-------
F. G. Mesich
           From the  results of this study, nomographs have also been prepared
 relating hydrocarbon concentration at the source (screening value)  to the
 percentage of  each  source type expected to have screening values above any
 selected value.   Other nomographs have been prepared relating screening values
 to  the  percentage of total mass emissions which can be expected from sources
 with  screening values greater than any given value.
          These nomographs for  the valves,  flanges,  and  pump  seals  (with
appropriate stream groups) are  presented  in Figures  15 through 21.  The "B"
figures relate the percent of total mass  emissions for a given source cate-
gory to screening values; the "A" figures relate  the percent  of sources to
screening values.  The screening values in  these  nomographs are the hydro-
carbon concentrations obtained  at the  source (0 cm)  with a Bacharach TLV
Sniffer calibrated with hexane.

          Confidence intervals  are included on each  of these  nomographs.
The confidence intervals for both types of  nomographs indicate how well the
cumulative function has been estimated from the data collected in this
program.

          The 95 percent confidence intervals for the cumulative percent
of sources can be interpreted as ranges of  values which  contain the actual
percent from the population of  sources studied.   Note that these intervals
apply to the entire population  of sources (i.e.,  a composite  of all United
States refineries), and are not necessarily applicable to a finite number
of sources at any particular refinery.  Because of the nature of the func-
tion, the confidence intervals will be approximately valid any time a random
sample of greater than 100 sources is  being considered.

          The 90 percent confidence intervals for the cumulative percent of
total emissions function can be interpreted as ranges of values which contain
the actual percent of total emissions  function for the entire population of
sources.  Again, these intervals describe how well the function has been
estimated for the entire population and are not directly applicable to a
particular refinery situation with a finite number of sources.  The varia-
tion of the function for a particular  sample of sources  is a  complex function
of the number of sources.

          The nomographs must be carefully  evaluated when comparing these
estimates to actual measured emissions (samples sources).  As discussed
earlier, the correlation between screening  values and actual  leak rates is
imperfect.  Because of this, values obtained from the nomographs for percent
of total emissions caused by a  specific percent of total sources may not
exactly match similar values for measured leak rates.  Table  8 gave
the distribution of total emissions as a  function of measured leak  rates.
In most cases, the nomographs will indicate a higher percentage than in
Table 8 of sources being responsible  for  a  given  percentage of total
emissions.  In this sense,  if actual  leak rates could be measured,  the
                                     167

-------
F. G.  Mesich
           u
           u
           u
           a
           o
           vt
           u
           u
TOO




 90




 80




 70




 60




 50




 40




 30




 20





 10




  0
                                  Upper Um1t of 951 Confidence  Interval
                                                Estimated Percent of Sources
                        Lower Limit of 955

                        Confidence Interval
                1  2  345  10      50 100       1,000      10,000      100,000    1,000,000



                             Screening Value  (pptnv)  (Log^Q Scale)
              Percent of Sources - Indicates  the percent of sources with screening


                                 values greater than the selected value.
                   Figure 15A.   Cumulative  Distribution of  Sources and

                                   Total  Emissions by Screening Values  for

                                   Valves - Gas/Vapor Streams.
                                           168

-------
F. G.  Mesich
            100




             90





             30

          *
          e
          o

          «  70
          •A


          UJ

          -  60
          Wt

          £


          •  50
          *«
          o
          t—


          o  40

          4J
          e
          «


          S  30
          &



             20





             10
                                        Upper Limit of 901

                                        Confidence Interval
              Estimated Percent of

              Total Mass Emissions
                   Lower Limit of the

                   90S Confidence Interval
                   i  i 1
                                  1   !
                1   2  345  10
50 100
1,000
10,000
100,000    1,000,000
                                Screening Value (ppmv) (Log^g Scale)
            Percent of Total  Mass Emissions - Indicates the percent of total emissions

                                           attributable to sources with  screening values

                                           greater than the selected value.
           Figure 15B.   Cumulative Distribution  of Source  and  Total

                          Emissions  by Screening Values for  Valves -

                          Gas/Vapor  Stream.
                                            169

-------
F. G.  Mesich
                                          Upper Limit of 95% Confidence Interval
                         Lower Limit of 95%
                         Confidence  Interval
                                                   Estimated Percent of Sources
                  1  2  345  10      50 100       1,000       10,000     100,000    1,000,000

                               Screening Value (ppmv) (Log10 Scale)

                Percent of Sources -  Indicates the percent of sources with screening
                                   values  greater than the selected value.
                    Figure  16A.   Cumulative  Distribution of  Source  and
                                   Total Emissions by Screening Values for
                                   Valves - Light Liquid/Two-Phase  Streams.
                                           170

-------
F.  G. Mesich
«
*J
o
            100


             90


             30


             70


             30


             50


             40


             30


             20


             10
                                                             Upper Limit of 90*
                                                             Confidence Interval
                          Estimated Percent df
                          Total Mass Emissions
                  111!   I
                                 I
                                    Lower Limit of the
                                    90S Confidence Interval
               1   2 345  10
                        50 100
1.000
                                                        10,000
                                                         100.000    1,000,000
                               Screening Value (ppav) (Log-|g Scale)
            Percent of Total Mass  Emissions  - Indicates the percent of total  emissions
                                          attributable to  sources  with screening values
                                          greater  than the selected value.
              Figure 16B.
                    Cumulative  Distribution of  Source and  Total
                    Emissions by Screening Values for Valves  -
                    Light  Liquid/Two-Phase Streams.
                                            171

-------
F. G.  Mesich
             V
             u
             £
             V
100

 90


 80


 70


 60


 SO


 40


 30


 20


 10
                               Upper  Limit of 95% Confidence  Interval

                               ^ /f
                                        Estimated Percent of Sources
                    Uoper Limit of
                    Confidence Interval
                   1  2 345  10     50 100       1,000      10,000     100,000    1.000,000

                                Screening Value  (ppmv)  (Log-|Q Scale)

                 Percent of Sources - indicates  the percent of sources with screening
                                   values greater than the selected value.
                   Figure  17A.   Cumulative Distribution  of Sources  and
                                   Total  Emissions by  Screening  Values for
                                   Valves - Heavy Liquids Stream,
                                           172

-------
F. G.  Mesich
                                                    Upper Limit of 90S
                                                    Confidence Interval
          a
          u
                                        \     \
                   Estimated Percent of —*-\  \
                   Total Mass Emissions
                                              \  \  \
                      Lower Limit of the         \  \  \
                      90S Confidence Interval      \   \ \
                                                \  \  \
               1  2  345 10
    50 100
1,000
10,000
100,000    1,000,000
                                Screening Value (ppav)  (Log-|0 Scale)
            Percent of Total  Mass Emissions - indicates the  percent of total  emissions
                                           attributable to sources with screening values
                                           •greater than the selected value.
             Figure 17B.
Cumulative Distribution of  Source and  Total
Emissions by  Screening  Values for Valves -
Heavy Liquid  Streams.
                                           173

-------
F.  G.  Mesich
              100


               90


               30


               70


               60
          jo    4Q
               30


               20


               10
                     Upper Limit of 95S Confidence Interval
                                  Estimated Percent of Sources
Lower Limit of the 955
Confidence Interval
                  1 2 345  10      50 100       1,000      10,000


                               Screening Value (ppmv) (Logi0 Scale)
                                           100,000    1,000,000
              Percent of Sources  - Indicates the percent of sources with screening
                                values  greater than the selected value.
                    Figure  ISA.   Cumulative  Distribution of  Sources and
                                    Total  Emissions  by Screening Values  for
                                    Valves - Hydrogen Service.
                                            174

-------
F.  G- Mesich
             100


              90



              30



              70


              60


              50



              40


              30


              20


              10
Estimated Percent of
Total Mass Emissions
                   t  n i   i
                                  j	i
                                         Upper Limit of 90S
                                         Confidence Interval
            Lower Limit of the 901
            Confidence Interval
                1   2 345  10
       50 100
1,000
10,000
100,000    1,000,000
                                Screening Value (ppmv) (Log-|Q Scale)
             Percent of Total Mass Emissions -  indicates  the percent of total emissions
                                            attributable to sources with screening values
                                            greater than the selected value.
                    Figure 18B.   Cumulative  Distribution of  Source and
                                    Total  Emissions  by Screening Values  for
                                    Valves  - Hydrogen Service.
                                             175

