United States      Office of Air Quality       EPA-450/3-88-010
           Environmental Protection  P'anning and Standards      October 1988
           Agency        Research Triangle Park NC 27711
           __
&ER&     Protocols for
          Generating
          Unit-Specific
          Emission Estimates
          for Equipment
          Leaks of VOC and
          VHAP

-------
                                               EPA-450/3-88-010
 Protocols for Generating Unit-Specific Emission
Estimates for Equipment  Leaks of VOC and VHAP
                         Emission Standards Division
                       U S. Environmental Protection
                       Region 5, Library (PL-12J)
                       77 West Jackson Boulevard, 12th Fi«r
                       Chicago. IL  60604-3590
                    U.S. ENVIRONMENTAL PROTECTION AGENCY
                          Office of Air and Radiation
                     Office of Air Quality Planning and Standards
                     Research Triangle Park, North Carolina 27711

                             October 1988

-------
This report has been reviewed by the Emission Standards Division of the Office of Air Quality Planning and Standards, EP
and approved for publication. Mention of trade names or commercial products is not intended to constitute endorsement
recommendation for use. Copies of this report are available through the Library Services Office (MD-35), U.S. Environment
Protection Agency, Research Triangle Park NC 27711, or from National Technical  Information Services 5285 Port Rov
Road, Springfield VA 22161.

-------
                               TABLE OF CONTENTS

Section                                                                   Eifli

List of
Figures   	 v

List of
Tables    	v1

GLOSSARY	vii

FOREWORD	x1

  1.0     INTRODUCTION	1-1

  2.0     DEVELOPMENT OF EMISSION ESTIMATES 	 2-1
          2.1  Introduction	2-1
          2.2  Use of the EPA's Average Emission Factors	2-1
          2.3  The Leak/No-Leak Approach	2-4
          2.4  Application of Stratified Emission Factors  	 2-8
          2.5  Leak Rate/Screening Value Correlations  	 2-9
          2.6  Unit-Specific Correlations  	 2-13
               2.6.1  Generation of Mass Emissions Data 	 2-13
               2.6.2  Development of Leak  Rate/Screening Value
                        Correlations	2-15
               2.6.3  Statistical Considerations of Leak Rate/Screening
                        Value Correlations	"	2-16
                      2.6.3.1  Graphical Comparison of EPA vs. Other
                                 Correlations	2-16
                      2.6.3.2  Statistical Comparison of EPA vs. Other
                                 Correlations  	 2-23
                      2.6.3.3  Other Statistical Analyses  	 2-23
               2.6.4  Generation of Emission Estimates	2-24
               2.6.5  Statistical Considerations for  the Predicted
                        Emission Estimate  	 2-24

  3.0     SOURCE SCREENING	3-1
          3.1  Screening Survey Procedures	3-1
          3.2  Monitoring Instrument	 3-3
               3.2.1  Operating Principles and  Limitations of Portable
                        VOC Detection Devices  	 3-4
               3.2.2  Performance Criteria and  Evaluation  for Portable
                        VOC Detectors	3-5
                      3.2.2.1  Response Factor	3-7
                      3.2.2.2  Response Time	3-8
                      3.2.2.3  Calibration Precision	3-9
                      3.2.2.4  Safety 	 3-9
               3.2.3  Monitoring Devices For Difficult Situations  .... 3-10

-------
                               TABLE OF CONTENTS
                                   (cont'd.)
Section
Page
          3.3  Screening Procedures 	  3-12
               3.3.1  Calibration Procedures	3-12
               3.3.2  Procedure for Screening Equipment  	  3-12
          3.4  Data Handling	3-19
          3.5  Calibration Procedures For Quality Assurance  	  3-22

  4.0     MASS EMISSIONS SAMPLING	4-1
          4.1  Vacuum Method	4-1
          4.2  Blow-Through Method	4-6
          4.3  Source Enclosure	4-10
               4.3.1  Valves	  .  4-12
               4.3.2  Pumps and Agitators	4-13
               4.3.3  Compressors	4-14
               4.3.4  Flanges	4-15
               4.3.5  Relief Valves	4-15
          4.4  Accuracy Checks, 	  4-16
          4.5  Analytical Techniques. ..........  	  4-16
               4.5.1  Analytical Instrumentation	4-16
               4.5.2  Calibration . . .	4-17
               4.5.3  Analytical Techniques for Condensate	4-17
               4.5.4  Quality  Control for Analytical Techniques  	  4-18

  5.0     REFERENCES	"	5-1

Appendix A - Reference Method  21

Appendix 8 - Response Factors

Appendix C - Calculations of Mass Emission Rates

Appendix D - Leak Rate/Screening Value Equations

Appendix E - Selection of Sample Size for Flange Screening

Appendix F - Determination of  Emission Rates for Default Zero  Screening  Values

Appendix G - Development of a  Default Emission Rate  for  Equipment  That Does
             Not Screen Above  the Background Concentration
                                        IV

-------
                                LIST OF FIGURES

Figure                                                                    £§££
 2-1   Strategy for Estimating Emissions from Equipment Leaks 	 2-2
 2-2   Modeled relationship-leak rate vs. OVA for valves in gas service  . 2-17
 2-3   Log1Q leak rate vs. log1Q OVA reading for pump seals	2-18
 2-4   L°9in l6^ rate vs- ^°9io OVA rea<^in9 ^or valves-gas service . .  . 2-1.9
 2-5   Log1Q leak rate vs. log1Q OVA reading for valves-liquid service.  . 2-20
 2-6   Comparison of SOCMI and refinery leak rate correlations	2-22
 3-1   Gate valves	3-14
 3-2   Globe valves	3-15
 3-3   Lubricated plug valve	3-16
 3-4a  Ball valve	3-17
 3-4b  Butterfly valve	3-17
 3-5a  Weir type diaphragm valve	3-18
 3-5b  Check valves	3-18
 3-6a  Vertical centrifugal pump	3-20
 3-6b  Horizontal centrifugal  pump	3-20
 3-7   Spring loaded relief valve	.'  . 3-21
 4-1   Sampling train for baggable sources of hydrocarbon emissions using
         a diaphragm	.-	4-2
 4-2   Mylar plastic sample bag	.-	4-4
 4-3   Equipment required for the blow-through sampling technique .... 4-7
 4-4   Tent construction around the seal area of a vertical pump	4-11

-------
                                 LIST OF TABLES

Table                                                                      Page
 2-1   Average Emission Factors for Fugitive Emissions  	   2-3
 2-2   Leaking and Non-Leaking Emission Factors for Fugitive Emissions
         (kg/hr source)	2-6
 2-3   Estimate of "Uncontrolled" Fugitive Emissions for a Hypothetical
         Case	2-7
 2-4   Stratified Emission Factors for Equipment Leaks  	   2-10
 2-5   Estimate of Fugitive Emissions Using Stratified  Emission Factors
         for a Hypothetical Case	2-11
 3-1   Fugitive Emission Sources 	 	   3-2
 3-2   Performance Criteria for Portable VOC Detectors  	   3-6
 3-3   Portable VOC Detection Instrument Performance Specifications.  .  .   3-11
 3-4   Example Data Sheet	3-23
 4-1   Calculation Procedure for Tented Leak Rate	4-8
 B-l   Response Factors with 95% Confidence Intervals Estimated
         at 10,000 ppmv Response	B-l
 B-2   Tested Compounds Which Appear to be Unable to Achieve an
         Instrument Response of 10,000 ppmv at any Feasible
         Concentration .	  .   B-7
 D-l   Prediction Equations for Nonmethane Leak Rate for Valves,  Flanges,
         and Pump Seals in SOCMI Processes	0-1

-------
                                   GLOSSARY


Censored source - a source whose screening value is known to be above
     100,000 ppmv but whose actual screening value is unknown.

Emission factor - the per component mass emission rate applicable to
     populations of sources, not individual component measurements.

Inaccessible source - a component which cannot be reached because it is
     beyond the reach of a 6-foot ladder.

Leak definition - monitoring instrument reading selected as the trigger value
     for initiating some action such as maintenance; e.g., 10,000 ppmv is the
     leak definition used by EPA.

Leak frequency - the percentage of sources (a particular equipment type and
     service) determined to be leaking based upon a chosen leak definition.

Leak rate - see mass emissions rate.

Leaking emission factor - the per component mass emission rate associated with
     the population of sources with screening concentrations at or above the
     leak definition.

Leaking source - a source whose screening concentration is greater than or
     equal to 10,000 ppmv.

Mass emissions rate - the quantity of volatile ocganic compound(s) released to
     the atmosphere in terms of total mass per unit time.  .

Monitoring instrument - portable hydrocarbon analyzer meeting the performance
     specifications given in Method 21.      '

Non-leaking emission factor - the per component mass emission rate associated
     with the population of sources with screening concentrations less than
     the leak definition.

Non-emitting source - a source whose screening value is 0 ppmv; a zero
     emission rate is assumed.

Non-leaking, emitting source - a source whose screening concentration is
     greater than 0 ppmv but less than 10,000 ppmv.

Screening value - raw monitoring instrument reading.

Screening concentration - the screening value (raw instrument reading)
     corrected by the appropriate response factor.

Unsafe-to-monitor equipment - equipment that exposes monitoring personnel to
     an immediate danger while screening the equipment.


                                      vii

-------
  EF - average emission rate per source
 EF, - average emission rate per leaking source
 EF2 - average emission rate per non-leaking emitting source
 X-,. - measured emission rate for the ith leaking source
 X2- - measured emission rate for the ith non-leaking emitting source
  X- • average measured emission rate for non-leaking emitting sources
  N, - number of leaking sources
  N2 - number of non-leaking emitting sources
  N. « number of non-emitting sources
   N - total number of sources (including inaccessible sources)
  n, - number of bagged leaking sources
  n2 * number of bagged non-leaking emitting sources
   2
 s-  - estimate of the variance of the emission rates of the non-Teaking
       emitting sources
        n2
     ' £ 
-------
  s  }"  »  variance  of the predicted emission  rates  of the leaking sources
  yz     screening <100,000 ppmv
 s   2 - variance of the predicted emission  rates  of the leaking sources
  y      screening  <100,000 ppmv
          N,_
 s ,u2 * variance of the measured emission rates of the censored sources
  ylb
                          /(n
                             ib
     The symbol  t is the value of the normal  deviate corresponding to the
desired confidence probability.  If the number of bagged sources, n,  is
greater than or equal to 50 then the value of t approaches 1.96.  If the
sample size is less than 50, the percentage points may be taken from Student's
t table with (n-1) degrees of freedom.  The t distribution holds exactly only
if the emission rates are normally distributed.  Moderate departures from
normality do not affect it greatly.

-------
                                   FOREWORD

     The EPA's protocol for estimating equipment leak emissions  is the
result of detailed information gathering and data analysis.  The protocol
was written to provide a thorough understanding of acceptable approaches to
generating process unit-specific emission estimates.  In preparing this
document, EPA has encouraged industry experts and trade association
representatives to comment on the document's content.  Chief among these
contributors has been the Chemical Manufacturers Association (CMA), who
has written a guidance document similar to this EPA protocol.

     The CMA document is consistent with this guide.  Both documents present
acceptable approaches to generating emission estimates.  Minor differences
that exist between the two documents should not affect the acceptability of
the emission estimates, provided the approaches to generating acceptable
emission estimates are thoroughly understood.  The CMA protocol  includes
a provision for the addition of bagging data from a single process unit to
the EPA correlation curves to modify the leak rate/screening value
correlation. While the EPA agrees with the technical premise that additional
data could be added to the existing database, the Agency believes that, for
practical reasons, this approach could create interpretation problems for
control agencies.  Therefore, the Agency does not support this approach.
Currently, EPA is working with CMA to expand the existing database of
equipment screening and bagging data.  Ultimately, the Agency will consider
the addition of the data currently being collected to the existing database.
This effort may provide a better understanding of the magnitude  of equipment
leaks  and appropriate controls for them.  The EPA will consider  appropriate
modifications to this protocol when additional data collected during this
interim period are analyzed.

-------
                               1.0  INTRODUCTION

     The purpose of this document is to present standard protocols for
estimating mass emissions from equipment leaks.  The EPA's interest in
developing standard protocols is to assist states and industry in their
efforts to estimate equipment leak emissions.  Emission estimates are being
developed for a variety of purposes, including permitting, inventories, and
for the Superfund Amendments and Reauthorization Act (SARA).
     The EPA previously developed equipment leak emission factors in previous
work supporting standards.  These factors represented emissions from typical
facilities and were used to develop industry-wide emission estimates.  The
EPA's standards-setting process will continue to rely upon the average
emission factors for industry-wide estimates.  However, those estimates do not
yield the accuracy that facilities are looking for today; therefore, there is
now a need for more accurate emission estimating techniques.
     Recognizing that the significance of our protocols is in providing a
consistent approach to data collection and emissions estimates, EPA believes
there is need for certain minimal requirements in any protocol.  Our desire  is
to develop standard protocols for general use in generating unit-specific
emission estimates for permitting and inventories.
     This document describes methodologies the EPA considers appropriate for
development of unit specific emission estimates for equipment leaks of organic
compounds:  volatile organic compounds (VOC) and volatile hazardous air
pollutants (VHAP).  Estimates generated using this protocol would be specific
only to process units  (or groups of sources) for which an estimate was made.
EPA has made provision for extending such estimates beyond the limits of that
group of sources.
     Five methods for  estimating emissions from equipment leaks from a
specific chemical processing unit are included in the protocol:
     •    Average emission factor method;
     t    Leak/no-leak emission  factor method;
     •    Three-strata emission  factor method;
     •    Application  of  EPA correlations; and
     t    Development  of  new correlations.
                                       1-1

-------
They range from simply applying accurate equipment counts to the  EPA's  average
emission factors to the more complex project of developing unit-specific
correlations of mass emission rates and screening values.
     All methods start with obtaining an accurate identification  and count of
equipment to be included in the emission estimate.  The equipment counts can
simply be used with the EPA's previously developed emission factors.  The next
step in complexity and refinement is the use of a portable organic analyzer to
find the number of leaking and nonleaking sources.  Leaking and nonleaking
emission factors previously developed by the EPA can then be applied to
generate the emission estimate.
     A final refinement in a method employing discrete emission factors is
that of applying emission factors to represent three different ranges of
screening values.  This has been called the stratified emission factor
approach, or the three-strata approach.  Applying the stratified  emission
factors requires more rigorous measurement of organic vapor concentrations
with a portable instrument because actual concentration readings  must be
recorded instead of noting whether a piece of equipment is classified as
leaking or not leaking.
     The remaining two methods make use. of correlation equations  relating mass
emissions to organic concentrations measured with a portable organic analyzer.
The EPA's previously developed correlations are offered for use,  and finally,
if a process unit's emissions are statistically different from those
represented by the EPA's correlations,  provision  is made for development of
correlations specifically for that process unit.
     Guidance and recommendations are offered for statistical analyses  and for
gathering the data required to execute  the methods.  The information is
provided  in the following major sections of this  protocol:
     •    Section 2.0  Emission Estimate Development;
     •    Section 3.0  Source Screening;
     •    .Section 4.0  Mass Emissions Sampling.
                                       1-2

-------
Section 2.0 details the types of data needed for making emissions estimates
and how that data is to be applied.  Sections 3.0 and 4.0 offer more detail on
collection of data using a portable organic analyzer and using a technique
known as "bagging," in which an enclosure is made around a potential leak
point and emissions are actually measured.  Additional reference material is
included in the Appendices.
                                       1-3

-------
                    2.0  DEVELOPMENT OF EMISSION ESTIMATES

2.1  INTRODUCTION
     In this section, five methods for estimating VOC and VHAP emissions from
equipment leaks in chemical processing units are described.  The methods
increase in complexity and in the amount of data collection and analysis
required.  However, all five methods require some data collection, data
analysis, and/or statistical evaluation.  The five estimation methods and the
options available for collecting and analyzing the data are shown in
Figure 2-1.  As the flow chart shows, the methods increase in rigorousness and
complexity.  The first method, the average factor method, is the least complex
and demanding and the last method, developing site specific correlations, is
the most complex and demanding.  The end product of each of the procedures
described is an emissions inventory for equipment leaks by type of equipment
and by service (i.e., light liquid, gas, heavy liquid).
2.2  USE OF THE ERA'S AVERAGE EMISSION FACTORS
     All of the methods require an accurate count of equipment components by
type of equipment and by service.  The most basic approach for estimating
emissions merely requires application of average emissions factors developed
by EPA to the equipment counts for the unit.  EPA's average emission factors
are shown in Table 2-1.  The product of the emission factor and the number of
equipment components yields the emission rate per source type, and the
emission rates for all source types are summed to generate the unit-specific
emission estimates.
     The method EPA used for development of emission factors for individual
equipment leak emission sources was described in reports on the Petroleum
Refinery Assessment Study.10'11  In this study, "screening" data were gathered
using a portable organic vapor analyzer, and mass emissions were measured by
enclosing individual pieces of equipment with bags or tents and measuring the
organic material emitted.
     These data permitted the development of leak rate/screening value
correlations and emission factors for sources in petroleum refineries.
Subsequently, EPA coordinated the SOCMI 24-unit Study,   a study of 24
individual chemical process units representing a cross-section of the SOCMI
                                      2-1

-------
Count Equipment Components
    (By type and service)
         1
     Apply Average Emission Factors
                and
       Composite Total Emissions
^ Inventory  Section 12
 Conduct Complete ScrMnmg
         Survey
 (Response Factor Adjustment
        Optional)
          i
          o
      Apply Leak/No LMK Emission
             Factors and
       Composite Total Emissions
       Apply Three Strata Emission

       Composite Total Emissions
   inventory   Section 2.3
   inventory   Section 2.4
                            Adlust Screening
                               Value to
                           OVAJMethane Format
  3ag Components for Eacn
 Equipment Type ana Service
    Develop individual
      Correlation*
   Comoant Statistically
  to Existing Correlations
   Apply N«w Corrwation
    and Composite Total
        Emissions
                         Apply EPA OVA
                         Correlations to
                        Screening Values
                        Apove tne Oerault
                       Zero Screening Value
                        Apply Default Zero
                        Smisawn Rate to
                        Screening. Values
                        3*ow me Oerault
                       Zero Screening" Value
                            inventory
                   Section 2.S
Inventory   Section 2.8
Figure 2-1.  Strategy for  Estimating Emissions from
                           Equipment  Leaks
                                          2-2

-------
TABLE 2-1.  AVERAGE EMISSION FACTORS FOR FUGITIVE EMISSIONS
Equipment
Valves


Pump Seals

Compressor Seals
Pressure Relief Seals
Fl anges
Open-Ended Lines
Sampling Connections
Service
Gas
Light Liquid
Heavy Liquid
Light Liquid
Heavy Liquid
Gas/Vapor
Gas/Vapor
All
All
All
Emission Factor
(kg/hr/source)
0.0056
0.0071
0.00023
0.0494
0.0214
0.228
0.104
0.00083
0.0017
0.0150
                            2-3

