducted on a John Zink flare facility.
Since this program was conducted to develop the measurement techniques for
use on flares, the results have questionable applicability to the SOCMI-
Fugitive VOC Emissions NSPS development. This is not a comprehensive study
of flare efficiency; rather, limited testing was conducted to evaluate
emission measurement techniques.
4-66
-------
These studies were conducted on two different flare systems and
employed two techniques for measuring the destruction efficiency of each
system. The major portion of testing was conducted using an air-augmented
single flare burning commercial grade propane at between 3000 and
5000 pounds per hour. A limited number of tests were conducted to determine
the horizontal and vertical profiles of the emissions envelope, the
emissions under normal operating conditions, and the emissions under
nonsmokeless operation (no supplemental air provided). The second flare
system tested consisted of three burners mounted in-line; these burners were
of the type generally used in multistage flare systems. Natural gas was
burned in the second system at between 2000 and 3000 pounds per hour. For
normal smokeless operation, both flare systems demonstrated destruction
efficiencies greater than 99 percent. The nonsmokeless operation of the
single flare resulted in destruction efficiencies greater than 90 percent
for an average efficiency of about 96 percent.
Agency Conclusions -- Reviewing available data and literature, EPA
determined that control devices in new installations should be capable of
attaining better than 95 percent efficiency in reducing VOC emissions.
EPA's review of thermal incinerators concluded that new incinerators should
achieve at least 98 percent efficiency or 20 ppmv VOC in the exhaust stream,
102
whichever is the less stringent requirement. Similarly, carbon
adsorption systems in new installations can be designed and operated to
attain at least 95 percent removal of VOC. The use of carbon adsorption
would be more limited in this application than incineration, but would
provide a recovered product.
Furthermore, EPA has determined that smokeless flares may be capable of
achieving 90 percent or better control efficiency. Although most data
reported show 95 percent (or over) combustion efficiency for flares, there
is considerable uncertainty whether this efficiency is applicable to plant
flares. These uncertainties arise from the following areas:
4-67
-------
1) Composition and heat content of the flared stream - the tests that
have been conducted were on streams with heat contents of about
1,000 Btu/scf. Some plant flares burn gases with heat contents
below this and occasionally go out due to a low heat content (less
than 160 Btu/scf).
2) Size of flares tested - the flares tested were small units, the
largest burning less than 7,000 Ib/hr of gas. Since some
industrial flares are sized to combust 900,000 Ib/hr, the air
required for complete combustion may not be as effectively
supplied by thermal draft effects as it is in the smaller flares.
3) State of tested flares - the plant-sized flares tested were highly
maintained state-of-the-art flares. This degree of technical
attention is not expected for most plant flares. And reduced
maintenance could result in decreased efficiency due to incomplete
mixing and by-passing of the flame.
4-68
-------
4.5 REFERENCES
1. U.S. Environmental Protection Agency. VOC Fugitive Emissions in
Synthetic Organic Chemicals Manufacturing Industry - Background Infor-
mation for Proposed Standards. Research Triangle Park, N.C.
Publication No. EPA-450/3-80-033a. November 1980.
2. Letter and attachments from Bottomley, F.R., Union Oil Company, to
Feldstein, M., Bay Area Air Quality Management District. April 10,
1979. 36p. Information about valve repairability.
3. Honerkamp, R.L., et al. Valve Screening at Six San Francisco Bay Area
Petroleum Refineries. Radian Corporation. Final Report.
nrM-tf "7n oin o~7n nc r«u«..-,^,, c. imn
HuiiciisaMiy, r, .L., c i, a i . vaivt JUICCMI
Petroleum Refineries. Radian Corporat
DCN# 79-219-370-05. February 6, 1979.
4. Wetherold, R. and L. Provost. (Radian Corporation.) Emission Factors
and Frequency of Leak Occurrence for Fittings in Refinery Process
Units. (Prepared for U.S. Environmental Protection Agency.) Research
Triangle Park, N.C. Publication No. EPA-600/2-79-044. February 1979.
5. Letter and attachments from Thompson, R.M., Shell Oil Company, to
Feldstein, M., Bay Area Air Quality Management District. April 26,
1979. 46p. Information about valve repairability.
6. Reference 2.
7. Reference 4.
8. Radian Corporation. Assessment of Atmospheric Emissions from Petroleum
Refining, Appendix B: Detailed Results. (Prepared for U.S. Environ-
mental Protection Agency.) Research Triangle Park, N.C. Publication
No. EPA-600/2-80-075c. April 1980.
9. Reference 5.
10. Reference 8.
11. U.S. Environmental Protection Agency. Air Pollution Emission Test at
Phillips Petroleum Company. Research Triangle Park, N.C. EMB Report
No. 78-OCM-12E. December 1979.
12. Labadie, G.P. (Chevron U.S.A. Inc.) Fugitive Hydrocarbon Emission
Control at Chevron U.S.A.'s El Segundo Refinery. (Presented at
American Petroleum Institute Operating Practice Committee,
San Francisco, California. May 14, 1979.)
13. Reference 11.
4-69
-------
14. Langley, G.J. and R.G. Wetherold. (Radian Corporation.) Evaluation of
Maintenance for Fugitive VOC Emissions Control. (Prepared for
U.S. Environmental Protection Agency.) Research Triangle Park, N.C.
EPA No. 600/2-81-080. May 1981.
15. Langley, G.J., et al. (Radian Corporation.) Analysis of SOCMI VOC
Fugitive Emissions Data. (Prepared for U.S. Environmental Protection
Agency. Research Triangle Park, N.C. EPA No. 600/2-81-111.
June 1981.
16. Reference 14.
17. Blacksmith, J.R., et al. (Radian Corporation.) Problem Oriented
Report: Frequency of Leak Occurrence for Fittings in Synthetic Organic
Chemical Plant Process Units. (Prepared for U.S. Environmental
Protection Agency.) Research Triangle Park, N.C. EPA
Mo. 600/2-81-003. September 1980.
18. Harvey, C.M. and A.C. Nelson. (PEDCo Environmental, Inc.) VOC
Fugitive Emission Data - High Density Polyethylene Process Unit.
(Prepared for U.S. Environmental Protection Agency.) Research Triangle
Park, N.C. EPA No. 600/2-81-109. June 1981.
19.
20.
21.
22.
Reference 14.
Reference 15.
Reference 18.
Reference 18.
23. Honerkamp, R.L. and M.L. Schwendemen. (Radian Corporation.)
Evaluation of the Maintenance Effect on Fugitive Emissions for
Refineries in the South Coast Air Quality Management District. Draft
final report. (Prepared for U.S. Environmental Protection Agency.)
Research Triangle Park, N.C. December 1981.
24. Reference 23.
25. Williamson, H.J., et al. (Radian Corporation.) Model for Evaluating
the Effects of Leak Detection and Repair Programs on Fugitive
Emissions. Technical Note DCN 81-290-403-06-05-03. September 1981.
26. Reference 25.
27. Reference 25.
28. Reference 17.
4-70
-------
29. Reference 14.
30. Reference 1.
31. Reference 17.
32. Reference 14.
33. U.S. Environmental Protection Agency. Air Pollution Emission Test at
Dow Chemical Company. Research Triangle Park, N.C. EMB Report
No. 78-OCM-12C.
U.S. Environmental Protection Agency. Air Pollution Emission Test at
Union Carbide Corporation. Research Triangle Park, N.C. EMB Report
No. 78-OCM-12A.
U.S. Environmental Protection Agency. Air Pollution Emission Test at
Stauffer Chemical Co. Research Triangle Park, N.C. EMB Report
No. 78-OCM-12D.
U.S. Environmental Protection Agency. Air Pollution Emission Test at
Phillips Petroleum Co. Research Triangle Park, N.C. EMB Report
No. 78-OCM-12E.
U.S. Environmental Protection Agency. Air Pollution Emission Test at
Refinery E. Research Triangle Park', N.C. EMB Report Mo. 78-OCM-12F.
Reference 11.
34. Letter and attachment from Cox, J.B., Exxon Company, to Walsh, R.T.,
EPA:CPB. March 21, 1979. 4p. Information about cyclohexane unit.
35. Reference 12.
36. Meeting Report. Honerkamp, R., Radian Corporation to Hustvedt, K.C.,
EPA:CPB, and distribution list. June 12, 1979. 14p. Minutes of
meeting between EPA and DuPont representatives about fugitive emissions
sampling.
37. Reference 18.
38. Reference 2.
39. Reference 5.
40. Reference 8.
41. Reference 11.
4-71
-------
42. Memo from Hustvedt, K.C., EPA:CPB, to Durham, J.F., EPAtCPB.
December 2, 1980. 170p. MRC SOCMI Fugitive Testing.
43. Schwanecke, R., "Air Pollution Resulting from Leakage from Chemical
Facilities." Luftverun-reinigung, 1970, pp.9-15. Translated for the
U.S. Environmental Protection Agency by SCITRAN. Santa Barbara,
California.
Kremer, H. "Leakages from Static and Dynamic Seals in Chemical and
Petrochemical Plants." 4th Meeting OG EW/DGMK. Salzburg.
October 1976. Translated for the U.S. Environmental Protection Agency
by SCITRAN, Santa Barbara, California.
Bierl, Alois, et al. "Leakage Rates of Sealing Elements." Chem. Ing.
Tech. 49_ (No.2) 1977, pp.89-95. Translated for the U.S. Environmental
Protection Agency by SCITRAN. Santa Barbara, California.
Schwanecke, R. "Air Pollution Through Small Leakages for Equipment of
the Chemical Industry and Ways for Their Prevention." Translated for
the U.S. Environmental Protection Agency by SCITRAN. Santa Barbara,
California.
44. Reference 23.
45. Reference 14.
46. Reference 18.
47. Reference 4.
48. Reference 17.
49. Reference 33.
50. Reference 34.
51. Reference 36.
52. Reference 18.
53. Reference 15.
54. Reference 14.
55. Reference 15.
56. Reference 43.
57. Reference 12.
4-72
-------
58. Reference 23.
59. Reference 14.
60. Reference 18.
61. Reference 1.
62. Reference 25.
63. Reference 1.
64. Reference 14.
65. Reference 15.
66. Reference 23.
67. Reference 14.
68. Reference 23.
69. Reference 23.
70. Reference 1.
71. Reference 17.
72. Reference 14.
73. Reference 35.
74. Reference 36.
75. Letter with enclosures from Schroy, Jerry M., Monsanto Company, to
Central Docket Section, Docket No. A-79-32. March 26, 1981. Technical
Comments on Proposed VOC Emission Regulations.
76. Reference 45.
77. Reference 44.
78. Reference 38.
79. Reference 23.
80. Reference 14.
81. Reference 1.
4-73
-------
82. Reference 25.
83. Reference 1.
84. Reference 23.
85. Reference 1.
86. Basdekis, H.S. and C.S. Parmele. (IT Enviroscience.) Control Device
Evaluation - Carbon Adsorption. In: U.S. Environmental Protection
Agency. Organic Chemical Manufacturing Volume 5: Adsoprtion,
Condensation and Absorption Devices. EPA No. 450/3-80-027.
December 1980. Report 1, p. 11-22.
87. Burklin, C.E., et al. (Radian Corporation.) Control of Hydrocarbon
Emissions from Petroleum Liquids. (Prepared for U.S. Environmental
Protection Agency.) Research Triangle Park, N.C. Publication
No. EPA-600/2-75-042. September 1975. p.16.
88. Memo from Farmer, Jack R., EPA:CPB, to Distribution. August 22, 1980.
29p. Thermal incinerators and flares.
89. U.S. Environmental Protection Agency. Draft Background Information for
Proposed Standards for Benzene Emissions from the Ethylbenzene/Styrene
Industry. Research Triangle Park, N.C. Publication No.
EPA-450/3-79-035a. October 1979.
90. Rolke, R.W., et al. (Shell Development Company.) Afterburner Systems
Study. (Prepared for U.S. Environmental Protection Agency.)
Washington D.C. EPA-R2-72-062. August 1872.
91. Reference 88.
92. Blackburn, J.W. (IT Enviroscience.) Control Device Evaluation -
Thermal Oxidation. In: U.S. Environmental Protection Agency. Organic
Chemical Manufacturing Volume 4: Combustion Control Devices.
Publication No. EPA-450/3-80-026. December 1980. Report 1, pp.II-3 -
II-8.
Radian Corporation. Control Techniques for Volatile Organic Emissions
from Stationary Sources. (Prepared for U.S. Environmental Protection
Agency.) Research Triangle Park, N.C. Publication No.
EPA-450/2-78-022. May 1978. p.34.
93. Howes, J.E., et al. (Battelle Columbus Laboratories.) Development of
Flare Emission Measurement Methodology. Draft Final Report. (Prepared
for U.S. Environmental Protection Agency.) Research Triangle
Park, N.C. August 1981.
4-74
-------
94. Siege"! , K.D. Degree of Conversion of Flare Gas in Refinery High
Flares. PhD. Thesis, Fridericiana University, Karlsruhe, FRG, 1980.
95. Reference 94.
96. Palmer, P.A. A Tracer Technique for Determining Efficiency of an
Elevated Flare. Wilmington, Delaware, E.I. duPont de Nemours and
Company, 1972.
97. American Petroleum Institute. Guide for Pressure Relief and
Depressuring Systems. API RP 521, First Edition. September 1969.
98. Reference 94.
99. Energy and Environmental Analysis, Inc. Evaluation of "Degree of
Conversion of Flare Gas in Refinery High Flares" by K.D. Siege!.
(Prepared for U.S. Environmental Protection Agency.) Research Triangle
Park, N.C. August 1980.
100. Lee, K.C. and G.M. Whipple. (Union Carbide Corporation.) Waste
Gaseous Hydrocarbon Combustion in a Flare. (Presented at the 74th
Annual Meeting of the Air Pollution Control Association, Philadelphia,
Pennsylvania. June 21-26, 1981.)
101. Reference 93.
102. Reference 88.
103. Reference 86.
4-75
-------
5. COST ESTIMATES
This section presents the cost estimates and the input parameters
affecting the cost estimates for each of the fugitive emission sources
presented in the Background Information Document (BID). Also discussed are
the new information and the public comments regarding these estimates and
any change's in the estimates due to the new information or the comments.
All costs are presented on a last quarter 1978 basis.