-------
F. G.  Mesich
 100

  90

  80

  70
I
1  60
I
l
;  so

  40

  30

  23

  10

   3
                                   Upper Limit of 95X Confidence Interval
                                                 Estimated Percent of Sources
                      Lower Limit of the 95
                      Confidence Interval
    1  2 345  10     50 100        1,000      10,000

                 Screening Value  (ppmv)  (LogjQ Scale)
                                                               100,000    1,000,000
           Percent of Sources - indicates the percent of sources with screening
                             values greater than the selected value.
             Figure  19A.   Cumulative Distribution  of Sources  and
                             Total  Emissions by  Screening Values for
                             Pump Seals -  Light  Liquid/Two-Phase Streams,
                                          176

-------
G. Mesich
        100


         90


         80
      e
     «   70
     M
     5
     *   60
50


40


30


20


10
                                                      Upper Limit of 90S
                                                      Confidence Interval
                         Estimated Percent of
                         Total Mass Emissions
              lit;   i
                                         Lower L1ro1t of the
                                         90S Confidence Interval
           1   2  345  10      50 100       1,000      10,000     100,000


                           Screening Value  (ppav) (Logio Scale)
                                                                1,000,000
        Percent of Total Mass Emissions - Indicates the percent of total emissions
                                      attributable to  sources  with screening values
                                      greater than the selected value.
          Figure 19B.   Cumulative Distribution of  Source and  Total
                         Emissions by  Screening  Values for Pump Seals
                         Light Liquid/Two-Phase  Streams.
                                       177

-------
F. G.  Mesich
                         Upper Limit of 95% Confidence  Interval
             _   Lower  Limit  of     N
                 the  95%  Confidence
                                       Estimated Percent of Sources
             1  2  345  10      50 100       1,000      10,000     100,000   1,000,000


                          Screening Value  (ppmv)  (log-jo  Scale)

           Percent of Sources - indicates the percent of sources with  screening

                             values greater than the selected value.
              Figure  20A.  Cumulative  Distribution  of Sources and
                            Total Emissions by Screening Values  for
                            Pump Seals  - Heavy Liquids
                                          178

-------
F.  G. Mesicti
          e
          o
          •
         *>
          o
          
-------
F.  G.  Mesich
           e
           0)
           o
 100



  90




  80




  70


I


I  60



  50




  40




  30




  20




  10
                   Upper Limit of 95





                	•—-	""""" •~ —
Jpper Limit of 951 Confidence Interval




                 .Estimated Percent of Sources
                   1111   i
                                               •«-1 - Lower Limit of 95S Confidence  Interval
                                                     - ^   i           i	i
                 1  2 345  10
                      50  100
                           1,000
10,000
100,000    1,000,000
                              Screening Value (ppav)  (Log-|g Scale)
              Percent of Sources - Indicates the percent of sources with  screening

                                 values  greater than the selected value.
               Figure  21A.   Cumulative Distribution  of Sources and

                               Total Emissions by  Screening Values  for

                               Flanges.
                                             180
-------
F. G.  Mesich
                                                          Upper Limit of 905
                                                          Confidence Interval
                        Lower Limit of the
                        905 Confidence Interval
1   2 345  10
                               50 100
1,000
                                                      10,000
                                                    ICO,COO    1,000,000
                              Screening Value (ppmv) (Log-jQ Scale)

          Percent of Total Mass Emissions - Indicates the percent of total emissions
                                         attributable to sources with screening  values
                                         greater than the selected value.
             Figure  21B.   Cumulative Distribution  of Source  and
                             Total Emissions by  Screening  Values
                             for  Flanges.
                                           181
-------
F. G. Mesich
 nomographs will be conservative (i.e., they will identify more sources to
 achieve a given level of reduction on total emissions than would be identi-
 fied through sampling).   In a practical sense, however, it is unreasonable
 to expect that every source with a screening value exceeding a specific level
 could be bagged and sampled.  Since, at this time, there is no better method
 than screening for identifying sources for maintenance, the nomographs are
 appropriate for evaluating maintenance and control options.

          The nomographs are therefore useful  in evaluating the potential
effectiveness of maintaining and repairing sources for reducing emissions.
For example, approximately 50 percent of valves in gas vapor stream  service
can be expected to have screening values above 50,000 ppmv.  However,  these
5 percent £>f the valves are  responsible for an estimated 95 percent  of  the
mass emissions.  Similarly,  for a screening value of 10,000 ppmv,  the  per-
cent of sources and percent  of emissions are 9 percent and 99 percent,
respectively.

COOLING TOWERS AND WASTEWATER TREATMENT SYSTEMS

          During the  course  of this  program, extensive effort was  expended
in an attempt to directly determine  the hydrocarbon emissions from open
sources such as cooling towers and wastewater  treatment systems.  This was
an exceedingly difficult task in that the composition of materials within
these sources was highly variable and the sources consisted of large areas
exposed to the atmosphere.   Enclosing of these sources was either imprac-
tical from a size standpoint or hazardous from an explosion standpoint.
A material balance technique was used in an attempt to quantify the  loss of
volatile hydrocarbons.


Cooling Towers

          Thirty-one  (31) cooling towers were  sampled, eight (8) of which
had statistically significant emissions.  Streams from five  (5) towers were
analyzed by both TOG analysis and purge analysis; therefore, streams from a
total of 21 towers were analyzed by  TOG and 15 by purging.  Because  purge
values were judged to be the more precise, they were chosen to represent
the towers analyzed by both methods  in the calculations of mean emissions
for all towers.  A summary of the emissions and the  A  ppm values for these
towers is given in Table 9.

          The magnitude of the sampling/analytical, variation caused  some
problems in quantifying the  low levels of emissions from the towers.   The
standard deviation for replicate TOG analyses was 4.2 ppm.  If two tests
were run each day, the standard deviation for  the average would be 3.0 ppm.
The between day standard deviation (after averaging replicate samples
and analyses) using the TOG  analyses was 3.61 ppm.  Since this is
close to the analytical standard deviation when replicate samples  are  aver-
aged, it appears most of the variation in the  TOG data is due to  the analyti-
cal technique or the homogenity of replicate samples.
                                     182
-------
                              TABLE 9.  SUMMARY OF COOLING TOWER EMISSIONS
               Cooling  Towers  Sampled
               Cooling  Towers  Having Statistically  Significant  Emissions
               Range  of Cooling  Tower Circulation Rates	
      31
      8
      714 to 58.000 GPM
                                                                                                                       o
                                                                                                                en
                                                                                                                H-
                                                                                                                O
                                                   Results  (estimate with 95% confidence interval)
oo
       Mean Cooling  Tower  A HC  Concentration
            From Emitting  Towers         n  ,„,  .  ~  , n
                 „ .,  .  ,               0.101  ±  0.19  ppm
                 Both  Analyses                        r
     From All Towers Sampled
          TOC Analysis
          Purge Analysis
          Both Analyses3

Mean Cooling Tower Emissions
     From Emitting Towers
          Both Analysis

     From All Towers Sampled
          TOC Analysis
          Purge Analysis
          Both Analyses
                                         1.25  ±  1.24  ppm
                                         0.0130  ±  0.0299 ppm
                                         0.0173  ±  0.058  ppm
                                         0.00088  ±  0.0016  lb/1000  gal
                                         0.0124  ±  0.0123  lb/1000  gal
                                         0.000108  ±  0.00025  lb/1000 gal
                                         0.000151  ±  0.00051  lb/1000 gal
        Range of Measurable Emissions    0.36  to  8.46  Ib/hr
(negligible, 0.29 ppm)
(0.01, 2.5 ppm)
(negligible, 0.043 ppm)
(negligible, 0.075 ppm)
(negligible,  0.0025 lb/1000 gal)
(0.0001,  0.025 lb/1000 gal)
(negligible,  0.000261 lb/1000 gal)
(negligible,  0.00066 lb/1000 gal)
         Calculated for 15 towers analyzed by TOC only  plus  16  towers analyzed by purge.  The 5  towers
         analyzed by both methods were represented only by the  purge values, considered more accurate
         than TOC values.
-------
 F.  G.  Mesich
          The analytical standard deviation for the purge method is 80 percent
of the concentration (averaging about 0.1 ppm).  The between day standard devi-
ation calculated here was 0.12 ppm so again most of the variation in the purge
data is due to the analytical method.  But, since the levels reported by the
purge method were at least an order of magnitude smaller than the TOG values,
the absolute variation is much smaller for towers evaluated using the purge
techniques.