-------
population, and a six-unit Maintenance Study.    As a result of these chemical
process unit studies, leak rate/screening value correlations and emission
factors were generated for valves in gas/vapor service in SOCMI.  Leak
rate/screening value correlations were also generated for light liquid valves
and light liquid pumps in SOCMI, but industry average emission factors were
not developed.
2.3  THE LEAK/NO-LEAK APPROACH
     The second method offers some refinement over the average emission factor
method, thereby allowing some adjustment to individual unit conditions and
operation.  This method and all the remaining methods require screening of all
equipment to be included in the inventory using a portable analyzer.
Equipment that is dangerous to screen may be omitted from the set of equipment
components to be screened.  Their emission rates may be estimated using the
EPA's average factors presented in Table 2-1.
     Insulated equipment can be considered difficult-to-monitor equipment.
The decision to remove insulation to facilitate screening is left to
individual judgement.  If insulation is not removed, the insulated component
is assumed to leak at the same rate as a similar uninsulated component.
     For flanges, a reduced number of components may be screened.  Appendix E
describes how to select the sample size for flanges.  In compiling screening
values for use in this technique (or any of those that follow), a response
factor may be applied to correlate the screening values measured for the
chemical in the line to the screening values observed in calibrating the
instrument to another chemical.  For a discussion of response factor
adjustment, see Section 3.2.2.1.
     The leak/no-leak approach to estimating emissions is based on the
assumption of only two emission rates and two populations of equipment
components:  sources that "leak" (with screening concentrations greater than
or equal to 10,000 ppmv) and sources that do not "leak" (with screening
concentrations less than 10,000 ppmv).
     The basis of this approach is as follows:  when a group of sources leak,
they leak  at a certain emission rate on the average.  Similarly, as a group,
non-leaking sources have, on the average, a certain mass emission rate
                                      2-4

-------
associated with them.  Thus, the overall emission estimate for a population of
emission sources consists of two components:  leaking source emissions and
non-leaking source emissions.
     Table 2-2 presents leaking and non-leaking emission factors previously
determined by EPA for equipment leaks.  EPA generated leaking and non-leaking
emission factors using:  1) emission factors generated from empirical
screening distribution data and leak rate/screening value correlations, 2) the
leak frequencies associated with the emission factors, and 3) the percent of
mass emissions associated with the leaking sources.  The detailed procedure
for generating leaking and non-leaking emission factors has been previously
published in the "Additional Information Document" (AID)13 for fugitive
emission sources of organic compounds and is also described in a document
entitled "Emission Factors for Equipment Leaks of VOC and HAP."
     The application of the leak/no-leak approach for estimating emissions is
shown in the following example for a hypothetical chemical process unit.
     Table 2-3 presents the data for this hypothetical process unit needed to
apply the leak/no-leak procedure for estimating emissions.  The first column
shows the number of sources in the process unit by source type.  The second
column shows the number of sources with screening values greater than or equal
to 10,000 ppmv (i.e., leaking sources).  The percentage of sources leaking is
shown in the third column.  The emission estimates for this hypothetical
process unit can then be computed using the leaking and non-leaking emission
factors in Table 2-2.  For example, three of the 47 pump seals in light liquid
service were found to be leaking.  Using the leak/no-leak approach, the
emission estimate is generated:
      (0.437 kg/hr/source)(3) + (0.012 kg/hr/source)(47-3),
or 1.84 kg/hr.  The emission estimate per source for pumps in light liquid
service could  then be computed by dividing the unit-specific emission estimate
by the equipment count.  The last column in Table 2-3 shows the annual
emission estimates by source type for the hypothetical unit.
     While this example illustrates how total VOC emissions per source type
are calculated, a similar procedure can be used to estimate emissions for a
particular species in the line.  For example, consider this same hypothetical
case where the light liquid pumped contained 20 weight percent of Compound A.
                                      2-5

-------
      TABLE 2-2.  LEAKING AND NON-LEAKING EMISSION FACTORS FOR
                  FUGITIVE EMISSIONS (kg/hr source)
Equipment
Valves


Pump Seals

Compressor Seals6
Pressure Relief Valves
Fl anges
Open-Ended Lines
Service
Gasa
LLb
HLC
LL
HL
Gas
Gas
All
All
Leaking
(>10,000 ppm)
Emission Factor
0.0451
0.0852
0.00023d
0.437
0.3885
1.608
1.691
0.0375
0.01195
Non-leaking
(<10,000 ppm)
Emission Factor
0.00048
0.00171
0.00023
0.0120
0.0135
0.0894
0.0447
0.00006
0.00150
 The leaking and-non-leaking emission factors for valves in gas/
 vapor service are based upon the emission factors determined for
 gas valves in ethylene, cumene, and vinyl acetate units during the
 SOCMI Maintenance Study,  References 3 and 15.

 LL - light liquid service.

CHL - 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.

Emission factor reflects existing control level of 60 percent
 found in the industry; control  is through the use of barrier
 fluid/degassing reservoir/vent-to-flare or other seal leakage
 capture system.
                                 2-6

-------
               TABLE  2-3.   ESTIMATE OF "UNCONTROLLED" FUGITIVE
                            EMISSIONS FOR A HYPOTHETICAL CASE
Source
Pump Seals
Light Liquid
Heavy Liquid
Valves
Gas/Vapor
Light Liquid
Heavy Liquid
Pressure Relief
Gas/Vapor
Open -Ended Lines
Compressor Seals
Number
Screened

47
3

625
1180
64
Valves
31
278
4
Number
Leaking

3
1

19
13
0

1
9
0
Percent
Leaking

6.4
33.3

3.0
1.1
0

3.2
3.2
0
Sampling Connections 70 -
Flanges
2880
20
0.7
Computed
Emission Estimate
(per Source)
kg/hr/source

0.0391
0.1385

0.0018
0.0026
0.00023

0.0978
0.0018
0.0894
0.0150
0.00032
Annual
Emissions
Mg/yr

16.1
3.6

10.1
27.2
0.1

26.6
4.5
3.1
9.2.
8.1
aBased on values from Table 2-2, using EE - [LEF * PCL + NLEF * (100-PCL)]/100
 where
          EE   = Emission Estimate (per Source)
          LEF  =• Leaking Emission Factor
          NLEF = Non-leaking Emission Factor
          PCL  » Percent of sources found leaking

bThis hypothetical process unit is assumed to be in continuous operation, so
 it operates for 8,760 hours per year.  Batch or campaign processes may
 operate fewer hours per year, so the annual emissions would be prorated to
 account for the hours the equipment contained the chemical.
                                       2-7

-------
The emission estimate for Compound A for light liquid pumps Is computed  by
applying the weight percent (20 percent in this case) to the emission estimate
generated above:
     (0.20)(1.84 kg/hr) - 0.368 kg/hr.
As another example, consider the hypothetical case where the light  liquid
pumped contained only 80 weight percent VOC and Compound A accounted for
20 percent by weight.  The emission estimate for Compound A for light liquid
pumps is computed as:
     (0.2/0.8)(1.84 kg/hr) - 0.460 kg/hr
In other words, compound A accounts for one-fourth of the total VOC emissions
of 1.84 kg/hr.  As before, emissions per source can be determined by dividing
the estimate by the number of components involved.  Emission factor estimates
calculated in this manner may then be applied to the equipment counts  (where
the material in the process line contained 20 weight percent of compound A) to
estimate emissions of compound A for that specific process unit.
2.4  APPLICATION OF STRATIFIED EMISSION FACTORS
     Another approach to generating emission estimates involves a refinement
of the leak/no-leak approach discussed above.  The leak/no-leak approach is
based on two emission rates and two populations (i.e., leaking and  non-leaking
sources).  The  stratified emission factor approach, on the other hand,  is
based on several population and emission factors spanning several discrete
screening value ranges.
     Screening  values  in the EPA SOCMI data base are distributed widely  from
0 ppmv to over  100,000 ppmv.  The mass emissions are correspondingly
distributed.  The  stratified emission factor approach for estimating emissions
segments this distribution  into discrete intervals to account  for different
ranges of  screening  values.  The following ranges are used:
          *     0  -. 1,000 ppmv;
          •     1,001  - 10,000 ppmv; and
           •     over  10,000  ppmv.
     Using data gathered during previous EPA studies, emission factors  for
each screening  value  range  have been  generated.  These stratified emission
factors  represent  the leak  rates measured during fugitive emissions testing.
                                       2-8

-------
Their development incorporated the statistical methods used by the  EPA  in
developing other emission factors.  The emission factor for each discrete
interval is presented in Table 2-4 by equipment type and service.
     This method requires that screening of all equipment  (except that  which
is hazardous to screen) be conducted in accordance with methods described  in
Section 3 of this report.  All screening values must be recorded according  to
the applicable ranges.  Then, as with the leak/no-leak approach, the total
emission estimate is generated by applying the number of components in  each
screening value range to the appropriate emission factor and summing each
subtotal.  An example of this procedure, shown in Table 2-5, illustrates how
to implement the stratified emission factor approach.
2.5  LEAK RATE/SCREENING VALUE CORRELATIONS
     Mathematical correlations offer an additional refinement to estimating
emissions from equipment leaks by providing a continuous function over  the
entire  range of screening values  instead of discrete intervals.  The EPA has
published correlations relating screening values to mass emissions  rates.   As
shown in Appendix D, there are four correlations which can be applied to all
equipment types and services.  The EPA's correlations are  based upon OVA
measurements taken using Method 21.  The instrument was calibrated  to methane
(see Method 21).  Screening value measurements used with these published
correlations should be in this same format.   For example,  if a TLV  instrument
calibrated to hexane were used to gather screening data, the data must  be
transformed to represent measurements gathered using an OVA, calibrated to
methane, before using the published correlations.  If a detector fails  to  meet
Method  21 specifications or published transformations are  not available, an
instrument response curve must be developed to relate screening values  to
actual  concentrations  in the  appropriate format  (OVA/methane).  Development  of
this screening value correction curve should  be  done in the laboratory  before
the detector is used  in  the field.
     Another correction  factor that can be  applied to Method 21 and
non-Method 21  instruments is  the  response factor.  For many compounds,  the
instrument response  is non-linear with  increasing screening value.   In  terms
of the  basic leak/no-leak approach described  in  a previous section,  a  single
                                       2-9

-------
TABLE 2-4.  STRATIFIED EMISSION FACTORS FOR EQUIPMENT LEAKS
                      (kg/hr/source)
Emission Factors (kg/hr/source) for
Screen 1 no Value Ranees. DDITIV
Source
Compressor seals
Pump seals
Valves
Flanges, connections
Pressure relief devices
Open-ended lines
Service
Gas/vapor
Light liquid
Heavy liquid
Gas/vapor
Light liquid
Heavy liquid
All
Gas/vapor
All
0-1,000
0.01132
0.00198
0.00380
0.00014
0.00028
0.00023
0.00002
0.0114
0.00013
1,001-10,000
0.264
0.0335
0.0926
0.00165
0.00963
0.00023
0.00875
0.279
0.00876
Over 10,000
1.608
0.437
0.3885
0.0451
0.0852
0.00023
0.0375
1.891
0.01195
                             2-10

-------
     TABLE 2-5.  ESTIMATE OF  FUGITIVE  EMISSIONS USING STRATIFIED EMISSION FACTORS FOR A HYPOTHETICAL CASE
Source
Compressor seals
Gas/vapor
Pump seals
Light liquid
Heavy liquid
Valves
Light liquid
Heavy liquid
Flanges
Pressure relief devices
Gas/vapor
Open-ended lines
Number
Screened
4
47
3
1 ,180
64
2.880
31
278
Number of Sources Screening (ppmv)
0-1 ,000
3
32
1
1,020
63
2,600
25
236
1,001-10,000
1
12
1
147
1
160
5
33
Over 10.000
0
3
1
13
0
20
1
9
Computed Emission Estimate
>er Source Type8 per Source5
(Kg/hr) (Kg/hr/source)
0.298
1.776
0.485
2.81
0.01472
2.20
3.37
0.427
0.0745
0.0378
0.1616
0.00238
0.00023
0.00076
0.1087
0.00154
aBased on emission factors from Table  2-2,  where EE - (NLj  * SEFj)  +  (NL2  * SEF2)  + ...
 EE = Emission Estimate
 NLj , NL2, etc. - Number leaking in 1st  range  (0-1,000).  Number leaking in 2nd  range (1,001-10.000),  etc.
 SEFj, SEF2, etc. - Stratified emission  factor for 1st range, stratified emission  factor for 2nd range, etc.
^Computed Emission Estimate per source = Computed Emission Estimate per Source  Type/Number Screened.

-------
point response factor adjustment at the leak definition of 10,000 ppm is
adequate.  If, however, the correlations are to be used, the best estimate of
screening concentration over the entire range is needed.  Therefore, a
correction for non-linear response must be made.  This can be accomplished in
the laboratory by generating a response curve.
     The flow chart in Figure 2-1 shows a separate treatment of "zero"
screening values.  The function describing the correlation of leak rate and
screening value becomes discontinuous for zero and near zero values.  The
correlation function mathematically predicts zero emissions for zero readings
on the portable instrument.  Empirically speaking, however, the EPA's data
shows this prediction to be incorrect.  Mass emissions have been measured from
equipment showing no screening concentration above zero.  The reason for this
involves the detector accuracy.  In the Refinery Assessment Study mass
emissions corresponding to a screening value of 200 ppm or less could not be
quantified because the accuracy limit of the detector was 200 ppm.  To handle
this discontinuity at the low end of the correlation, the EPA has derived a
"default zero" screening value (8 ppm) and an associated mass emission rate
for the  screening values between zero and that default zero reading.  These
"default zero" values and emission rates shown below were derived as shown in
Appendix-F from mass emissions data gathered in chemical plants and the
published leak rate/screening value correlations shown in Appendix D:
                                                            Zero Screening
                                 "Default Zero"              Value Emission
Equipment Type/Service       Screening Value, oom         Rate  (kq/hr/source)
Valves,  gas                           8                         0.000033
Valves,  light liquid                  8                         0.000451
Flanges                               8                         0.000093
Pumps  and all other components        8                         0.000039
These  emission factors should be applied to equipment components screening
between  0 and 8  ppm.   The  published correlations would be applied to all
screening concentrations above 8 ppm.  The total emissions estimate for
equipment leaks  is generated by  totaling emissions estimates for all "default
zeroes"  and  adding that total to the  total estimates generated  using the
correlations.  Appendix G  presents  an alternative methodology  for generating
unit-specific default  zero  emission rates.

                                      2-12

-------
2.6  UNIT-SPECIFIC CORRELATIONS
     A process unit may develop its own correlations if leak rates are
statistically different from EPA's rates.  The steps involved in generating  a
unit-specific emissions estimate using correlations developed specifically for
that unit are:
     1.   Gathering of mass emission data and calculation of mass emission
          rate (leak rate);
     2.   Development of leak rate/screening value correlations;
     3.   Statistical consideration of leak rate/screening value correlations;
     4.   Application of leak rate/screening value correlation to the
          empirical screening data; and
     5.   Prediction of emissions.
These steps are discussed individually in the following sections.
2.6.1  Generation of Mass Emissions Data
     After complete screening of the process unit is done for all source types
(except possibly flanges), selected sources can be bagged to measure the mass
emission rate.  The sources selected for mass emission measurement should be
rescreened at the time of bagging..  The mass emisston rate determined by
bagging and the rescreening value can then be used to validate the application
of EPA's correlations to the process unit.  Alternatively, the measurements
can be used in development of leak rate/screening value relationships for that
specific unit.  The screening value can be affected by the portable screening
instrument used and the gas used to calibrate the instrument.  The EPA  leak
rate/screening value correlations are based on data gathered using an OVA
instrument calibrated with methane.
      If other instruments or calibration gases are used, then the screening
value  should  be adjusted  (using theoretical or empirical correction factors)
to be  equivalent to values measured using an OVA calibrated with methane prior
to any comparisons to the EPA equations.  Instrument readings above the
saturation point of the detector must be bagged to quantify the emission rate.
     The amount of leak rate testing  (bagging) that should be done depends on
the objective of the data collection.  If one wishes to check the fit of EPA
published equations to a  particular process unit, as few as four leak rate
measurements  of a  particular source type in a particular service may be
                                      2-13

-------
adequate.  If it is desirable to develop new equations, 30 leak rate
measurements should be obtained.  Because of the inherent variability of leak
rate/screening data, it is difficult to detect differences between
correlations with fewer than 30 data pairs.  However, the statistical goal  is
to generate estimates that are within 50 percent of the mean value with 95
percent confidence.  If this statistical goal can be achieved with fewer data
pairs, then fewer data pairs are acceptable.
     Consider a hypothetical process unit having a large population of sources
with screening values well distributed over the range of 0 to 100,000+ ppmv.
To develop statistically valid leak rate/screening value correlations, mass
emissions data must be collected from individual sources that have screening
values distributed over the entire range.  For each source type (i.e., valves,
pumps, etc.) and service (i.e., gas, light liquid, etc.), a random sample of
six sources should be chosen for bagging from each of the following screening
value ranges:
                         Screening Value Range (Domv)
                                    1-100
                                  101 - 1,000
                                1,001 - 10,000
                               10,001 - 100,000
                                   > 100,000
      If the maximum response of the screening instrument is 100,000 ppmv, then
20 or all  (whichever is less) of the sources screened at 100,000 ppmv should
be bagged.  If six sources are not available in a particular screening value
range, additional sources from the nearest range should be tested.  If
screening  values greater than 10,000 ppmv are not found in the process, the
following  five groups can be used:
                                 1 - 100 ppmv
                               101 - 300 ppmv
                               301 - 1,000 ppmv
                             1,001 - 3,000 ppmv
                             3,001 - 10,000 ppmv
                                     2-14

-------
Similar groupings can be developed if all sources in the unit screen less than
1,000 ppmv.
     If a statistical determination is made that estimates within 50 percent
of the mean value with 95 percent confidence can be achieved with fewer than
30 data pairs, the following bagging strategies are recommended:
                               TOTAL NUMBER OF LEAK RATE MEASUREMENTS
RANGE (oomv) 4
1 -
101 -
1,001 -
10,001 -
> 100,
100 2
1,000
10,000
100,000 2
000
8
2
1
1
2
2
12
3
2
2
3
2
20
4
4
4
• 4
4
     These groupings and recommended number of sources are given as
guidelines.  They are based on experience in measuring leak rates and
developing leak rate/screening value equations.  Other source selection
strategies can be used if an appropriate rationale is given.
2.6.2  Development of Leak Rate/Screening Value Correlations
     With appropriate mass emission data and screening values,  leak
rate/screening value correlations can be generated.  Least-squares regression
analyses are done for each source type/service, regressing the  logarithm  of
the  nonmethane leak rate on the  logarithm of the screening concentration,
according to
     Log (leak rate) • BQ + BI Log  (screening concentration),
where
     B  , Bj « model parameters.
All  analyses should be done using logarithms of both the  leak rate and
screening concentration.  Confidence intervals should be  calculated  for  the
estimated equation  (Reference  19, pages 28-31).  A scale  bias correction
factor  (Reference 20) is required in transforming the equation  in the
log-scale back to the original units.  A statistician should be consulted
unless  one  is familiar with this type of analysis.  Bagged sources whose
screening values  are known to  be above 100,000 ppmv but whose actual  screening
                                      2-15

-------
values are unknown should not be used to fit the regression line described
above.  For those sources censored at 100,000 ppmv and which were not bagged,
leak rates should be estimated from the average leak rate for those censored
sources that were bagged.
2.6.3  Statistical Considerations of Leak Rate/Screening Value Correlations
     The predictive equations resulting from the least-squares regression
analyses discussed above must be statistically evaluated.
     If new correlation equations are developed, they must be shown to be
statistically different from the existing correlations developed during the
refinery and SOCMI studies.  If the new correlations are statistically
different from the existing correlations, the new correlations may be used in
the development of the unit-specific emission estimates for that specific
class of sources.  If not, the EPA's correlations should be used.  The
correlations can be compared graphically or mathematically.
     2.6.3.1  Graphical Comparison of EPA vs. Other Correlations.  If enough
data are collected to develop a process specific leak rate/screening equation,
then this equation should be compared to the published EPA equation.  The
EPA's predictive equation should be used unless the equation developed for the
specific process unit is found to give significantly different estimates.
This comparison can be done graphically by drawing the process specific
equation on a graph of the EPA equation.  Figure 2-2 (from Reference 14,
page 66) is an example of such a graph.
     A graphical comparison may also be made by plotting the leak
rate/screening data from the process on a log-scale plot of the EPA equations.
The log scale is preferred so that linear comparisons can be made.  Figures
2-3 through 2-5  (from Reference 14, pages 57-62) show such plots of the EPA
equations.  The  leak rate/screening value measurements used to develop the
equations are also shown on the plots.  These plots were developed from draft
versions of the  correlations.  The correlations were refined when censored
sources were  included  in modeling the leak rate/screening value relationships.
Only  small differences exist between the old and new models (see Reference 8,
pages  14-18).
                                      2-16

-------
ro
i
             0.08-
             0.07-
     0.06-
x:
To



0)


DC


ca
        0)
             0.05"
             0.04-
        T3
        0)
        r?    0.03 -
             0.02 -
             0.01-
                0
                                                                             X
                                                                           X
                                                                         X
                                                                   X
                              (95% Confidence Limiis of the Mean)
                                                                 x
                       —,	1	1	1	

                       20,000           40,000          60,000          80,000


                                 Maximum Screening Value (ppmv)
                                                                                             100,000
                 Figure  2-2.   Modeled relationship - leak rate vs.  OVA for valves In gas service.

