&ER&
                                   United States
                                   Environmental Protection
                                   Agency
                                   Industrial Environmental Research
                                   Laboratory
                                   Research Triangle Park NC 27711
                                   Research and Development
                                   EPA-600/S2-81 -111  Sept. 1381
Project  Summary
                                   Analysis of  SOCMI  VOC
                                   Fugitive  Emissions  Data

                                   G. J. Langley, S. M. Dennis, J. F. Ward, and L. P. Provost
                                    The report gives results of an exami-
                                   nation of fugitive emissions data from
                                   Synthetic Organic Chemical Manufac-
                                   turing Industry (SOCMI) processing
                                   units (collected  under earlier EPA
                                   studies)  for correlations between
                                   process variables and leak frequency.
                                   Although  line temperature  did not
                                   have a consistent relationship with
                                   leak frequency, the data showed that
                                   leak frequency increased with increas-
                                   ing  line pressure. Also, emissions
                                   factors for three process types (vinyl
                                   acetate, cumene, and ethylene) were
                                   developed and presented. Increases in
                                   mass emissions due to occurrence and
                                   recurrence of leaks for these three
                                   process  types  are  also estimated.
                                   Finally, the effect of adjusting portable
                                   hydrocarbon readings by chemical
                                   response factor curves on  leakage
                                   frequency estimates is investigated.
                                   Despite the wide range of response
                                   factors encountered, the adjusted leak
                                   frequencies were essentially the same
                                   as the unadjusted frequencies.
                                    This Project Summary was devel-
                                   oped by EPA's Industrial Environmen-
                                   tal Research Laboratory. Research
                                   Triangle Park, NC. to announce key
                                   findings of the research project that is
                                   fully documented in a separate report
                                   of the same title (see Project Report
                                   ordering information at back).

                                   Introduction
                                    The contribution of fugitive leaks from
                                   process unit components are  being
                                   investigated as a potential source of
                                   Volatile Organic Compound (VOC) emis-
                                   sions in the Synthetic Organic Chemical
                                   Manufacturing Industry (SOCMI). The
                                   purpose of this study was to provide an
                                   in-depth analysis of data on emissions
                                   collected under EPA contracts 68-02-
                                   3171-1,68-02-3173-2 and -11, 68-02-
                                   3174-5, and 68-02-3176-1 and -6, and
                                   68-03-2776-4. These data were col-
                                   lected  by Radian, PEDCo, TRW, and
                                   Acurex and are summarized  in EPA
                                   reports EPA-600/2-81-003  (NTIS PB
                                   81-141566) and 600/2-81-080 (NTIS
                                   PB 81 -206005). The results of this study
                                   will be available for use in evaluating
                                   VOC fugitive emissions.
                                    The study design and test procedures
                                   for the data analyzed in this report are
                                   described in the above-mentioned EPA
                                   reports. The 24 process units studied in
                                   the data collection programs were
                                   selected to represent a cross-section of
                                   the population of the SOCMI. Several
                                   factors were considered during process
                                   unit  selection, including total  annual
                                   production in volume,  number of
                                   producers, process  conditions, cor-
                                   rosivity, volatility, toxicity, and value of
                                   the final chemical product. Factors
                                   varied widely from unit type to unit type,
                                   so that the selected process unit types
                                   represented a reasonable sample of the.
                                   variety of chemical process .units en-
                                   countered in SOCMI.
                                    The leak frequency in SOCMI was
                                   evaluated by the collection of screening
                                   data  from 24 process units, where a
                                   screening value is the maximum re-
                                   peatable concentration of  total hydro-
                                   carbons detected at a source  with a
                                   portable hydrocarbon detector. Main-
                                   tenance was evaluated by measurement
                                   of fugitive emission leak rates (Ip/hr) at