          Since sampling was only done on five to seven days for most towers,
and emissions from the towers were found to be relatively low, it was not sur-
prising to get some negative values as estimates of emissions for a particular
tower.  The negative estimates are as follows:

Analytical            Number of                  Towers with Negative Estimate
  Method                Towers                       Number        Percent

TOC                       21                            7            33.3

Purge                     15                            2            13.3

Combined                  31                            8            25.8

          The negative estimates are due primarily to the analytical variation.
In order not to bias the average emission calculation for cooling towers, these
negative values have been used rather than setting the estimate to zero.

          The mean emissions for the 16 towers analyzed by TOC only and the
15 analyzed by purge were 0.00015 lb/1000 gal with 95 percent confidence inter-
val of ± 0.00051  (negligible, 0.00066 lb/1000 gal).  Mean emissions for the
eight towers with statistically significant emissions were 0.00088 ± 0.0016 lb./
1000 gal (negligible, 0.0025 lb/1000 gal).  Mean emissions for the 21 towers
analyzed by TOC were 0.0124 ± 0.0123 lb/1000 gal (0.0001, 0.025 lb/1000 gal).
For the 15 towers analyzed by the purge method, mean emissions were 0.000108 ±
0.00025 lb/1000 gal (negligible, 0.00026 lb/1000 gal).

          Where values obtained by TOC analysis and values obtained by purging
were combined to obtain a mean value, the confidence limit was sometimes larger
than the obtained value.

          Because of the varying precision of the methods, the upper confidence
limit for each estimate may be a more useful value than the estimated average
for many purposes.  These values which give a "worst-case" estimate for the
magnitude of hydrocarbon emissions from cooling towers are as follows:

          Analytical                         "Worst-Case" Estimate of Average
         Method Used                           Emissions from Cooling Towers

             TOC                                  0.025 lb/1000 gal
             Purge                                0.0003 lb/1000 gal
             Combined                             0.0007 lb/1000 gal
                                     184
-------
F. G. Mesich
Even  these values  are  small relative to other sources of emissions from
refineries.


WASTEWATER SYSTEMS

         Wastewater  treatment is usually accomplished in three stages:  pri-
mary,  secondary, and  tertiary treatment.  Primary treatment facilities are
principally  involved  in physically upgrading the wastewater by removal of oil,
oily  sludge, and grit.  Thus, primary treatment facilities will be the princi-
pal sources  of  fugitive hydrocarbon emissions from the waste treatment plant.
Oil removal  equipment  includes API separators, corrugated plate interceptors,
flocculation units, and dissolved air flotation units.  The latter are also
used for suspended solids removal.

          Table 10 summaries the average emissions per gallon of material
throughout for  all sampled devises by refinery.  Unfortunately, the data from
the cooling  tower  and  wastewater treatment systems are not sufficiently repro-
ducible to develop usable emission factors.  Cooling towers appear to be minor
sources of emissions  while oil/water separators require more work to determine
the significance of the emissions.
                                     185
-------
                               TABLE 10.   DESCRIPTION OF SAMPLED DEVICES - WASTE OIL/WATER SYSTEMS
00
Average Hydrocarbon Emissions m
Refinery
1
2
3
4
5
6
7
B
Device
R Rectangular API Separator
Circular DAF
Rectangular API Separator
Corrugated Plato Interceptor
Corrugated Plate Interceptor
Rectangular API Separator
Forebay Covered
Surge Tank
Two Rectangular Separators
Rectangular DAF
Rectangular API Separator
Rectangular API Separator
Rectangular PAF
Circular Separator
Circular DAF
Covered/Uncovered
C
U
C
C
C
U
U
U
U
I!
U
U
U
U
Losses from
Oil Phase,
Ib/gal slop oil
1.6 + 2
0.073 ± 0.4
1.84 + 1.11
-1.5 + 0.08
-0.11 + 0.06
0.12 + 1.3
0.45
-1.1 + 0.74
0.14 ± 0.4
0.48 + 0.61
Losses from H1
Water Phasn, y
Ib/gal water
2.7ilO~* +
8.2xlO~* +
-3.01xlO~*
—
2.2x10"* +
-2.4xlO~5~H
1.5x10"* +
6.5x10"* -t
1.1x10"* +
3.4xlO~* +
1.4xlO"s +
1.8x10"*
1.5xlO~*
+ lxlO~s

2.7xlO~*
3xlO~6
h 2.7x10 s
2.4xlO~*
1.9xlO~*
1.3x10"*
1.8xlO~*
1.7x10 !
-------
James J.  Morgester
                                  REVIEW

                                    by

                            James J. Morgester
                      California Air Resources Board
                          Sacramento, California


                                    on


              RESULTS OF MEASUREMENT AND CHARACTERIZATION OF
                        ATMOSPHERIC EMISSIONS FROM
                           PETROLEUM REFINERIES
                                  RESUME
         James Morgester attended the University of Washington and the
University of California at Berkeley specializing in physical science
and environmental law.  Jim is the author of over 30 technical papers on
air and water pollution control and has 20 years experience in the fields,
He is presently chief of enforcement for the California Air Resources
Board.
                                    187
-------
James J. Morgester
                                  REVIEW

                                    by

                            James J. Morgester
                      California Air Resources Board
                          Sacramento, California

                                    on

              RESULTS OF MEASUREMENT AND CHARACTERIZATION OF
                        ATMOSPHERIC EMISSIONS FROM
                           PETROLEUM REFINERIES
         The thirteen-refinery study that has been conducted by Radian
Corporation for the Environmental Protection Agency (EPA) and summarized
in  this paper is probably the best work, on the national level, done to
date on fugitive emission quantification in general and valve, flange,
pump, and compressor leakage in particular.  However,  since the sample
population for which the results are published in this paper was taken
from refineries in various regions of the U.S. (four refineries on the
West Coast), it is questionable whether the study statistics are directly
applicable to refineries in California or any specific region.  In addi-
tion, the sample size for each source (valve, flange,  pump, etc.) was less
than that sampled in previous California studies (e.g. 13,685 valves were
inspected in the Air Resources Board's 1978 California study, whereas
Radian examined 2,244 valves).  This means that generalizing from Radian's
sample results is not as well supported as generalizing based on a larger
data base.

         The most startling findings of the Radian study have been the high
occurrence of leakage found in valves and flanges (27 percent for valves
and 3 percent for flanges, overall), and the high average mass emission
rates for valves (approximately 0.55 Ib/day/valve depending on the assumed
line-service profile at the refinery).  The Air Resources Board's 1978 study
of valve and flange leakage in California refineries indicated nominal leak
frequencies of about 9 percent for valves and about 0.4 percent for flanges.
Furthermore, in 3 previous valve and flange studies in California (includ-
ing the Air Resources Board study), the overall mass emission rate for
valves was calculated to be in the 0.11-0.15 Ib/day/valve range.  Compari-
son of this emission rate with that found by Radian for valves indicates
the possibility of significant differences in inspection methods and/or the
                                    188
-------
James J. Merges ter
existing  preventative maintenance programs of California refineries versus
other  U.S.  refineries.

         The  specific results of Radian's California refinery inspections
have not  been published to date so it is impossible to make any comparison
between those results and the data obtained in other regions of the U.S.
It has been our experience that significant differences can exist among
refineries  in overall valve and flange leakage and in the leakage at similar
process units,  indicating a cause for leaks other than the nature of valve
and flange  service.   It was evident to the Air Resources Board field
personnel during our 1978 inspections of 7 major refineries and 6 chemical
plants in California that variations of 6 percent to 18 percent valve leak-
age in refineries and 0.3 percent and 22 percent in chemical plants were
largely due to the priority and emphasis given to routine maintenance of
valves and  flanges by each facility.  In chemical plants it was clear that
the priority  given to routine maintenance was directly influenced by the
costs  of  products being lost to leakage.