5.1 VALVES
5.1.1 Technical Basis in the BID
The cost for leak detection and repair programs for valves were
calculated in the BID. The factors affecting the total cost of a leak
detection and repair program are (1) monitoring time, (2) repair time for
on-line and off-line repair, and (3) fractions of leaks repaired on-line and
off-line. The following estimates were used for the above:
(1) Monitoring time: The time required to monitor leaking
valves was available from several sources. Exxon
Company, U.S.A. conducted an in-depth,study to determine
the monitoring manpower requirements. Five Exxon
refineries conducted field surveys using Bacharach and
Century Systems analyzers to determine the time required
for monitoring various fugitive emissions sources. The
average monitoring time for a leak detection survey for
valves was found to be 2 man-minutes per valve. In the
EPA 24-unit study the actual monitoring time was found
to be considerably less than that estimated using the
2 man-minute monitoring time for valves and monitoring
times for other emissions sources, e.g. pumps, compres-
sors etc. estimated by the Exxon Company and other
studies. Table 5-1 compares the estimated and actual
monitoring times for each of the 24 units in the EPA
study. Overall, the actual time was about 75 percent of
the estimated time. In another study conducted by Union
Carbide Corporation in one of their chemical units it
was estimated that 400-500 sources could be screened in
one day. Although this estimate is based on a three-
man team, the third person is a unit operator to provide
process data. Therefore, for a two-man monitoring team
this corresponds to 1.9 - 2.4 man-minutes per source.
5-1
-------
TABLE 5-1. ESTIMATED VERSUS ACTUAL MONITORING TIMES FOR
SOCMI PROCESS UNITS IN THE 24 UNIT STUDY
Unit
Number
1
2
3
4
5
6
11
12
20,21
22
28,29
31
32
33
34
35
60,61,62
64
65
66
TOTALS
Chemical
Vinyl Acetate
Ethyl ene
Vinyl Acetate
Ethyl ene
Cumene
Cumene
Ethyl ene
Acetone/Phenol
Ethyl ene Di chloride/
Vinyl Chloride
Formaldehyde
Ethyl ene Di chloride/
Vinyl Chloride
Methyl Ethyl Ketone
Methyl Ethyl Ketone
Acetaldehyde
Methyl Methacrylate
Adipic Acid
Chlorinated Ethanes
Adipic Acid
Acrylonitrile
Acrylonitrile
Total
Number of EPA-Estimated
Sources Monitoring Time
Monitored (hours)
1391
5078
2713
5278
1025
1573
3685
3207
2298
230
3363
585
679
1148
2019
1577
3332
664
1406
1864
43,115
54
176
98
182
36
55
143
128
91
9
123
22
26
44
77
53
121
26
51
68
1583
Actual
Monitoring
Time
(hours)
46
110
42
132
15
26
117
171
100
7
90
16
25
23
30
18
89
21
59
59
1196
Estimated monitoring time is based on the equipment counts in each unit
and the monitoring time estimates for each source presented in the BID.
Source: Reference 4.
5-2
-------
Other data were provided by Phillips Petroleum Company.
In a study conducted at a refinery and natural gas
processing center which included three ethylene units,
Phillips screened about 70,000 sources with a 2 man team
in about 936 manhours for an average of 0.8 man-minutes
per source. A monitoring time estimate of 2 man-
minutes per valve was used in the BID.
.(2) Repair time: The California Air Resources Board
conducted a study of fugitive emissions. The repair
time estimate of 10 man-minutes for on-line repair of
valves was based on this study. The repair time
estimate for off-line repair was 4 man-hours per valve
and was based on the Exxon study.
(3) Fraction of leaks repaired on-line and off-line: This
estimate waSr,also based on the California Air Resources
Board Study. 75 percent of all valves were estimated
to be repaired on-line while 25 percent were estimated
to be repaired off-line.
5.1.2 New Information
Allied Chemical Company conducted a study of fugitive emissions in a
g
high density polyethylene unit. A two-man team was used to conduct the
screening. One man performed the screening while the other recorded the
data. The unit, area, the tag number, and the screening value were
recorded. All leaking sources were tagged. The overall monitoring time for
six inspections averaged 2.9 man-minutes per source. However, by the sixth
inspection the monitoring time was down to 1.9 man-minutes per source.
The on-line repair time requirement for valves was determined in six
process units during the EPA maintenance study. The repair time require-
ment determined in this study was 9.6 man-minutes. The summary of the
maintenance study repair time data is presented in Table 5-2.
5.1.3 Public Comments
Three major comments were received regarding the cost estimates for
valves.
1. Several commenters expressed disagreement with the monitoring time
estimate. One commenter said that the monitoring time estimate was low by a
minimum factor of two. Another commenter stated that the data provided by
the industry had been used out of context to come up with the estimate. He
5-3
-------
TABLE 5-2. SUMMARY OF ON-LINE REPAIR TIME DATA
(Six Unit Maintenance Study)
Maintenance
Unit Period
1 1
2 1
2
3 1
4 1
5 1
2
6 1
2
Total
Number Total Time for
of Maintenance
Attempts (minutes)
10
3
13
8 (2)*
2
2
8
5 (2)*
4
18
10 (2)*
10 (1)*
8
3 (1)*
29 (3)*
8 (6)*
13
1
149**(17)*
180
201
287
48
45
15
125
60
80
143
70
129
90
125
1,598
Average Maintenance
Time Per Valve
(minutes)
18.1
N.A.
13.4
28.7
15.0
N.A.
6.0
6.3
3.7
6.9
5.0
7.3
17.9
17.5
4.0
6.4
9.6
N.A
9.6
*Valves were not maintained.
**Does not include valves for which no time information was available.
(N.A. - No information on time of maintenance available).
Source: Reference 11.
5-4
-------
said that the 2 man-minutes figure was generated as a ball-park number for
initial comparison purposes. Commenters claimed that more recent detailed
data show monitoring times of 3 to 4 minutes. An industry report was cited
which shows that the monitoring time varies from a minimum of 3 minutes to a
maximum of 12 minutes per valve.
2. One commenter was concerned that the cost estimates did not include
any unit downtime for repair of valves which must be taken off-line for
repai r.
3. Another commenter asked that further consideration of administra-
tive manpower be made. As an example, he estimated that a unit with 2,800
valves would require about 15 to 20 man-days of engineering time and 250 to
300 man-days of drafting time to establish the program. He said that added
to this would be drafting time for showing minor revisions in the plant or
monitoring schematics.
5.1.4 EPA's Conclusions
The estimate used for monitoring time was based on information provided
by Exxon Company, USA. This information was presented as the result of "an
12
in-depth study to determine the monitoring manpower requirements." In
absence of data to the contrary, a monitoring time of 2 man-minutes is the
most reasonable estimate available. In addition, information from other
studies shows that the monitoring time may even be less than 2 man-minutes.
It should be noted that valves are expected to be clustered in groups. It
would be possible to monitor one group in a short time and then move on to
another. The 2 man-minutes estimate is an average for all valves in a unit.
The 3 to 12 man-minutes estimate made in a document cited in the public
comments was inaccurately cited as a monitoring time estimate for valves.
These are actually estimates of time required to monitor pump seals,
compressor seals, valves, drains, and pressure relief valves. Furthermore,
these estimates also include time required for doing some minor valve
repair. The repair time requirement was accounted for separately in the
RID. It should be noted that the repair time requirement determined by the
maintenance study i.e. 9.6 man-minutes has verified the 10 man-minutes
estimate used in the BID.
b-5
-------
There would be no need for a unit to shut down for repair. Allowance
is made for repairing such critical valves at the next unit turnaround.
Because shutdowns are not required, allowance for cost of unit downtime is
not necessary.
The engineering and draft time requirement mentioned by one commenter
was accounted for in the BID. The labor charge estimates used in the BID
include allowance for administrative and support costs to implement the
regulation. The details of these estimates are discussed later in
Section 5.7 (Other Costs).
13
Based on the results of the LDAR Model (discussed in Section 4) the
annual cost of monitoring and repair of valves have been estimated for the
following monitoring intervals: (1) monthly, (2) quarterly, (3) quarterly
with monthly follow-up of leaking valves, (4) semi-annual, and (5) annual.
An example cost calculation (for monthly monitoring) is presented in
Tables 5-3 through 5-6 for model units A, B, and C (see Section 3 on Model
Units). The input parameters e.g. occurrence rate, initial leak frequency,
etc. are discussed in the Emission Reduction section (Section 4).
A similar procedure is followed to calculate the net annualized costs
of monthly/quarterly, quarterly, semi-annual, and annual leak detection and
repair programs for valves. These costs are summarized in Table 5-7.
The costs shown are for the second turnaround period. The second
turnaround period (the time between turnarounds 1 and 2) was used to
minimize the uncertainties in determining when in the first turnaround cycle
the leak detection and repair program began. Furthermore, the second
turnaround period average approximates the average for turnaround periods
after a leak detection and repair program has been in place for a long time.
5.2 PUMPS
5.2.1 Technical Basis in the BID
The costs for both equipment control and leak detection and repair
programs were calculated in the Background Information Document. These
costs are presented below.
5-6
-------
TABLE 5-3. INITIAL LEAK REPAIR LABOR-HOURS REQUIREMENT
FOR VALVES BY MODEL UNIT
No. of Valves Initial Leak Estimated No. of Repair Time, Labor-Hours
Per Model Unit Frequency Initial Leaks Man-Hours Required
ABC ABC ABC
99
131
402
524
1232(G)
1618(LL)
0.114
0.065
11.3
8.5
45.8
34.1
140.4
105.2
1
1
.13
.13
12.8
9.6
22.4
51.8
38.5
90.3
158.7
118.9
277.6
Based on 75 percent valves repaired on-line in 10 man-minutes and
25 percent repaired off-line in 4 man-hours.
TABLE 5-4. TOTAL COSTS FOR INITIAL LEAK REPAIR
FOR VALVES BY MODEL UNIT
Initial Leak Repair Labor
Admin. & Support Costs (0.
Total Costs
Annual ized charges for ini
(0.163 x total costs)
Charges ($15/hour)
4 x labor charges)
tial leak repair
A
$336
134
$470
$77
B
1355
542
1897
309
C
4164
1666
5830
950
alnitial leak repair costs amortized over 10 years at 10 percent interest
(CRF = 0.163).
5-7
-------
TABLE 5-5. ANNUAL MONITORING AND LEAK REPAIR LABOR REQUIREMENTS
(Monthly Leak Detection and Repair Program for Valves)
en
i
oo
No.
Per
A
99
131
of Valves
Model
B
402
524
Unit
C
1232(G)
1618(LL)
Type of
Monitoring
Instrument
Instrument
Monitoring
Time ,
Man-Mi na
2
2
Times
Monitored
Per Year
12
12
Monitoring labor-hours
A
39.6
52.4
92.0
Required
B
160.8
209.6
C
492.8
647.2
No. of leaks Repair
Per Year Time,
ABC Man-hours
18.9 76.8 235.3 1.13
25.0 99.9 308.4 1.13
370.4 1140.0
Leak Repair
Labor-hours
Required
ABC
21.4 86.8 265,
28.3 112.9 348,
.9
.5
39.7 299.7 614.4
Instrument monitoring time is 1 minute for a 2 man team.
Average number of leaks found over turnaround 2 from the LORP model, based on monthly occurrence rate of 1.3 percent.
-------
TABLE 5-6. ANNUAL MONITORING AND LEAK REPAIR COSTS FOR
MONTHLY MONITORING OF VALVES BY MODEL UNIT
Monitoring labor-hours
Repair labor-hours
Total labor-hours (Monitoring & Repair)
Labor charges ($15 x tota
Admin. & Support costs (0
charges)
Annual ized charge for ini
repai r
Total costs ($/year)
Product recovery credit
1 labor-hours)
.4 x labor
tial leak
($/year)
Net annual ized costs ($/year)
A
92.0
49.7
141.7
$2126
850
77
$3043
(2490)
$553
B
370.4
199.7
570.1
8552
3421
309
12282
(9990)
2292
C
1140.0
614.4
1754.4
26316
10526
950
37792
(31020)
6772
Product recovery credit is calculated at $300/Mg. The emission reductions
are shown in Section 4.
Note: Figures in parenthesis indicate credits.
5-9
-------
TABLE 5-7. NET ANNUAL MONITORING AND REPAIR COSTS OF LEAK DETECTION
AND REPAIR PROGRAMS FOR VALVES BY MODEL UNITS
Monitoring IntervalNet Annualized Costs
ABC
Monthly 550 2300 6800
Monthly/Quarterly (290) (1060) (3300)
Quarterly (270) (1050) (3200)
Semi-annual 140 450 1400
Annual 1300 5100 15600
Note: Figures in parenthesis indicate a credit.
5-10
-------
Equipment Costs: The equipment for pumps consisted of the following.
(1) Dual mechanical seals, (2) barrier fluid systems for dual mechanical
seals, and (3) closed vents for degassing reservoirs of dual seal pumps.
The cost estimates are in last quarter 1978 dollars and were based on data
from the Hydroscience (now ITE) report. The cost items are as follows:
Dual Mechanical Seals $575/pump
Barrier fluid system for $1500/pump
dual seals
Closed vents for degassing $3265/pump
reservoirs
Total Cost $5340/pump
(last quarter 1978 dollars)
[Seal cost = $560. Single seal
credit - $225. Shop installa-
tion = $240.]
[Pressurized reservoir system =
$700. System cooler = $800.
Pumps that have dual mechanical
seals without regulatory
requirement may not have the
cost of a barrier fluid system
added. The barrier fluid
system is assumed to be an
integral part of the seal
system.]
[Based on installation of a
122m length of 5.1 cm.
diameter, schedule 40 carbon
steel pipe = $5200. Three
5.1 cm. cast steel plug valves
and one metal gauze flame
arrestor - $1300. These costs
include connection of the
degassing reservoir to an
existing enclosed combustion
device or vapor recovery
header. Cost of a control
device added specifically to
control the degassing vents is,
therefore, not included. Two
pumps assumed connected to a
single degassing vent.]
Leak Detection and Repair Costs: The factors affecting the costs of a leak
detection and repair program are (1) monitoring time and (2) repair time.
Both these estimates were based on information provided by Exxon Company,
5-11
-------
USA. These estimates were (1) 10 man-minutes for instrument monitoring,
(2) 0.5 man-minutes for visual monitoring, and (3) 80 man-hours for repair
[based on 2 man-days/pump for field repair (25 percent) and 12.5 man-days/
pump for shop repair (75 percent)].
5.2.2 New Information
No new information was developed regarding cost estimates for pumps.