-------
IN)
I
00
0)
•*-•

OC


(0
0)


0>
c



Q>
 c
 o
z
 en
 o
              1 -
              0-
            -1-
            -3-
            -4-
                             Legend:  • = 1 Observations
                                     • = 2 Observations

                                     A = 3-10 Observations
             -5-
                    1.5
                      2.0
2.5
                                             "T"
                                              3.0
                                                3.5
nr
 4.0
                                            Log Mean OVA Reading
                                  T"
                                   4.5
                                                                          5.0
5.5
6.0
                            Figure  2-3.   LogjQ  leak rate vs.  logjg OVA reading for pump  seals.

-------
Q)
(0

3

Q)
C
to


0>
C
o
z

O)
o
       0-
     -2-
     — 3 -
      _ 4 _
     -5-
                               Legend: • =  1 Observations

                                       • =  2 Observations

                                       A =  3-10 Observations

                                       O -  11-14 Observations
     -6-
           1.25
                    1.75
2.25
—T

 2.75
                                                T
      3.25     3.75


Log Mean OVA Reading
 1—

4.25
4.75
                                                                                   5.25
                                                                                           5.75
             Figure 2-4.  Log10 leak  rate vs. Iog10 OVA  reading for valves  in  gas  service.

-------
ro
i
0)


flC


CO
0)


0)
c
(0

*-•

-------
     If the points fall on both sides of the line representing the equation
and are within the range of data points used to develop equations, then  it is
reasonable to use the published equations for estimating emissions from  the
process.  Statistical tests, such as the Wilcoxon Paired Replicate Rank  test
(Reference 21, pages 401-403) can be used to aid in the decision of
reasonableness of the equations.
     As an example, Figure 2-6 (from Reference 14, page 79) shows two
comparisons of equations developed by EPA from Petroleum Refineries to the
SOCMI equations.  The first comparison (valves in gas service) shows the
refinery equation consistently estimates greater leak rates throughout the
curve.  The confidence limits for the two equations do not overlap.  For this
comparison, it is obvious that different equations are required for the  two
different types of process units.
     A second example of graphical comparison is shown in Figure 2-6 for
valves in light-liquid service.  In this case the refinery equation estimates
greater leak rates than the SOCMI equation, but the refinery equation is
within the confidence bands for the SOCMI equation..  The SOCMI equation  falls
below the lower confidence band for the refinery equation for valves screening
greater than 30,000 ppmv.  The comparison is not as clear as the valves  in gas
service.  Since there were about 300 data pairs used.to develop the SOCMI
equation, it should be used to develop estimates for SOCMI units.
     For comparisons of process specific equations to the EPA equations  there
will usually be fewer data sets used to develop the specific equations.  This
will result in much wider confidence bands for the process specific equations.
If 30 pairs of data are selected, the upper confidence band should be less
than a factor of two greater than the upper confidence band predicted by the
equation.
     The form of the graph can be modified to facilitate the comparison  of the
equations.  If most of the sources screen less than 10,000 ppmv, the equations
can be graphed on a scale from 0 to 10,000 ppmv.  The equations can also be
graphed on a log-scale to make the comparisons linear.  However, the
scale-bias correction factors would not be incorporated if the comparisons are
made in this way.  Therefore, this type of comparison is not recommended.
                                     2-21

-------
                 20,000      40,000      60,000      80,000
                  Maximum TLV Sniffer Reading (ppmv)

                            Valves-Gas
                                                      100,000
"35
a
CO

-------
     2.6.3.2  Statistical  Comparison of EPA vs. Other Correlations.  A more
formal  approach to comparing leak rate/screening value equations is to use
statistical tests to compare the parameters estimated by the equations.  The
important parameters of the leak rate/screening value equations (developed on
a log-scale) are the following:
          1)  the intercept (Bo);
          2)  the slope (Bl);
          3)  the standard error (Se).
Statistical tests for the slope and intercept are also available (Reference
19, page 24-33).
     Ninety-five percent confidence intervals for the slope and intercept were
developed for each of the SOCMI leak rate/screening value equations
(Reference 8, pages 24-25).  For example, the slope of the equations for
valves in gas service using the OVA was estimated to be 0.693.  The 95 percent
confidence interval for this slope is given as 0.53 to 0.85.   If a slope
estimated from a particular SOCMI process unit is within this  confidence
interval, then the EPA equation provides a reasonable slope for a predictive
equation.  If the slope estimate is outside the confidence interval for the
EPA correlation, the confidence interval for the slope estimated by the
unit-specific equation must be evaluated.  Reference 19 indicates how  to make
these evaluations.
     A statistical test to compare the  standard errors for two equations can
be done  using an F-test (Reference 20,  page 31 or Reference 21, pages
111-116).   Since the standard  error is  the key component of the scale-bias
correction  factor, this comparison of  standard errors will also serve  as a
comparison  of the  scale bias correction factors incorporated  in the two
correlations.
     2.6.3.3  Other Statistical Analyses.  There are other types of
statistical  analysis that  can  be used  in the development, evaluation,  and
comparison of  leak rate/screening value equations.  A General  Linear Models
analysis of the  leak/rate  screening value data  (Reference 20,  Chapter  2) can
 be used  to develop equations and to test for different leak rate/screening
 relationships  for  different  groups  of data.  This approach was used  in
 Reference 14 (pages 51-53)  to  evaluate differences  between the cumene,
 ethylene,  and  vinyl acetate  units.

                                      2-23

-------
2.6.4  Generation of Emission Estimates
     The generation of the emission estimate using correlations developed  for
specific process units makes use of the empirical screening data and the leak
rate/screening value correlations.  The leak rate/screening value correlations
are applied to the screening data to calculate predicted emission rates.
     Emission rates are estimated for all emitting sources (including those
sources which were bagged) by:
     Predicted emission rate - exp [B  + B, log (screening concentration)]
                                       x (scale bias correction factor)
     B  and B, are the coefficient estimates either published by EPA
(Appendix D) or developed from a linear regression analysis, if site specific
correlations are used.  The results presented in arithmetic scale are adjusted
to represent mean emission rates using the scale bias correction factor.
     The sum of the individual predicted leak rates represents the total mass
emission estimate for a given type of source and service being considered.  An
emission estimate per source can be derived by averaging the total emission
estimate over the number of sources in the distribution.
2..6.S  Statistical Considerations for the Predicted Emission Estimate
     This section presents formulas for calculating the 95 percent confidence
intervals for the predicted emission estimate.  For clarity, formulas for
calculating the emission estimate per source and total emission estimate are
presented also.  Definitions for the formulas are presented in the Glossary.

For the Average Emission Rate Per Leaking Source (EE,):
                                   Nl
          EE, without censoring - i^l  li  ,

with a 95 percent confidence interval of:
          EE1 ± ASvl
                                      2-24

-------
          EE,, with censoring • Nla Yla * NlbYlb  ,
            1                          "l
with a 95 percent confidence interval of:
                 l                 lb
For the Average Emission Rate Per Non-leaking Emitting  source  (EF,.):
                 &
          EE, - fi Y2i ,
                  N2
with a 95 percent confidence interval of:
          EE, -i-  tsv2
For the Average Emission Rate Per Source  (EE):
          EE . N1EE1 * N2EE2  ,
                Nj + N2 + N3
with 95 percent confidence  intervals of:
          EE ±_L__ {[N^s.]2 +  [N,sv2]2}1/2 without  censoring.
               NT + N2 + N3    * yi        * y*
Total  Emissions + N x EE (use the EE confidence  limits  above  multiplied  by N)
                                      2-25

-------
                             3.0  SOURCE SCREENING

     This section presents protocols and methodologies for conducting
screening of equipment components with a portable organic analyzer.  The
overall survey procedure is presented to give the perspective of a unit-wide
screening plan.  In addition, screening protocols are given for the different
equipment types.  The selection of an appropriate portable monitoring
instrument is discussed and a section on data handling is also included.
3.1  SCREENING SURVEY PROCEDURES
     For development of a unit-specific emission estimate, the concentration
of organic materials at seals, shafts, and other potential leak points must be
measured.  In preparing to screen all equipment in a unit, all equipment to be
included in the emission estimate unit needs to be identified.  A listing of
equipment types that are potential sources of fugitive emissions is given in
Table 3-1.
     The first step in the screening survey is to define precisely the process
unit boundaries.  This is usually quite straightforward, but occasionally
multiple units may be built on the same pad and share some common facilities.
A process unit can be defined as the smallest set of process equipment which
can operate independently and includes all operations necessary to achieve its
process objective.  The exact basis for the unit definition should be
documented.  A plot plan of the unit should be obtained and marked with the
appropriate boundaries.
     The next  step is to obtain a simplified flow diagram of the process and
note the process streams.  The actual screening and data collection can be
done most systematically by following each stream.  For instance, a logical
starting point would be where one of the  feed lines enters the process
boundary.  The screening team would follow that line, screening all sources,
until  the line terminates at the flanges  of a reactor or separation step.
Screening would then continue on the outlet side of the reactor or separation
equipment.  Minor loops like a bypass around a control valve, pump, or heat
exchanger should be screened on the initial pass.  Larger loops of process
equipment, such as parallel passes and processing alternatives, are more
                                       3-1

-------
TABLE 3-1.  FUGITIVE EMISSION SOURCES

           Equipment Types
             Pump seals
          Compressor seals
               Valves
       Pressure relief devices
        Sampling connections
 Flanges, screwed connections, etc.
          Open-ended lines
        Drains, vents, doors
           Agitator seals

               Service
              Gas/vapor
            Light liquid
            Heavy liquid
                 3-2

-------
effectively treated as separate streams.  If the emission estimate  is being
generated in support of control strategies for individual chemical  compounds,
this approach offers the advantage of screening groups of equipment components
with roughly the same composition of materials in the line.
     Each source should be uniquely identified to indicate that it  has been
screened.  For example, sources can be tagged.  Tags can consist of any form
of weatherproof and readily visible identification.  Alternatively, a process
unit can be considered appropriately tagged if the unit has a system of
identifying markings with an associated diagram allowing easy location of
marked sources.  Once all the equipment along the major streams has been
screened, the unit should be divided into a grid to search for fittings missed
on the initial survey.  The unit survey is completed when all sources in the
unit have been either screened or identified as a nonhydrocarbon fitting.
Leak!ess equipment and equipment not in VOC (or VHAP) service should not be
included in the screening survey.  Equipment documented as inaccessible should
be included in the survey; such equipment must be screened annually under
regulations for fugitive emissions.
     Consistent with equipment leaks standards for hazardous air pollutants,
unsafe-to-monitor equipment do not need to be included in the survey.
Documentation must be provided, however, to substantiate the unsafe nature of
such equipment.
3.2  MONITORING INSTRUMENT
     A number of portable VOC detection devices are capable of measuring leaks
from process units.  These devices operate on a variety of principles, the
three most common being  ionization, infrared absorption, and combustion.  Any
analyzer can be used provided  it meets the specifications and performance
criteria set forth in EPA Reference Method 21.   Reference Method 21 is
included in this document as Appendix A.  Any analytical instruments are
permitted provided they  are shown to measure the organic compounds  of
interest and the results are related to EPA's data base generated using a
flame ionization detector, calibrated to methane.  Analytical instruments
referenced to compounds  other  than methane must develop response factors to
relate screening values  to concentration.  This will allow use of many
instruments that cannot  be calibrated with methane.
                                      3-3

-------
3.2.1  Operating Principles and Limitations of Portable VOC Detection Devices
     lonization detectors operate by Ionizing the sample and then measuring
the charge (number of ions) produced.  Two methods of ionization currently
used are flame ionization and photoionization.
     A standard flame ionization detector (FIO) usually measures the total
carbon content of the organic vapor sampled.  Although carbon monoxide and
carbon dioxide do not produce interferences, FID analyzers do react to water
vapor, although at a low sensitivity.  Furthermore, if water condenses in the
sample tube, erratic readings may result.  A filter is used to remove
particulate matter from the sample.  When equipped with options, certain
portable FIO instruments are capable of measuring total gaseous nonmethane
organics or individual organic components.  Certain organic compounds
containing nitrogen, oxygen, or halogen atoms when sampled with an FID give a
reduced response; and some organics may not give any response at all.  For
this reason, it is necessary to develop response factors for each compound
that is to be measured.  Response factors are discussed in Section 3.2.2.1.
     Photoionization detectors use ultraviolet light (instead of a flame) to
ionize organic vapors.  As with FID's, the detector response varies with the
functional group in the organic compounds.  Photoionization detectors have
been used to detect leaks  in process units used in the synthetic organic
chemical manufacturing industry, especially for certain compounds such as
formaldehyde that will not give a response on an FID or combustible detector.
     Nondispersive  infrared (NDIR) instruments operate on the principle of
light  absorption characteristics of certain gases.  NDIR instruments are
usually subject to  interference because other gases such as water vapor and
carbon dioxide may  also absorb light at the same wavelength as a compound of
interest.  These detectors are generally  used only for the detection and
measurement of single components.  For the detection and measurement of single
components, the wavelength at which a certain compound absorbs infrared
radiation  is predetermined and the device is preset for that specific
wavelength  by using optical filters.  For example, if set to a wavelength of
3.4 micrometers,  infrared  devices can detect and measure petroleum fractions,
including  gasoline  and naphtha.
                                       3-4

-------
     Combustion analyzers are designed either to measure the thermal
conductivity of a gas or to measure the heat produced by combusting the gas.
The most common method used in portable VOC detection devices is measuring the
heat of combustion.  These devices are referred to as hot wire detectors or
catalytic oxidizers.  Combustion analyzers, like most other detectors, are
nonspecific for gas mixtures.  In addition, combustion analyzers exhibit
reduced response (in some cases, no response) to gases that are not readily
combusted, such as formaldehyde and carbon tetrachloride.
3.2.2  Performance Criteria and Evaluation for Portable VOC Detectors
     As previously stated, any portable VOC detector may be used as a
screening device directly provided it meets the performance criteria specified
in Reference Method 21 (see Appendix A).  Although this approach has been
shown to be applicable to many organics, it cannot be universally applied.
Facilities may need to develop an appropriate method for testing inorganic and
some organic compounds.  A discussion of the performance criteria and detector
evaluation procedure is presented below and summarized in Table 3-2.
     In addition to the performance criteria, Reference Method 21 also
requires that the analyzer meet the following specifications:
     •    The VOC detector shall respond to those organic compounds being
          processed (determined by the response factor).
     •    The analyzer shall be capable of measuring the leak definition
          specified in the regulation, (i.e., 10,000 ppmv or "no detectable
          limit").
     t    The scale of the analyzer shall be readable to +5 percent of the
          specified leak definition concentration.
     •    The analyzer shall be equipped with a pump so that a continuous
          sample is provided at a nominal flow rate of between 0.5  and 3.0
          liters per minute.
     •    The analyzer shall be intrinsically safe for operation in explosive
          atmospheres.
     Also, criteria for the  calibration gases to be used are specified.  Two
or more calibration gases  are  required for analyzer performance evaluation.
One  is  a  zero gas  which  is air with less than 10 ppmv VOC.  The other
calibration gases,  or reference gases, use reference compounds in air
mixtures.  The  concentration of the reference gas should represent  the range

                                      3-5

-------
         TABLE 3-2.   PERFORMANCE CRITERIA FOR PORTABLE  VOC DETECTORS3
   Criteria
  Requirement
    Time Interval
Instrument response
  factor
Instrument response
  time
Calibration precision
Must be <10 unless
correction curve is
used.

Must be <30 seconds
Must be <10% of
calibration gas value
One time, before detector
is put in service.
One time, before detector
is put in service.  If
modification to sample
pumping or flow configura-
tion is made, a new test
is required.