         If the Radian study results which are summarized in this paper by
Mr. Mesich  are indicative of results to be expected on the regional level
(although I have reservations about this being true), then past estimates
of hydrocarbon emissions from valves and flanges in refineries may have
been understated by as much as factor of five.  Using your emission factors
for valves  and flanges, I calculate that emissions from valves in refineries
in the South  Coast Air Basin would be 64.1 tons/day and from flanges 3 tons/
day.  Accordingly, previous estimates of the emission reduction and cost
savings achieveable by the implementation of effective and enforceable
valve  and flange emission control rules may have also been vastly under-
estimated.  In the coming era of higher and higher costs for petroleum
products, such rules may very well come to be viewed as the most cost
effective ever implemented.

         If such rules have been logically implemented 20 years ago, after
the 1958  joint study, the cumulative product savings in California refin-
eries  alone would have amounted to about 880,000 tons, or the equivalent
of over 300,000,000 gallons of gasoline.  At today's wholesale prices for
gasoline  the  total cost savings would be $220,000,000.

         Obviously,  the Radian study has been long overdue, and the data
generated from it probably will provide air quality analysts and control
strategists working on the national level with a valuable tool in the years
to come.
                                    189
-------
F. G. Mesich
                           QUESTIONS AND ANSWERS
Q.  James Stone/Louisiana Air Control Commission - On slide 23 you show that
your flanges don't have any leaks greater than 1 pound per hour and at some
other point in the talks they told me that you had included heat exchangers
and things like that in the flanges and if I have seen a large leak that
was not going to be stopped from a flange it was a heat exchanger.  Does
that mean that you did not include these in your study?

A. - No I  All flanges measured are included in the data base.  I would like
to state for the record that visual inspection is not a good way to esti-
mate the magnitude of a leak.  For example, in several instances that I was
personally involved in, things that had liquid leaks were screening and
bagging very low in terms of air emissions, again showing the dependence on
the volatility of material that was leaking from the line.  And further,
one of the largest sources our study identified was not recognized by the
crew until they screened it and put a bag around it and the bag just about
blew out.  It was not visually or audibly apparent.

Q.  James Stone/Louisiana Air Control Commission - The one refinery I am
thinking about has a whole bank of exchangers.  I have given them viola-
tions on opacity from smoke because of the drips from the exchangers.  And
I think those would probably qualify as bigger than one.

A. - All we can speak from is the data base of the thirteen refineries that
we were in.  And every flange that we measured is in that data base
somewhere.

Q.  Paul Harrison/Engineering-Science - I think that both the CARE study and
the Radian study may be accurate, depending upon how you count.  For
example,  I've seen an oven with maybe 3,000 components on one oven.  A
hydrogen oven where you have a lot of process gas entering; each one has a
burner,  each one has a valve, each one has a union and has them on both
sides.   And,  if you start counting that way your percentage of leakers
goes down.  The value 0.5 pounds per day for unmaintained valves is not bad.
I have found that the percentage of leakers x^as something like 2 percent
for a new refinery and 6 percent for an old refinery, but that is including
everything, including flanges, valves, compressors seals all in the same
pot.   I did not differentiate with valves.  So, I could easily believe the
GARB study.  In any one unit I could believe the Radian study.  In certain
process units it is probably easily 25 percent with that stringent kind of
study.   In addition, my study was conducted at 5 cm as opposed to the
1 cm so I could easily double it as well up to 12 percent which is more
like the GARB study.  One more comment about heat exchangers.  I caution you
                                    190
-------
F. G. Mesich
that just because you have visible  smoke  coming  out  of  a heat  exchanger
does not mean that it is detectable as VOC.   That  could be  particulates
and "heavies" which are dripping  out.  Many  times  those "heavies" aren't
very volatile and you can barely  see them on a detector.  So,  they would
not get through the screening.  The thing about  visual  leaks is  that many
times there are "heavies" that  get  into the  ground,  and maybe  they are
volatilizing over a long period.  But many times screening  devices will
not pick them up.

A. - That is correct.  Two other  very brief  comments.   One  is  that I don't
know whether our percentage of  leakers is really different  from  California's
or not.  In that ours is based  on the 200 ppm level  at  the  surface.  We
presented a little data looking at  the soap  bubble technique.  The data are
really insufficient at this point to correlate between  the  two methods of
detecting leaks.  I tend to prefer  the direct measurement to a somewhat
subjective interpretation of  the  formation of bubbles,  but  I certainly
would not make the statement  that the latter method  would be ineffective.
Jim Morgester evidenced a concern that there would be large differences
between refineries.  Lloyd presented a slide this morning that pretty
graphically shows that among  the  thirteen refineries, we found twelve that
the differences between refineries  were not  a variable  in terms  of influ-
encing our data, and I believe  one  refinery  did  have a  small effect in the
variance analysis.  So, we basically did  not see large  differences, when
you aggregate the data base.

COMMENT/Rosebrook - I think there are two other  things  that should be said.
The first, deals with the differences in  the data.   I recall during some
testimony about six weeks ago in  San Francisco when  we  introduced the
screening data which was collected  at the six Bay Area  refineries.  This
included the screening of some  25,000 valves.  Our data then began to
differ when we used a different cutoff.   We  used the California  screening
approach of one centimeter with the proper calibration  gas  and so forth
and reported it as they wished  to have it reported.  I  do not remember
whose data was which but one  of us  found  10  percent  leaking and  the other
one found 8 percent leaking.  So, basically  we are talking  in terms of how
we define a leak.  Once we came close to  using the same definition, I
think we found approximately  the  same percentage of  leaking fittings.  Thus,
we are not talking about the  difference between  6  and 30.   Given the proper
basis these numbers do agree.   The  second point  is that anyone having
access to the raw data could  compare two  processing  units or two refineries
(if one takes only two of them) and find  that there  is  indeed a  difference.
But insofar as its effect on  our  overall  data base no refinery had a signifi-
cant contribution which would skew  the data  base.  We definitely found
refineries where our random samples tended to give emission rates which were
much higher than they were in other refineries.

Q.  Thomas C. Ponder, Jr./PEDCo Environmental, Inc.  - Based on what you were
showing us Frank, you are saying  that if  we  had  a  50,000 barrel  a day
refinery and 300,000 barrel a day refinery and they  had the same number of
process units we should have  the  same amount of  emissions from each one?
                                    191
-------
F. G. Mesich
Because the number of valves is pretty constant like a FCC has the same
number of valves, a little FCC and a big FCC has the same number of valves.

A. - Within the confidence limits yes.

COMMENT/Dan Martin/Union Carbide Corporation - The data presented here is
for existing units in refineries or in chemical plants.  It has only been
in the last 12 to 18 months that the concern for fugitive losses has come
out.  On units now being built, should there be much concern about this
because now we are addressing ourselves to valve and piping specifications
that we never did before.  We found a lot of valves in the past that came
in improperly packed and were put in the system.  Under the new standards
I wonder when building a new unit if fugitive losses would be a problem in
a permitting process rather than going back as some sort of a RACT or back
up and correct the existing units.  In addition, a plant that is now
operating today has to be more concerned about occupational health.  We
are already going in now and doing a lot of correction on fugitive losses
where you have employee exposure.  We have never been concerned about that
in the past and are tightening up things.  Is fugitive loss really going
to be that big of a problem in the future?  And I have one last comment and
answer.  If a person has over 30,000 ppm coming out of something particu-
larly a VOC you better be careful because you've got a flammable situation.
You have got a time bomb.  And even though the wind is blowing, some day
you are going to have a fire on your hands.

COMMENT/K. C. Hustvedt/USEPA-RTP - From the plants that I have been in, the
benzene unit is the one area where I think occupational health was already
a factor.  We haven't seen much difference in the leak incidence from our
testing in those plants or elsewhere.  It is a problem where you can't see
the leaks.  The plants are doing what they can for the ones they know about,
but it is the ones they don't know about that cause the leaks that have
created the emission factors we now have.  Another point that will be shown
in some of the correlations we see tomorrow is that it is not so much a
factor of the type of equipment put in and how it is installed, but more
how it is maintained once it is inline.  You can see that the percent of
sources, creating 90 percent of the emissions is a very small percent of
the sources.   It is the few that aren't properly maintained where something
has gone wrong and they haven't been able to detect it.  So, I think with
even better design in the future and better installation you are still
going to have problems.   Vibration; you will still have incorrect specifica-
tion; and, still have to do some level of monitoring to find these as they
happen and correct the problem.