5.2.3 Public Comments
Comments were received on the equipment cost estimates, on the costs
associated with a control device, and retrofit costs.
1. One commenter gave estimates of $3900 and $7000 per pump for
equipment cost estimates. The BID estimate was said to be lacking in costs
of three components: a spare flushing oil pump, strainers, and instrumen-
tation. The average cost of a dual seal was said to be $1250 (1980
dollars).
2. Several commenters said that even though a control device may
already exist at a facility, it cannot be considered a cost-free utility.
Such equipment was said to be a major portion of the total capital invest-
ment for implementation of the proposed rules. The commenters asked that
the cost of the control device be included in the analysis.
3. One commenter said that equipment layout considerations had been
ignored and that the cost of real estate or adequate transmission lines had
not been included.
4. Two commenters said that energy costs associated with enclosed
combustion devices had not been included in the cost estimates. One of the
commenters also said that another energy cost which had been overlooked was
energy required for a high pressure barrier fluid system.
5. There was also some concern that retrofit costs had been under-
estimated. For example, it was said that an existing plant may not have
enough room for installing a barrier fluid degassing system and combustion
device close to the pump. Locating the equipment away from the pump will
increase construction and operating costs by requiring more piping and
increasing the energy requirements for the system.
5-12
-------
6. One commenter said that labor costs for replacement items such as
pump seals would be higher than the cost for initial installation. The
reason for this was said to be the higher time requirement for disassembly
of equipment and the inefficiencies of working in the field.
5.2.4 EPA's Conclusions
The cost of the barrier fluid system includes the costs for the compo-
nents considered lacking by the commenters. However, the cost of the dual
seal has been revised to reflect the estimate provided by the commenters.
In 1978 dollars the installed cost of a dual seal (after credit for the
single seal) is estimated to be $1030. The BID estimate of this cost was
$575.
No need was seen for the addition of the cost of an enclosed combustion
device. The cost to a facility due to a regulation is the incremental
expenses incurred in order to comply with the regulation. Combustion
devices or vapor recovery systems are expected to already exist at the plant
site. There will, therefore, be no incremental cost associated with the use
of the device. The capital cost of such an existing device need not be
accounted for in the cost analysis. Transmission lines have also been
adequately accounted for. The cost of the degassing reservoir vents
includes 122m of piping. The real estate will be available at the site.
As discussed in Section 7.4 of the BID there are no energy costs
associated with the proposed standards. In fact, the use of combustion
devices could result in a net energy credit if heat energy is recovered.
However, since it is not possible to quantify the number of process units
that may choose to recover heat or the extent of the recovered energy, these
credits were ignored for the purpose of the cost analysis. The equipment
for pumps includes a low pressure barrier fluid system. There is,
therefore, no cost for the energy requirements of a high pressure barrier
fluid system. If a high pressure barrier fluid system were used, there
would be no need for the degassing vent piping.
Shortage of space at an existing plant may make it difficult to install
a barrier fluid system and combustion device close to the pump. However, in
such cases it is expected that most units would choose to enclose seal areas
5-13
-------
and vent the captured emission to an existing control device. Only piping
and a block or check valve would then be required.
As shown in Section 5.2.1, the initial cost of installation of pump
seals is based on labor requirements for replacing an old seal. As such,
instead of replacement cost being understated, the initial installation
costs are overstated.
Section 4 (Emission Reductions) presented the comments made by the
industry regarding use of flares for control of emissions from pumps and
compressors. An analysis of costs for a dual seal/barrier fluid system
vented to a flare has been performed for the purpose of comparison with
costs of a dual seal system vented to an enclosed combustion device or a
vapor recovery system. In addition, leak detection and repair costs have
been recalculated with some changes in estimates. The details of these
analyses are presented below.
Equipment Costs: Two control levels were identified in order to provide a
range of control costs:
(1) Dual pump seal/barrier fluid systems provide no fugitive
VOC emission reduction;
(2) Dual pump seal/barrier fluid systems provide 95 percent
reduction of fugitive VOC emissions.
For the purpose of this cost analysis, it was assumed that existing flares
achieve 90 percent control while enclosed combustion devices (including
devices such as condensers, absorbers, adsorbers, etc.) achieve 95 percent
control. As discussed earlier, no heat recovery credit is given for VOC
controlled by enclosed combustion devices since the credit is hard to
account for, is generally small, and may be intermittent.
Capital costs of a VOC emissions control system for a pump seal include
the costs of dual seals, a barrier fluid system, and a degassing vent
system. The costs of the flare or the enclosed combustion device are not
included as they are assumed to be available at the plant site. Further,
the cost of a vent system servicing an enclosed combustion device is much
greater than the cost of a vent system servicing a flare because a flame
arrestor and additional piping and valves are required. Enclosed combustion
5-14
-------
devices do not have existing common duct systems (header) in the process
unit to which fugitive emissions may be vented. Therefore, individual
piping systems are provided to vent the VOC to the enclosed combustion
device. To avoid flame propagation from the combustion device to the
degassing reservior, each line is equipped with a flame arrester. To allow
in-service repair of the flame arrestor, each arrester assembly is outfitted
with a bypass system of piping and valves. Usually, a flare header runs
throughout a process unit to collect gases from minor process vents and
pressure relief devices. Therefore, only short, lengths of piping would be
needed to connect the degassing reservoir to the flare system. Each pipe
that is connected to the flare header is equipped with a check valve to
avoid back pressure into the degassing reservoir. A block valve for
isolating the degassing reservoir from the flare system is also provided to
allow positive shutoff from the flare header during seal maintenance. The
cost analysis is presented in Tables 5-8 and 5-9.
Leak Detection and Repair Costs: Leak detection and repair costs have been
recalculated to reflect some changes in estimates. Changes were made in the
estimate of repair time. The monitoring time estimate remains the same.
In the BID, a repair time estimate of 80 man-hours per pump seal was
used. This estimate was based on information provided by Exxon Company,
USA. It was assumed that 25 percent of the pumps would be repaired in the
field at the rate of 2 man-days/pump and 75 percent would be repaired in the
shop at the rate of 12.5 man-days/pump.
Further consideration of these estimates indicates that shop repair
would only be needed for cases where the problem is related directly to the
pump itself. Field repair would be sufficient for the purpose of seal
replacement. The repair time estimate has, therefore, been changed to
16 man-hours per pump. In addition, the cost of the replacement seal was
not included in the total cost, of repair in the BID. This cost has now been
included in the analysis of leak detection and repair costs. The credit for
the old seal is estimated to be half the purchase price of the new seal.
18
Finally, it is assumed that the life of a pump seal, is 2 years.
Therefore, 1/24 (4.2 percent) of all pump seals will be replaced as routine
5-15
-------
TABLE 5-8. EQUIPMENT COST FOR CONTROL OF EMISSIONS FROM A PUMP SEAL
(last quarter 1978 dollars)
Degassing Reservior
Degassing Reservoir Vented to an Enclosed
Vented to a Flare Combustion Device
Capital Costs
Dual seal9 , 1030 1030
Barrier fluid0 1500 1500.
Vent system 700C 3265°
Total 3230 5795
Annual Costs
Dual. seald
Barrier fluid
Vent system
Maintenance & f
-r
Miscellaneous charges
Total9
500
245
115
250
1110
500
245
530
480
1755
Based on $1250/seal (1980 dollars) average cost quoted by industry.
1978 cost = $1030 (seal cost = $1015, installation cost = $240, single
seal credit - $225).
DFrom Chapter 8 of the BID.
/-
Based on installed capital cost of 20m of 5.1cm piping ($850), one 5.1cm
check valve ($125), and one 5.1cm block valve ($415) per vent system and
one vent per pair of pump seals.
Based on 2 year seal life and 10 percent interest (CRF = 0.58) for dual
seal and 10 year amortization period and 10 percent interest (CRF =
0.163) for installation charges.
p
Based on 10 year equipment life and 10 percent interest (CRF = 0.163).
Based on 9 percent of total capital costs (Chapter 8 of the BID).
9Costs do not include credits for recovered product.
5-16
-------
TABLE 5-9. NET ANNUALIZED COST OF EQUIPMENT FOR CONTROL OF
EMISSIONS FROM PUMP SEALS (last quarter 1978 dollars)
Degassing Reservoir Vented
Degassing Reservoir Vented To an Enclosed Combustion
To a Flare Device
(1) (2) (1) (2)
95 Percent 95 Percent
Control by No Control
Seal by Seal
Gross annual ized
cost ($/year) $1110 1110
Product recovery
credit ($/year) (123)a 0
Net annual ized
cost ($/year) 987 1110
Control by
Seal
1755
(123)a
1632
No Control
by Seal
1755
0
1755
Credit from 95 percent of uncontrolled VOC fugitive emissions that are
retained in the process due to operation of a pressure sensor between the
dual seals to immediately detect seal failure. Product valued at $300/Mg.
Note: Columns (1) and (2) show a range of net annualized costs depending
upon the degree of control by the seal.
5-17
-------
maintenance every month. On the average, half of routinely maintained
seals, i.e. 2.1 percent of all seals are assumed to be leaking seals. No
cost is attributed to the program for the leaking seals that will be
replaced due to routine maintenance.
Leak detection and repair costs have been calculated for monthly,
quarterly, semi-annual, and annual programs for pumps. An example calcula-
tion (for monthly monitoring program) is presented in Tables 5-10 through
5-14 for model units A, B, and C (see Section 3 on Model Units). The input
parameters e.g. occurrence rate, initial frequency, etc. are discussed in i
the Emission Reduction section (Section 4).
A similar procedure is followed to calculate costs of quarterly, semi-
annual, and annual leak detection and repair programs for pumps. These
costs are summarized in Table 5-14.
5.3 SAFETY/RELIEF VALVES
5.3.1 Technical Basis in the BID
Cost of control of fugitive emissions from safety/relief valves were
calculated for both leak detection and repair programs and equipment
control.
Leak Detection and Repair Costs: Costs were calculated for quarterly and
monthly monitoring of gas service safety/relief valves. The monitoring time
19
estimate was based on information provided by Exxon Company. This
estimate was 16 man-minutes. Also, as suggested by Exxon, it was assumed
that leaks would be corrected by routine maintenance with no additional
labor requirements.
Equipment Costs: Costs were computed for the installation of a rupture disk
upstream of a safety/relief valve in gas service. These costs were based on
20
estimates from the Hydroscience (now IT Environscience) report. The cost
estimates were for the following items.
One 7.6cm stainless steel rupture disk $195
One 7.6cm carbon steel rupture disk holder 325
One 0.6cm dial face pressure gauge 15
One 0.6cm carbon steel bleed valve 25
Installation 240
Subtotal $800
5-1B
-------
TABLE 5-10. INITIAL LEAK REPAIR LABOR-HOURS REQUIREMENT
FOR PUMP SEALS BY MODEL UNIT
No. of Pump Seals
Per Model Unit
ABC
8 29 91
Initial Leak
Frequency
0.088
Estimated No. of Repair
Initial Leaks Time,
ABC Man-hours
0.7 2.6 8 16
Labor-hours
Requi red
ABC
11.2 41.6 128
TABLE 5-11. TOTAL COSTS FOR INITIAL LEAK REPAIR
FOR PUMP SEALS BY MODEL UNIT
Initial Leak Repair Labor Charges
($15/hour)
Admin. & Support Costs (0.4 x labor
charges)
Seal Costs ($113/single seal)3
Total Costs
Annual ized charges for initial leak
repair (0.163 x Total costs)
A
$168
67
79
$314
51
B
624
250
293
1167
190
C
1920
768
900
3588
585
Includes 50 percent credit for old seal.
Initial leak repair costs amortized over 10 year at 10 percent interest
(CRF = 0.163).
5-19
-------
TABLE 5-12. ANNUAL MONITORING AND LEAK REPAIR LABOR REQUIREMENTS
(Monthly Leak Detection and Repair Program for Pump Seals)
No. of Pumps Seals Type of Monitoring
Per Model Unit Monitoring Time
ABC Man-Mina
Instrument 10
o on qi
K ^ yi Visual 0.5
Times
Monitored
Per Year
12
52
Monitoring labor-hours
Required
A
16.0
3.5
TO
B
58.0
12.5
7075
C
182.0
39.4
22T73"
Leak Repair
No. of leaks Repair Labor-hours
Per Year Time, Required
ABC Man-hours ABC
3.3 11.8 37.1 16 52.8 188.8 593.6
Instrument monitoring time is 5 minutes for a two-man team.
Based on 5.5 percent monthly occurrence rate less 2.1 percent credit for routine maintenance. These values were computed
using the LDAR model. Fractions of leaking pumps were included in both cost and emission estimates.
ro
o
-------
TABLE 5-13. ANNUAL MONITORING AND LEAK REPAIR COSTS FOR PUMP
SEALS BY MODEL UNIT
Monitoring labor-hours
Repair labor-hours
Total labor-hours (Monitoring & Repair)
Labor charges ($15 x total labor-hours)
Admin. & Support costs (0.4 x labor
charges)
Annual ized charge for initial leak
repair
Seal costs ($113/seal)a
Total costs ($/year)
Product recovery credit ($/year)
Net annual ized cost ($/year)
A
19.5
52.8
72.3
$1085
434
51
372
$1942
(632)
$1310
B
70.5
188.8
259.3
3890
1556
190
1328
6964
(2290)
4674
C
221.4
593.6
815.0
12225
4890
585
4174
21874
(7183)
14691
alncludes 50 percent credit for old seal.
Product recovery credit is calculated at $300/Mg,
are shown in Section 4.
Note: Figures in parentheses indicate a credit.
The emission reductions
5-21
-------
TABLE 5-14. NET ANNUAL MONITORING AND REPAIR COSTS OF LEAK DETECTION
AND REPAIR PROGRAMS FOR PUMP SEALS BY MODEL UNIT
Net Annualized Costs
Monitoring Interval A B C__
Monthly $1300 4700 15000
Quarterly 1300 4800 15000
Semi-annual 1600 5800 18000
Annual 2200 8000 26000
5-22
-------
It was assumed that no piping modification was required and that the disk
and its holder simply could be inserted between the flanges of the relief
valve and the system it protects. To allow in-service disk replacement, a
block valve was assumed to be installed upstream of the rupture disk. In
addition, to prevent damage to the relief valve by disk fragments, it was
assumed that an off-set mounting would be required. The total installed
cost of a new rupture disk system was estimated to be $1730 (last quarter
1978 dollars). The rupture disk life was assumed to be 2 years.