Before detector is put in
service and at 3-month
intervals or next use,
whichever is later.
 Sourc.e:  Reference Method 21.
                                      3-6

-------
of responses measured.  For the purpose of developing unit-specific emission
estimates, a reference gas for the appropriate range should be used.
     3.2.2.1  Response Factor.  The sensitivity of an analyzer varies,
depending on the composition of the sample and concentration detected.  The
response factor (RF) helps to quantify the sensitivity for each compound.  The
response factor is defined by:

                                     Actual concentration of compound
               Response Factor  -  observed concentration from detector

     It is important to note that the response factor may be used as a guide
to selecting an appropriate monitoring device.  For example, from the data
presented in Appendix B, it can be readily seen that in screening equipment in
a process unit containing cumene, an FID can be used (RF - 1.87) directly,
with no correction for response factor; while the catalytic oxidation detector
cannot (RF has no value).  Similarly, from the same data, neither of these
devices would be capable of detecting leaks from a source containing carbon
tetrachloride if response factor adjustments are not used.
     Using response factors as a general guide to analyzer applicability  is
especially important when dealing with chemical mixtures, since many process
streams in industrial plants are composed of mixtures of chemicals..  One  EPA
study concluded that  analyzer response factors for a chemical mixture fall
between the  responses expected for the pure components.   Therefore, if
desired,  an  interpolated or weighted average can be used to predict the
response  for mixtures based on known responses for individual chemicals.   If
sufficient  information on pure components  is not available or  if desired,  the
response  factor of  a  mixture  for a specific instrument may be determined
experimentally in the laboratory.  Such results should be well documented.
For  further  information see EPA-600/2-81-110, Response of Portable  VOC
Analyzers to Chemical Mixtures,5 and EPA-600/2-8-051, Response Factors  of VOC
Analyzers at a Meter  Reading  of  10,000 ppmv for Selected Organic Compounds.
      Response factors also may be used to  correct screening value measurements
gathered  to  estimate  emissions when using  an emission factor  approach  (e.g.,
leak/no-leak approach or  three-strata  approach).  If using leak rate/screening
value correlations,  however,  a standard curve covering the entire range  of
                                       3-7

-------
screening values measured must be used.  A response factor must be determined
for each compound that Is to be measured, either by testing or from  reference
sources.*'5'6  An analyzer's response factor for Individual compounds  to  be
measured must be less than 10 at or near the concentrations defining emission
factors (e.g., 1,000 and 10,000 ppmv) to be acceptable.  The response  factor
tests are required before placing the analyzer In service, but do not  have to
be repeated at subsequent Intervals.
     Response factors can be determined by the following method.  First,  the
analyzer Is calibrated using the reference gas.  Then, for each organic
species that 1s to be measured, a known standard In air Is obtained  or
prepared.  This mixture Is then Injected Into the analyzer and the observed
meter reading Is recorded.  The analyzer is then zeroed by injecting zero air
until a stable reading is obtained.  The procedure is repeated by alternating
between the mixture and zero air until a total of three measurements is
obtained.  A response factor is calculated for each repetition and then
averaged over the three measurements.
     Alternately, if response factors have been published for the compounds of
interest for the combination of detector and calibration gas desired,  the
response factor determination is not required, and existing results  may be
referenced.  Results of a study developing response factors of FID (Foxboro
OVA-108 and OVA-128) and catalytic oxidation (J. W. Sacharach TLV Sniffer)
                                                   A
analyzers are presented in Table B-l of Appendix B.   These response factors
can be used when determining if a screening concentration is above or  below
10,000 ppmv.  The values are for pure organic chemicals only.  Table B-2  of
Appendix B presents tested compounds that appear unable to achieve an
instrument response of 10,000 ppmv at any feasible concentration unless
response factors are used.   These single response factors are adequate for
response factor adjustments when using one of the emission factor approaches
for estimating emissions.  When applying correlations, however, an accurate
correction for the non-linear response of an instrument is needed.   Response
factors must be evaluated at several concentrations to establish a correction
curve  for use  in translating instrument readings.
     3.2.2.2   Response Time.  The response time of an analyzer is defined as
the time interval from a step change in VOC concentration at the input of a
                                       3-8

-------
sampling system, to the time at which 90 percent of the corresponding  final
value is reached as displayed on the analyzer readout meter.  The response
time must be equal to or less than 30 seconds.  The response time must  be
determined for the analyzer configuration that will be used during testing,
and the response time test is required before placing an analyzer in service.
If a modification to the sample pumping system or flow configuration is made
that would change the response time, a new test is required before further
use.
     3.2.2.3  Calibration Precision.  Calibration precision is the degree of
agreement between measurements of the same known value.  To ensure that the
readings obtained are repeatable, a calibration precision test must be
completed before placing the analyzer in service, and at three-month
intervals, or at the next use, whichever is later.  The calibration precision
must be equal to or less than 10 percent of the calibration gas value.
     To perform the calibration precision test, a total of three measurements
is required for each nonzero concentration.  Measurements are made by  first
introducing zero gas and adjusting the analyzer to zero.  The specified
calibration gas (reference) is then introduced and the meter reading is
recorded.  The average algebraic difference between the meter readings  and the
known value of the calibration gas is then computed.  This average difference
is then divided by the known calibration value and multiplied by 100 to
express the resulting calibration precision as percent.
     3.2.2.4  Safety.  Portable instruments to detect VOC emissions from
equipment leak sources are required to be used in hazardous locations  such as
petroleum refineries and bulk gasoline terminals.  The National Electrical
Code requires that instruments to be used in hazardous locations be certified
to be explosion-proof, intrinsically safe, or purged.
     Hazardous locations are divided into three classes:  Class I, Class II,
and Class III.  Each class is divided into two divisions (Division 1 or 2)
according to the probability that a hazardous atmosphere will be present; and
also into seven groups depending on the type of hazardous material exposure.
Groups A through 0 are flammable gases or vapors, and Groups E, F, and G apply
to combustible or conductive gases.  Class I, Division 1, Groups A, B, C, and
D locations are those in which hazardous concentrations of flammable gases or
                                      3-9

-------
vapors may exist under normal operating conditions.  Class I, Division 2,
Groups A, B, C, and 0 locations are those in which hazardous concentrations of
flammables may exist only under unlikely conditions of operation.
     As of 1982, five manufacturers produced portable VOC detection
instruments that are certified to be intrinsically safe.  Table 3-3 lists
these manufacturers, approved instrument model numbers, instrument
certification categories, and performance specifications.  Additionally, newer
instruments may be available that meet the performance requirements for
generating emission estimates.
3.2.3  Monitoring Devices For Difficult Situations
     In some cases, a monitoring device may not be available that meets all of
the performance specifications of Method 21.  For example, there are several
cases (e.g., phosgene) where the response factor at 10,000 ppmv is greater
than ten (10).  The instrument may meet all other requirements, but fails as  a
Method 21 instrument because it cannot meet the response factor requirement.
The instrument can still be used to estimate emissions from equipment leaks,
provided the instrument is shown to be sufficiently, reliable in providing data
that can be related to EPA's data collected using an OVA calibrated to
methane.
     Several steps must be taken initially to document the viability of a
device that fails to meet the Method 21 requirements.  First, a laboratory
program must be undertaken to demonstrate the response of the monitoring
instrument to the compounds being measured.  This response must be documented.
The second step involves relating the instrument response (i.e., screening
value) to actual concentrations, or in other words, developing an instrument
response curve.  The screening value response curve must be developed for the
entire screening value range and documented so that screening values taken  in
the field can be adjusted to actual concentrations.  Third, the testing
program  should be sufficiently well documented to demonstrate how the
instrument will be used in screening equipment.  For example, if the response
time of  the candidate instrument exceeds the Method 21 performance
specification, the test plan should reflect added screening time at each
potential leak point to be screened.  Once this laboratory demonstration has
been completed and the  screening value correction curve has been established,
the screening  can begin.
                                     3-10

-------
                                           TABLE 3-3.  PORTABLE VOC DETECTION  INSTRUMENT PERFORMANCE SPECIFICATIONS
Manufacturer
Bacharach Instrument Co.,
Santa Clara, California

Pollutant(»)
Model No. Detected
L Combustible gases
TLV Sniffer Combustible g*<«*
Principle of
Operation
Catalytic
Combustion
Catalytic
Combustion
Range
(PP»)
0-100X LEL
0-1,000 and
0-10,000
Certification
Intrinsically safe. Class I,
Division 1, Groups C and D
Intrinsically safe, Clas* I,
Division 1, Groups C and Di and
             Fbxboro,
             S.  Norwalk,  Connecticut
                                               OVA-128      Total hydrocarbons
                                               OVA-108      Total hydrocarbons.
                             Class I, Division 2, Groups A
                             and B

  tlDIGC       0-1,000       Intrinsically safe. Class I,
                             Division 1, Groups A, B, C and D

  FID/GC       0-10,000      Intrinsically safe, Class I,
                           •  Division 1, Croups A, B, C and D
CO
 I
             UNU Systems,  Inc.
             Newton Upper  Falls,
             Massachusetts
             Mine Safety Appliances  Co.
             Pittsburgh, Pennsylvania
             Survey and Analysis,  Inc.,
             Northboro, Massachusetts
PI-101       Chlorinated hydrocarbons,
             aromatlcs, aldehydes,
             ketones, any substance
             which UV light lonlces

  40         Combustible gases
  Photo-
lonlxatlon
Catalytic
Combust Ion
On Mark      Combustible gases and      .  Thermal
Model 5      vapors                    Conductivity
                                                       0-20,
                                                       0-200,
                                                       0-2,000
                                                       0-10* and
                                                       0-100X LEL
               0-5 and
               0-1001 LEL
Intrinsically safe. Class I,
Division 2, Groups A, B, C and D
Intrinsically safe, Class I,
Division 1, Group DI and Class I,
Division 2, Groups A, B, and C

Intrinsically safe. Class I,
Division 1, Groups A, B, C and D

-------
3.3  SCREENING PROCEDURES
3.3.1  Calibration Procedures
     Pretest procedures to be followed include calibration of the monitoring
instrument.  The VOC analyzer should be assembled and started up according to
the manufacturer's instructions.  After the appropriate warm-up period and zero
internal calibration procedure, the calibration gas is introduced into the
sample probe.  The instrument meter readout is then adjusted to correspond to
the calibration gas value.  If the meter readout cannot be adjusted to the
proper value, a malfunction of the instrument is indicated and corrective
measures should be taken before use.
3.3.2  Procedure for Screening Equipment
     The mechanics of the screening operation outlined in Reference Method 21
are summarized in the following discussion.  The probe inlet is placed at the
surface of the leak interface where leakage could occur.  (The leak interface
is the boundary between the process fluid and the atmosphere.)  Care must be
taken to ensure that the probe is held perpendicular, not tangential, to the
leak interface.  Otherwise, inaccurate readings will result.  The probe must
then be moved along the interface periphery while observing the instrument
readout.   If an increased meter reading is observed, slowly move the probe
along the  interface where leakage is indicated until the maximum meter reading
is obtained.  The probe inlet should be left at this maximum reading location
for approximately two times the instrument response time.  The maximum reading
is recorded as the screening value.
     The instrument measurement may exceed the scale of the instrument.  For
example, the reading may exceed 10,000 ppmv when using an OVA analyzer.  For
the purposes of generating an emission estimate, these higher readings also
must be recorded.  A dilution probe should be employed to allow measurement of
concentrations greater than the instrument's normal range.  The OVA can be
equipped with a dilution probe that permits measurement of concentrations up
to 100,000 ppmv.  It is important to note that extending the measurement range
also necessitates the calibration of the instrument to the higher
concentrations.
     Care  should  be taken to avoid  fouling of the probe with grease, dust, or
even liquids.  A  short piece of Teflon tubing can be used as a probe tip
                                      3-12

-------
extender and snipped off as the tip fouls.  In areas with a noticeable
particulate loading, this tubing can be packed with untreated fiberglass to
act as a filter.  (Note that the instrument must also be calibrated with this
filter in place).  If a surface to be screened is obviously dirty, hold the
probe tip just over the surface to avoid scooping up contaminants.  Some
fouling is unavoidable, so it is recommended that the probe tip sintered steel
filter be cleaned at least daily and the side-pack filter on a weekly basis.
Normally, these filters can be cleaned by just rapping them lightly on a table
top, but if the deposits are wet and caked on, washing with an aqueous
solution of soap and alcohol is recommended.  This can also be used to wash
the probe and transfer line periodically.  Care should be taken to blow the
equipment dry before reuse.
     This general procedure can be used to screen equipment such as valves,
flanges, pumps and compressors, pressure relief devices, and other potential
sources of VOC leakage such as process drains and open-ended lines or valves.
     For valves, the most common source of leaks is at the seal between the
stem and housing.  To screen this source, the probe, is placed where the stem
exits the packing gland.  The probe is moved around the stem circumference.
The maximum reading is recorded as the screening value.  Also, the probe is
placed at the packing gland take-up flange seat and the probe is moved along
the periphery.   In addition, valve housings of multipart assemblies should be
screened at the  surface of all points where leaks could occur.  Figures 3-1
through 3-5 illustrate screening points for several different types of valves.
     For flanges, the probe is placed at  the outer edge of the flange-gasket
interface and the circumference of the flange is sampled.  For screwed
flanges, the threaded connection interface must also be screened.  Other types
of nonpermanent  joints,  such as threaded  connections, are sampled with a
similar traverse.
     Pumps  and  compressors are screened with a circumferential traverse at the
outer  surface of the pump or compressor shaft and seal  interface where the
shaft  exits the  housing.   If the source is a rotating shaft, the probe inlet
is positioned within 1 cm of the shaft-seal interface.  If the housing
configuration prevents a complete traverse of the shaft periphery, all
accessible  portions must be sampled.  All other joints  on the pump or
                                      3-13

-------
                                                                       aut
                                           Figure 3-lb.   Non-rising stem type.
Figure 3-la.  Rising stem type.
                            Figure 3-1.  Gate  valves.
                                    3-14

-------
Figure 3-2a.   Manual Globe Valve
                                            Figure 3-2b.  Globe type control valve
                     Figure 3-2.  Globe valves.



                                   3-15

-------
Figure 3-3.   Lubricated plug valve,
                      3-16

-------
 «o
CO


  •
 fO
**•
 I
                                                            tt»
                                                           H-

                                                            
-------
     Sum
Figure 3-5a.  Weir type diaphragm valve.
              Figure 3-5.
                                      Figure 3-5b.   Check  valves,
                            3-18

-------
compressor housing where leakage could occur should also be sampled.
Figure 3-6 illustrates screening points for two types of centrifugal pumps.
Agitator seals should be screened in a similar manner.
     The configuration of most pressure relief devices prevents sampling at
the sealing seat.  Because of their design and function, pressure relief
devices must be approached with extreme caution.  These devices should not be
approached during periods of process upsets, or other times when the device  is
likely to activate.  Similarly, care must be used in screening pressure relief
devices to avoid interfering with the working parts of the device (e.g., the
seal disk, the spring, etc.)  For those devices equipped with an enclosed
extension, or horn, the probe inlet is placed at approximately the center of
the exhaust area to the atmosphere.  Again, only the probe should be placed  in
the horn; personnel conducting the screening should not place hands, arms, or
any parts of the body into the horn.  Figure 3-7 illustrates the screening
points for a spring-loaded relief valve.
     Fugitive leaks from most other sources such as process drains, open-ended
lines or valves, and seal system degassing vents and accumulator vents, are
emitted through  a regularly shaped opening.  If that opening is very small (as
in sampling lines of less than one inch diameter), a single reading in the
center is sufficient.  For larger openings  (like a six  inch drain mouth),  it
is necessary to  traverse the perimeter of the opening.  The concentration  at
the center must  also be read.  For even larger sources  (like a wash-up drain
grate), a grid of readings should be taken on about six inch centers.  For
access door seals, the probe inlet is placed at the surface of the door seal
for the peripheral traverse.   For all of these types of equipment,  it  is the
maximum value that is recorded as the screening concentration.
3.4  DATA HANDLING
     To handle the screening data uniformly, it is recommended that data be
recorded on prepared data sheets.  The data collected should include the
following:
      1.   Date.
      2.   Hydrocarbon detector type.
      3.   Source identification.   (If permanent ID's  are  not in place, assign
           ID'S consecutively as  each source is screened.   The first source
          screened is assigned  ID1, the second source screened, ID2, and so
          on.)
                                     3-19

-------
Figure 3-6a.  Vertical centrifugal pump.
                                        Figure  3-6b.   Horizontal  centrifugal  oumo
                            Figure  3-6.
                                    3-20

-------
Figure 3-7.   Spring loaded relief valve.
                   3-21

-------
     4.   Screening value recorded in ppmv.
     5.   Source type (i.e., type of valve, pump, compressor, flange, etc.)
     6.   Service.  For example, gas, light liquid, and heavy liquid.  Liquids
          are classified based on their most volatile component present at
          20 weight percent or more.  If the components have a total vapor
          pressure equal to or greater than 0.04 psi at 20 C, the material
          (containing greater than or equal to 20 percent VOC) is classified
          as a light liquid; if not, it is classified as a heavy liquid.
          Classification is based upon the actual process conditions, not
          ambient conditions.
     7.   Comments.  If any explanation is required, it should be noted in a
          "comments" section.  An example data sheet is given in Table 3-4.
     In some cases, it may be necessary or desirable to adjust the screening
values for response factor.  In these cases, the data sheet should be designed
to accommodate extra columns for response factor and corrected screening
values.
3.5  CALIBRATION PROCEDURES FOR QUALITY ASSURANCE
     Calibration procedures must be used for quality control to ensure the
collected data are of high quality and can be compared to the data already
gathered by EPA.
     Each screening instrument should be checked for calibration before each
use and readings from these checks recorded.  As an example, instruments
should be calibrated each morning before use and before use in the afternoon.
The calibration should also be checked periodically (such as during breaks in
the daily testing schedule) to ensure that calibration has not drifted.  The
instruments should be calibrated if the reading is off by more than +5% on the
high standard, or +20 on the low standard.  If more than one instrument is
being  used at one process unit, the calibration readings for all instruments
must be recorded.
                                     3-22

-------
TABLE 3-4.  EXAMPLE DATA SHEET

Date





V
CJ









Hydrocarbon Detector
Type
















Source
ID
















Screening Value
(ppow)





.

'








Proceaa
Unit










,





Service
















Print ry
Material
















Comment a
















-------
                                  SECTION 4.0
                           MASS EMISSIONS SAMPLING

     The process of measuring organic concentrations at potential leak points
of equipment when combined with the EPA's leaking and non-leaking emission
factors gives a good estimate of the leak rate from a process unit.  More
rigorous procedures may be used to determine the mass emission rate more
accurately.  Measuring mass emission rates requires source enclosure combined
with sampling of the gas in the enclosure.  The dilution or vacuum method is
the preferred technique for sampling the emissions from sources that can be
enclosed.  An alternate sampling technique is the blow-through method which is
usually used for sources with very large leak rates (in excess of the vacuum
pump capacity).  An important step in either of these methods, however, is
source enclosure, a procedure called bagging.
4.1  VACUUM METHOD
     In the past, the method preferred for sampling leaks from baggable
sources has been the dilution or vacuum method.  The- sampling train used in
this method is shown in Figure 4-1.
     The train can be mounted on a portable cart, which can be easily moved
around the process unit from source to source. -The major equipment items in
the sampling train are the vacuum pump used to draw air through the system,
and the dry gas meter used to measure the flow rate of gas through the train.
In previous studies that EPA conducted, a 4.8 CFM Teflon-ring piston type
vacuum pump equipped with a 3/4-horsepower air-driven motor was used.  Low
pressure air (-100 psig) is available at or near most process units.  An
example of a suitable dry gas meter is a Rockwell Model 1755 Test Gas Meter
                            12
with a Number 83 Test Index.    Other equipment that EPA used in the train
includes Whitey valves, copper and stainless steel tubing, Teflon tubing, 100
cc glass airtight syringe, thermometers, mercury and water manometers, a cold
trap, and an air-driven diaphragm sampling pump.
     The leak source is shown as a valve in Figure 4-1.  However, the same
sampling train can be used for all equipment to be sampled using the vacuum
technique.  As described in Section 4.3, the size and shape of the enclosure
(tent) around the potential leak point is adjusted to fit the particular shape

                                      4-1

-------
Magnahellc
  Tent
                  Thla line ahould
                    be aa abort
                    aa poaalble
                   Cold Trap
                   (Ice Bath)
                                                       Trap
&
                                                    a
                                                   Meter
             3t,
                                                                      Hg Manometer
                Control
                Valve
Tf
ZTP3
&«
                                                                                         Vacuum Pump
                                                                             Filter
      Small
    Diaphragm
      Pump
                                                                              Sample Bag
                                                                         Two Way Valve
      Figure 4-1.   Sampling train for baggable  sources of hydrocarbon emissions using a diaphragm.