         One other point you were talking about.  The 30,000 ppm having an
explosive problem.  I think when you get right down to a source, 30,000 ppm
sounds like a big number but it really isn't much.  Sometimes you can have
30,000 ppm at the source and step back a foot and not see anything, if it
is a pinhole type leak and you have a lot of wind.
                                    192
-------
F. G. Mesich
COMMENT/Rosebrook - Jim Morgester would you  care to  address  that  question
also, from a state's standpoint and whether  you  see  any  difference  in  the
future on the types of enforcement problems  and  the  types  of regulations
and so forth.

COMMENT/James J. Morgester/California Air  Resources  Board  -  Well, you  have
two different issues you have to look at.  California's  standpoint  is,
number one, we have a number of refineries there on-line now and  it is
clear to me that this problem was demonstrated in 1958.  The problem is
still with us twenty-two years later.  Maybe it  is worse than what  we
estimated in 1958, so there has got to be  a  clear motivating mechanism to
make the people that are the decision makers aware of  the  issue and the
problem and take care of it.  That is for  existing facilities.  I think
that it is a little optimistic to think that any significantly different
type of valve arrangement is going to be used in a new unit.  I think  that
the valves are shelf items and that most design  engineers  simply pull  this
valve off the shelf and unfortunately I think you are  going  to see  the same
type of valve configuration twenty years in  the  future that  we have seen
twenty years in the past.  So, there is going to be  a  need for this type
of regulations, strictly looking at it from  a parochial  narrow vision
enforcement standpoint.  My only objective is to make  the  stakes high
enough that the management of the oil refineries cause these emissions
sources to be looked at.

Q.  Thomas C. Ponder, Jr./PEDCo Environmental, Inc.  -  About  two weeks  ago
we had inspected a vinyl chloride plant with a VOC instrument, even though
they had an ambient network in the plant to  pick up  fugitive emissions
leaks.  We found leaks over a 1,000 ppm, which was as  high as this  machine
had been calibrated to go that day.  They  were picking up  nothing on their
ambient network.  We found nothing coming  out of the pump  seals, which
have double mechanicals with oil flush for each  shaft,  and  the compressor
seals and the relief valves, which have rupture  disks.   These people were
very shocked to find leaks out of flanges  and sampling valves, well over
1,000 ppm.  Obviously since their networks weren't picking it up they
weren't checking.

COMMENT/ Jim A. Mullins/ Shell Oil Company - I would like  to respond to the
comment on the vinyl chloride plant.  In particular  I  think  it is totally a
function of not only the placement of area monitors  but  the  level of
detection those area monitors are set to detect  and  the  option levels  where
repairs are done.  In our particular plants  we have  such an  action  level
that the area monitors will detect.  For example, in a period of six months
they detected 150 leaks, which were repaired.  At the  end  of that six
month period every flange., valve and seal  in the plant was checked  item by
item and we found two leaks.  And that was over  5,000  valves, pumps,
flanges.  So, I think it is totally a function of how  the  system  is
designed and should not be interpreted as  a  general  indictment of area
monitoring.
                                   193
-------
F. G. Mesich
Q.  Michael Scherm/Union Carbide Corporation - Did you make  any  distinction
in your data base of the difference between a daily operated type valve and
a valve that is simply in-line and rarely turned or rarely used?

A.  (By Rosebrook) - There is a distinction which rests primarily on the
emission rates which we found for control valves, as opposed to  any type
of block valves.  We found that if there is a difference  in  the  emission
rates for control valves, it is not statistically significant.  We made
no attempt to determine for each of the valves that we monitored, any
frequency of use, other than breaking them up by putting  control valves
in a separate category.

Q.  Greg David/Dow Chemical Oyster Creek Division - I would  just like to
point out that I think that on your examples or predictions  for hypotheti-
cal refineries that you should have put somewhere on that piece of paper
that that is worst case.  You have the number of valves to be 90 and you
estimate emissions in pounds per hour to equal 5.  However,  in your
screening phase you found somewhere around 27 percent of  those valves to
be leakers, and I think that you should put a qualifier on that page.

A. - Actually I wish that were true, but the emission factors are based on
the total valve population and not only on leaking valves.   In other words,
the zeroes are counted in and used as a devisor for determining emission
factors.  Otherwise the emission factor wouldn't be very  useful.
                                   194
-------
W. R. Phillips
                REFINERY AIR EMISSIONS CONTROL TECHNOLOGY

                             W.  Robert Phillips
                             Radian Corporation
                              Austin, Texas
                                  ABSTRACT

           Selected  refinery process and fugitive emission sources and
 controls are  discussed.   Emission data from a recent study by Radian
 Corporation of  selected  emission sources in thirteen refineries are dis-
 cussed.  Pollution  control technologies are described and evaluated.
 Specific topics discussed include sulfur recovery,  catalyst regeneration,
 process boilers and heaters; valve and pump seals and packings; wastewater
 and cooling water systems.
                                   RESUME

           Mr.  Phillips (B.S.  - Chemical Engineering,  Texas Tech,  1955)  is
 employed by Radian Corporation.   He has twenty years  process industry  exper-
 ience which began with his return from military service in 1959.   Experience
 includes (chronologically) refinery production engineering (3+ years);
 chemical company research, development and project engineering of principally
 oxidation reaction systems (12+ years); environmental and energy  system
 research and development,  Radian Corporation (3 years).  Professional
 Affiliation:   American Institute of Chemical Engineers.
                                     195
-------
 W.  R.  Phillips
                 REFINERY AIR EMISSIONS CONTROL TECHNOLOGY


INTRODUCTION

Background

          Radian's refinery emissions study was conducted in thirteen
refineries.  Although fugitive emission testing within process battery limits
was emphasized, twenty process stacks (process heaters, CO boilers, SRU tail
gas, etc.) and off-site cooling towers and primary wastewater facilities
were surveyed.

Objectives

          Objectives of this paper are to:

          •    Review state-of-the-art of process and fugitive emission
               controls.

          •    Discuss available control technology.
          Discussion of process emissions for purposes of this paper has
been narrowed relative to that which appears in Radian's final Refinery
Emissions Report.  Focus will be on sulfur recovery, catalytic cracking
regenerator control, and process heater control.

          Certain low-impact fugitive air emission sources will be discussed
only briefly.  Emissions from loading, unloading and storage tanks were not
measured as part of the original field study, so are omitted from discussion
here.

PROCESS EMISSIONS

          Table 1 lists process emissions by source and type.  Emission
sources in this table are coded to indicate which were field-measured and
reported in the recent refinery study.

          The major sources of atmospheric process emissions are sulfur
recovery, regeneration of fluid catalytic cracker catalyst and process heaters
and boilers.   This section focuses on these sources.

          The major types of atmospheric process emissions from refineries
are hydrocarbons, sulfur oxides, particulates and carbon oxides.
                                    196
-------
.  R.  Phillips
            TABLE  1.   PROCESS EMISSIONS BY  SOURCE AND TYPE
 Source
Sulfur
Recovery
        a
Catalyst
Regeneration
(CO Boiler Vent)
           3,
Boilers and
Process Heaters

      b
Vacuum
Distillation
                       HC
X
X
X
X
X
                                             Emissions
                SO,
                                       X
X
X
Coking
b
Air Blowing
Chemical ,
Sweetening
Acid Treating
Slowdown
Compressor
Engines
X
X
X
X
X
X
X


X

X X
                CO  Aldehydes  NH3
                         X
X
X
                                                X
X
X
                               NOX
                X
                                                                        X
 Detailed results are  in Refinery Assessment Report.
  Emission not measured in this study.
                                     197
-------
W. R. Phillips
          Process heaters and boilers are used in a number of different
refinery processes.  Therefore, they will be discussed, not by process, but
collectively as a separate emission source.