5.3.2 New Information
No new information was developed regarding cost estimates for safety
relief valves.
5.3.3 Public Comments
Several comments were received on the relief valve/rupture disk cost
estimates.
1. One commenter said that the cost of repair and. monitoring of relief
valve/rupture disk systems should be included in the analysis. He estimated
that it would take 10 minutes to monitor the relief valve after repair. The
cost of the replacement disk would be $200. Four hours were estimated for
repair. He further estimated 3 discharges per year.
2. Another commenter said that the capital cost of the relief valve/
rupture disk systems had been underestimated. He cited assumption of very
modest piping modification to be the reason for the low cost estimate. His
estimates were $3800 per disk for a new installation including disk,
flanges, a tee downstream, and a pressure gauge.
3. On the other hand, another commenter said that the capital cost for
relief valves/rupture disk systems have been overstated. He said that new
developments have eliminated the requirement for offset monitoring and the
cost of tees, elbows, etc. have been eliminated.
5.3.4 EPA's Conclusions
As pointed out by one commenter, some relief valves may have 3
discharges per year. On the other hand, according to information from Exxon
21
Company, USA, "most safety valves on light ends towers never discharge."
On the average, a rupture disk may be expected to require replacement every
5-23
-------
2 years. This estimate of useful life is assumed in the cost analysis, i.e.
an allowance for replacement of the rupture disk every 2 years has been made
in the analysis.
Equipment cost estimates for control of fugitive emissions from safety/
relief valves have been calculated for four different systems. It is
expected that each of the four systems would be used for a certain
proportion of relief valves in the industry. These costs are s-hown in
Table 5-15.
5.4 SAMPLING SYSTEMS
5.4.1 Technical Basis in the BID
Equipment costs were computed for closed loop sampling connections.
The cost estimates were based on information from the Hydroscience (now ITE)
24
report and are shown in Table 5-16.
5.4.2 New Information
The Texas Chemical Council (TCC) has estimated that 75 percent of the
25
sampling systems in new plants are currently closed loop sampling systems.
5.4.3 Public Comments
Carbon steel construction was said to be inadequate for many sampling
applications in SOCMI. The cost of a stainless steel closed loop sampling
connection was said to be $1600.
5.4.4 EPA's Conclusions
EPA has seen no evidence that carbon steel construction may not be
adequate for most sampling applications in SOCMI. In addition, since
75 percent of the sampling systems in new plants are closed loop type the
cost estimates in the BID have overstated the cost of regulation to the
industry. EPA expects that the 75 percent of the sampling systems which are
closed purge systems are in toxic or corrosive service which might require
stainless steel. EPA further believes that the remaining 25 percent can use
carbon steel materials.
To reflect the new information concerning the number of closed loop
sampling systems currently in use, a change has been made in the uncon-
trolled equipment count for model units. For details see Section 3 on Model
Units.
5-24
-------
TABLE 5-15. RELIEF VALVE CONTROL COSTS, FOUR SYSTEMS
(1) Rupture disk systems Total Capital Cost = $1730/relief valve
with block valves (Details presented earlier)
Annualized Costs
Rupture disk9. $112
Holders, etc. 250
Maintenance & Misc. 156
Total ($/year) $519
(2) Rupture disk?systems with
3-way valve.
Assembly
One 7.6cm stainless steel rupture disk $195
One 7.6cm carbon steel disk holder 325
One 0.6cm dial face pressure gauge 15
One 7.6cm safety/relief valve 1240
Two 7.5cm elbows 25
Subtotal $1825
Three-way Valve
One 7.6cm, 3-way, 2 port valve 1120
Installation 540
Subtotal $1660
Total Cost (last quarter 1978 dollars) $3485
, Annualized Costs
Rupture disk $113
Holder, valve, etc.f 536
Maintenance & Misc. 314
Total ($/year) $963
(3) 0-rings Total Costs = $200/relief valve
(last quarter 1978 dollars)
Annualized Costs
0-ringg , $33
Maintenance & Misc. 18_
Total ($/year) $51
5-2o
-------
TABLE 15. (CONTINUED)
(4) Closed vent system to $1960/relief valve [Based on installation
transport the discharge of 15m of 10.2cm diameter schedule 40
or leakage of safety/ carbon steel pipe. Cost (1967) = $700.
relief valves to a flare. Cost index = 278.1/113. Installation =
$60. One.,10.1cm carbon steel check valve
= $180.T
Annualized Costs
Vent System1 . $320
Maintenance & Misc. 176
Total ($/year) $496
aBased on 2 year equipment life and 10 percent interest (CRF = 0.58).
Based on 10 year equipment life and 10 percent interest (CRF = 0.163).
°Based on 9 percent of total capital costs.
Based on 2 year equipment life and 10 percent interest (CRF = 0.58).
p
Based on 10 year equipment life and 10 percent interest (CRF = 0.163).
Based on 9 percent of total capital costs.
ABased on 10 year equipment life and 10 percent interest (CRF = 0.163).
Based on 9 percent of total capital cost.
Based on 10 year equipment life and 10 percent interest (CRF = 0.163).
JBased on 9 percent of total capital cost.
5-26
-------
TABLE 5-16. COSTS FOR CLOSED LOOP SAMPLING SYSTEMS
One 6m length of 2.5cm diameter schedule 40, $190
carbon steel pipe and three 2.5cm carbon
steel ball valves.
Installation 270
Total/sampling connection (last quarter 1978 dollars) $460
Annualized Costs
Closed Loop Connection3 $75
Maintenance & Misc. 41
Total ($/year) $116
Based on 10 year equipment life and 10 percent interest (CFR = 0.163)
Based on 9 percent of total capital cost.
5-27
-------
5.5 OPEN-ENDED LINES
5.5.1 Technical Basis in the BID
Equipment costs were calculated for caps on open-ended lines. The
/
estimate was based on information in the text "Plant Design and Economics
9/:
for Chemical Engineers". The costs are shown in Table 5-17.
5.5.2 New Information
The Texas Chemical Council (TCC) has estimated that all open-ended
27
lines in new units are already capped as routine procedure.
5.5.3 Public Comments
One commenter disagreed with the cost estimates for caps for open-ended
lines. He felt that an average size of 3.8cm to 5cm (1.5in. to 2in.) was
more realistic than the 2.5cm estimate in the BID. The commenter estimated
an average cost for screwed valves of $150 each. He further stated that
many small lines are constructed of stainless steel instead of carbon steel.
Also, many lines, he said, 2.5cm or larger, are flanged, not screwed.
5.5.4 EPA's Conclusions
po
Information obtained from the Hydroscience (now ITE) study shows that
29
92 percent of open-ended lines are 5cm. or less in diameter. Based on
this information, the 2.5cm. estimate for the average valve size seems
reasonable. Larger lines would be expected to use blind flanges which cost
about the same as small valves.
As in the case of sampling systems, the equipment count for model units
has been corrected to reflect the new information regarding the current
.practice of capping open-ended lines in SOCMI units. The corrections are
discussed in Section 3 on Model Units.
5.6 COMPRESSORS
5.6.1 Technical Basis in the BID
The costs for both equipment control and leak detection programs were
calculated in the BID. These costs are presented below.
Equipment Costs: The cost of control equipment for compressors was based on
installation of closed vents for degassing reservoirs of compressors. The
estimate was based on information contained in the Hydroscience (now ITE)
30
report and was for the following items:
5-28
-------
TABLE 5-17. COSTS FOR AN OPEN-ENDED LINE CAP
One 2.5cm screwed valve $30
Installation 15
Total (last quarter 1978 dollars) $45
AnnualIzed Costs
Caps for Open-Ended Lines $7
Maintenance & Misc. _4
Total ($/year)$ll
aBased on 10 year equipment life and 10 percent interest (CRF = 0.163)
Based on 9 percent of total capital cost.
5-29
-------
122m length of 5.1cm diameter, schedule 40 $5200
carbon steel pipe
Three 5.1cm cast steel plug valves and one 1330
metal gauze flame arrestor
Total (last quarter 1978 dollars) $6530/compressor
The above costs include connection of the degassing reservoir to an existing
enclosed combustion device or vapor recovery header. Cost of a control
device added specifically to control the degassing vents, is therefore, not
included.
Leak Detection and Repair Costs: The factors affecting the costs of a leak
detection and repair program are monitoring time and repair time. Both
31
these estimates were based on information provided by Exxon Company, USA.
The estimates were (1) 20 man-minutes for monitoring time and
(2) 40 man-hours for repair time.
5.6.2 New Information
No new information was received regarding cost estimates for
compressors.
5.6.3 Public Comments
As with pumps, concern was expressed about the cost of existing control
device not being included in the analysis. Commenters said that even though
a control device may exist at a facility, it cannot be considered a
cost-free utility. Each user, therefore, should assume a share of the
capital and operating costs of the device.
5.6.4 EPA's Conclusions
As discussed in the section on pumps, no need is seen for the addition
of the cost of an existing combustion device since it is not an incremental
cost due to the regulation.
Also, as in the analysis for pumps, costs have been calculated for
degassing reservoir of compressors vented to flares as well as enclosed
combustion devices. These costs are presented below.
Equipment Costs: Two control levels are identified in order to provide a
range of control costs:
(1) Compressor seals provide no fugitive VOC emissions reduction;
5-30
-------
(2) Compressor seals provide 95 percent reduction of fugitive VOC
emissions. For the purpose of this cost analysis, it was assumed that
existing flares achieve 90 percent control while enclosed combustion devices
(including condensers, absorbers, etc.) achieve 95 percent control. No heat
recovery credit is given for VOC controlled by enclosed combustion device
(as discussed in Section 5.2.4 on Pumps). The cost analysis for compressors
is presented in Tables 5-18 and 5-19. The reasons for the difference in the
costs for flares and enclosed combustion devices were also previously
discussed in Section 5.2.4 on pumps.
The cost estimates presented in Tables 5-18 and 5-19 are for control of
an uncontrolled compressor seal; they should not be applied directly to the
model unit or the industry compressor population. As discussed in
Section 3, 60 percent of the compressors in the industry are assumed to be
controlled currently by effective systems. Therefore, in estimating costs
of controlling compressor seals on a model unit basis, and extrapolating to
the industry, 40 percent of the costs listed in these tables are applied to
the compressor seal counts for each model unit.
Leak Detection and Repair Costs: The costs of monthly and quarterly leak
detection and repair programs for compressors were presented in the BID. No
further calculations of costs for leak detection and programs for compres-
sors were performed. Further analysis seemed unnecessary for two major
reasons. First, a large percentage of the compressors tested in SOCMI had
control equipment installed. Second, if a compressors were found leaking at
10,000 ppm, the expected repair technique would be the installation of
control equipment. Furthermore, since compressors are not generally spared,
repair to a non-leaking status would have to be postponed until the next
turnaround, thereby eliminating almost completely the effectiveness of the
leak detection and repair program.
5.7 OTHERS
5.7.1 Technical Basis in the BID
The estimates of other input cost parameters used in the BID are
presented below.
5-31
-------
TABLE 5-18. EQUIPMENT COST FOR CONTROL OF EMISSIONS FROM
COMPRESSOR SEALS (last quarter 1978 dollars)
Degassing Reservoir
Degassing Reservoir Vented to an Enclosed
Vented to a Flare Combustion Device
Capital Costs
Vent System 1400a 6530b
Annual Costs
Vent System0 230 1060
Maintenance & .
Miscellaneous charges 130 590
Total6 360 1650
Based on installed capital cost of 20m of 5.1cm piping ($860), one 5.1cm
check valve ($125), and one 5.1cm block valve ($405) per unit system and
one vent per compressor seal.
bFrom Chapter 8 of the BID.
/"»
Based on 10 year equipment life and 10 percent interest (CRF = 0.163).
Based on 9 percent of total capital costs (Chapter 8 of the BID).
p
Costs do not include credits for recovered product.
5-32
-------
TABLE 5-19. NET ANNUALIZED COST OF EQUIPMENT FOR CONTROL OF EMISSIONS
FROM A COMPRESSOR SEAL (last quarter 1978 dollars)
Degassing Reservoir
Degassing Reservoir Vented to an Enclosed
Vented to a Flare Combustion Device
(1) (2) (1) (2)
95% Control No Control 95% Control No Control
by Seal by Seal by Seal by Seal
Gross annualized
cost ($/year) $360 360 1650 1650
Product recovery
credit ($/year) (1395)a 0 (1395)a 0
Net annualized
cost ($/year) (1035) 360 (255) 1650
Credit from 95 percent of uncontrolled fugitive VOC emissions that are
retained in the process due to operation of a pressure sensor between the
dual seals to immediately detect seal failure. Product recovery credit is
based on emissions from one uncontrolled compressor. Emissions from
uncontrolled compressor - 0.57 kg/hr. The composition of this emission
factor was discussed in Section 3 (Model Units). Product valued at
$300/Mg.
Note: Columns (1) and (2) show a range of net annualized costs depending
upon the degree of control by the seal.
5-33
-------
1. An hourly labor rate of $15 was used for 1978. This estimate
included wages plus 40 percent of wages for labor-related administrative and
overhead costs. These estimates were based on information presented in the
32
control technique guidelines (CTG) for Petroleum Refineries.
2. Administrative and support costs to implement the regulation were
estimated to be 40 percent of monitoring and maintenance labor. This
33
estimate was also based on the Refinery CTG.
3. Two monitoring instruments per model unit were assumed to be
required at a cost of $4250 per instrument. In addition, an annual
maintenance cost of $2700 was assumed for the monitoring instruments. These
34 35
estimates were provided by Century Systems and the Refinery CTG.
4. The product recovery credit was estimated to be $360 per Mg. The
credit estimate was based on the average price of all SOCMI chemicals and
was based on information obtained from the U.S. International Trade
Commission
5.7.2 New Information
No new information has been received regarding the above estimates.
5.7.3 Public Comments
1. Some commenters felt that labor costs had been underestimated. One
commenter said that the labor charge should be $18 per hour. This estimate
was said to be more representative of the Texas Gulf Coast experience. The
commenter cited the recent wage settlements including employee benefits
improvements in support of his arguments. Another commenter, while agreeing
that $15 per hour was quite accurate as a base labor rate, expressed concern
that overhead costs, labor benefits, social security taxes, and vacations
had not been included. He estimated that the resulting loaded labor cost
would be $25 per hour.