-------
and operating condition of the equipment.  When the full sampling train  is
connected, the vacuum pump is able to maintain a maximum flow rate of
approximately 2.5 cubic feet per minute (0.001 cubic meters per second).
     The tent is connected by means of a bulkhead fitting and Teflon tubing to
the sampling train.  A separate line is connected from the tent to a
magnahelic to allow continuous monitoring of the pressure inside of the  tent.
If a significant vacuum exists inside the tent when air is being pulled
through, a hole is made in the opposite side of the tent from the outlet to
the sampling train.  This allows air to enter the tent more easily and thus
reduces the vacuum in the enclosure.  In practice, it has been found that only
a very slight vacuum (0.1 inches HgO) is present in the tent during most of
the sampling, even in the absence of a hole through the tent wall.  Sufficient
air enters around the seals to prevent the development of a significant vacuum
in the tent.  Sample bags are used to collect gas samples and transport them
to the laboratory for analysis.  As with the tent material, sample bags must
be constructed of material that is impermeable to hydrocarbons so that the
process components are contained within the sample bag.  Mylar plastic,
Teflon, and Tedlar are materials well suited to this function.  A typical
sample bag is shown in Figure 4-2.
     A cold trap is placed in the system to condense water and heavy organics,
thus preventing condensation in downstream lines and equipment.   The cold trap
can be simply a 500 ml flask in an ice bath placed as close as possible to the
tent.  While not always necessary, this ice bath cold trap is very effective
in preventing condensation in the remainder of the sampling train and in the
gas sample bag.  Any organic condensate that collects in the cold trap must be
measured for later use in calculating total leak rates.  On some streams the
use of a cold trap is critical.
     The flow rate through the system can be varied by throttling the flow
with a control valve immediately upstream of the vacuum pump.  As the flow
rate is decreased, the concentration of hydrocarbon increases in the gas
flowing through the sampling system.  This allows considerable flexibil-'cy in
avoiding operations with an explosive mixture of hydrocarbon in  the air.  If
an explosive mixture is present, the blow-through method should  be used.
                                      4-3

-------
Figure 4-2.   Mylar plastic sample bag.
                  4-4

-------
     In summary,  the vacuum sampling procedure consists of the allowing
steps:
     •     Obtain  and record a screening value with the portable hydrocarbon
          detector,
     0     Enclose the source in a tight shroud,
     •     Connect the tent to the sampling train,
     •     Immerse the cold trap in an ice bath,
     •     Note the initial reading of the dry gas meter,
     •     Start the vacuum pump and a stopwatch simultaneously,
     •     Record  the temperature and pressure at the dry gas meter,
     t     Observe the VOC concentration at the vacuum pump exhaust with the
          hydrocarbon detector,
     t     Record  the temperature, pressure, dry gas meter reading, outlet VOC
          concentration and elapsed time every 2 to 5 minutes,
     •     Fill a  gas sample bag from the discharge of the Teflon-lined
          diaphragm pump when the outlet VOC concentration stabilizes (i.e.,
          the system is at equilibrium),
     •     Fill another bag with ambient air near the tent area to detect
                                           *
          the background VOC concentration,
     •     Take a  final set of readings and stop the vacuum pump,
     •     Remove, seal, and transport the cold trap is to the laboratory
          along with the two bag samples and the data sheet,
     •     Remove  the tent,
     •     Rescreen the source with the portable hydrocarbon detector and
          record.
All of the above  data and any pertinent comments should be recorded in a
permanent laboratory notebook.
      This bag sample may be omitted in favor of the hydrocarbon detector
reading if the background VOC concentration is less than 5 percent of the
lower limit of the screening range (i.e., <50 ppm for the 1,000 to 10,000 ppm
class). When bagging a "zero" screening value component, it is extremely
important to take a background bag.  Data obtained without a background  ag
will bias emission factors on the high side unless this correction is rr  e.
                                      4-5

-------
4.2  BLOW- THROUGH METHOD3
     Blow-through refers to blowing nitrogen  through  a  flexible  tent to create
a constant VOC concentration  inside the tent.  Nitrogen is  metered into the
tent through  1 or 2  polyvinyl chloride tubes.  The  temperature  and oxygen
concentrations are measured inside the tent with  a  platinum-RTD  thermocouple
and an oxygen/combustible gas monitor.  The flow  of nitrogen  is  monitored in a
gas rotameter calibrated to nitrogen.  The nitrogen passes  though  activated
charcoal  and  drierite  to remove  any organics  and  moisture.  The  pressure in
the tent  never exceeds 1 psig.   Figure 4-3 illustrates  the  equipment required
for the blow- through method.
     At least two OVA  (Foxboro 108 Organic Vapor  Analyzer  ) measurements are
made inside the  tent,  each measurement at two or  more differing  nitrogen flow
rates  to  increase accuracy.   [The OVA must be attached  to  a dilution probe and
calibrated to nitrogen-diluted gases to allow its use in the  nitrogen
atmosphere in the tent.]  The OVA readings are converted to emission rates of
VOC via response factors for  the OVA; the calculation is shown  in  Table 4-1.
Alternatively, gas samples from  the tent can  be collected  with  a portable
.sampling  pump and transported to a lab for chemical speciation  and
concentration measurement.  The  calculation in Table  4-1 is also appl i cable
for concentrations measured in this manner  (except  that the response factor is
set equal to  1).
     The  oxygen  concentration is also measured to determine the  total gas flow
rate through  the tent.  However, oxygen concentration is also used to rate the
quality of the tent, and except  for a few extremely difficult situations, VOC
concentrations are not measured  until the oxygen  concentration  in  the tent is
reduced below 5%.
     Air  can  replace nitrogen as a dilution gas  if  the  hydrocarbon
concentration is hot expected to be high enough  to  cause an explosive
atmosphere  inside the tent.   However, calculations  based on air tent data must
assume a  nominal tent leakage rate, i.e., extra  flow  through  the tent due to
air entering  the tent that  is not metered through the tubing.
     Safety  has  been a key  factor  in the development  of this  procedure.  Over
400 valves  have  been tented with this technique  without incident.   Nitrogen is
 aText adapted from Chemical  Manufacturers Association,  Guidance for Estimating
  Fugitive Emissions, July,  1987.
                                       4-6

-------
                    1/4" Intulatad Tallon
                         Tubing
              Pressure
Plant
Nitrogen
(or Air|
Oessicent
Activated
Charcoal
1




j
i
j
{
!
I!


                      Rotametar
                                Foil or
                                Mylar
                                Enclosure
                                                                 Valve Stem
                                                                 Tape or Adjustable Metal Band
                                                                                     Sample Port for Collecting Data on
                                                                                       — Temperature
                                                                                       — Hydrocarbon Concentration
                                                                                       — Oxygen Concentration
                                                                              Tape or Compressed Foil
Figure 4-3.   Equipment required  for  the blow-through sampling  technique.

-------
            TABLE 4-1.  CALCULATION PROCEDURES FOR TENTED LEAK RATE
     Tent rate    4.836 x IP'5 fOHMWHOVAHRF)
      (Ib/hr)   *            T + 460

where

     Q - flow rate into tent in cubic meters/hour

              N, flow rate in liters/minute            _3 m  L     liter-hour-1
          1 " (             21               '

      MW*    - molecular weight of gas in pounds/pound-mole
      T      » temperature in tent in °F
      OVA    » instrument reading minus background reading, in vppm
      RF**   - response factor for leaking gas relative to calibration gas
               a conversion factor taking into account the gas
               assuming a pressure in the tent of 1 atmosphere
4.836 x 10   » a conversion factor taking into account the gas constant and
 For mixtures:
   gas molecular weight • (mole fraction-gas 1)(molecular weight-gas 1) +
                          (mole fraction-gas 2)(molecular weight-gas 2) + ...

**
  For mixtures:
   gas response factor =« (mole fraction-gas 1)(response factor-gas 1) -*•
                         (mole fraction-gas 2)(response factor-gas 2) + ...

  where response factor is calculated at the OVA concentration.  For liquid
  mixtures, a detailed analysis based on vapor-liquid equilibrium must be done
  to use OVA concentrations.
                                        4-8

-------
used as a dilution gas instead of.air to prevent an explosive atmosphere

within the tent.  All of the instruments used are battery-operated and

approved for Class I, Division I use.  In summary, the blow-through method

consists of the following steps:

     t    Interview the unit operator to determine the composition of the
          material in the designated equipment component (in weight or volume
          %), and the operating conditions of the component.

     •    Screen the component by placing a Tygon nozzle on the end of the
          Foxboro/Century OVA (Organic Vapor Analyzer), holding the end of the
          nozzle within 1 centimeter of the leak interface, and recording the
          highest concentration seen on the OVA readout.

     •    Cut a tent from appropriate material (see section 4.3 - Source
          Enclosure) that will easily fit over the equipment component.

     •    Connect tubing from the nearest low pressure nitrogen station to a
          rotameter  stand, which includes a regulator, dessicant, activated
          charcoal, and a rotameter in series.

     t    Run tubing from the rotameter outlet to a "Y" that splits the
          nitrogen flow into two pieces of tubing.  Insert the tubes into
          openings located on either side of the tent.

     •    Turn on nitrogen at the utilities station and regulate  it at the
          rotameter to approximately 40 liters/minute.

     •    After the nitrogen is flowing, wrap aluminum foil around those parts
          of the equipment component where air could enter the tent-enclosed
          volume.

     •    Use duct tape, wire, and/or rope to secure the tent to  the
          component.

     •    Put a third hole in the tent roughly equidistant from the two
          nitrogen-fed holes.

     •    Measure the oxygen concentration in the tent by inserting the lead
          from an 0- meter into the third hole.  Adjust the tent  (add
          additional tape, foil,  rope, etc.) until the 0- concentration is
          less than 5%.                                 c

     t    Measure the temperature in the tent with a platinum-RTD
          thermocouple.

     t    Calibrate the OVA to methane or hexane at a known concentration in
          nitrogen using the OVA  dilution probe.  Remember to correct for the
          dilution before inserting the OVA concentration reading into the
          calculation in Table 4-1.


                                      4-9

-------
     t    Check the VOC concentration at several points in the tent with the
          OVA to insure that the tent contents are at steady state.

     •    Measure the hydrocarbon VOC concentration in the tent with the OVA
          at three different nitrogen flow rates.  Typically the flow rates
          will be 40, 30 and 20 liters/minute if there is no OVA response at
          the higher rates (i.e., the mass leak rate from the component is
          very low).  After each adjustment of the nitrogen flow, check the 0-
          concentration to ensure it stays below 556.  Alternatively, collect
          samples in Tedlar or aluminized sample bags by drawing sample out of
          the bag with a portable sampling pump.

     0    Remove the tent and any plugs from the component and collect any
          condensate on the inside of the tent in a plastic graduated
          cylinder.  Record the amount collected and the elapsed time the tent
          was on the component.

     •    If there is liquid dripping from the component, collect the drips
          for a timed period which produces enough collected material for
          accurate volume measurement.  Record the amount in the report, but
          do not add it to the vapor leak rate.

4.3  SOURCE ENCLOSURE
     To measure the leak rate from any given fitting accurately, it is
necessary to isolate that fitting from the ambient air.  This is accomplished
by an enclosure (or tent) formed around the leak source.  Appropriate tent
material must be selected, so that the substances found in the process fluid

are contained within the tent.  For example, because polyethylene is permeable

to hydrocarbons, it is an inappropriate tent material.  By comparison, Mylar
plastic (polyethylene terephthalate) is well suited to this function as it

does not absorb significant amounts of hydrocarbons, it is very tough, and it
has a high melting point (250°C).  Other suitable materials are Tedlar (PVF
film) and Teflon (fluorocarbon resin).  For convenience, Mylar will be used as
an example of tent material in the following discussions.  The thickness of
the tent material can range from 1.5 to 15 mil depending on the type of
equipment being bagged.  A typical tent is shown in Figure 4-4.  The
enclosures should be kept as small as practical.  This has several beneficial
effects:

     •    the time required to reach equilibrium is kept to a minimum,

     •    the time required to construct the enclosure is minimized,
     •    a more effective seal results from the reduced seal area, and
                                     4-10

-------
                                                                   Surgical
                                                                    tubing
                                                                  pressure tap
Figure 4-4.  Tent construction around the seal  area  of a vertical pump.
                                     4-11

-------
     •    condensation of heavy hydrocarbons  Inside  the  enclosure is minimized
          or prevented due to reduced  residence  time and decreased surface
          area available for heat transfer.
Tent construction  for individual sources  is discussed in the  sections 4.3.1
through 4.3.5.
4.3.1  Valves
     The most important property of  the valve that affects  the  type of
enclosure (tent) selected for use is the  metal skin  temperature of the area
enclosed by the tent and around which  the seal is made.
     At skin temperatures of 400°F or  less, the  valve stem  and/or stem support
can be wrapped with 1.5-2.0 mil Mylar  and sealed at  each end  and at the seam
with duct tape.  A leak-tight seal is  not required when  the vacuum method of
leak measurement is used.  Actually, it is better to allow  for  some areas of
incomplete sealing to provide access for  air  being drawn through the tent by
the vacuum pump in the sampling train.
     Two bulkhead  fittings are then  attached  to  the  Mylar tent.  One is for
the water manometer or differential  pressure  gauge .connection,  and the other
is for the line to the sampling train.  If, after starting  the  sampling,  a
vacuum greater than 0.1  inches of water is found to  exist in  the tent, a hole
.can be added on the side of the tent opposite the outlet to the sampling
train.  This provides an additional  entrance  for the dilution air, and reduces
the vacuum  in the  tent while  sampling.  The Mylar tent must be  constructed to
enclose the valve  stem seal and the  packing gland seal.
     When  skin temperatures are in excess of  400°F,  other methods of tenting
the valves  can be  used.   In one method, adjusted metal bands  or wires are
wrapped  around all hot points that are in contact with the  Mylar tent
material.   Seals  are  then made  against the insulation using duct tape or
adjustable  metal  bands of  stainless  steel.
     At  extremely  high temperatures, metal foil  can  be wrapped  around the
valve  leak area.   In  a previous study, seals  were made using  adjustable metal
 bands.     At  points where  the shape  of the equipment prevented  a satisfactory
 seal with  metal  bands,  the  foil was  crimped  to make  a seal.  It was necessary,
 in some  of these  instances,  to  use  a relatively  high capacity vacuum pump to
 insure a constant  inflow of air through  all  seal areas.
                                      4-12

-------
     Mixed-phase valve leaks are enclosed with the same type of tent described
above.  The rate of leakage of liquid is measured by collecting it over  a
measured length of time.  A low point is formed in the tent to collect the
liquid so Its volume can be measured and converted to a mass rate.  The  total
mass leak rate can then be estimated by adding the liquid mass leak rate to
the gas mass leak rate.
4.3.2  Pumos and Agitators
     As with valves, the property of most concern when preparing to sample a
pump or agitator is the metal skin temperature at areas or points that are in
contact with the tent material.
     At skin temperatures below 400°F, Mylar plastic and duct tape are
satisfactory materials for constructing a tent around a pump or agitator seal,
The vast majority of centrifugal pumps in VOC service have a housing or
support that connects the pump drive (or bearing housing) to the pump itself.
The two supports normally enclose about half of the area between the pump and
drive motor, leaving open areas on the sides.  It is usually a relatively
simple matter to cut panels to fit these remaining open areas.  In a previous
study, the panels were cut from 14 mil Mylar.    Bulkhead fittings for the
outlets to the water manometer and sampling train are placed through one Mylar
panel and sealed.  An opening (hole, bulkhead fitting) is made in the opposing
panel, if necessary, to allow easier flow of dilution air into the enclosure
around the seal.
     In many horizontal pumps, there is a line from the bottom of the lower
metal support to a drain.  This line serves as a drain for coolant, sealant,
and/or process liquid leaking from the pump seal.  This line (and all other
lines from the enclosure) should be sealed off to avoid drawing air and
hydrocarbon vapors from the drain back up to the sealed enclosures.  If  there
is no liquid flowing through this drain line, it can be plugged off.  If
liquid is going to the drain, a short length of hose or tubing can be attached
to the end of the drain line and looped upward to form an effective liquid
seal.
     In the cases where supports are absent or quite narrow, a cylindrical
enclosure around the seal can be made so that it extends from the pump housing
to the motor or bearing support.  This enclosure can be made of 14 mil  Mylar,
since this thickness provides considerable strength and rigidity.
                                     4-13

-------
     Reciprocating pumps present a somewhat more difficult tenting problem.
If supports are present, the same type of two-panel Mylar tent can be
constructed as that for centrifugal pumps.  In many instances, however,
sufficiently large supports are not provided, or the distance between pump and
driver is relatively long.  In these cases, a cylindrical enclosure similar to
that used for centrifugal pumps can be constructed.  If it is impractical to
extend the enclosure all the way from the pump seal to the pump driver, a seal
can be made around the reciprocating shaft.  This can usually be best made by
using heavy aluminum foil and crimping it to fit closely around the shaft.
The foil is attached to the Mylar plastic of the enclosure and sealed with
duct tape.
     If the temperature is too high or the potential points of contact are too
numerous to insulate, an enclosure made of aluminum foil can be constructed.
This enclosure is sealed around the pump and bearing housing using si 11 cone
fabric insulating tape, adjustable metal bands, or wire.
     In cases where liquid and vapors are leaking from a pump, the enclosures
described above for valves can be used.  The outlet from the tent to the
sampling train is placed at the top of the enclosure and as far away from
spraying leaks as practical.  Thus, entrainment of the liquid into the
sampling train (and cold trap) is avoided.  The rate of leakage of liquid is
measured by collecting  it over a measured length of time.  A low point is
formed in the tent to collect the liquid so its volume can be measured and
converted to a mass rate.  The total mass leak rate can then be estimated by
adding the liquid mass  leak rate to the gas mass leak rate.
4.3.3  Compressors
      In general, the same types of tents that are suitable for pumps can be
directly applied to compressors.  The construction and application of these
enclosures have been described in the preceding discussions of pump sampling
and will not be repeated  in this section.
      Compressors generally handle light gases, and  in many cases, the seals
are enclosed.  The seal  enclosures are vented to the atmosphere at a
high-point vent or may  be vented to the blowdown/flare system.
      If  the  seals are vented  to a  high-point vent, this vent line can be
sampled.  A  Mylar bag can be  constructed  and sealed around the outlet of the
                                      4-14

-------
vent and connected to the sampling train.  The leak rate from the vented
compressor seals is then measured using the normal flow-through method.
     If the high-point vents are inaccessible, the vent lines from the
compressor seal enclosures can be disconnected at some convenient point
between the compressor and the normal vent exit.  Sampling is then done at
this intermediate point.
     When enclosed compressor seals are vented by means of induced draft
blowers or fans, the outlet from the blower/fan can be sampled.  In a previous
study, a pitot tube was used to determine the air flow rate.    A sample of
the outlet air was taken and returned to the laboratory for methane and
nonmethane hydrocarbon analysis.  Ambient air samples were taken around the
compressor seal enclosure area at the same time and were analyzed for
hydrocarbon content.  The compressor seal leak rate was determined from the
knowledge of air flow rate, its hydrocarbon content, and the hydrocarbon
content of the ambient air.
4.3.4  Flanoes
     In most cases, the physical configurations of .flanges lend themselves
well to the determination of leak rates.  For small to moderately large leaks
from flanges with metal skin temperatures up to about 400°F, a narrow section
of Mylar film can be used to span the open distance between the two flange
faces of the leaking source.  The Mylar is attached and sealed to each flange
with duct tape.  Connections (bulkhead fittings) for the water manometer and
sampling train are attached to the Mylar.
     When testing flanges with skin temperatures above 400°F, the outside
perimeter of both sides of the flange connection are wrapped with asbestos
insulating tape.  Then, a narrow strip of aluminum foil can be used to span
the opening between the flange faces.  This narrow strip of material can be
sealed against the asbestos tape using adjustable bands of stainless steel.
4.3.5  Relief Valves
     Relief devices in gas/vapor service generally relieve to the atmosphere
through a large diameter pipe which is normally located at a high point on the
process unit that it serves.  The "horns" can be easily bagged by placing a
Mylar plastic bag over the opening and sealing it to the horn with duct tape.
Because many of these devices are high above grade, accessibility to the
                                     4-15