-------
selected sources before  and after
maintenance at six process units
representing three chemical processes.
The  results  of these two programs
provide the background  necessary for
the current study:
  • Source population data.
  • Screening value profiles for each
    source type.
  • Screening-to-emission  rate rela-
    tionships.
  The screening procedures began with
the definition of the  process unit
boundaries. All feed streams, reaction/
separation facilities,  and product and
by-product delivery lines were identified
on process flow diagrams and in the
process unit. Process data, including
stream compositions, line temperatures,
and line pressures, were obtained for all
flow streams.
  The Century Systems  Models OVA-
108 and OVA-128 hydrocarbon detectors
were  used for screening. The detector
probe of the instrument was placed
directly on those areas of the sources
where leakage  would typically occur.
For example, gate valves were screened
along the circumference  of the annular
area around the valve stem where the
stem exits the packing gland and at the
packing-gland/valve-bonnet interface.
The actual  leak rates were measured
using a flow-through method described
in report EPA-600/2-80-075a (NTIS PB
80-225253), and  were  analyzed on
Byron Total Hydrocarbon Analyzer.
  All accessible sources of the follow-
ing source types were screened:
  • Process valves.
  • Pump seals.
  • Compressor seals.
  • Agitator seals.
  • Relief valves.
  • Process drains.
  • Open-ended lines.
Also,  a randomly selected subset of
flanges were screened. Originally, only
5 percent of all flanges were screened.
The subset was increased to 20 percent
of all  flanges when initial results indi-
cated a higher frequency of  emitting
flanges than  had been encountered in
previous programs. The  important var-
iables  available  from this  study are:
screening value, source category, stream
service, source type, chemical produced,
ambient  temperature elevation, line
temperature, and line pressure. For the
purposes of  this report, a source  is
defined as "leaking" if  its screening
value is greater than or equal to 10,000
ppmv.
  This summary presents conclusions
 based on four distinct data analysis
 tasks: 1) an analysis of important corre-
 lating process  parameters (line pres-
 sure, etc.); 2) emission factor develop-
 ment for three specific chemical
 processes (seven units); 3) an evalua-
 tion of the increase in mass emissions
 due to  occurrence and recurrence  of
 leaks; and 4) an analysis of the impact
 on  leak frequency from adjusting
 screening values by  VOC detector
 response factors.
Conclusions
1) Relationship of Leak Frequency to
Process Parameters
  The  process  parameters that  were
examined for  their  effect  on leak
frequency were:  process,  service,
material in the  line, line pressure, line
temperature, ambient temperature, and
source elevation. Data on four source
types (valves, pump seals, flanges, and
open-ended lines) were used to exam-
ine the effects of these parameters. The
sources were grouped into 32 categories
based  on source type, process  type,
stream service, and primary chemical in
the  line.  These groupings were  for
statistical  reasons and were not based
on engineering  reasoning.
  Stream  service was defined as gas,
light liquid,  or heavy liquid. Heavy
liquids were  not  included  in any
analyses,  since  they leaked so rarely
regardless of the other conditions. Gas
stream service  generally  had  a higher
leak frequency than light liquid service.
Proceeding with four source types and
two stream service types, the data were
then categorized by process unit  as
ethylene processes,  high  leaking  proc-
esses,  or  low leaking  processes. The
ethylene units were analyzed separately
because of the large number of sources
in ethylene processes  and their high
leak frequency.  The high leaking group
consisted of all other units with greater
than 1  percent of all source  types
leaking. The low leaking group consisted
of all units with less than 1 percent of all
source- types  leaking.  Since very few
sources leaked,  the low leaking process
units were not  considered in further
analyses.  Within these process unit
groups, the data were further subdivided
by primary materials in the line. Caution
should  be used in these evaluations,
however, since  other chemicals in the
line  may also have an effect on leak
frequency.
  Examination of the data within these
categories resulted in  the following
conclusions for this data set:
  •  Leak frequency was affected not
     only by the  type of  chemical
     process  but also by the type  of
     primary material in the line.
  •  Control valves had  a higher  leak
     frequency than block valves.
  •  For block valves, gate valves had a
     higher leak frequency than most of
     the other types, and plug and ball
     valves had lower leak frequencies.
  •  On-line pump  seals had an overall
     leak  frequency of  13.1 percent
     versus 4.9 percent  for  off-line
     pump seals.
  •  These data did not show a differ-
     ence in  leak  frequency between
     double and single mechanical
     pump seals, although the type  of
     barrier fluid  was unknown  and
     therefore unaccounted for in this
     analysis.
  •  Line pressure  was seen to have a
     statistically significant effect  in
     almost every case, with higher
     levels of pressure associated with
     higher leak frequencies.
  •  Line temperature had no consistent I
     effect on  leak frequency.  The
     combined effect of line pressure
     and temperature was important in
     some cases.
  •  Ambient temperature had a con-
     sistent effect  on leak frequency;
     however, the effect was not statis-
     tically significant for a majority  of
     the cases. Higher leak frequencies
     tended to be  associated with the
     higher ambient temperature cate-
     gory.
  •  Elevation had  no consistent effect
     on  leak  frequencies. In the  four
     cases where a statistically signifi-
     cant effect was observed, sources
     at ground level had a higher  leak
     frequency than sources at higher
     elevations.
2) Emission Factor Development
  The sources included in the develop-
ment of the emissions factors are all
valves and pump seals screened in the
seven ethylene, cumene, and vinyl
acetate process  units, or 51.2 percent
(16,575)  of all valves and pump seals
screened in  the screening program.
Since leak rate screening value models
were only developed for  these three
process types, emissions factor estima-
tion was limited to these three processes..
  The emissions factors  developed  inJ
this study are reported in Table 1. The"