2. Disagreement was also expressed with the estimates of monitoring
instruments costs. One commenter verified the cost of the instruments by
contacting the manufacturer, but disagreed with the estimate of the number
of instruments which would be required. He felt that one instrument for
1200 monitoring points with one for a spare was a good estimate. But he
estimated that a large model unit with over 3000 points to monitor would
5-34
-------
require four instruments, thereby doubling EPA's cost estimate for the large
units.
3. Another commenter was concerned that the cost of instrument
calibration and maintenance had not been considered. Looking at EPA
contractor data for SOCMI plants, he said that about 20 percent of the total
time expended by a contractor crew was devoted to instrument calibration and
maintenance.
4. Finally, one commenter said that the product recovery credit was
overstated. He estimated $220 per Mg instead of $360 per Mg. The commenter
said that the average market value of $360 was based on very pure finished
products. Emissions reduction will occur on new materials and semi-finished
streams which have a lower product value.
5.7.4 EPA's Conclusions
Considering the fact that all BID estimates are in 1978 dollars, EPA
feels that $15 per hour for labor costs is adequate. Also, it is not likely
that the monitoring and maintenance team would be at the upper end of the
industry wage scale. For workers at the lower end of the wage scale the
estimate probably overstates the labor cost. It should also be noted that
the $15 estimate includes allowance for overhead and administrative costs
equal to 40 percent of the wages. Based on a review of the wage information
for 1978, $11 per hour for direct wages and $4 per hour for overhead and
administrative costs seems quite reasonable.
Even the most complex facilities are not expected to need more than two
monitoring instruments. For example, the largest number of valves expected
in a facility would be about 6,000. This means that for such a facility a
maximum of 6,000 valves would be monitored during a given month. At the
rate of 1 minute monitoring time per valve it will take 100 hours to monitor
all 6,000 valves. This corresponds to 2 1/2 work weeks. Two monitoring
instruments, therefore, will be sufficient.
As shown in Chapter 8 of the BID an allowance has been made for
material and labor for monitoring instrument calibration and maintenance.
This allowance is equal to $2700 per year. In addition, as discussed in
Section 5.1.1 for valves, actual field tests in the 24 unit study have shown
b-35
-------
that the present monitoring time estimates include sufficient time for
instrument calibration and maintenance.
The average product value used in the cost analysis is based on an
average of costs for all SOCMI chemicals (including raw materials and
finished products), weighted by the corresponding production volume of that
chemical. Thus, the higher volume chemicals, which are generally the lower
value chemicals, have the heaviest impact on the average product value. On
the average, though, this estimate should closely reflect the value of
products saved by emissions reduction because, if chemicals are not lost at
various stages of the process, they would become higher-valued products at a
very minimal incremental cost. In reviewing this estimation procedure, an
error in the mathematics was found and corrected. Therefore, the recovered
product value was revised to $300 per Mg.
5-36
-------
ERRATA
Page 1 of 3
5.8 REFERENCES
1. Letter and attachments from Johnson, J. M., Exxon Company, to
Walsh, R. T., EPA:CPB. July 28, 1977. 14p. Review of "Control of
Hydrocarbon from Miscellaneous Refinery Sources" report.
2. Blacksmith, J. R., et al. (Radian Corporation.) Frequency of Leak
Occurrence for Fittings in Synthetic Organic Chemical Plant Process
Units. (Prepared for U. S. Environmental Protection Agency.) Research
Triangle Park, N. C. Publication No. EPA 600/2-81-003.
September 1980. p.6.
3. Lee, K. C., et al. (Union Carbide Corporation.) A Fugitive Emission
Study in a Petrochemical Manufacturing Unit. Presented at the 73rd
Annual Meeting of the Air Pollution Control Association. Montreal,
Quebec. June 22-27, 1980. 16p.
4. Reference 2.
5. Letter and attachments from Ballard, B. F., Phillips Petroleum Company,
to White, J. C., EPA:Region IV. September 8, 1977. "Phillips
Petroleum Company Sweeney Refinery and NGL Process Center Crude
Processing Expansion and Modernization Fugitive Emission Control
Program" report.
6. State of California Air Resources Board. Emissions from Leaking
Valves, Flanges, Pump and Compressor Seals, and Other Equipment at Oil
Refineries. April 1978. p. V-18.
7. Reference 1.
8. Reference 6.
9. Harvey, C. M. and A. C. Nelson. (PEDCo Environmental, Inc.) VOC
Fugitive Emission Data - High Density Polyethylene Process Unit.
(Prepared for U. S. Environmental Protection Agency.) Research
Triangle Park, N. C. Publication No. EPA-600/2-81-109. June 1981.
10. Langley, G. J. and R. G. Wetherold. (Radian Corporation.) Evaluation
of Maintenance for Fugitive VOC Emission Control. (Prepared for
U. S. Environmental Protection Agency.) Research Triangle Park, N. C.
Publication No. EPA-600/2-81-080. May 1981. p. 81.
11. Reference 10.
12. Reference 1.
13. Williamson, H. J., et al. (Radian Corporation.) Model for Evaluating
the Effects of Leak Detection and Repair Programs on Fugitive
Emissions. Technical Note DCN 81-290-403-06-05-03. September 1981-.
5-37
-------
ERRATA
Page 2 of 3
14. En'ckson, D. G. and V. Kalcevic. (IT Enviroscience.) Fugitive
Emissions in: U. S. Environmental Protection Agency. Organic
Chemical Manufacturing. Volume 3: Storage, Fugitive and Secondary
Sources, Research Triangle Park, N. C. Publication No.
EPA-450/3-80-025. December 1980. Report 2, p. V-9.
15. Reference 1.
16. Reference 1.
17. Telecon. Pat Patterson, Chesterton Seals, with Syed Shareef, Radian
Corporation. January 7, 1982. Information about credit for old pump
seals.
18. Ramsden, J. H. How to Choose and Install Mechanical Seals. Chemical
Engineering (85)(22):97-102. October 9. 1978.
19. Reference 1.
20. Reference 14.
21. Letter and attachments from Kronenberger, L., Exxon Company, USA to
Goodwin, D. R., EPA:ESED. February 2, 1977. Response to EPA request
for information on miscellaneous hydrocarbon emission sources from
refineries.
22. Memo from Cole, D. G. (Pacific Environmental Services.) to
Hustvedt, K. C., EPArCPB. July 29, 1981. 4p. Estimated Costs for
Rupture Disk Systems with a 3-Way Valve.
23. Richardson Engineering Service, Inc. The Richardson Rapid Construction
Cost Estimating System. San Marcos, California, Volume 3, 1981.
24. Reference 14.
25. Meeting Report. Dimmick, F., EPA:SDB, to Wyatt, S., EPArSDB.
September 4, 1980. lOp. Minutes of meeting between EPA and Texas
Chemical Council representatives about TCC comments on recommended NSPS
for fugitive VOC emissions in SOCMI.
26. Peters, M. S. and K. D. Timmerhaus. Plant Design and Economics for
Chemical Engineers, Second Edition. New York, McGraw-Hill Book
Company, 1968.
27. Reference 25.
28. Reference 14.
5-33
-------
ERRATA
Page 3 of 3
29. Memo from Hustvedt, K. C., EPArCPB, to file. January 5, 1982. Summary
of HI Data Compiled on 15 July 1980: Open-ended Valves By Size. 9 p.
30. Reference 14.
31. Reference 1.
32. U. S. Environmental Protection Agency. Control of Volatile Organic
Compound Leaks from Petroleum Refinery Equipment. Research Triangle
Park, N. C. Publication No. EPA-450/2-78-076. June 1978.
33. Reference 32.
34. Letter and attachments from Amey, G. C., Century Systems Corporation to
Serne, J., Pacific Environmental Services. October 17, 1979. Cost
data for VOC monitoring instrument.
35. Reference 32.
36. Letter from Smith, V. H., Research Triangle Institute, to
Honerkamp, R., Radian Corporation. November 30, 1979. Ip.
Information about baseline projections.
5-39
-------
5.8 REFERENCES
1. Letter and attachments from Johnson, J.M., Exxon Company, to
Walsh, R.T., EPArCPB. July 28, 1977. 14p. Review of "Control of
Hydrocarbon from Miscellaneous Refinery Sources" report.
2. Blacksmith, J.R., et al. (Radian Corporation.) Frequency of Leak
Occurrence for Fittings in Synthetic Organic Chemical Plant Process
Units. (Prepared for U.S. Environmental Protection Agency.) Research
Triangle Park, N.C. EPA Contract No. 68-02-3171. September 1980. p.6.
3. Erikson, D.G. and V. Kalcevic. (Hydroscience, Inc.) Emissions Control
Options for the Snythetic Organic Chemicals Manufacturing Industry.
(Prepared for U.S. Environmental Protection Agency.) Research Triangle
Park, N.C. EPA Contract No. 68-02-2577. February 1979. p.B-12.
4. Reference 2, p.7.
5. Reference 3.
6. Langley, G.J. and R.G. Wetherold. (Radian Corporation.) Evaluation of
Maintenance for Fugitive VOC Emission Control. (Prepared for
U.S. Environmental Protection Agency.) Research Triangle Park, N.C.
EPA Contract No. 68-03-2776. May 1981. p.81.
7. Reference 6, p.82.
8. Reference 3, pp. IV-3, 8, 9.
9. Reference 1.
10. Reference 1.
11. Reference 1.
12. Reference 3, p.IV-8.
13. Meeting Report. Dimmick, F., EPA:SDB, to Wyatt, S., EPAtSDB.
August 11, 1980. lOp. Minutes of meeting between EPA and Texas
Chemical Council representatives about TCC comments on recommended NSPS
for fugitive VOC emissions in SOCMI.
14. Peters, M.S. and K.D. Timmerhaus. Plant Design and Economics for
Chemical Engineers, Second Edition. New York, McGraw-Hill Book Company,
1968.
15. Reference 13.
16. Reference 3, data sheets.
5-37
-------
17. Memo from Hustvedt, K.C., EPA:CPB, to files. July 15, 1980. Ip.
Summary of Hydroscience survey data.
18. Reference 3, pp. IV-8, 9.
19. Reference 1.
20. U.S. Environmental Protection Agency. Control of Volatile Organic
Compound Leaks from Petroleum Refinery Equipment. Research Triangle
Park, N.C. Publication No. EPA-450/2-78-036. June 1978. pp.4-45.
21. Reference 3, p. IV-3.
22. Reference 3, pp. IV-9, 10.
23. Letter and attachments from Amey, G.C., Century Systems Corporation, to
Serne, J., Pacific Environmental Services. October 17, 1979. 3p. Cost
data for VOC monitoring instrument.
24. Reference 3, p. IV-9.
25. Letter from Smith, V.H., Research Triangle Institute, to Honerkemp, R.,
Radian Corporation. November 30, 1979. Ip. Information about baseline
projections.
5-38
-------
APPENDIX A
METHODOLOGY FOR ECONOMIC ANALYSIS
Appendix A presents information used in responding to public comments
relating to the potential impact of fugitive emission standards for the
synthetic organic chemicals manufacturing industry (SOCMI). Section A.I
describes the weighting procedures used to estimate historical production,
quantity of sales, and value of sales data. The weights are based on a 1978
International Trade Commission Report. The prices of SOCMI Chemicals during
1978 and the computation of the average price of SOCMI Chemicals are also
explained in Section A.I. The methodology used to project SOCMI replacement
investment is described in Section A.2. In section A.3, the process used to
estimate the cost of capital is described. Section A.4 explains the
methodologies and assumptions used to estimate the price and rate of return
impacts. Section A.5 lists the references.
A.I ESTIMATION OF SOCMI PRODUCTION, SALES, AND PRICE VALUES
SOCMI chemicals may be used as primary feedstocks, chemical inter-
mediates, or end use chemicals. Primary feedstocks are produced from crude
raw materials and used in the manufacture of other chemicals. Chemical
intermediates are products of primary feedstocks and are also used to
produce other chemicals. End use chemicals are products of chemical
intermediates and/or primary feedstocks and are used either as final goods
or as inputs to production processes outside the chemical industry. Data on
production, quantity of sales, and value of sales for the synthetic organic
chemicals are reported annually by the United States International Trade
Commission (formerly known as the United States Tariff Commission) in the
report, Synthetic Organic Chemicals: U.S. Production and Sales. Data on
production, quantity of sales, and value of sales of the SOCMI chemicals for
which data are available are reproduced in Table A-l from the ITC report for
the year 1978.
Reasons for the nonavailability of data on production, quantity of
sales, and value of sales for SOCMI chemicals not listed in Table A-l are
outlined in the Introductory Section, pages 1 and 2, of the ITC report.
A-l
-------
TABLE A-l. U.S. PRODUCTION AND SALES OF SYNTHETIC'
ORGANIC CHEMICALS, 1978
Sales
Price
Chemicals
I. Primary Products
from Petroleum and
Natural Gas for
Chemical Conversion
Benzene5 (1° and 2°)
Cumene
Cyclohexane
Ethyl benzene
Naphthalene, all
grades
Styrene.
Toluene ,
Xylenes, mixed
o-Xylene
p-Xylene
All other aromatics
and naphthenes
Acetylene
Ethyl ene
Propylene
Butadiene and buty-
lene fractions
1,3-Butadiene, grade
for rubber (elasto-
mers)
1-Butene
Isobutylene, 2-butene
and mixed butyl enes
Isoprene (2-Methyl-l,
3-butadiene)
Production
(1,000 Ib)
10,503,883
3,380,322
2,331,665
8,385,482
156,801
7,186,193
7,542,434
6,412,745
1,013,131
3,515,869
4,123,482
245,670
25,954,627
13,013,529
482,789
3,515,206
91,655
909,301
184,117
Quantity
(1,000 Ib)
5,144,345
1,579,161
2,194,849
365,102
84,642
2,882,387
4,948,978
3,375,066
939,180
2,178,563
3,526,357
8,784,177
5,679,958
261,011
2,505,025
75,867
597,712
116,969
$1,000
518,323
174,249
245,860
38,616
13,234
500,786
377,409
249,435
103,775
269,139
163,977
1,095,624
526,528
50,984
486,944
13,793
65,925
18,195
$
per Ib
0.101
.110
.112
.106
.156
.174
.076
.074
.110
.124
.047
.125
.093
.195
.194
.182
.110
.156
$
per Mg
220
240
250
230
340
380
170
160
240
270
100
*
280
210
430
430
410
240
340
(continued)
A-2
-------
TABLE A-l. (Continued)
Sales
Price
Chemicals
Dodecene (Tetrapropy-
lene)
Nonene (Tripropylene)
Subtotal
II. Cyclic Intermediates
Anil ine (Anil ine oil )
Benzoic acid, tech
Biphenyl
Cresols, total
Cresylic acid,
refined
Cyclohexanone
o-Dichlorobenzene
p-Dichlorobenzene
a-Methylstyrene
Nitrobenzene
Nonyl phenol
Phenol , total
Salicylic acid, tech
Terephthalic acid,
dimethyl ester
Toluene-2 ,4-diamine
(4-m-Tolylene-
diamine)
Subtotal
III. Miscellaneous Cyclic
and Acylic Chemical
Benzyl alcohol
Caprolactam
p-Hydroxybenzoic acid,
methyl ester
Production
(1,000 Ib)
359,705
431,525
99,740,131
605,772
85,175
63,527
96,869
45,003
1,161,712
41,140
41,224
75,571
575,523
125,167
2,681,603
47,149
5,954,216
139,250
11,738,901
s
8,572
918,660
Quantity
(1,000 Ib)
97,144
283,539
45,620,032
187,767
36,822
17,450
94,932
46,078
36,992
44,028
38,062
61,176
20,192
49,294
1,431,536
6,335
2,070,664
5,642
. . .