-------
sampling train may be limited or prevented.  It is sometimes possible to run  a
long piece of tubing from the outlet connection on the bag to the sampling
train located at grade level or on a stable platform.  And as discussed
previously in the section on screening, the purpose of pressure relief devices
makes them inherently dangerous to sample, especially over a long period of
time.  If these equipment are to be sampled for mass emissions, special care
and precautions should be taken to ensure the safety of the personnel
conducting the field sampling.
4.4  ACCURACY CHECKS
     An accuracy check is normally performed in the laboratory by sampling an
artificially induced leak rate of a known gas.  This effectively points out
any bias in the test methods and defines the variance component due to
sampling.  If the result is outside the 80 percent to 120 percent recovery
range, the problem should be investigated and corrected before sampling
continues.  The problems and associated solutions should be noted in the test
report.
4.5  ANALYTICAL TECHNIQUES
     The techniques used in the laboratory analysis of the bagged samples will
depend on the type of processes sampled.  The following sections describe the
analytical instrumentation, calibration, analytical techniques for condensate,
and quality control.  These are guidelines for the field sampling team and are
not meant to be a detailed protocol for the laboratory personnel.  Laboratory
personnel should be well versed in the analysis of hydrocarbon mixtures and
should design their specific analyses to the samples being examined.
4.5.1  Analytical Instrumentation
     The selection of analytical instrumentation is critical to identification
of  individual organics in the sample.  The instrument of choice depends on the
type of sample being processed.  Organic components of the streams of some
process units may produce widely differing responses with an FID.  For this
reason, more comprehensive analyses by gas chromatography must be performed to
identify individual components.  FID gas chromatographs or electron capture
gas  chromatographs are commonly used to identify individual components of a
sample.  Other considerations besides  instrument choice are the type of column
used,  and whether or not temperature programming is necessary to separate out

                                     4-16

-------
individual components of the process stream with sufficient resolution.  An
example of a gas chromatograph that has been used in previous studies  is the
                                                 12
Hewlett Packard 573-A Dual FID Gas Chromatograph. *•
     For some process streams such as ethylene and cumene, total hydrocarbon
analyses are satisfactory.  The response of the FID to these compounds is
similar;  therefore an instrument such as the Byron Total Hydrocarbon Analyzer
(THC) can be calibrated with a single representative hydrocarbon.  This
instrument has an FID and resolves samples with methane and nonmethane
components.  The emission streams can then be analyzed for the total
hydrocarbon content with reasonable accuracy.
4.5.2  Calibration
     Proper calibration of both the laboratory and field instruments for the
wide range of organic species is different.  Each instrument must be
calibrated with the appropriate species each day it is used.
     Calibration of the portable hydrocarbon detector should be done with
standards of hexane in air.  A hexane standard of approximately 2,000  ppmv is
available in a bottle supplied and certified to ±2 .percent by the
manufacturer.  Higher concentrations than this may lead to problems involving
condensation at ambient temperatures.  Lower concentrations must be generated
by dilution.
     The total hydrocarbon analyzer can be calibrated with the same hexane
standard.  The standard is related to an NBS propane standard by measuring the
concentration of the NBS standard using the total hydrocarbon analyzer after
the  analyzer has been calibrated with the hexane standard.
     The gas chromatograph is calibrated with either gas standards generated
from calibrated permeation tubes containing individual VOC components or
bottled standards of common gases.
4.5.3  Analytical Techniques for Condensate
     Any condensate collected by the ice bath in the sampling train should be
brought to the laboratory sealed in the cold trap flask.  This material is
transferred to a graduated cylinder to measure the volume collected.   If there
is enough volume to make  it feasible, the organic layer should be separated
from the aqueous layer (if present) and weighed to determine its density.
Both the aqueous and organic phases should be analyzed by gas chromatography
if water-miscible organic compounds are present.
                                     4-17

-------
4.5.4  •Quality Control for Analytical Techniques
     Blind standards should be prepared and submitted for analysis.  The blind
standards are prepared by diluting or mixing primary standards in a prescribed
fashion so that the resulting concentrations are known.  The calculated
concentrations and the analytical results should be logged into a laboratory
notebook.  If the results are not within 25 percent of the certified level,
further analyses must be performed to determine the reason.  Such a procedure
not only defines the analytical variance component and analytical accuracy,
but it serves to point out equipment malfunctions and/or operator error before
many data are affected.
                                      4-18

-------
                                 SECTION  5.0
                                  REFERENCES


1.    Code of Federal Regulations, Title 40, Part 60, Appendix A.  Reference
     Method 21, "Determination of Volatile Organic Compound Leaks."

2.    Harris, G.E. and G.J. Langley. (Radian Corporation.)  Detailed Screening
     Plan for Fugitive Emissions from SOCMI Process Units.  (Prepared for U.
     S. Environmental Protection Agency.) Research Triangle Park, N.C.
     December 1979.

3.    Joseph, 6.T. and M. Peterson. (Northrop Services, Inc.)  APTI Course
     51:417  Controlling VOC Emissions from Leaking Process Equipment Student
     Guidebook.  (Prepared for U.S. Environmental Protection Agency.) Research
     Triangle Park, N.C. Publication No. EPA-450/2-82-01.  1982.

4.    U. S. Environmental Protection Agency.  Response Factors of VOC Analyzers
     at a Meter Reading of 10,000 ppmv for Selected Organic Compounds.
     Publication No. EPA-600/2-81-051.  1981.

5.    U. S. Environmental Protection Agency.  Response of Portable VOC
     Analyzers to Chemical Mixtures.  Publication No. EPA-600/2-81-110.  1981.

6.    Brown, G.E., et al. (Radian Corporation.)  Response Factors of VOC
     Analyzers Calibrated with Methane for Selected Organic Compounds.
     (Prepared for U. S. Environmental Protection Agency.)  Research Triangle
     Park, N.C.  Publication No. EPA-600/2-81-022.  September 1980.

7.    Anastas, M.Y., and H.J. Bel knap. (PEDCo Environmental, Inc.)  Summary of
     Portable VOC Detection Instruments.   (Prepared for U. S. Environmental
     Protection Agency.)  Washington, D.C. Publication No. EPA-340/1-80-010.
     March  1980.

8.    Langley, G.J., and L.P. Provost. (Radian Corporation.)  Revision of
     Emission Factors for Nonmethane Hydrocarbons from Valves and Pump Seals
     in SOCMI Processes.  (Prepared for U. S. Environmental Protection
     Agency.)  Research Triangle Park, N.C.  November 1981.

9.    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.  Publication No. EPA-600/2-81-111.
     June  1981.

10.  Radian Corporation.  Assessment of Atmospheric Emissions from Petroleum
     Refining: Volume 4. Appendix  C, D, E.  (Prepared for U.S.  Environmental
     Protection  Agency.) Research  Triangle Park, N.C.  Publication No. EPA-
     600/2-80-075d.  July 1980.
                                       5-1

-------
11.   Wetherhold, R.G., L.P. Provost, and C.D. Smith (Radian Corporation.)
     Assessment of Atmospheric Emissions from Petroleum Refining: Volume 3.
     Appendix B. (Prepared for U.S. Environmental Protection Agency.)
     Research Triangle Park, N.C. Publication No.  EPA-600/2-80-075c.  April
     1980.

12.   Radian Corporation.  Test Plan for Control of Fugitive Emissions  from  the
     Synthetic Organic Chemical Manufacturing Industry.  (Prepared for U. S.
     Environmental Protection Agency.)  Research Triangle Park, N.C.
     October 1979.

13.   U.S. Environmental Protection Agency.  Fugitive Emission Sources  of
     Organic Compounds - Additional Information on Emissions, Emission
     Reductions, and Costs.  Research Triangle Park, N.C.  Publication No.
     EPA-450/3-82-010.  April 1982.

14.   Langley, G.J. and R.S. Wetherhold.  (Radian Corporation.)  Evaluation  of
     Maintenance for Fugitive VOC Emissions Control.   (Prepared for U.S.
     Environmental Protection Agency.) Research Triangle Park, N.C.
     Publication No. EPA-600/52-81-080.  May 1981.

15.   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. Publication No.  EPA-600/2-81-003.
     September 1980.

16.  Stelling, J.H.E.  (Radian Corporation.)  Emission  Factors for Equipment
     Leaks of VOC and HAP.   (Prepared for U.S. Environmental Protection
     Agency.)  Research Triangle Park, N.C.  Publication No. EPA-450/3-86-002.
     January 1986.

17.  Wetherhold, R.S., and  L.P.  Provost.  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.  1979.

18.  Draper, N. and H. Smith. Applied Regression Analysis.  John Wiley and
     Sons,  1966.

19.  Draper, N.R.  and  H.  Smith.  Applied Regression Analysis.  Second  Edition,
     John Wiley and Sons,  New  York,  1981.

20.  Finney, D.J.  "On  Distribution  of a Variate  Whose  Logarithm  is  Normally
     Distributed", Journal  of  the  Roval Statistical Society. Series  B, 7
      (1941),  155-161.

21.  Rickmers,  A.D. and H.N. Todd,  Statistics. An  Introduction.  McGraw-Hill
     Book Co.,  New York,  1967.
                                       5-2

-------
    APPENDIX A
REFERENCE METHOD 21

-------

-------
  Mxnoo 21—DmquuHATiow OP VOIATBJ
        Oftounc COMFOOHBS LCAKS

  1. XppticaoUUv and Principle,
  1.1  XppacaotiU* Thia method applies to
the determination of volatile organic coo*.
pound (VOC) leak! from process equipment.
These sources Include, but an not Umitee:
to.  valves, flanges and other connection*.
pumps and compressors,  presaun relief de-
vices, process  drains, open-ended  valves.
pump and compressor seal system <
venta.  •"*""••"'•«'••  vessel venta.
scale. *p*i access tteur sseJa»
  1.2  Principle. A  portable
used to detect  VOC leaks from Individual
sources. The Instrument detector type la net
specified,  but It must meet  the  speafioa.

Section  3. A leak definition  concentration
baaed oa a reference compound la •r"*"*fltil
la each applicable regulation. Thia pnea»
dun la intended to locate sad classify leak*
only, sad la not to be used sa a direct mesa.
un of mass imKsslnn rates from Individual
  i
  tl  UaJt Oe/tnttlon Concentration. The
local VOC uiuceuuatiou at the surface of a
leak source that indleatea that a VOC ems*
ston (leak) la present. The leak definition It
an Instrument meter reading baasd on a ref-
  2.2  Ae/trrence Compound. The VOC spa-
das selected sa an instrument calibration.
beats for specification of the leak definition
concentration. (For example: If a leak defi-
nition  concentration  la   10.000 ppmv  aa
methane, then any source emission that re-
sults In a local concentration that yields a
meter reading of 10.000 on an instrument
calibrated with methane would be classified
sa a leak. In tola example, the leak deflm-
tion la  10.000 ppmv. and the reference com-
pound la methane.)
  2.1  Calibration Gas. The VOC eompouad
used to adjust the instrument meter reading
to a known value. The calibration CM is usu-
ally the reference compound at a concentra-
                                                tion approximately equal to the leak defini-
                                                tion concentration.
                                                 2.4  No  Detectable ATnttssion.  The  local
                                                VOC concentration at the surface of a leak
                                                source that Indicates that a VOC emission
                                                (leak) la not present. Since background VOC
                                                concentrations may exist, and to account
                                                for  Instrument  drift and Imperfect repro-
                                                dudbUlty. a difference between the source
                                                surface concentration and the local ambient
                                                concentration la determined. A difference
                                                based on  meter readings of lesa than a con-
                                                centration CTfTTspfttnltng to the minimum
                                                readability specification  Indicates that  a
                                                VOC emission (leak) la not present. (For ex-
                                                ample, if  the leak definition in a regulation
                                                la 10.000 ppmv. then the allowable Increase
                                                in surface concentration versus local ambi-
                                                ent concentration would be 900 ppmv baaed
                                                on the instrument meter naiUngi)
                                                 tS  Xatpoaw  factor.  The ratio of the
                                                (mown concentration of a VOC compound
                                                to the observed meter reading when  meae-
                                                ured using aa Instrument calibrated  with
                                                the reference compound specified In the ao-
                                                 it  CKtoracloa frteition. The
                                                                                            b. The Instrument response time must be
                                                                                           equal to or less than 30 seconds. The  re-
                                                                                           sponse time must be determined for the in-
                                                                                           strument configuration  to  be used dunng
                                                    i known value,
                                                percentage  of the
                                                tween the
                                                                     	and the  known

                                                 2.7 Aetponst Tun*  The  time Interval
                                                from a step change la VOC concentration at.
                                                the  input  of  the templing system  to  the
                                                time at which 90 percent of the correspond-
                                                lag  final value la reached at displayed oa
                                                the instrument readout meter.
                                                 3.1
                                                 3.1.1
                                                 a. The VOC Instrument detector shall re-
                                                spond to the compounds h^^g processed.
                                                Detector  types  which may  meet thia re-
                                                quirement Include, but an  not limited to.
                                                catalytic TirtilatliTiit flame lonmation.  Infra-
                                                red absorption, and photol
                                                 b.  The instrument
                                                                          be capable  of
                                                measuring the leak definition concentration
                                                specified la the regulation.
                                                 c. The scale of the Instrument meter shall
                                                be readable to £9 percent of the specified
                                                leak definition concentration.
                                                 d. The instrument shall be equipped with
                                                t pump so that a continuous sample la pro-
                                                vided to the detector. The nominal sample
                                                flow rate shall bo  vt to 3 liters per "<«"M"
                                                 c  The  instrument shall be intrinsically
                                                «fe for operation  la exple
                                                          by the applicable U.S-A.
                                                wds (e*.  National Electrical Code by the
                                                National Fin Prevention Association).
                                                 3.1.3  Performance Criteria.
                                                 a. The instrument response factors for the
                                                mdlvidal compounds to  be  measured must
                                                be less than 10.
  c. The calibration precision must be equal
to or lesa than 10 percent of the calibration
fas value.
  d.  The evaluation procedure for each pa-
rameter la given in Section 4.4.
  J.L3  •Performance  Evaluation  Require-
ment*.
  a.  A response factor must be determined
for each compound that Is to be measured.
either by testing or from reference sources.
The  response  factor  testa  we  required
before placing the analyzer into service, but
do not nave to be repeated sa subsequent in-
tervala.
  b.  The calibration precision test must be
completed prior to placing the analyzer Into
service, and at subsequent 3-month intervals
or at the next use whichever Is later.
  e. The response time teat is required prior
M placing the Instrument Into service. If  «,
modification to the sample pumping system
or flow configuration la  made that would
change the  response time, a new test is re-
quired prior to further use.
  3.2 Calibration  Gases. The monitoring
Instrument  la calibrated In terms of parts
per million  by volume (ppmv) of the refer-
ence compound specified la the applicable
regulation. The calibration cases  required
for monitoring and Instrument performance
evaluation an a aero cas (air. lesa than 10
ppmv VOC) and a calibration cas in air mix-
tun approximately equal to the leak defini-
tion specified In the regulation. If cylinder
calibration gas mixture are used, they must
be analysed and certified by the ratnufac-
turer to be within s2 percent accuracy, and
a  shelf  life must be  specified.  Cylinder
standards must  be either reanalyzed or re-
placed at the end of the specified shelf life.
Alternately, calibration cases may bs  pre-
pared by the user according to any accepted
caseous  standards  preparation  procedure
that will yield a mixture accurate to within
=2 percent. Prepared standards must be re-
placed each  day of use unices it can be dem-
onstrated that degradation does not occur
during storage.
  Calibrations may be  performed using  a
compound other than the reference com-
pound If a conversion factor  is determined
for that alternative compound so that  the
resulting meter readings during source sur-
veys can be converted to reference com-
pound results.
  4. Procedure*.
  4.1  Pretest /"reparation*. Perform the in-
strument  evaluation  procedures  liven in
Section 4.4 if the evaluation requirements of
Section 3.1.3 have not been met.
                                                               A-l

-------
  4.2  Caliontion  Pmcedwn.
and start up the VOC analyser according to
the manufacturer's instructions.  After  the
appropriate warmup period and zero inter-
nal calibration  procedure, introduce  the
calibration gas into the instrument sample
probe. Adjust the Instrument meter readout
to correspond to the calibration gas value.
  Notm,-If the meter readout cannot be ad-
Justed to the proper value, a malfunction of
the analyser Is Indicated aad corrective ac-
tions an necessary bef on use.
  4.3  Individual Source Surveys.
  4.2.1  T»pe I-UfJt Definition  Basest on
Conetntmtton. Place the probe inlet at the
surface of the component Interface where
leakage could occur. Move the probe alone
the interface periphery while obeervtnc the
instrument readout. U an increased meter
readme I* observed, slowly sample the Utter'
face where leaks** la indicated until the
^..^•...tii oMter readlnc la obtained. Leave
the probe Inlet at this  ™««n»Mm mailing lo-
cation for approximately two  times the In*
                       Is grsatsr than the
leak definition in UM applicable regulation.
record aad report the results as specified in
the regulation reporting requirements,
        of *rtvt aBBttcaKloa of
technique to specific equipment types are:
  a. Valvee—The most mnimon  source of
leaks from valves is at the seal between UM
stem sad aouamc.  Place UM probe at UM
IntcrYaca where the stem exists UM paektnc
ftmrnt ^ftt tiimtlt **** steal  circumference.
Also, place the probe at UM interface of the
paektac fiand  take-up  name  seat  aad
sample the periphery.  la addition, rarrey
valve housings of multipart sessmrily at the
surface of all interfaces where a  leak could
  b.  Manges aad Other Connections—Por
welded fiances, place the probe at the outer
      of  UM  flange-gasket interface  aad
         >i^ drcumference of  the flange.