Sulfur Recovery

          The amount of sulfur in various product streams depends directly
on the sulfur content of the crude oil.  As the oil is fractionated initially
sulfur tends to become more concentrated in the heavier cuts because of the
low volatility of its various compounds.  The sulfur content of crude can
vary from less than 0.1 weight percent to more than 5 weight percent.  Any
crude oil with more than 0.5 weight percent sulfur is generally considered to
be a sour crude and its products are subjected to sulfur removal processing.1
If not removed, the sulfur can cause corrosion, pollution and catalysis
problems during refining or when the products are used as fuel or as petro-
chemical feedstocks.

          Sulfur removal from whole crude is not generally economical.2
Intermediate stock streams routinely subjected to sulfur removal include the
outlet streams from crude distillation and cracking units.3  Sulfur components
in these streams are converted to hydrogen sulfide by hydro-processing with
hydrogen over a nickel-molybdenum catalyst at an elevated temperature.
Resulting H2S boils between ethane and propane, so must be selectively
removed from the sour gas stream and concentrated by one of several means,
the most common of which is absorption by monoethanolamine (MEA) or diethanol-
amine (DBA) followed by steam stripping.

          With increasing use of hydro-processing of ever increasing sulfur-
containing crude stocks, it has become environmentally and economically sound
to introduce a process for removal of EzS generated by hydro-processing.  The
Glaus process presently dominates.  Tail gas from a Glaus unit can be a major
source of SOa emissions in a refinery.  In the Glaus process, some H2S feed
is oxidized to form SOa and water.  Additional HzS reacts with 502 to form
elemental sulfur and water.

          Glaus unit tail gas contains H2S, SC>2, CS2, COS and Sx.  The
emission rates of these sulfur compounds depends on the concentration of the
HzS stream to the Glaus unit and the efficiency of the unit.  Tail gas from
a typical three-stage Glaus unit, 95 to 96 percent efficient, can be expected
to contain about 7000-12,000 parts per million by volume sulfur compounds. '*'5
The tail gas also contains carbon monoxide formed from small amounts of
hydrocarbons and carbon dioxide in the feed stream.  Typical compositions of
Glaus unit feed and product gases are found in Table 2.

Catalyst Regeneration

          Catalysts are used in several petroleum refining operations,
including fluid catalytic cracking, moving bed catalytic cracking  (known as
Thermofor catalytic cracking or TCC), catalytic hydrocracking, reforming, and
various oil desulfurizations.  These catalysts become coated with carbon and
                                     198
-------
W. R. Phillips
         TABLE 2.   TYPICAL COMPOSITIONS OF FEED STREAM AND TAIL CAS
                   FOR A 94 PERCENT EFFICIENT GLAUS UNITn
                                    Sour  Gas  Feed,            Glaus Tail Gas,
 Component                          Volume  Percent            Volume Percent
H2S
S02
Se Vapor
SB Aerosol
COS
CS2
CO
CO 2
02
N2
H2
H20
HC

Temperature, °F
Pressure, psig
89-9
0.0
0.0
0.0
0.0
0.0
0.0
4.6
0.0
0.0
0.0
5.5
0.0
100.0
104
6.6
                                                                   0.85
                                                                   0.42a
                                                                 0.10 as Si
                                                                 0.30 as Si
                                                                   0.05
                                                                   0.05
                                                                   0.22
                                                                   2.37
                                                                   0.00
                                                                  61.04
                                                                   1.60
                                                                  33.00
                                                                   0.00

                                                                 100.00
                                                                 284

                                                                   1.5

 Total Gas Volumeb                       -                      3.0 x feed
                                                               gas volume
 SNSPS requires an emission  of  less  than  250  ppmv  (0.025  percent) S02, zero
  02, dry basis if Claus unit  tail gas  is oxidized  last as a  control step,
  or, 300 ppmv S02 equivalent  reduced compounds  (H2S,  COS, CS2) and 10 ppm
  S02 if the tail gas  is reduced  as  the last  process  step.

  Gas volumes compared at  standard conditions.
                                      199
-------
W. R. Phillips
metals and must be regenerated to restore their activity.  During regeneration,
the carbon is oxidized to carbon monoxide and carbon dioxide and the hydro-
carbons are burned incompletely.

          In most applications, a catalyst must be regenerated only a few
times a year.  Emissions during these episodes may include catalyst fumes,
oil mist, hydrocarbons, ammonia, SOx, chlorides, cyanides, NOX, CO, and
aerosols.7  Though there may be significant emissions during the regeneration
of one of these catalysts, the total emissions over the course of the year
are not significant.

          Catalytic cracking catalyst regeneration is a continuous process.
Uncontrolled cracking catalyst regeneration is one of the major sources of
air pollution in a petroleum refinery.  Flue gases from catalytic cracker
regenerators contain particulates, SOX, carbon monoxide, hydrocarbons, NOX,
aldehydes and ammonia.

          Emission factors for uncontrolled regeneration of FCC and TCC
catalysts are reported in AP-42 and are listed here in Table 3.  These
factors are from a 1956 stack sampling survey of FCC and TCC units in Los
Angeles County.8  The survey involved six FCC units and nine TCC units.

          TABLE 3.  EMISSION FACTORS FOR UNCONTROLLED REGENERATION
                     OF THE CATALYTIC CRACKING CATALYST9

               _ Emission Factor, lb/1000 bbl Fresh Feed _
               Particu-   SOX as            Total      NOX as
 Process         late     SOa       CO   Hydrocarbon   NOa    Aldehydes  NH3
  Fluid Catalytic  242     493     13,700    220         71       19      54
  Cracking  (FCC)

  Moving Bed        17      60      3,800     87          5       12       6
  Catalytic
  Cracking  (TCC)
          These factors indicate that the uncontrolled emissions from FCC
units are several times greater than from TCC units.  The  term "uncontrolled
emission" here implies conventional regeneration without any external control.
Radian believes that these 1956 emission factors should be reviewed because
of advances in technology, especially the FCC particulates emission factor
(242) because of the following:

          1.   FCC regenerator cyclone technology has advanced since the 1956
               survey.  Catalyst losses from properly designed two-stage
               regenerator cyclone systems in the range of 80-100  lb/103 bbl
               of fresh feed are typical.
                                     200
-------
W.  R.  Phillips
          2.   Five  out  of  six of the FCC units surveyed in 1956 had emission
              factors of 50 lb/103  bbl fresh feed or less.  The sixth may
              have  been troubled by condensation at low stack temperature
              of  high concentrations of S03.  The individual emission factors
              were  181, 50,  43,  35, 27, and 24 (average:  60).

          3.   For several  of the TCC units in the survey,  emissions for the
              entire unit  were extrapolated from measurements made on one
              stack.  This method can produce errors, because TCC units
              employ several dissimilar stacks.

Boilers  and Process  Heaters

          Most refineries use steam boilers to provide steam for direct use
in various processes, for heating and for driving steam turbines.   Large
amounts  of steam are needed for light ends strippers, vacuum steam ejectors,
process  heat  exchangers  and reactors.  About 40 pounds of steam are required
by a typical  refinery per barrel of refining feed.  This steam demand requires
a boiler size of 53,000  Btu per barrel of refining feed.10   Some steam is also
generated in  waste heat  boilers,  the largest of which is, in some refineries,
a carbon monoxide  boiler used to control emissions from the regeneration of
the catalytic cracking catalyst.   Another carbon monoxide control technique
is high  temperature  catalyst regeneration at approximately  1300°F minimum.
Most process  steam generated is low pressure steam.

          Process  heaters are the largest combustion source of hydrocarbons
in a refinery.   Total process heater demand in a modern refinery is approxi-
mately 270,000 Btu per barrel of refining feed.  Older, less efficient
refineries may require 600,000 Btu per barrel of refining feed.11

          Refining boilers  and heaters are fired with the most available fuel,
usually  purchased  natural gas, refinery fuel gas (mostly methane), or residual
fuel oil.  Ordinarily, the  refinery gas supplies approximately one-half the
fuel needs; natural  gas  is  used in the summer months and residual oil in the
cooler months when natural  gas supplies go to preferred residential customers.
These estimates  vary with the individual refinery.  Emission factors for
burning  of natural gas and  residual fuel are found in Table 4. -

          In  addition to combustion emissions, there are also emissions
associated with  the  decoking of heaters.  At intervals of about  six months
to three years,  each heater must be flushed with a steam-air mixture to
remove interior  coke deposits.  Emissions are similar to those from decoking
operations on delayed coking units,  but less.