848
$1,000
12,018
30,805
4,955,619
41,865
8,686
4,224
47,630
12,991
12,226
11,810
10,311
10,780
4,302
11,177
231,622
5,418
. . .
413,042
5,400
. . .
2,320
-* - ' F
$
per 1 b
.124
.109
.109
.220
.240
.24
.50
.28
.33
.27
.27
.18
.21
.23
.16
.86
0.200
.96
. . .
2.74
$
per Mg
270
240
240
490
530
530
1,100
620
730
600
600
400
460
510
350
1,900
. . .
440
210
. . .
6,040
(continued)
A-3
-------
TABLE A-l. (Continued)
Sales
Price
Production
Chemicals (1,000 Ib)
Maleic anhydride
Butylamines, total
n-Butylamine, mono-
Di-n-butylamine
All other butyl amines
Dimethyl ami ne sulfate
Ethyl amines, total
Diethylamine
Monoethylamine
Triethylamine
Isopropylamine, mono-
Ethanol amines, total
Acetonitrile
Acrylonitrile
Acetic acid, 100%
Acetic anhydride, 100%
Acrylic acid
Adi pic acid
Fumaric acid
Propionic acid
Butyraldehyde
Formaldehyde (37% by
weight)
Acetone:
From cumene
From isopropyl
alcohol
2-Butanone (Methyl
ethyl ketone)
4-Methyl -2-pentanone
(Methyl isobutyl
ketone)
4-Methyl -3-penten-2-
one (Mesityl oxide)
All other
n-Butyl alcohol (n-
Propylcarbinol )
341,127
55,804
4,097
4,921
46,768
5,904
65,623
15,169
36,151
14,303
45,844
362,027
44,942
1,752,302
2,775,520
*
325,318
1,621,219
27,993
83,078
782,653
6,380,959
2,051,811
467,602
660,835
232,691
33,143
312,586
755,855
Quantity
(1,000 Ib)
271,469
51,442
4,103
4,009
43,330
48,021
5,712
31,950
10,359
41,886
320,236
586,816
823,274
132,078
46,503
24,402
62,848
68,168
2,241,958
1,083,662
456,699
669,341
155,944
15,797
101,384
395,494
$
$1,000 per Ib
66,405
26,927
2,318
2,466
22,143
24,488
3,458
13,244
7,786
14,824
109,401
134,965
120,263
32,032
15,058
. *
9,941
10,879
11,962
105,917
145,869
71,366
127,007
42,703
4,991
35,560
66,985
.24
.52
.57
.62
.51
*
.51
.61
.41
.75
.35
.34
.23
.15
.24
.32
.
.41
.17
.18
.05
.13
.16
.19
.27
.32
.35
.17
$
per Ib
530
1,150
1,260
1,370
1,120
1,120
1,340
900
1,650
770
750
510
330
530
710
900
370
400
110
290
350
420
600
710
770
370
(continued)
A-4
-------
TABLE A-l. (Continued)
Sales
Price
Production
Chemicals (1,000 Ib)
Methanol , synthetic
n-Butyl acetate,
unmixed
Ethyl acetate (85%)
Ethyl acrylate
Isobutyl acetate
Vinyl acetate
All other
Ethylene glycol
Glycerol , synthetic
only
2-Methoxyethanol
(Ethylene glycol
monomethyl ether)
2-(2-Methoxyethoxy)
ethanol (Diethylene
glycol monomethyl
ether)
Polyethylene glycol
Polypropylene glycol
Triethylene glycol
Carbon tetrachloride
Chloroethane (Ethyl
chloride)
Chloroform
Chloromethane (Methyl
chloride)
1 ,2-Dichloroethane
(Ethylene
di chloride)
Dichloromethane
(Methylene chloride)
1,2,-Dichloropropane
(Propylene
di chloride)
6,443,242
122,106
181,944
299,306
*
1,691,969
1,504,442
3,903,889
133,907
114,381
17,066
89,849
26,829
119,944
737,030
539,793
349,169
453,810
11,000,619
570,098
74,112
Quantity
(1,000 Ib)
3,080,747
128,161
177,973
147,541
48,433
942,659
718,456
3,137,188
116,612
106,400
12,567
84,603
20,871
83,901
363,406
159,079
302,114
200,797
1,033,313
490,678
33,382
$
$1,000 per Ib
181,027
32,790
33,744
44,295
12,409
167,107
287,853
546,690
54,448
29,869
3,816
31,074
7,799
26,110
41,698
23,428
53,423
28,977
82,645
114,342
2,120
7
.06
.26
.19
.30
.26
.18
.40
.17
.47
.28
.30
.37
.37
.31
.11
.15
.18
.14
.08
.23
.06
$
per Mg
130
570
420
660
570
400
880
370
1,040
620
660
820
820
680
240
330
400
310
180
510
130
(continued]
A-5
-------
TABLE A-l. (Continued)
Sales
Price
Chemicals
Production Quantity $ $
(1,000 Ib) (1,000 Ib) $1,000 per Ib per Mg
Tetrachl oroethyl ene
(Perchloroethylene) 725,457
1,1,1-Trichloroethane
(Methyl chloroform) 644,475
Trichl oroethyl ene 298,986
Vinyl chloride, monomer
(Chloroethylene) 6,941,123
All other chlorinated
hydrocarbons 1,208,509
Chlorodifluoromethane
(F-22) 205,612
Di Chlorodifluoromethane
(F-12) 327,097
Tri chl orof 1 uoromethane
(F-ll) 193,735
Carbon disulfide 476,175
Epoxides, ethers, and
acetals, total 7,584,748
Ethylene oxide 5,012,419
Ethyl ether, absolute 12,098
Propylene oxide 2,046,843
All other epoxides,
ethers, and acetals 513,388
Phosgene (Carbonyl
chloride) 1,296,941
Subtotal 75,973,179
Grand Total 187,452,211
aData presented here are reproduced from
549,111
631,243
298,557
4,885,688
94,649
139,797
316,864
166,898
375,962
1,658,547
525,113
»
1,133,434
29,768,656
77,459,352
53,589
135,388
46,588
618,407
31,756
106,236
134,743
57,341
31,645
400,521
123,079
*
277,442
. .
5,007,662
10,376,323
.10
.21
.16
.13
.34
.76 1
.43
.34
.08
.24
.23
.24
0.168
0.134
220
460
350
290
750
,670
950
750
180
530
510
*
530
370
295
Table 1 of
Synthetic Organic Chemicals, U.S. Production and
Sales, 1978,
United
States
International Trade Commission, Washington, DC20436.
""Production and sales of benzene, toluene, and xylenes by coke-oven
operators represent less than 2 percent of total benzene, toluene, and
xylenes production and sales and are not included in the table.
A-6
-------
First, the ITC reports data for only those chemicals produced by more than
three producers; second, producers report data for only those chemicals for
which the volume of production exceeds 1,000 pounds or the value of sales
exceeds $1,000; and third, many chemicals are manufactured by the industry
as intermediate products to produce other chemicals and are never sold in
the market.
For the year 1978, the totals of production, quantity of sales, and
value of sales of SOCMI chemicals listed in Table A-l are 85 million Mg,
35 million Mg, and $10 billion, respectively. The average price of SOCMI
chemicals for the year 1978 is computed as about $300/Mg (total value of
sales/total quantity of sales). The average price of primary SOCMI products
from petroleum and natural gas for the year 1978 is $240 per Mg, which is
smaller than the average prices of "cyclic intermediates" ($440 per Mg) and
"miscellaneous cyclic and acyclic" ($370 per Mg) SOCMI chemicals. Prices of
individual chemicals such as benzene, ethylene, propylene, phenol,
formaldehyde, and vinyl chloride are $220, $280, $210, $390, $110, and
$290 per Mg, respectively. The prices of individual SOCMI chemicals during
1978 ranged between $100 and $6,040 per Mg.
The average price of all chemicals listed in the ITC report under the
three major categories "Primary Products from Petroleum and Natural Gas,"
"Cyclic Intermediates," and "Miscellaneous Cyclic and Acyclic" is computed
as $350 per Mg. SOCMI chemicals listed in Table A-l represent more than
70 percent of the production of all chemicals listed in the ITC report.
To derive appropriate estimates of the historical data (1968-1977),
production, quantity of sales, and value of sales in 1978 were computed for
all the SOCMI chemicals included in each of the three categories. The com-
puted sums of production, quantity of sales, and value of sales of SOCMI
chemicals were divided, respectively, by the corresponding totals for all
chemicals listed in the ITC report to get fixed weights for the three
categories. These fixed weights were used to compute estimates of
production, quantity of sales, and value of sales of chemicals in each
category over the period 1975-1978. For example, for 1975, the production
of SOCMI chemicals was estimated by a weighted sum of 1975 production data
A-7
-------
on Primary Products from Petroleum and Natural Gas for Chemical Conversion,
Cyclic Intermediates, and Miscellaneous Cyclic and Acyclic Chemicals, as
reported in the 1975 ITC report. Table A-2 presents the estimated ratios
used to weight the ITC data.
Prior to 1975, the chemicals included in the category Miscellaneous
Cyclic and Acyclic Chemicals were reported as Miscellaneous Synthetic
Organic Chemicals. A weighting scheme based on 1974 data for this category
was developed using the procedure described above and was used to estimate
production and sales of SOCMI chemicals in this category for the period
1968-1974. Data on production and sales of SOCMI for the remaining two
categories for the period 1968-1974 were estimated using the 1978 weights.
A. 2 REPLACEMENT INVESTMENT PROJECTIONS
The projections are based on two key theoretical assumptions: (I) the
historical growth rate of capacity, p, has been constant over time; and (II)
model units have a fixed life of L years. These assumptions are summarized
in the following equations:
KT = (1+P) KT_L (1)
IT = PKT + RT (2)
RT = IT_L (3)
where
K = industry capacity,
I = gross investment,
R = replacement investment,
T = time subscript,
and K, I and R are measured in terms of model units.
Equation (1) is an algebraic restatement of assumption I. Equation (2)
is simply a mathematical definition of gross investment, that is, gross
investment, IT, is equal to additions to new capacity, pIC, plus replacement
investment, R,.. Equation (3) is an algebraic restatement of assumption II.
A-8
-------
TABLE A-2. WEIGHTS USED TO ESTIMATE HISTORICAL PRODUCTION,
SALES AND PRICES OF SYNTHETIC ORGANIC CHEMICALS3
Primary Products from
Petroleum and Natural Gas
Total SOCMI Chemicals0
ITC Grand Total0
Weight
Cyclic Intermediates
Total SOCMI Chemicals0
ITC Grgnd Total0
Weight
Miscellaneous Cyclic
and Acyclic Chemicals
Production
(Mg)
45,241,826
58,489,843
77%
5,324,731
9,042,805
59%
Quantity of
Sales
(Mg)
20,693,111
29,157,773
71%
939,249
4,015,536
23%
Value of
Sales
($ 1,000)
4,955,619
6,159,507
80%
413,042
2,803,327
15%
Price5
($/Mg)
240
210
440
700
Total SOCMI Chemicals0 34,461,208 13,502,974 5,007,612 370
ITC Grand Total 41,776,760 17,660,815 8,581,663 490
Weight6 82% 77% 58%
Miscellaneous Chemicals
1974 Figures
Total SOCMI Chemicals0 33,416,656 15,793,169 3,607,825 230
ITC Grand Total0 45,633,845 21,514,545 7,815,487 360
Weight6 73% 73% 46%
1978 figures except where indicated.
Price is computed by dividing the value of sales by the quantity of sales.
°Total SOCMI chemicals consist of all the SOCMI chemicals listed in
Table A-l.
ITC Grand Total includes SOCMI chemicals and other chemicals listed in
Table 1 of the respective categories in the ITC reports.
eWeight is computed as the ratio of total SOCMI chemicals to ITC grand
total.
Source: Synthetic Organic Chemicals: U.S. Production and Sales, 1978,
1974, United States International Trade Commission, Washington, DC
20436.
A-9
-------
Appropriately lagging Equation (2) and back substituting from (2) into (3),
it can be shown that
oo
RT = p E KT_iL (4)
1=1
Further, by substituting for the various 1C -. in Equation (4) using
Equation (1), and rearranging terms, the following result is obtained:
RT = KT_L I p_ (5)
1=0 (1+p) 1L
The expression I p is a constant and, if p is assumed to be 0.06
i=0
and L to be 20, approximately equal to 0.087. Equation (5) can be used to
project replacement investment in any year, T, if an estimate of the capital
stock in the (T-L)th year is available. For SOCMI , capital stock data are
available for 1976. This information, together with an assumed historical
growth rate of 6 percent, was used to estimate the capital stock for the
years 1961 to 1965 by means of Equation (1). The resulting capital stock
estimates are then used in Equation (5) to project replacement investment in
SOCMI for each of the five years following proposal of any regulatory
alternatives (1981-1985), on the basis of the empirical assumption that each
model unit has a life of 20 years. The annual projections of replacement
investment are then summed to obtain a projection of the number of
replacement facilities subject to the provisions of any regulatory alterna-
tive in the fifth year following its proposal. The projections of
replacement investment obtained by applying this methodology are presented
in Table A-3.
A-10
-------
TABLE A-3. PROJECTIONS OF REPLACEMENT INVESTMENT
Number of replacement capacity units
Year
1981
198?