-------
 Table 1    Estimated Emission Factors for Nonmethane Hydrocarbons from Valves and Pump Seals

                                                       Emissions Factor (95% Confidence Interval)
         Source Type
                                            (Ib/hr)
           (kg/hr)
     Valves
     — Gas Service
       Ethylene processes
       Cumene processes
       Vinyl acetate processes
     — Light Liquid
       Ethylene processes
       Cumene processes
       Vinyl acetate processes
     Pump Seals
     — Light Liquid
       Ethylene processes
       Cumene processes
       Vinyl acetate processes
                                  0.024(0.008, 0.07)
                                  0.011(0.003, 0.05)
                                  0.0046(0.001, 0.03)

                                  0.020(0.007, 0.06)
                                  0.0056(0.002, 0.02)
                                  0.0003(0.0001, 0.002)
                                  0.069(0.006, 0.8)
                                  0.052(0.001, 2.7)
                                  0.0043(0.0001, 0.1)
 0.011(0.004, 0.03)
 0.0052(0.001, 0.02)
 0.0021(0.0004,  0.01)

 0.010(0.003, 0.03)
 0.0025(0.001, 0.01)
 0.0001(0.00003, 0.001)
 0.031(0.003, 0.4)
 0.023(0.0004,  1.2)
 0.0020(0.00006, 0.06)
 emissions factors for the ethylene
 process are consistently higher than the
 factors for the cumene and vinyl acetate
 processes.  The vinyl acetate process
 tends to have the  lowest  emissions
 factors of the three process types.
   Cumulative distributions of screening
 values and mass  emissions as a
 function of screening values were also
I developed for each of  the three proc-
 esses. Table 2 gives the estimates and
 confidence intervals from these  curves
 for a 10,000 ppmv screening value.
 3) Increase in Mass Emissions Due to
 Occurrence and Recurrence
   Data collected during the EPA SOCMI
 maintenance program were further
 analyzed to estimate the effects  of leak
 occurrence and recurrence on mass
                             emissions. The following conclusions
                             are based on these analyses:
                               • The increase in emissions for
                                 valves for which a leak occurred
                                 over a 1  to 6 month  period was
                                 estimated to be  530 percent (95
                                 percent confidence interval of 200
                                 to 900 percent).
                               • Not enough  data was available to
                                 accurately quantify the effect on
                                 emissions from  leak  occurrence
                                 from pump  seals.  However,  the
                                 percent increase estimate was 75
                                 percent with a 95 percent confi-
                                 dence interval of -100 to  6000
                                 percent.
                               • The percent  increase in emissions
                                 for valves with a leak recurrence
                                 within the 6  month period  was
    estimated to be 510 percent (95
    percent  confidence interval of
    -100 to 1700 percent).
  • Further analysis of the  effect of
    valve maintenance  on emissions
    showed a 98 percent reduction in
    emissions for valves which were
    "repaired" (screening value <10,000
    ppmv after maintenance) and a 63
    percent  reduction for  sources
    which were "not repaired"(screen-
    ing value remained >10,000 ppmv
    after simple, on-line maintenance).