(such as  threaded connections) with'a stmt-
 lari

 eumfarinrial traveneat  the outer  surface

 interface, II the source is a rotating shaft.
         the probe inlet within 1 cat of the
           Interface for the survey. If the
       _ eonflcurattoa  prevents a complete
 traverse of the shaft periphery, sample all
 accessible portions, flamnlc all other Joints
 on tne pump or compressor houauc where
 leakage could occur.
   d. Pressure Relief Devices  The eonflgura-
 Uon of most prnssiirs relief devices prevents
 •^^p'^T at the ssallin seat interface. Por
 those devices equipped with aa tncloaed ex-
 tension, or horn, place the probe Inlet at ap-
                                           proximately the center of the exhaust are*
                                           to the atmosphere.
                                             e. Process Drains—Por open drains. oi»e»
                                           the probe inlet at approximately the center
                                           of the area, open  to  the atmosphere.  PQ,
                                           covered drains, place  the probe at the  sur-
                                           face of the cover interface and conduct a p*.
                                           rlpheral traverse.
                                             f. Open-Cnded Unes or Valves—Place the
                                           probe Inlet at approximately the center of
                                           the opealac to the atmosphere.
                                             c. Seal System Degassing Vents and Acco-
                                           mulator Vents—Place  the probe inlet u to.
                                           proxlmately the center of the openinc te
                                           the atmosphere.
                                             h. Access Door  Seals Place  the probe
                                           inlet at the surface of the door seal inter.
                                           face and conduct a peripheral traverse.
                                             4.3.2  lYp» //— "No  Dctcctoote £fl*UMo*"
                                             Determine  the  local ""*««•"»  coneenov
                                           Uon around the source by moving the orobe
                                           inlet randomly upwind sad downwind  u »
                                           <««»«~t of one to two meters  from  tbe
                                           source.  If aa Interference  exists with  this
                                           determination due to a nearby emission or
                                           leak, the  local ambient concentration  may
                                           be  determined at distances closer to tbe
                                           source, but la no ease shall the distance be
                                           leas than M centimeters.  Then  move tbe
                                           probe inlet to the surface of the source aas
                                           determine the eoneemntion described la
                                           4J.L The difference between these coneea.
                                           trations determines whether there ire as
                                           rlsTsrTshls imlsslnns Record aad report the
                                           results as specified by the regulation.
                                             For those eases where the regulation re>
                                           quins a rpeclfle device installation, or tost
                                           iperlfled vents be ducted or piped to a con-
                                           trol device, the twit*-—"— of these condltteos
                                           •hall be visually confirmed. When the rego-
                                              ioa also requires that no i
                                            suns exist, visual observations and stapling
                                            surveys an required. Iframnles of this teen-
                                            niqueare:
                                               Pump or Compnasor Seals—U applica-
                                            ble, determine the type of abaft seal. Pre-
                                            form  a survey of the local
                                            VOC concentration aad determine if <
                                                              -. as described above.
                                             (b) Seal System Degassing Vents, Accumu-
                                           lator Vessel Vents. Pressure Relief Devices-
                                           If applicable, observe whether  or not the
                                           applicable duetlnc or piping exists. Also, de-
                                           termine If say sources axlst in the ducting
                                           or piping when emissions could occur prter
                                           to the control device. If the required duet.
                                           Ing or piping exists and there an no sources
                                           when UM omissions could be vented to the
                                           atmosphere prior to the control device. Una
                                           It is presumed that no detectable emlartom
                                           an present If then are sources in the duct-
                                           lac  or piping  when emissions could BS
                                           vented or sources when leaks could occur.
                                           the tnmp""t surveys described la this para-
                                           graph shall be used to determine If detects.
                                           ble emissions exist.
                                      A-2

-------
   4.3.3 Alternative Screening Procedure. A
 jcreening procedure based on the formation
 of bubble* in a soap solution that la sprayed
 OR a potential leak source may be used for
 tnose sources that do not have continuously
 moving part*, that  do not have surface tem-
 peratures greater than the boiling point or
 lest than the freezing point of the soap solu-
 tion, that do not have open area* to the at-
 mosphere that  the soap solution  cannot
 bridge, or that  do  not exhibit evidence of
 liquid leakage. Source* that have these con-
 dition* present must be surveyed "ting the
 instrument technique* of o.l or O.Z
  Spray a soap  solution over all potential
 teak source*. The soap solution may be a
 commercially available leak detection solu-
 tion or may be prepared using concentrated
 detergent and water. A pressure sprayer or
 t squeeM  bottle may be used to dispense the
 solution. Observe the potential leak sites to
 determine If  any bubbles are formed. If no
 bubbles are  observed, the  source  Is pre-
 sumed to  have no detectable emissions or
 leaks as Applicable,  if any bubbles are  ob-
 served, the Instrument techniques of O.1 or
 OJ shall be used  to determine if a leak
 txisu. or If the source has detectable
jiflfl*,, £( e|OOeUC*iVsUsl»
                                              Record  the  meter readings. Calculate the
                                              average  algebraic  difference between the
                                              meter readings and the known value. Divide
                                              this average difference by the known call-
                                              oration value and mutiply by 100 to express
                                              the resulting calibration precision as a per.
                                              centtge.                             w
                                               4.4.3 JtopoiiM  rime.  Introduce zero n*
                                              tow the  instrument  sample probe. When
                                              the  meter reading  has stabilized, switch
                                              quickly  to  the  speei/led  calibration  gas.
                                              Measure the time from switching to when
                                              90 percent of the final stable reading is at-
                                              tained.  Perform  this test  aequencV three
                                             times and record the  resultsTcalculau the
                                             average response time.

                                               S.1 OuOate.  OJL. on* O.X. Worm.  Re-
                                                                	     staMeter
                                                                                    -
                                                           0A Snvtronmental Protee-
                                                          .               ea
                                             uon Agency.  Btssarch Trlangl* Park.
                                             i2 Brown, a*.  H aL
                                              VOC Analyaen CallbratMl with
                                                       OrBu^ compounds
                                                                              Pactors
                 	1 ffocedwet. At
 —   —	 o* the instrument perform-
 Aoce evaluation test, isssmtils and start up
 the lns«nm»emAeeording to the manufae-

                     preliminary  adjust-
                factor.  Calibrate the  In-
         with the reference compound a*
specified In  the  MT"^M* regulation, Par
«*«npjnie,j^ u»at I, to be m*MuiS
dunag individual source survey*, obtain or
prepare a known standard In air at a con-
centration of Approximately 80 owremta*
the applicant leak definition umeMUmitad
by volatility or exploamty. In th«e  cue*
        a standard at 90 percent of the
         coneentratloa or
                                                         .•     •                 a-


                                                            . September       ^
 saturat
         ,,,
        thi* mixture to the analyser and
       the  iheened meter reading. Intro-
          ttr unto a stable reading i* *£
                     or detector type, the
         factor determination la not  nT
     .        e* of published response fac-
tors for flam* lontxation and catalytic aid
^f?? ^f1!!^0/* **• in«»««l«« in Sectton 3.
  1.4.3 Coliomtton fruition.  iSto
of three measuremenu by alternately urtng
zero ga* and  the jpeetfled  calibration  Ǥ*.
                                        A-3

-------

-------
                   APPENDIX B

                RESPONSE FACTORS
This appendix presents response factors estimated
for a 10,000 ppmv detector reading.  These
response factors were previously reported in
References 4 and 6.  They can be used to determine
whether or not a concentration is above or below
10,000 ppmv.  The response factors may or may not
be accurate at other concentrations.-

-------
TABLE 6-1.   RESPONSE FACTORS WITH 95* CONFIDENCE INTERVALS ESTIMATED AT 10,000 PPMV RESPONSE
	 : 	
CAS No.«
64-19-7
108-24-7
67-64-1
75-B6-5
75-05-8
98-86-2
75-36-5
74-86-2
79-10-7
107-13-1

107-18-6
71-41-Oc

lpO-66-3
100-52-7
71-43-2
100-47-0
98-88-4
100-44-7
10-86-1
106-99-0




106-98-9
123-86-4
141-32-2


	 , —
Compound Name
Acettc acid
Acetic anhydride
Acetone
Acetone cyanohydrtn
Acetonltrlle
Acetophenone
Acetyl chloride
Acetylene
Acrylic actd
Acrylonltrlle
All one
Ally! alcohol
Amyl alcohol, N-
Amylene
An 1 sole
Benzaldehyda
Benzene
Benzonttrlle
Benzoyl chlortde
Benzyl chlortde
Bromobenzene
Butadiene. 1,3-
Butane, N-
Butanol, N-
Butanol, Sec-
Butanol, Tert-
Butene, 1-
Butyl acetate
Butyl aery late, N-
Butyl ether, N-
Butyl ether, Sec-
Volatlllty
Class
LL
LL
LL
HL
LL
HL
LL
G
LL
LL
G
LL
HL
LL
LL
HL
LL
HL
HL
HL
L .
G
G
LL
LL
S
G
LL
LL
LL
LL
0V
Response Factor
1.64
1.39
0.80
3.51
0.95
18.70
2.04
0.39
4.59
0.97
0.64
0.96
0.75
0.44
0.92
2.46
0.29
2.99
22.10 D
15.30 D
0.40
0.57
0.50
1.44 I
0.76
0.53
0.56
0.66
0.70
2.60
0.35
A
Confidence
1.11.
1.09,
0.57.
0.69.
0.85,
5.52.
1.72,
0.36.
3.38.
0.80.
0.60.
0.76.
0.57.
0.34.
0.65.
1.38,
0.28.
1.18.
3.43.
3.96.
0.34,
0.54.
0.46.
0.89.
0.70.
0.38.
0.51.
0.54.
0.63.
0.81.
0.21,

Intervals
2.65
1.86
1.20
> 100. 00
1.06
>100.00
2.48
0.43
6.57
1.20
0.69
1.27
1.04
0.61
1.46
5.62
0.31
15.30
> 100. 00
> 100. 00
0.48
0.60
0.55
2.34
0.83
0.81
0.62
0.83
0.78
95.60
0.95
TL
Response Factor
15.60
5.88 I
1.22
21.00 N
1.18
B
2.72
B
B
3.49 I
15.00
X
2.14
1.03
3.91
B
1.07
B
B
B
1.19.
10.90
0.63
4.11 I
1.25
2.17
5.84
1.38
2.57 I
3.58 I
1.15
V
Confidence
7.05.
2.71,
0.81.
1.09,
0.94.

1.65,


0.44,
9.68,

0.45,
0.59.
0.52.

0.96,



0.27,
8.11.
0.58,
2.16.
0.99,
1.34,
4.20,
1.15,
1.17,
1.82,
0.75,

Intervals
46.20
. 12.80
2.00
> 100. 00
1.52

5.32


27.90
26.50

> 100. 00
2.59
> 100. 00

1.20



> 100. 00
15.40
0.70
7.83
1.66
4.43
8.89
1.70
5.68
7.04
2.17

-------
TABLE B-l.   RESPONSE  FACTORS WITH 95* CONFIDENCE INTERVALS ESTIMATED AT 10,000 PPMV RESPONSE  (cont'd.)


CAS No. «
109-73-9
13952-84-6
75-64-9

123-72-8
107-92-6
109-74-0
75-1-50

108-90-7

67-66-3
25167-80-0
03
IN}
108-41-8
95-49-8
106-43-4
95-48-7
4170-30-0
98-82-8
110-82-7
108-93-0
108-94-1
110-83-8
108-91-8

123-42-2




Compound Name
Butylanlne. N-
Butylanlne, Sec-
Butyl amlne, Tart-
Butyl benzene, Tert-
' Butyraldehyde, N-
Butyric acid
Butyronltrtle
Carbon dlsulflde
Chi oroacet aldehyde
Chlorobenzene
Chloroathane
Chloroforn
Chlorophenol, 0-
Chloropropene, 1-
Chloropropene, 3-
Chlorotoluene, M-
Chlorotoluene, 0-
Chlorotoluene, P-
Cresol, 0-
Crotonaldehyde
Cumene
Cyclohexane
Cyclohexanol
Cyclohexanone
Cyclohexene
Cyclohexyl amlne
Decane
Dl acetone alcohol
Diacetyl
Dlchloro-1-propene, 2,3-

Volatlllty
Class
LL
LL
LL
HL
LL
HL
LL
LL
LL
LL
G
L
HL
LL
LL
LL
LL
LL
S
LL
LL
LL
HL
LL
LL
LL
HL
HL
LL
LL

0V
Response Factor
0.69
0.70
0.63
1.32
1.29
0.60
0.52
B
9.10
0.38
5.38 I
9.28
4.56
0.67
0.80
0.48
0.48
0.56
0.96
1.25
1.87
0.47
0.85
1.50
0.49
0.57
0.09 N
1.45
1.54
0.75

A
Conf \ dence
0.53,
0.58.
0.58.
0.89.
1.07.
0.38,
0.40.

5.73.
0.32,
1.87.
5.19.
1.72.
0.61,
0.72,
0.45,
0.42,
0.52.
0.70,
0.82,
1.10,
0.39,
0.65.
0.97.
0.42,
0.42,
0.05,
0.96,
1.25,
0.56,


Intervals
0.98
0.87
0.70
2.20
1.61
3.14
0.74

16.20
0.47
26.40
20.00
27.20
0.73
0.90
0.51
0.55
0.61
1.45
2.24
3.71
0.58
1.20
2.76
0.57
0.86
>100.00
2.48
1.92
1.09

TL
Response Factor
2.02
1.56
1.95
B
2.30
10.70 I
1.47 I
3.92
5.07
0.88
3.90 P
B
18.30 I
0.87
1.24
0.91
1.06
1.17 I
4.36 I
B
B
0.70
B
7.04
2.17
1.38
0.16 I
0.98
3.28
1.75
~~
V
Confidence
1.14.
0.77,
1.42.

0.96.
6.53,
0.62,
1.87,
3.08.
0.77.
1.58.

6.50.
0.69,
1.08,
0.40,
0.33,
0.77.
0.40.


0.62.

1.59,
1.78,
1.28.
0.07,
0.44,
2.25.
1.14,


Intervals
4.97
5.24
2.91

12.80
17.60
3.48
12.60
9.79
1.00
14.10

51.50
1.16
1.42
7.47
> 100. 00
1.77
47.40


0.80

> 100. 00
2.74
1.48
0.35
5.93
5.12
3.18

-------
TABLE B-l.   RESPONSE FACTORS WITH 95* CONFIDENCE INTERVALS ESTIMATED AT 10.000 PPMV RESPONSE  (cont'd.)
CAS No.»
541-73-1
95-50-1

107-06-2




25167-70-8

68-12-2
57-14-7
67-68-5
123-91-1
106-89-8

64-17-5

141-78-6
141-97-9
140-88-5
105-39-5

100-41-4
74-85-1
75-21-8
107-15-3
64-18-6

Compound Name
Dlchlorobenzene, M-
Dtchlorobenzene, 0-
Dtchloroethane. 1,1-
Dlchloroethane, 1,2-
Dlchloroethylene, CIS1.2-
Dlchloroethylane, TRANS 1.2-
D 1 chl orome th an e
Dichloropropane. 1,2-
Dilsobutylene
Dimethoxy ethane. 1,2-
Dtmethylformamtde, N.N-
Dlmethylhydrazlne, 1,1-
Dtmethylsulfoxlde
Dloxane
Eplchlorohydrin
Ethane
Ethanol
Ethoxy ethanol, 2-
Ethyl acetate
Ethyl acetoacatate
Ethyl aery late
Ethyl chloroacetate
Ethyl ether
Ethyl benzene
Ethylene
Ethylene oxide
E thy lened lamina
Formic acid
Clycldol
Volatility
Class
HL
HL
LL
LL
LL
LL
LL
LL
LL
LL
LL
LL
HL
LL
LL
G
LL
LL
LL
HL
LL
LL
LL
LL
G
G
LL
LL
LL
OVA
Response Factor
0.64
0.68
0.78
0.95
1.27
1.11
2.81
1.03
0.35
1.22
4.19
1.03
0.07 I
1.48
1.69
0.65
1.78
1.55
0.86
3.82
0.77
1.99
0.97
0.75
0.71
2.46
1.73
14.20
6.88
Confidence
0.55,
0.47,
0.62,
0.77,
1.05,
0.98,
2.13,
0.82,
0.29,
0.64.
2.90.
0.77,
0.05.
1.04,
1.56.
0.44.
1.59,
1.26,
0.77,
1.89,
0.63.
1.70.
0.77.
0.52.
0.63,
1.95.
1.29,
10.60.
3.33,
Intervals
0.77
1.11
1.02
1.22
1.56
1.27
3.87
1.33
0.44
3.61
6.58
1.45
0.11
2.33
1.84
1.S8
2.01
1.96
0.95
10.70
0.97
2.36
1.30
1.11
0.82
3.29
2.46
19.80
19.70
TL
Response Factor
2.36
1.26
1.86
2.15
1.63
1.66
3.85
1.65
1.41
1.52
5.29
2.70
8.45 I
1.31
2.03
0.69 I
X
1.82
1.43
5.60
X
1.59
1.14
4.74 D
1.56
2.40
3.26
B
5.66
V

Confldepce Intervals
0.58,
0.35,
1.56,
1.66.
0.99,
0.67,
2.46,
1.06,
0.96.
0.65,
4.05.
0.51.
4.15,
0.70.
1.79.
0.21.

0.96,
1.07.
1.93.

0.40,
0.94,
1.38,
1.26,
0.96,
0.78,

2.08.
> 100. 00
> 100. 00
2.25
2.92
3.47
12.60
6.88
3.05
2.40
8.38
7.20
> 100. 00
17.20
3.60
2.33
2.30

5.12
2.00
38.80

>100.00
1.42
61.30
2.06
> 100. 00
>100.00

34.70

-------
TABLE B-l.  RESPONSE FACTORS KITH 95* CONFIDENCE INTERVALS ESTIMATED AT 10,000 PPMV RESPONSE  (cont'd.)
CAS No.«



115-11-7
78-79-5
67-63-0
108-21-4
75-29-6
141-79-7

79-41-4
67-56-1

79-20-9

74-87-3
78-93-3
107-31-3
80-62-6




100-61-8
108-87-2

Compound Name
Heptane
Hexane, N-
Haxene, 1-
Hydroxy acetone
Isobutane
Isobutylene
Isoprene
Isopropanol
Isopropyl acetate
Isopropyl chloride
I sovaler aldehyde
Mesityl oxide
Methacroleln
Methacryllc acid
Methanol
Methoxy-ethanol, 2-
Methyl acentate
Methyl acetylene
Methyl chloride
Methyl ethyl ketone
Methyl formate
Methyl methacrylate
Methy 1-2-pentanol, 4-
Methyl-2-pentanone, 4-
Methyl-3-butyn-2-OL. 2-
Methylal
Methyl aniline, N-
Methylcyclohexane
Methy Icyclohexene, 1-
Volatlllty
Class
LL
LL
LL
LL
G
G
LL
LL
LL
LL
LL
LL
LL
HL
LL
LL
LL
G
G
LL
LL
LL
LL
LL
LL
, LL
HL
LL
LL
OVA
Response Factor Confidence
0.41 I
0.41
0.49
6.90
0.41
3.13
0.59
0.91
0.71
0.68
0.64
1.09
1.20
0.82
4.39 P
2.25
1.74
0.61
1.44
0.64
3.11
0.99
1.66
0.56
0.59
1.37
4.64
0.48
0.44
0.28,
0.38.
0.39.
4.45,
0.29.
0.90.
0.46.
0.72.
0.62.
0.60.
0.57,
0.94.
0.90,
0.31.
3.61,
1.62.
1.46.
0.58.
1.22,
0.51,
2.42.
0.90.
1.27.
0.46.
0.44.
1.06.
3.91.
0.28,
0.36,
Intervals
0.60
0.45
0.66
12.10
1.04
38.50
0.80
1.20
0.83
0.77
0.74
1.29
1.71
14.70
5.60
3.34
2.13
0.64
1.76
0.84
4.14
1.10
2.32
0.69
0.86
1.83
5.57
1.39
0.54
TL
Response Factor
0.73
0.69
4.69 D
15.20
0.55
B
X
1.39
1.31
0.98
2.19 D
3.14
3.49 D
1.06 I
2.01
3.13
1.85
6.79
1.84
1.12
1.94
2.42
2.00
1.63
X
1.46
9.46
0.84
2.79
V
Cofif Idanca
0.33,
0.63.
0.85.
6.11.
0.41.