1983
1984
1985
Annual
49
51
55
58
61
Cumulative
49
100
155
213
274
A.3 METHODOLOGY FOR COMPUTING COST OF CAPITAL TO SYNTHETIC ORGANIC CHEMICAL
MANUFACTURERS
The cost of capital for any new project is the cost of equity, debt,
and preferred stock, weighted by the percentage of funds generated by each
type of financing, that is,
k = k i + k. ^ + k ^ (1)
c e I 1 I p I
where
k = cost of capital
k = cost of equity capital
k. = cost of debt capital
k = cost of preferred stock capital
E = the amount of equity used to finance a given investment
A-ll
-------
D = the amount of debt used to finance a given investment
P E the amount of preferred stock used to finance a given
investment
I E the total funds needed for the investment
The k variables are interest rates representing the aftertax return on
investment that is needed to pay stock dividends and interest on debt. Each
k term is a nominal interest rate in that it contains an implicit allowance
for inflation. However, the cost of capital computed with equation (1) is
treated in the text as the real dollar interest rate that would prevail in
times of economic stability. The nominal rate is used as though it were a
real rate partly to ensure that estimates of the cost and other adverse
economic effects of investment in air pollution controls will be biased
upward rather than downward, and partly to avoid miscalculations that could
result from using the wrong inflation rate to convert the nominal rate to a
real rate.
The first step in estimating Equation (1) is to determine the relevant
weights for the three types of financing. It is assumed that the proportion
of debt, equity, and preferred stock to be used on any new project will be
the same as currently exists in the firm's capital structure. This implies
that the firm is currently using the optimal mix of financing. Figures for
the three types of funds came from the COMPUSTAT tapes, supplied by Standard
& Poor's Corporation, for each firm's fiscal year ending in 1977. Common
equity included the par value of common stock, retained earnings, capital
surplus, self-insurance reserves, and capital premium, while debt included
all obligations due more than a year from the company's balance sheet date.
Preferred stock represented the net number of preferred shares outstanding
at year-end multiplied by the involuntary liquidating value per share.
The next step in calculating Equation (1) is to estimate the cost of
equity financing. Two approaches are commonly used: the results derived
from the capital-asset pricing model (CAPM) and the results derived from the
A-12
-------
dividend capitalization model (DCM). The CAPM examines the necessary
returns on a firm's stock in relation to a portfolio comprised of all
existing stocks, while the DCM evaluates the stream of dividends and the
discount rate needed to arrive at the firm's existing share price. The
required return on equity using the CAPM is:
ke = i * 6 (km-i) (2)
where
i = the expected risk free interest rate
the expected excess return o
the firm's beta coefficient.
k -i 5 the expected excess return on the market, and
The beta coefficient is an historical measure of the extent to which a
firm's stock price fluctuates in relation to an index of the stock market as
a whole. 8 takes on a value of zero for a stock whose price is constant, a
value of one for a stock whose price follows the same path as an index of
the whole stock market, and a value of greater than (less than) one for a
stock whose price fluctuates more (less) dramatically than does the general
index. The CAPM is thus a modified regression equation in which g is the
slope of a straight line relating k and k .
The required return on equity using the DCM is:
1
ke = P0 +9 (3)
where
D, = the dividend expected in period 1
P E the share price at the beginning of period 1
g = the expected rate of dividend growth, assumed to be constant.
A-13
-------
The DCM is developed on the assumptions that (1) the price of a stock is the
present value of anticipated dividends, and that (2) these dividends grow
each year by a fixed percentage that is less than the required return on
equity.
Figures for Equation (2) were developed in the following manner. The
expected risk-free rate was assumed equal to the yield on a 3-month Treasury
Bill, as reported in the October 1, 1979, Wall Street Journal. The current
yield was 10.46 percent. This corresponds to the yield from a bond with no
possibility of default and offering no chance of a capital loss and is
therefore riskless. The firm's beta coefficients came from the
September 24, 1979, Value Line Investment Survey. The expected excess
return equalled 2.9646 percent, the 5-year average (July 1974-June 1979) of
the monthly excess returns on the Standard & Poor's 500 Stock Index
multiplied by twelve.
Figures for Equation (3) came from two sources. Both share price and
expected yearly dividends came from figures reported in the October 1, 1979,
Wall Street Journal. The growth rate was calculated from data contained on
the COMPUSTAT tapes. Note that the use of historical data does not
necessarily make the estimated rate of return on capital inconsistent with
the fourth quarter 1978 cost data used in this study as both short- and
long-term interest rates are currently in a state of flux. Three different
growth rates were tried: the 5-year average growth of total assets, the
5-year average growth of per share earnings, and the 5-year average growth
of dividends.
A number of theoretical reasons exist for preferring the CAPM approach
to the DCM for estimating the required return on equity, but the figures
calculated revealed a more practical justification. Using growth estimated
from per share earnings or dividends resulted in a number of firms having
negative required returns with the DCM method. Although using the growth in
assets resulted in only one firm with a negative required return, several
firms had extremely low returns (less than 10 percent). It is unreasonable
to expect that stockholders would demand a return on their stock that is
less than the existing yield on Treasury Bills, yet all three variants of
A-14
-------
the DCM method led to this conclusion for a number of firms. From these
considerations it was decided to use the CAPM calculations as the required
return on equity.
The third step in estimating Equation (1) is calculating the cost of
debt financing. This would be a relatively easy estimation if interest
rates did not change over time. Past yields on old issues of bonds would
suffice. Since interest rates have been increasing, it was felt that a more
forward-looking rate was required. The method selected was to take the
average yield as given in the September 3, 1979, Moody's Bond Survey for the
firm's bond ratings class as the necessary yield the firm must offer on
long-term debt. The firm's ratings class came from the September 1979
Moody's Bond Record or the 1979 Moody's Industrial Manual. A small number
of firms were not rated by Moody's. One firm was ranked in Standard and
Poor's Bond Guide and this was used to approximate a Moody1s bond class.
For other firms, data concerning bank notes, revolving credit, or term-loan
agreements that tied the interest rate on these types of debt to the current
prime rate were obtained from 1979 Moody's Industrial Manual or the
Standard and Poor's Corporation Records. This data were taken to measure
the necessary yield on long-term debt for such firms. Table A-4 presents
TABLE A-4. YIELDS BY RATING CLASS FOR COST OF DEPT FUNDS, 1979
(prime rate = 13.50 %)
Ratings Class Yield (percent)
AAA 9.25
AA 9.59
A 9.72
BAA 10.38
BA 11.97
B 12.395
A-15
-------
the yields by ratings class and the prime rate (as of September 1, 1979)
used for the cost of debt funds.
The yield on long-term debt does not represent the aftertax cost of
debt financing since interest charges are tax deductable. To arrive at the
aftertax cost of debt capital, the yield must be multiplied by 1 minus the
marginal tax rate.
ki
= k(l - t)
where
k = the yield on bonds, and
t = the marginal tax rate.
It is assumed that the firms in the sample are profitable, so that taxes
must be paid, and that their marginal tax rate is 48 percent.
The last step in estimating Equation (1) is to arrive at the cost of
preferred stock financing. Unlike debt, preferred stock does not have a
maturity date, so that the current yield should approximate the yield on new
issues. The yield is:
where
D = stated annual dividend, and
P = the price of a share of preferred stock.*
Note that as preferred stock dividends do not increase over time the growth
factor required in the discounted cash flow model (equation 3) is omitted
here.
A-16
-------
The figures for dividends and share price came from the October 1, 1979,
Wai 1 Street Journal or, if not included in this source, from the January 1,
1979, listing in the Daily Stock Price Record. A number of firms did not
have their preferred stock listed in either source, yet had preferred stock
in their capital structures. All used less than 15 percent preferred stock,
with the majority less than 5 percent. For these firms the aftertax yield
on preferred stock was set equal to the pretax yield on long-term debt.
Table A-5 lists the cost of capital for all 100 firms in the sample,
and also includes some of the components of equation (1). These firms
represent the best available sample of the approximately 600 firms in the
industry. However, it is likely that on the average the firms included in
the sample are larger than the firms excluded, as many small firms do not
have to publish detailed financial records. This potential sample bias may
have resulted in a slight underestimate of the industry's cost of capital
because larger firms are often able to acquire investment funds more cheaply
than smaller firms. This is because larger firms are usually able to reduce
their transactions costs of borrowing and because they represent a less
risky investment due to product, diversification than small firms.
A.4 METHODOLOGICAL CONSIDERATIONS: PRICE AND RATE OF RETURN IMPACTS
Let P denote product price, Q denote unit output, TOC denote total
operating costs, K denote the amount of capital invested in the unit, r
denote the rate of return on capital and t denote the tax rate in a given
year. The aftertax rate of return on capital invested in the unit may then
be defined as:
- (1-t) (PQ - TOC) (1)
r _
where (PQ - TOC) is the unit's pretax net revenues from its operations in
that year. Now, assume that the unit is required to change its operating
costs and level of capital investment in order to comply with the implemen-
tation of some regulatory alternative. Under the full cost absorption
scenarios the unit will be unable to adjust the the price of its product or
A-17
-------
TABLE A-5. FINANCIAL DATA FOR 100 CHEMICAL FIRMS
1-11
3=>
I
i>
CO
Name
Abbott Labs
Akzona
Alco Standard Corp.
Allied Chem Corp.
American Cyanamid
Armco Steel Corp.
Atlantic Richfield
Beatrice Foods
Bendix Corp.
Bethlehem Steel Corp.
Borden, Inc.
Borg-Warner Chem.
Brown Co.
CPC International
Inc.
Celanese Corp.
Charter International
Oil
Cities Service Co.
Combustion
Engineering
Continental Oil
Crompton & Knowles
Dart Indust.
Dayco Corp.
De Soto, Inc.
Diamond Shamrock
Corp.
Dow Chemical
Du Pont De Nemours
Eastern Gas & Fuel
Associates
Essex Chem. Corp.
Cost of
capital
12.014
10.276
12.151
10.091
11.083
10.588
9.749
11.232
11.118
10.913
10.484
11.863
9.813
11.638
10.181
9.175
10.395
11.494
10.881
11.298
10.689
8.270
11.499
9.790
10.060
11.328
11.605
12.502
Return
on
equity
14.018
13.276
13.425
13.721
13.425
13.276
13.128
12.832
13.425
14.018
12.683
13.128
12.387
13.128
13.128
14.166
12.980
14.314
13.721
13.425
14.166
12.980
13.128
13.721
14.018
13.573
14.018
14.166
Return
on
debt3
9.590
10.380
15.120
9.720
9.590
9.720
9.590
9.250
9.720
9.720
9.590
9.720
12.395
9.590
11.970
12.395
9.720
9.720
9.590
14.450
9.720
11.970
13.750
9.720
9.590
9.250
14.180
12.395
Return
on
preferred
stock
--
--
--
6.461
--
7.429
3.333
--
--
10.084
2.564
4.211
6.071
--
--
8.654
--
--
Proportion
of
equity
.77262
. 61914
. 64134
. 58118
.72252
. 66880
. 51602
. 79803
. 72911
.65360
. 71317
. 82756
. 56680
.81691
. 53511
.27557
. 67388
.68700
.67568
. 53329
.63113
.30351
.72746
. 54639
. 56176
.72512
. 63681
. 78453
Proportion
of
debt
.216575
. 380859
.259343
. 418825
. 277480
. 306858
. 362174
.194329
. 248140
. 346402
.285155
. 145263
.433202
. 183087
. 396896
.623167
.326120
.296229
. 321308
.375634
.231645
. 666445
.272535
.453615
.438236
.232172
.363188
. 215465
Proportion
of
preferred
stock
. 010804
.000
. 099317
.000
.000
.024337
. 121802
. 007644
.022754
.000
. 001677
.027181
.000
.000
.067997
. 101265
.000
.016774
. 003009
. 091078
.137221
.030044
.000
.000
.000
. 042712
.000
.000
-------
TABLE A-5. (Continued)
Name
Exxon Corp.
FMC Corp.
Ferro Corp.
Firestone Tire &
Rubber
Ford Motor Co.
GAF Corp.
General Electric Co.
General Motors Corp.
General Tire & Rubber
Georgia-Pacific Corp.
Goodrich (B.F. ) Co.
Goodyear Tire &
Rubber Co.
Gulf Oil Corp.
Hercules, Inc.
Inland Steel
Insilco Corp.
Interlake, Inc.
International
Harvester
Kaiser Steel Corp.
Kraft, Inc.
Marathon Oil Co.
Martin Marietta Chem.
Mead Corp.
Merck & Co.
Minnesota Mining &
Manuf .
Mobil Oil Corp.
Monsanto Co.
Morton-Norwich
Products
Cost of
capital
11.875
10.183
12.369
10.610
12.069
9.398
12.130
12.798
11.440
10.793
10.430
10.101
11.745
11.177
10.092
9.339
11.331
10.534
11.688
10.774
9.582
11.238
10.000
12.309
12.572
10.868
10.970
10.726
Return
on
equity
13.276
13.573
13.276
12.980
13.276
13.573
13.721
13.425
13.276
13.573
13.276
12.980
12.980
13.869
12.980
13.276
13.128
13.573
14.018
12.683
13.128
13.276
13.869
13.573
13.869
13.128
13.573
13.721
Return
on a
debt3
9.250
9.720
9.720
9.720
9.250
10.380
9.250
9.250
11.970
9.590
10.380
9.720
9.250
9.720
9.590
11.970
9.720
9.720
14.000
9.250
9.720
9.720
9.720
9.250
9.250
9.250
9.590
9.720
Return
on
preferred
stock0
--
6.250
--
--
--
7.559
--
8.715
--
--
8.864
--
--
--
--
7.752
--
--
--
-
--
4.308
--
--
5.000
*
Proportion
of
equity
. 83450
. 59257
. 88968
.70096
.85743
. 44490
. 82148
. 91962
.73287
.67625
.62957
.63679
. 84880
.69461
.62702
. 41885
.77736
.63297
.63274
.75752
. 56074
.75212
. 56423
. 85481
.85677
. 72833
.69690
. 65441
Proportion
of
debt
.165504
.339730
. 110317
. 299038
. 142565
. 387035
.178521
.063516
. 258968
.323751
. 349707
.363210
.151203
.305394
.352735
.475634
.222640
. 348230
.345717
.242479
.439257
. 247882
. 398718
.143358
.143235
.271665
.300335
. 345589
Proportio
of
preferred
stock
.000
.067701
.000
.000
.000
. 168061
.000
.016862
.008163
.000
.020723
.000
.000
.000
.020249
.105511
.000
. 018796
.021539
.000
.000
.000
.037048
.001827
.000
.000
.002767
.uuO
-------
TABLE A-5. (Continued)
r\i
o
Name
National Distillers
& Chem.