4) Impact of Response Adjustments on
Leak Frequency Estimation
  Three different techniques were used
to adjust the original screening value for
each source:
Table 2.
Summary of Percent of Sources Distribution Curves and Percent of Mass Emissions Curves at Screening Value
of 10,000 PPMV
                                          Percent of Sources
                                       Screening >10,000 ppmv
                                                                     Percent of Mass Emissions
                                                                       Attributable to Sources
                                                                      Screening > 7 0,000 ppmv
Source Type
Valves
Gas
Ethylene
Cumene
Vinyl acetate
Light Liquid
Ethylene
Cumene
Vinyl acetate
Pump Seals
Light Liquid
Ethylene
Cumene
Vinyl acetate
Estimate


15
16
3.7

26
12
0.2


30
14
1.7
95% Confidence
Interval .


(14, 16)
(13, 19)
(2,5)

(24, 27)
(10, 13)
(0, 0.4)


(20, 39)
(1. 27)
(0,4)
Estimate


94
94
90

89
80
25


96
89
67
95% Confidence
Interval


(93, 95)
(90, 96)
(85. 94)

(87, 90)
(72, 86)
( 9. 47)


(90, 98)
(50, 98)
( 5, 92)

-------
       • The original OVA reading adjusted
         for the associated OVA response
         relationship of the primary chemi-
         cal compound in the line.
       • Weighted logarithmic average of
         response of primary and secondary
         chemicals.
      '• Weighted arithmetic average of
         response of primary and secondary
         chemicals.
       The  percent of leaking valves was
     calculated for  each of  the three esti-
     mates for  both gas and light liquid
     services. The three estimates were
     found to be similar in most cases to the
     leak frequency based on the original
     screening values.  Table 3 presents the
     overall results.
                                                                                                                      4
         Table 3.     Comparable Estimates for Percent Leaking (Valves/
                     (24 SOCMI Process Units)
Process
Stream
Gas
Light
Liquid
Number
Screened3
9,374
18,133
Percent
Leaking
Based on
OVA
Readings
11.3
6.1
Percent
Leaking
Based on
Method 1
Adjustments*
10.1
5-3
Percent
Leaking
Based on
Method 2
Adjustments0
10.2
5.6
Percent
Leaking
Based on
Method 3
Adjustments*
10.3
5.5
          *119 sources with soreening values = 1O.OO1 ppmv were excluded.
          ^Method 1 is the adjustment to the OVA reading based on the response of the primary
           chemical in the line.
          "Method 2 is the mixed chemical weighted logarithmic average technique.
          aMethod 3 is the mixed chemical weighted average technique.
                                                G. J. Langley, S. M. Dennis, J. F. Ward, and L. P. Provost are with Radian Corp.,
                                                 P.O. Box 9948, Austin, TX 78766.
                                                Bruce A. Tichenor is the EPA Project Officer (see below).
                                                The complete report, -entitled "Analysis of SOCMI VOC Fugitive Emissions
                                                 Data," (Order No. PB 81-234 270; Cost: $18.50,  subject to change) will be
                                                 available only from:
                                                        National Technical Information Service
                                                        5285 Port Royal Road
                                                        Springfield. VA 22161
                                                        Telephone: 703-487-4650
                                                The EPA Project Officer can be contacted at:
                                                        Industrial Environmental Research Laboratory
                                                        U.S. Environmental Protection Agency
                                                        Research Triangle Park,  NC 27711
                                                                                    
-------