0.94.
1.04.
0.82.
1.14.
1.43.
1.51.
0.24.
1.66,
1.13.
1.44.
4.86,
0.73,
0.93,
1.72,
1.39.
1.40.
1.22,

1.24,
2.55.
0.68,
1.79,

Intervals
6.10
0.76
> 100. 00
66.40
0.81

2.31
1.72
1.22
6.65
12.00
19.80
4.56
2.48
27.40
2.49
10.40
>100.00
1.38
2.21
5.38
3.15
2.35

1.74
35.20
1.09
5.12

-------
                             TABLE B-l.   RESPONSE  FACTORS  WITH  95* CONFIDENCE INTERVALS ESTIMATED AT 10.000 PPMV RESPONSE  (cont'd.)
Ul
CAS No.«
77-75-8
98-83-9
110-91-8
98-95-3
79-24-3
75-52-5
25332-01-4


109-66-0


123-38-6
79-09-4
71-23-8

115-07-1
75-56-9
110-86-1
100-42-5
79-34-Sc


108-88-3

71-55-6
79-00-5
79-01-6
96-18-4
Compound Name
Methyl pentynol
Methyl styrene, A-
Morpholtne
Nitrobenzene
Nttroethane
Nltromethane
Nltropropane
Nonane-N
Octane
Pant ana
Ptcoltne, 2-
Propane
Proplonaldehyde
Proplonlc actd
Propyl alcohol
Propyl benzene, N-
Propylene
Propylene oxtde
Pyrldlne
Styrene
Tetrachi oroathane. 1,1,1,2-
Tetrachloroethane, 1.1,2.2-
Tatrachl oroethy 1 ene
Tol uene
Trtchlorobenzene, 1,2,4-
Trlchloroethane, 1,1,1-
Trtchloroethane, 1,1,2-
Trlchloroethylene
Trlchloropropane, 1,2,3-
Volatlllty
Class
LL
LL
LL
HL
LL
LL
LL
LL
LL
LL
LL
G
LL
LL
LL
LL
G
LL
LL
LL
LL
LL
LL
LL
HL
LL
LL
LL
LL
OVA
' Response Factor
1.17
13.90
0.92
B
1.40
3.52
1.05
1.54
1.03
0.52
0.43
0.55 I
1.14
1.30
0.93
0.51
0.77
0.83
0.47
4.22
4.83 D
7.89
2.97
0.39
1.21 I
0.80
1.25
0.95
0.96
Confidence
0.71,
9.50.
0.67,

1.20.
3.03.
0.80,
0.94.
0.89,
0.42.
0.38,
0.46,
1.00,
1.03,
0.77,
0.45.
0.44,
0.74,
0.40,
3.45.
1.24,
5.01.
1.71,
0.36,
0.50,
0.72,
1.05.
0.83,
0.64,
Intervals
2.48
21.50
1.40

1.65
4.15
1.48
2.98
1.21
0.66
0.50
0.72
1.32
1.70
1.16
0.58
2.66
0.95
0.55
5.27
> 100. 00
13.80
6.11
0.43
2.94
0.90
1.50
1.09
1.78
TL
Responga Factor
3.42
B
2.59 I
0.01 I
3.45
7.60
2.02
11.10
2.11
0.63
1.18
0.60 P
1.71
5.08 D
1.74
B
1.74 I
1.15
1.16
B
6.91
25.40
B
2.68 D
0.47 I
2.40
3.69
3.93
1.99
V

Confidence Intervals
1.83,

0.64,
0.00.
1.56,
1.91.
1.17.
3.13,
1.68,
0.57,
1.08.
0.59,
1.11.
0.73.
1.06.

0.15.
0.69.
1.03.

3.14.
8.06.

0.79.
0.32.
1.81.
2.77,
2.68,
1.27.
8.54

10.50
82.80
13.00
>100.00
4.47
>100.00
2.75
0.70
1.29
0.69
3.06
> 100. 00
3.50

20.30
2.46
1.34

22.50
> 100. 00

> 100. 00
0.68
3.35
5.16
6.32
3.82

-------
                             TABLE B-l.  RESPONSE FACTORS KITH 95X CONFIDENCE INTERVALS ESTIMATED AT 10,000 PPMV RESPONSE  (cont'd.)

CAS No."
121-44-8
108-05-4
75-01-4

75-35-4
106-42-3

95-47-6

Compound Name
Trlethylamtne
Vtnyl acetate
Vinyl chlortde
Vinyl proplonate
Vlnylldene chloride
Xylene. P-
Xylene. M-
Xylene, 0-
Volatlllty
Class
LL
LL
G
LL
LL
LL
LL
LL
0V
Response Factor
0.51
1.27
0.84
1.00 I
1.12
2.12
0.40
0.43
A
Cgnf 1 dance
0.40,
0.95.
0.61,
0.57,
0.87,
1.71,
0.36,
0.28,

Intervals
0.70
1.82
1.38
1.74
1.52
2.68
0.46
0.85
TL
Response Factor
1.48
.5.91 0
1.06
1.21 I
2.41
7.87
5.67 0
1.40
V
Confidence
0.96,
1.26,
0.59,
0.46.
1.82,
3.49.
0.91,
0.61.

Intervals
2.76
> 100. 00
4.60
3.20
3.35
24.90
> 100. 00
9.33
CD
I
CTt
•CAS numbers refer to the Chemical Abstracts Registry numbers of specific chemicals, 1sowers, or mixtures of chemicals.
G = Gasj LL * light liquid! HL ° heavy liquid
I - Inverse estimation method
D - possible outliers In data
N * narrow range of data
X = no data available
8 - 10,000 ppmv response unachievable
P - suspect points eliminated

-------
TABLE B-2. TESTED COMPOUNDS WHICH APPEAR TO BE UNABLE TO ACHIEVE AN  INSTRUMENT
           RESPONSE OF 10,000 PPMV AT ANY FEASIBLE CONCENTRATION
 Instrument*
     OVA
                   TLV
   CAS1
Compound Name
CAS1
Compound Name
            Acetyl-1-propanol, 3-
 75-1-50    Carbon disulfide
 56-23-5    Carbon tetrachloride
            Dichloro-1-propanol, 2,3
            Dichloro-2-propanol, 1,3
            Oilsopropyl benzene, 1,3
            Dimethylstyrene, 2,4
    1221    Freon 12
 98-01-1    Furfural
            Methyl-2,4-pentanedio1,2
    1660    Monoethanolamine
 98-95-3    Nitrobenzene
108-95-2    Phenol
            Phenyl-2-propanol, 2-
                         98-86-2    Acetophenone
                                    Acetyl-1-propanol, 3-
                         74-86-2    Acetylene
                         79-10-7    Acrylic acid
                        100-52-7    Benzaldehyde
                        100-47-0    Benzonitrile
                         98-88-4    Benzoyl chloride
                        100-44-7    Benzyl chloride
                                    Butyl benzene, Tert-
                         56-23-5    Carbon tetrachloride
                         67-66-3    Chloroform
                       4170-30-0    Crotonaldehyde
                         98-82-8    Cumene
                        108-93-0    Cyclohexanol
                                    Dichloro-1-propanol, 2,3
                                    Dichloro-2-propanol, 1,3
                                    Diisopropyl benzene, 2,4
                                    Dimethylstyrene, 2,4
                         64-18-6    Formic acid
                            1221    Freon 12
                         98-01-1    Furfural
                        115-11-7    Isobutylene
                                    Methyl-2,4-pentanediol
                         98-83-9    Methylstyrene, A-
                            1660    Monoethanolamine
                        108-95-2    Phenol
                                    Phenyl-2-propanol, 2-
                                    Propylbenzene, N-
                       100-42-05    Styrene
                            2860    Tetrachloroethyl ene
 OVA and TLV are two portable hydrocarbon analyzers that have been used in
 previous studies of fugitive emissions.  Operating with a flame ionization
 detector (FID), OVA measures nonmethane hydrocarbon emissions; TLV measures
 total hydrocarbon emissions.

 CAS numbers refer to the Chemical Abstracts Registry numbers of specific
 chemicals, isomers, or mixtures of chemicals.
                                      B-7

-------

-------
            APPENDIX C
CALCULATION OF MASS EMISSION RATES

-------

-------
                                   APPENDIX C
     The calculation of emission rates is illustrated below for VOC:
                         E  - E  + E
                          T    V    L
     where Ej - total VOC emission rate,  Ib/hr,
           E,  - condensed organic liquid  component of the VOC emission rate,
                Ib/hr, and
           EV - vapor component of the VOC emission rate, Ib/hr.

     The emission rate of the vapor component is calculated from the following
equation:
                                     (cs-ca)
                         Ev - K - f"-

     where K  - 5.6 x 10"6, a factor incorporating conversion factors and
                standard temperature and pressure,
           Q  - flow rate of gas through the sample train, actual cubic
                feet/minute,
           P  « sampling system pressure at the dry gas meter, psia,
           MA - molecular weight of the air/VOC mixture,
           C  - concentration of VOC in the gas sample from the sampling
                train, ppm by weight,
           C  - concentration of VOC in the ambient air, ppm by weight, and
           T  - sampling system temperature at the dry gas meter, °R.

     The emission rate of the organic condensate component, EL, is determined
from the following equation:
                         E
                         tL
                                       C-l

-------
     where E,  * organic condensate rate, Ib/hr,
           V  - volume of condensate collected, ml, and
           t  - time over which the sample was collected, min.

This calculation assumes an average density of 0.75 g/cc for the organic
condensate.
     These equations can be modified as necessary to provide for
water-miscible compounds and to express emissions in terms of compounds other
than hexane.
                                       C-2

-------
             APPENDIX D
LEAK RATE/SCREENING VALUE EQUATIONS

-------
                       TABLE  D-l.   PREDICTION  EQUATIONS  FOR  NONMETHANE  LEAK  RATE  FOR
                                   VALVES,  FLANGES,  AND  PUMP SEALS  IN SOCMI  PROCESSES
SOURCE TYPE
Valves3
Gas Service
Light Liquid
*? Flanges
»-•
Pump Seals3
INSTRUMENT
OVA
Service OVA
OVA
OVA
LEAST - SQUARES EQUATION0
NMLK = 1.68 (10~5) (MXOVA)0'693
NMLK = 3.74 (10~4) (MXOVA)0'47
NMLK = 3.731 (10~5) (MXOVA)0'82
NMLK = 1.335 (10"5) (MXOVA)0'898
NUMBER
OF DATA
PAIRS
99
129
52
52
CORRELATION
COEFFICIENT
(r)
0.66
0.47
0.77
0.81
STANDARD
DEVIATION
OF ESTIMATE
0.716
0.902
0.520
0.650
 NMLK  - Nonmethane leak rate (Ib/hr)
 MXOVA = Maximum screening value (ppmv)  -  OVA instrument

aSource:  Langley,  G.  J.,  and R.  G.  Uetherold (Radian Corporation)  1981.   Evaluation of Maintenance for
          Fugitive  VOC Emissions Control.   EPA-600/52-81-080.   pp.  54-56.

 Source:  Wetherold,  R.G., L.P.  Provost,  and C.D.  Smith (Radian Corporation).   Assessment of Atmospheric
          Emissions from Petroleum Refining:  Volume 3.  Appendix B.   (Prepared for U.S. Environmental
          Protection Agency.)  Research  Triangle Park,  North Carolina.   Publication No. EPA-600/2-80-075c.
CNMLK is given in units of Ib/hr.   The units may have to be changed for reporting purposes.

-------
                 APPENDIX E
SELECTION OF SAMPLE SIZE FOR FLANGE SCREENING

-------

-------
                              APPENDIX E
     SELECTION OF SAMPLE SIZE FOR SCREENING FLANGES AND OTHER CONNECTORS

     In estimating emissions for a given process unit, all equipment
components must be surveyed for each class of components.  For example, all
valves in gas/vapor service must be screened to establish the number of
components in each of the given ranges of screening values.  The one exception
to this "total component screening" criterion is the category of flanges and
other connectors.  In typical process units, flanges and other connectors
represent the largest count of individual equipment components, making it
costly to screen all components.  The purpose of this appendix is to present a
methodology for determining how many flanges or other connectors must be
screened to constitute a large enough sample size to identify the actual
screening value distribution of flanges and other connectors in the entire
process unit.  Please note that the sampling is to be a random sampling
throughout the process unit.
     The basis for selecting the sample population to be screened is the
probability that at least one "leaking" flange will be in the screened
population.  The "leaker" is used as a representation of the complete
distribution of screening values for the entire class of sources.  The
following binomial distribution was developed to approximate the number of
flanges and other connectors that must be screened to ensure that the entire
distribution of screening values for these components is represented in the
sample:
               n > N * [1 -  (1 -P)1/D1                        (E-l)
where D may be taken as (fraction of leaking flanges) * N and p > 0.95.
     Currently available data gathered for flanges in VOC service in chemical
process units show an estimated 2.09 percent flanges leaking.  The  EPA
recommends starting with the factor 0.0209 for the fraction of leaking
flanges, since the actual leak frequency for flanges in a chemical process
unit probably will not be known prior to the selection of a sample size for
screening.  A larger sample  size will be required for units exhibiting a lower
fraction of leaking flanges  and other connectors.  If a lower leak frequency
                                       E-l

-------
for a process unit is known or can be estimated, it may be substituted into
the equation above in place of 0.0209.
     After 'n' flanges have been screened, an actual leak frequency should be
calculated as follows:
          Leaking Frequency - Number of leaking flanqes       (£.2)

Then, the confidence level of the sample size can be calculated using the
following equation, based upon a hypergeometric distribution:
          D   i   fN-D'H (N-nH
          K * i '  N! (N-O'-n)l                               (E-3)

where N  » Total population of flanges and other connectors
      n  » Sample size
      n/ , Number of leaking flanges and connectors * ».
                              n
If  'p' calculated in this manner is less than 0.95, then a less than
95  percent confidence exists that the screening value distribution has been
properly identified.  Therefore, additional flanges/connectors must be
screened to achieve a 95  percent confidence level.  The number of  additional
flanges required to satisfy the requirement for a 95 percent confidence  level
can be calculated by solving Equation (E-l) again,  using the leak  frequency
calculated  in  Equation  (E-2), and subtracting the original sample  size.  After
this additional number  of flanges/connectors have been screened, the revised
fraction of leaking components  and the confidence level of the new sample  size
(i.e.,.the  original sample size plus  the  additional flanges/connectors
screened) should be recalculated using Equation (E-3).  The Agency requires
sufficient  screening  to achieve a 95  percent confidence level, until a maximum
of 50 percent  of the  total number of  flanges and other connectors  in the
process unit  have  been  screened.  The EPA  believes  that 50 percent of the
total  flange/connector  population is  a reasonable upper limit for  a sample
size.   If half of  the total number of flanges and other connectors are
screened, no  further  flange screening is  necessary, even  if  a 95 percent
confidence  level  has  not been  achieved.
                                        E-2

-------
            APPENDIX F
DETERMINATION OF EMISSION RATES FOR
   DEFAULT ZERO SCREENING VALUES

-------
                                  APPENDIX F

                      DETERMINATION OF EMISSION RATES FOR
                         DEFAULT ZERO SCREENING VALUES

     In developing emission estimates using correlations, the emissions
associated with zero screening values are estimated using default emission
rates.   Due to their logarithmic form, the correlations estimate zero
emissions for zero screening values.  Field testing, however, indicates that
there are finite emissions associated with equipment screening at the
background concentrations.  To account for these emissions in the protocol,
EPA has established some default zero screening values and their associated
emission rates based on data gathered during the SOCMI Maintenance Study.
     Twelve data pairs for valves in gas/vapor service were bagged to
determine mass emissions during the Maintenance Study.  The mass emission
rates for these twelve sources were averaged to establish the emission rate
for gas/vapor valves screening at 0 ppm above the background level.  This
average emission rate is 0.000033 kilograms .per hour per source.  Using the
SOCMI/OVA correlation for gas/vapor valves shown in Appendix D and this
emission rate, the "default zero" screening value was established (8 ppm).
All components screening at or below this default screening value are assumed
to have an emission rate equal to the default zero emission rate.
     No mass emissions data for zero screening values were available for
equipment components other than valves in gas/vapor service.  The default  zero
emission rates for all equipment types and services were established using the
correlations shown in Appendix D and the default zero screening value (8 ppm)
calculated from the gas/vapor valves bagging data.
                                                            Zero Screening
                                "Default Zero"              Value Emission
Equipment Type/Service       Screening Value, pom         Rate fkq/hr/source)
Valves, gas                           8                        0.000033
Valves, light liquid                  8                        0.000451
Flanges                               8                        0.000093
Pumps and all other components        8                        0.000039
                                      F-l

-------

-------
                                APPENDIX G
                  DEVELOPMENT OF A DEFAULT EMISSION RATE
               FOR EQUIPMENT THAT DOES NOT SCREEN ABOVE THE
                         BACKGROUND CONCENTRATION
Text adapted from Guidance for Estimating Fugitive Emissions, published by the
Chemical Manufacturers Association.

-------

-------
                                  APPENDIX G
             DEVELOPMENT OF A DEFAULT EMISSION RATE FOR EQUIPMENT
            THAT DOES NOT SCREEN ABOVE THE BACKGROUND CONCENTRATION.

     If bagging data for equipment components that do not screen above
background do not exist, the EPA default emission rates in Appendix F
should be used to estimate emissions from these sources.  If a process unit
intends to develop its own default zero emission rates, it must bag samples
from each equipment type (e.g., gas/vapor valves^ light liquid valves) for
which it plans to determine default zero emission rates.  If bagging data
exist, the first step in developing a default emission rate is to adjust
for the background concentration by subtracting the background bag from the
source bag; that is, calculate
     where LR_ .  » leak rate from source bag i
             5,1
           LR« .,-  " leak rate from background bag i
     Next, take the logarithm of the data and add a constant (if necessary)
to avoid taking the logarithm of a. nonpositive  X.. The following procedure
is used:

     Case 1:  If all of the data are positive,  y. = log(Xi)
     Case 2:  If all of the data are nonnegative with at least one zero,
               y. - log(Xi + C)
     where C = largest power of 10 less than the smallest nonzero value of
               X (e.g., if the smallest nonzero value of X is 8 x 10" ,
               then C - 1 x 10'^).
     Case 3:  If some of the data are negative, the smallest value (Kl) is
first subtracted from each observation in the data set (note that Kl is a
negative number).  This results in all nonnegatives, with at least one zero
value.  Then K2 is added to each observation in the set, where K2 » largest
power of 10 less than the smallest nonzero value of the updated nonnegative
data set.  Hence, the logarithmic expression used is
     yi - log(Xi  + C)    where C - K2-K1

                                      G-l

-------
The average, standard deviation, and scale bias correction factor (SBCF)
are calculated for the Yi as follows:
                  n
          y -  1  £  y<
               n i«l  1

           9        "      - 2
          sz -_l_ Y (yryr
               n-l ^   1
                   1-1
          SBCF - g [(InlO)2 (S2/2)]
     where g(t) - 1 + (n-l) t + (n.-.l)3t2  +  (n-l)5t3          + ...
                        n       n22!(n+l)   n33!(n+l)(n+3)
(Approx. 10-15 terms required to obtain accuracy of + 10" )

     The estimate of emissions for components screening below the lower
limit of detection (LLD) for the screening instrument, (EF,*), is then:
     EFj* - max [(10ySBCF)-C, ]3]

     where C - 0 for Case 1, and previously stated for Case 2 and Case 3.

     A  statistical test  is used to compare the calculated EF,* with the
established  'default zero' in Appendix F.  The results of this test
determine which value  (EF,* or  'default zero') is used to represent
emissions for components screening below the LLD.  The test involves two
cases:
                                       G-2

-------
Case A:  If EF^Default zero', then use EF^ if


                     y+t(.95, n-l)Vs2/n
            EFUL - 10                      SBCF < 'Default zero'

          (Use xDefault zero' if EFUL > 'Default zero')

Case B:  If EFJA > 'Default zero', then use EFj* if
                     y-t(.95, n-l
            EF,,  « 10                      SBCF > \Default zero'

               (Use xDefault zero' if EFLL < xDefault zero')
A confidence interval for EFj can be calculated as:  .
               y+tj(.975, n-lhn          y+tj(.975, r\-
            (10                       SBCF, 10                     SBCF)
Note:  This  interval should only involve the nonnegative region.  In the
       case  where the above computation includes negative values, it should
       be truncated to zero.
                                      G-3

-------
                              eat reaa i
-------