National Steel Corp.
Northwest Indust.
Owens-Corning
Fiberglass
PPG Industries
Penwalt Corp.
Pfizer
Phillips Petroleum Co.
Procter & Gamble Co.
Quaker Oats Co.
Reeves Bros. Inc.
Reichold Chems.
Republic Steel Corp.
Riegel Textile Corp.
Rockwell International
Rohn and Haas Co.
SCM Corp.
Scott Paper Co.
Shakespeare Co.
Sherwin-Williams Co.
Squibb Corp.
A. E. Staley Mfg. Co.
Stauffer Chemical Co.
Sterling Drug
Sun Chem. Corp.
Sybron Corp.
Tenneco, Inc.
Texaco
Texfi Indust.
Textron Inc.
Union Camp Corp.
Union Carbide Corp.
Cost of
capital
11.037
9.909
8.015
11.653
10.596
9.013
11.244
11.670
11.824
10.946
10.629
10.647
11.305
11.201
9.589
10.739
10.835
10.784
11.229
9.617
11.266
10.428
10.188
12.595
10.427
10.786
9.155
11.230
10.090
10.085
11.359
10.775
Return
on
equity
13.128
12.683
13.869
13.425
13.276
13.276
14.018
13.721
13.276
13.573
12.535
13.425
13.425
12.980
12.535
13.721
14.018
13.721
13.276
12.980
14.018
13.573
13.425
13.276
13.573
13.869
12.980
12.980
13.275
13.425
13.276
13.573
Return
on a
debt3
9.720
9.590
10.380
9.720
9.590
9.720
9.590
9.250
9.250
9.720
10.380
10.380
9.720
11.970
9.720
9.720
10.380
9.590
14.000
10.380
9.590
9.720
9.720
9.590
12.395
9.720
10.380
9.250
16.000
9.720
9.590
9.590
Return
on
preferred
stock0
9.193
--
2.9412
7.529
--
--
9.008
--
5.398
--
10.00
--
--
--
3.887
--
6.222
--
Proportion
of
equity
.73310
.63946
.32561
. 78828
. 67661
. 41712
. 69289
. 76982
.82842
.651578
.732870
. 571986
. 746819
.736598
.602132
.655939
.630766
.660791
.658505
. 523981
.695345
. 629947
. 613351
. 917816
. 558689
. 616191
. 505890
. 785863
. 356904
.577353
. 768639
. 674170
Proportion
of
debt
.251565
. 360538
. 617085
.211721
. 323394
. 369200
. 307113
. 230179
. 171428
. 262094
.267130
.295871
.253181
. 263402
.309032
. 344061
. 369234
. 333680
. 341495
.422439
. 304655
. 368508
. 386649
. 082184
.441311
. 319517
.442129
. 214137
.643096
.252757
.231361
. 325830
Proportion
of
preferred
stock
.015334
.000
.057301
.000
.000
.213675
.000
.000
. 000153
. 086328
.000
. 132143
.000
.000
. 088836
.000
.000
.005529
.000
.053579
.000
. 001544
.000
.000
.000
. 064292
. 051981
.000
.000
. 1S9890
.000
.000
-------
TABLE A-5. (Continued)
Name
Union Oil, Calif.
Uni royal
U.S. Gypsum
U.S. Steel Corp.
Upjohn Co.
Vulcan Materials Co.
Walter (Jim) Corp.
Westinghouse Electric
Corp.
Weyerhaeuser Co.
Wheel i ng-Pi ttsburgh
Steel
Whittaker Corp.
Wit Chem. Corp.
Cost of
capital
10.577
10.514
10.726
10.919
11.052
10.675
9.019
12.596
10.402
11.238
10.070
10.736
Return
on
equity
13.128
13.425
13.276
13.573
13.573
12.980
13.721
14.018
14.166
13.869
14.314
13.573
Return
on
debt3
9.590
11.970
9.590
9.590
9.590
9.720
11.970
9.720
9.590
14.000
11.970
9.720
Return
on
preferred
stock
--
16.000
5.539
--
--
4.444
8.837
5.957
12.739
--
3.313
Proportion
of
equity
.663994
.521603
.686341
.690912
.706383
.709218
.398726
.838775
. 583685
.512893
.457808
.673790
Proportion
of
debt
.295934
.423786
.223477
. 309088
.293617
.290782
.491966
.155115
.357341
. 381136
.517470
.292825
Proportion
of
preferred
stock
. 040072
. 054611
.090182
.000
.000
.000
.109308
.006110
.058973
.105972
.024722
.033385
The return on debt data represent pretax estimates and are multiplied by 0.52 to obtain the aftertax rates
used in computing the cost of capital.
Dashes indicate missing data. In these cases the pretax returns on debt were used to compute the cost
of capital.
-------
unit output. Consequently, the rate of return on investment, r, will
change. The formula used to estimate this impact is obtained by totally
differentiating Equation (1) with respect to TOC and K; that is,
, _
dr - -
(1-t) dTOC . (1-t) (PQ-TOC) dK
+
Substituting in (2) from (1) and rearranging terms, it follows that:
, _ (1-t) dTOC + rdK (3)
-or -
Equation (3) is the formula used to calculate the full cost absorption rate
of return impacts.
Price impacts are estimated on the basis of the assumption that firms
will be able to maintain the preregulation rate of return (r) by increasing
product prices. Thus, r is now a constant and P a variable. Rearranging
terms in Equation (1), it may be shown that:
p _ TOC + r K/(l-t) (4)
" Q
In full cost pass through scenarios, changes in TOC and K leave r and Q
unaffected but result in a change in P. The formula for estimating this
change in P may be obtained by total differentiating Equation (4) with
respect to TOC and K; that is,
HP = dTOC + rdK /(1-t) (5)
ar - Q
Equation (5) is the formula used to estimate the full cost pass through
price impacts.
A-22
-------
A.5 REFERENCES
1. COMPUSTAT. New York: Standard & Poor's Corporation, 1978.
2. Daily Stock Price Record. New York: Moody's Investors Service, Inc.,
September 1979.
3. Moody's Bond Record. New York: Moody's Investors Service, Inc.,
September 1979.
4. Moody's Bond Survey. New York: Moody1s Investors Service, Inc.,
September 3, 1979.
5. Moody's Industrial Manual. New York: Moody1s Investors Service, Inc.,
1979.
6. Scherer, F. M., et al. The Economics of Multi-Plant Operation.
Cambridge, Mass.: Harvard University Press, 1975.
7. Standard & Poor's Bond Guide. New York: Standard & Poor's
Corporation, September 1979.
8. Standard & Poor's Corporation Records, (Quarterly Dividend Record).
New York: Standard & Poor's Corporation, September 1979.
9. Standard & Poor's Statistical Service, Basic Statistics. New York:
Standard & Poor's Corporation, September 1979.
10. Value Line Investment Survey. New York: Arnold Bernhard & Co., Inc.,
September 24, 1979.
11. The Wall Street Journal. New York: Dow Jones & Company, October 1,
1979.
A-23
-------
APPENDIX B
AGGREGATION OF MODEL UNIT IMPACTS
Sections 2,4, and 5 present data for fugitive emissions on a source-
by-source basis. This information can be combined with the data on model
units given in Section 3 to estimate impacts on a model unit basis.
For example, in computing the estimated fugitive emissions from model
unit B, the emission factors for SOCMI developed in Section 2 are applied to
the equipment counts for unit B given in Section 3. The sum of emissions
for each source type represents the estimated emissions for the model unit.
This procedure is demonstrated in Table B-l for model unit B. Similarly,
emission reductions achieved by various control options computed for each
source type and presented in Section 4 can be aggregated to estimate
emission reductions achievable on a model unit basis. Table B-2 presents a
summary of estimates of emission reductions achievable for the three model
units. Cost estimates may be aggregated in the same manner. The capital or
annual ized costs for any fugitive emission source type in a model unit is
determined by applying the per source estimate to the number of sources in
the model unit. Examples of this technique are shown in Table B-3 (Capital
Costs) and Table B-4 (Annualized Costs).
B-l
-------
TABLE B-l. EXAMPLE OF EMISSIONS ESTIMATED FOR MODEL UNIT B
IN ABSENCE OF STANDARDS
CD
I
rv>
Emissions
Source
Pump seals
Light Liquid
Heavy Liquid
Valves
Gas
Light Liquid
Heavy Liquid
Safety/relief valves
Gas
Open-ended lines
Compressors
Sampling connections
Flanges
Total
Number of
Sources
29
30
402
524
524
11
410
2
26
2400
tmission Factor,
kg/hr/source
0.0494
0.0214
0.0056
0.0070
0.00023
0.1040
0.0017
0.228
0.0150
0.00083
Annual Emissions, Mg/yra
12.55
5.62
19.72
32.59
1.06
10.02
0
3.99
3.42
17.45
99.51
For estimating purposes, one operating year was assumed to be 8/60 hours.
-------
TABLE B-2. ESTIMATE OF EMISSION REDUCTIONS ACHIEVABLE FOR MODEL SOCMI UNITS IN MG/YR
ra
i
CO
Source
Pumps
Light Liquid
Heavy Liquid
Valve?
Gas
Light L ic'iid
Heavy Liquid
Safety/rel ief
valves
Gas
Open-ended 1 inesc
Flanges
Sampl ing
connections
Compressors
Total
^Reduction
Control
Method
Leak detection
and repair
Lest, detection
and repair
Rupturp disk
Plugs & Caps
Mo control
Closed purge
systems
Vented seal
areas
Basel
A
3.46
1.31
5.86
8.15
0.27
2.73
0
4.36
0.92
2.00
28.06
ine Emissions Control
8 C Efficiency
12.55
5.62
19.72
32.59
1.06
10.02
0
17.45
3.42
3.99
106.42
39.
17,
60.
100,
3,
30
0
53
10
15
331
,38 0.61
,43
,44 0.73
.63 0.59
.26
.06 1.0
1.0
.80
.51 1.0
.98 1.0
.49
Controlled Emissions
ABC
1.36
1.31
1.3!
3.34
0.27
0
0
4.36
0
0
11.95
4.92
5.62
5.32
13.36
l.Ofi
0
0
17.45
0
0
47.73
15.44
17.43
16.32
41.26
3.26
0
0
53.80
0
0
147.51
Emissions Reduction
ABC
2.10
0
3.55
4.81
0
2.73
0
0
0.92
2.00
16.11
57
7.63
0
14.40
19.23
0
10.02
0
0
3.42
3.99
58.69
55
23.94
0
44.12
59.37
0
30.06
0
0
10.51
15.98
183.98
56
Baseline emissions means emissions in the absence of standards.
Monthly monitoring program presented.
GAs discussed in Section 3 of this document, open-ended lines are assumed to be controlled at baseline.
-------
TABLE B-3. SUMMARY OF AGGREGATED CAPITAL COST ESTIMATES FOR MODEL SOCMI UNITS, 1978$
Emission Source
Pump seals, light liquid
Valves, gas and light liquid
Safety/relief valves, gasa
Compressor seals
Sampling connections
Control
Method
Leak detection and
repair program
Leak detection and
repair program
Rupture disks
Vented seal areas
Closed purge systems
A
320
470
7,820
1,590
3,220
8,500C
21,910
Model Unit
B
1,170
1,900
28,680
3,170
11,960
8,500C
55,380
C
3,590
5,830
86,050
12,690
36,800
8,500C
153,460
Assumes a 50/50 split between systems using 3-way valves and systems using block valves.
"'Assumes a 50/50 split between systems tied to an enclosed combustion device and systems tied to a
flare; also assumes 60 percent of the compressors in the industry are already controlled.
'Monitoring instrument cost.
-------
TABLE B-4. ANNUALIZED COST ESTIMATES FOR SOCMI MODEL UNITS, 1978$
Control
Emission Source Method
Pump seals, light liquid Leak detection and
Annual ized capital cost repair program3
Annual, operating cost
Valves, gas and light liquid Leak detection and
Annual ized capital cost repair program3
Annual ized operating cost
Safety/relief valves, gas Rupture disks
Annual ized capital cost
Annual operating cost
Compressor seals Vented seal areas
Annual ized capital cost
Annual operating cost
Sampling connections Closed purge systems
Annual ized capital cost
Annual operating cost
Total annual ized costs
Product recovery credit
Net annual ized costs
A
50
1,890
80
2,970
1,520
700
260
140
530
290
1,960^
3,040C
13,430
(4,830)
8,600
Model Unit
B
190
6,770
310
11,970
5,570
2,580
520
280
1,950
1,070
1,960^
3,040C
36,210
(17,610)
18,600
C
590
21,290
950
36,840
16,700
7,750
2,060
1,140
6,000
3,280
1,960^
3,040C
101,600
(55,200)
46,400
aMonthly monitoring program presented
Monitoring instrument - annualized capital cost
cMonitoring instrument - annual operating cost.
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
. REPORT NO.
EPA-450/3-82-010
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Fugitive Emission Sources of Organic Compounds
Additional Information on Emissions, Emission
Reductions, and Costs
5. REPORT DATE
April 1982
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Trianale Park, North Carolina 27711
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-02-3058
12. SPONSORING AGENCY NAME AND ADDRESS
DAA for Air Quality Planning and Standards
Office of Air, Noise, and Radiation
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
PA/200/04
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Standards of performance to control fugitive emissions of VOC from new, modified,
and reconstructed Synthetic Organic Chemical Manufacturing Industry (SOCMI) process
units were proposed on January 5, 1981 (46 FR 1136). This document contains the data
and methodologies which EPA believes most accurately characterizes SOCMI fugitive
emission rates of VOC, effectiveness of control techniques, and control costs.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
Air pollution
Pollution control
Standards of performance
Volatile organic compounds
Organic chemical industry
fie. DISTRIBUTE STATEMENT
Release unlimited. Available from EPA
Library (MD-35), Research Triangle Park,
North Carolina 27711
b.IDENTIFIERS/OPEN ENDED TERMS
Air pollution control
19. SECURITY CLASS (Tins Report/
Unclassified
This page I
20 SECURITY CLASS(
Unclassified
c. COSATl Field/Group
>1. NO. OF PAGES"
258
22. PRICE
EPA Form 2220-1 (Re*. 4-77) PB^/IOUS ED> TION is OBSOLETE
------- |