Existing Control Technology

          Most of  the process emissions described previously can be con-
trolled.   This section describes control methods that are now in use or
might be adapted to  refinery use.
                                     201
-------
W. R. Phillips
         TABLE 4.  EMISSIONS FROM REFINERY BOILERS AND HEATERS12

                                                   Fuel
                                   Natural Gas,                Fuel Oil,
  Pollutant                       lb/106 std ft3               lb/103 gal
Hydrocarbons (as


Particulates                           5-15


SOX as S02  '                            0.6b                     157


CO                                      17                          5


NOX as N02                           120-230d                      6Q6
 A function of fuel oil grade and sulfur content
     For Grade 6:  lb/103 gal = 10 S + 3
     For Grade 5:  10 lb/103 gal
     For Grade 4:  7 lb/103 gal

 Based on average sulfur content of natural gas of 2000 gr/106 Std Ft .
Q
 S equals percent by weight of sulfur in fuel.
 Uses first number for tangentially fired units, second for horizontally
 fired units.
p
 Strongly dependent on the fuel nitrogen content.
                                   202
-------
W. R. Phillips
          "Existing" controls included are  those considered to be in
relatively common usage in refining.  "Available" controls, which will be
discussed following this section, are those which have had only limited
application or which have not yet been applied.  Those controls which have
been used in other industries and which might be applicable to the refining
industry, are included in the section following available controls.

Sulfur Recovery

          The Glaus unit is the accepted method for sulfur removal in a
modern refinery.  However, because  it is not totally efficient in producing
elemental sulfur, it is a major source of emissions.  Much progress has been
made in recent years in the control of emissions from Glaus units.  This
discussion will consider first the  Glaus unit itself, then methods for
cleaning up the Glaus unit tail gas.  Incineration is an integral part of
several of these methods, so it is  discussed immediately after the Glaus
unit.

          More than 70 methods have been proposed for treatment of the Glaus
unit tail gas.    These methods may be continuations of the Glaus reaction or
add-on processes with chemistry quite different from that of the Glaus reac-
tion.  Incineration is sometimes used alone to clean Glaus unit tail gas,
sometimes to prepare the tail gas for further treatment, and sometimes after
that treatment.

The Glaus Process—

          The Glaus process is recognized as a very effective control device.
Since implementation of the Environmental Pollution Act in 1970, Glaus has
been used to remove sulfur from refinery process streams at an average
efficiency exceeding 95 percent.

          The overall Glaus reaction is as  follows:

          H2S 4- % 02 = ( — ) Sn + H20                                     (1)

where n represents the various molecular forms of sulfur vapor.  The two
most popular designs of Glaus units are illustrated in Figure 1.  In the
"once-through" design, the incoming H2S-rich stream is burned in a limited
amount of air to convert one-third  of the H2S to S02 according to the
following reaction:

          2H2S + 202 = S02 + S + 2H20                                    C2)

The hot gases from this reaction are then passed over a bauxite, alumina, or
cobalt-molybdenum catalyst, to react the sulfur dioxide with unburned H2S
according to the following reaction:
          2H2S + S02 = 3S +  2H20
                                                                          (3)
                                     203
-------
ACID [LAS
    AIR
                                                      FEED WATER
                                               CONDLNSERS
                                                   L.P. STEAK
                                                            gar
                                                                                TAIL  GAS
                                    1	»- LIQUID SULFUR

                                   STRAIGHT THROUGH CLAUS  PROCESS
                                                               SUITOR PIT
     ACID GAS t
                                                                              •>• TAIL GAS
                                                                 SULFUR PIT
                                             LIQUID SULFUR
                                       SPLIT now CLAUS PROCCSS

                            Figure  1.   The  Glaus Process.

                                            204
-------
W. R.  Phillips
If the  "split-stream",  or "by-pass", design is used, one-third of the incoming
stream  is  separated and burned more completely according to the following
reaction:

          H2S  + 3/202 =  S02 + H20

The remaining  H2S is reacted over a catalyst with the hot gas from the
furnace to form elemental sulfur according to reaction (3) above.

          The  "direct oxidation" design is for streams with lower concentra-
tions of H2S.   In this  design the incoming stream is preheated, mixed with
air, and then  passed over the bauxite or alumina catalyst.

          The  Glaus designs described above with one pass through the
catalytic  reactor convert 80 to 86 percent of the H2S to elemental sulfur.^ '1S
This efficiency can be  greatly enhanced by repeating the catalytic stage one
or more times.  Thus, two-stage Glaus units can achieve 92 to 95 percent
efficiency;  three stages, 95 to 96 percent; and four stages, 96 to 97 per-
cent.   Conversion is  ultimately limited by the reverse reaction.  Recovery
rates for  various feed  compositions are found in Table 5.

          These efficiencies, once considered sufficient, do not meet new
regulations.   Further treatment of. the Glaus unit tail gas is discussed in
succeeding sections.

Incineration of Glaus Unit Tail Gas—

          The  tail gas  from the Glaus unit is often  incinerated before it
either  passes  to the atmosphere or is subjected to further treatment.  This
incineration takes place at temperatures of about 1200°F or above in refrac-
tory-lined vessels with one or more burners.

          Auxiliary fuel such as natural gas or fuel oil provides the heat
necessary  for  incineration since the heating value of the tail gas is low.
Excess  air levels of 20 to 30 percent are used.

          The  objective of tail gas incineration is to convert all sulfur
compounds  in the tail gas to S02, but this conversion is incomplete.  Typical
compositions of a sour  gas feed stream and the corresponding Glaus unit tail
gas before and after incineration are given in Table 6.

Tail Gas Clean-Up—•

          Each of the six tail gas clean-up methods listed in Table 7 has
three or more  commercial installations to its credit.  The first three pro-
cesses—Amoco's CBA process, the Sulfreen process, and the IFF process—are
continuations  of the Glaus reaction under more favorable conditions.  The
second  three processes—the Beavon process, the SCOT process, and the Wellman-^
Lord process—are add-on units with higher efficiencies than the first three.
                                     205
-------
                   TABLE 5.   TYPICAL CLAUS  PLANT SULFUR RECOVERY  FOR  VARIOUS  FEED  COMPOSITIONS18
o
O\
Hydrogen Sulfide in Sulfur
Plant Feed (Dry Basis) , %
20
30
40
50
60
70
80
90

Two Reactors
92.7
93.1
93.5
93.9
94.4
94.7
95.0
95.3
Calculated Percentage Recovery3-
Three Reactors
93.8
94.4
94.8
95.3
95.7
96.1
96.4
96.6

Four Reactors
95.0
95.7
96.1
96.5
96.7
96.8
97.0
97.1
        o
         Assumes  1 mole  percent hydrocarbon contamination,  conventional temperatures and reheat techniques,

         average  organic by-products and entrainment allowance.
                                                                                                                     !*
                                                                                                                     >-ti

                                                                                                                     H-
                                                                                                                     M
                                                                                                                     M
                                                                                                                     H-
                                                                                                                     -d
                                                                                                                     en
-------
W. R. Phillips
     TABLE 6.  TYPICAL COMPOSITIONS OF FEED STREAM AND TAIL GAS STREAMS
           FROM A 94 PERCENT EFFICIENT GLAUS UNIT AND INCINERATOR19
Sour Gas Feed,
Component Volume Percent
H2S
SO 2
Se Vapor
SB Aerosol
COS
CS2
CO
C02
02
N2
H2
H20
HC
Temperature
°C
°F
Pressure
Kilopascals
Psig
Total Gas Volume3
89.9
0.0
0.0
0.0
0.0
0.0
0.0
4.6
0.0
0.0
0.0
5.5
0.0
100.0

40
104

150
6.6
-
Thermally Incinerated
Glaus Tail Gas, Tail Gas,
Volume Percent Volume Percent
0.85
0.42
0.10 as Si
0.30 as Si
0.05
0.05
0.22
2.37
0.00
61.04
1.60
33.00
0.00
100.00

140
284

110
1.5
3.0 x Feed
Gas Volume
0.001
0.89
0.00
0.00
0.02
0.01
0.10
1.45
7.39
71.07
0.50
18.57
0.00
100.00

400
752

100
0
5.8 x Feed
Gas Volume
  Gas volumes  compared  at  standard  conditions.
                                     207
-------
                         TABLE 7.   ESTABLISHED METHODS FOR REMOVAL OF SULFUR FROM GLAUS TAIL GAS
?c
to
o
00
Name
CBA





Sulf reen



IFP-1500


BSRP




SCOT


Wellman-
Lord


Developer
Amoco





SNPA/Lurgi



Institut
Francais
du Petrole
Ralph M.
Parsons &
Union Oil
Co. of
California
Shell


Wellman
Power Gas


Efficiency Cost (Percent Commercial
Description (Glaus + Add-on) Product of Glaus _ ,
Glaus reaction continued at 98-99.5 percent S0 50-150 3
low temperature; removal of 1500 ppmv S
condensed sulfur drives
reaction. Bed regenerated
with hot gas from Glaus
unit.
Glaus reaction continued at 99 percent S0 50-150 19
low temperature as in CBA. 1500-2000 ppm S
Bed regenerated with hot
nitrogen.
Glaus reaction occurs in a 1000-2000 ppm S So Variable 25
solvent.

a
All sulfur compounds reduced 250 ppm S or less So 100 36
to HaS which is processed
in a Stretford unit.


All sulfur compounds reduced 200-500 ppmv HaS Feed to 75-100 35
to HaS which is recycled to Glaus
Glaus.
S02 in incinerator gas 200 ppmv S02 NaSOi*/ 130-150 for 7
contacted with NaS03 to form NaS03 100 It/d
NaHSOa. NaSOs regenerated in Crystals Glaus unit
evaporator /crystallizer.
                                                                                                                          "S
        1Figure includes plants in operation, under construction or being designed.
-------
W. R.  Phillips
         These  six processes are discussed in more detail in Radian's
refinery  emissions  report.

FCCU Catalyst Regeneration

         Regeneration of the catalyst in fluid catalytic cracking units
(FCCU's)  produces three principal types of atmospheric emissions:   SOX,
particulates, and COX. • Lesser emissions include hydrocarbons, NOX, aldehydes,
and ammonia.  SOX is typically controlled by feedstock desulfurization;
particulates by  cyclones and electrostatic precipitators; and CO by a CO
boiler.   No single  process can control all three.

SOX Emissions—

         Hydrodesulfurization (HDS)  of feedstock to FCCU's has been
practiced for years, since it increases the yield of salable products.

         In HDS of FCCU feedstock, the ratio of weight percent sulfur in
the coke  over the weight percent sulfur in the desulfurized feedstock
increases with  the  degree of desulfurization.  The result is that  very high
levels of hydrodesulfurization are needed to achieve 90 percent or higher
reduction in SOX emissions.  For a feedstock with 2.3 weight percent sulfur,
for example, 92  to  95 percent desulfurization of the feed is necessary for
a 250 ppm SOX concentration in the flue gas.21

Particulates  (Catalyst Fines)—

         Before exiting the regenerator, gases pass through a series of
cyclones  that remove the small catalyst particles (fines) present  in the exit
gas.  Some refineries have additional cyclones downstream of the regenerator.

         Particles smaller than 5 microns are ordinarily not collected by
cyclones. The majority of refineries use electrostatic precipitators to
remove these catalyst fines from the flue gas.

         The collecting efficiency of ESP's for catalyst fines is commonly
99.5 percent of  the particles that escape the cyclones.22  In most cases,
final disposal  of the waste particles is by burial in a sanitary landfill.

CO Emissions—

         All methods of controlling the CO content of flue gas from FCC
regeneration involve combustion of CO to C02.  A typical unit with "conven-
tional" regeneration burns off the coke from spent catalyst to, roughly, a
50-50 mixture of CO and C02-

         The majority of refineries—66 percent in 1976— used a CO boiler
to recover part  of  the energy from hot FCCU flue gases and to reduce CO
emissions.23  The flue gas goes to the furnace of a CO boiler and external
heat is applied  to  raise the temperature high enough (vL300°F) to achieve
                                     209
-------
W. R. Phillips
near complete combustion (99.5 percent or more).  The heat  of  combustion is
recovered as steam, often used to drive the regenerator air blower as well
as for other refinery operations.

          In all but small refineries, the cost of CO boilers  can be recovered
in a few years.  Small refineries may find it more economical  to control CO
emissions with flares, even though no energy recovery is possible.

Other Catalyst Regeneration

          Because emissions from TCC catalyst regeneration  are significantly
less than those from FCCU catalyst regeneration, use of a CO boiler may not
be justified.  Flue gases from TCC catalyst regenerators are usually released
directly to' the atmosphere.

          Flue gases from other catalyst regenerations may  also be incinerated
in a process heater or flared, but use of these control methods is not wide-
spread because these emission sources are typically insignificant.

Control Technology Available in Refineries

          Controls with limited application and those which have not yet been
applied are included in this section.  Information available on new tech-
nologies is often limited.

          Controls chosen for inclusion in this section are those which have
been proposed for consideration by the industry.  Inclusion here merely indi-
cates that it is worthy of consideration, and not necessarily  a good choice.

Sulfur Recovery

          A number of alternatives to the Glaus method of sulfur removal
have been proposed in recent years.  Some are applicable only  to Claus tail
gas treatment, while others may be applied to other problem sulfur-bearing
streams as well.  Also being tested are one alternative to  the Claus unit
and an integrated Claus tail gas process.  These alternatives  would produce
no objectionable tail gas stream.  Tail gas treatment methods  are found in
Table 8.

UOP Sulfox Process2k

          The UOP Sulfox process is an alternative to the Claus process.
Initially, aqueous ammonia, instead of an amine solution, is used to scrub
H2S from refinery feed gas.  The ammonia is then scrubbed from the gas with
purified water.

          The rich solution is mixed with air and sour water and passed over
a catalyst.  Elemental sulfur is formed according to the following reaction:

          NH4HS + h 02 = S0 + H20 + NH3                                   (5)
                                    210
-------
TABLE 8.  AVAILABLE METHODS FOR REMOVAL  OF  SULFUR FROM GLAUS TAIL GAS
Name
IFP-15025
Cleanair26'27
9 fl
Trencor-M

Aqua-Glaus29'30

Sulf oxide31
Topsoe32
Developer
Institut
Franca is
du Petrole
Pritchard
Trentham

Stauffer

Alberta
Sulfur
Research,
Ltd.
SNPA/
Topsoe
Efficiency or
Description Outlet Concentration
Gas from IFP-1500 scrubbed <200 ppm S02
with ammonia; S02-laden
ammonia mixed with H2S in a
glycol to form elemental
sulfur and water.
Glaus reactors operated at 50 ppmv S
high temperature to reduce
COS and CS2 levels; S02 and
elemental sulfur removed by
aqueous scrubbing; H2S
removed by Stretford Process.
All sulfur compounds reduced 100-200 ppmv S02
to H2S. H2S absorbed by
amine solution and returned
to Glaus.
S02 from incinerator mixed <100 ppmv S02
with H2S-rich Glaus feed;
Glaus reaction occurs in
aqueous phase.
Glaus reaction occurs in an <1000 ppmv S,
organic sulfoxide medium. Typically
<500 ppmv S
S02 from incinerator 90 Percent
oxidized to SO 3 which is
converted to H2SOij.
Cost (Percent
Product of Glaus)
So Variable
S0 100
Feed to 150
Glaus

So 125-135
Na2S04

S0 Not
Available
H2SOit Not
Available
                                                                                                 D"
                                                                                                 H-
                                                                  Continued
-------
TABLE 8.  Continued
Name
SFGD 3 3 ' 3 4
Westvaco35'36
Ammonium
Bisulfate/
Ammonium
Thiosulfate37
BSR/
Selectox I38
Limestone
Slurry39
Catalytic
Incineration
Efficiency or
Developer Description Outlet Concentration
Shell SOa in gas from incinerator 90 Percent
absorbed by CuO bed which is
regenerated with hydrogen.
Westvaco S02 in gas from incinerator <200 ppmv S02
removed and catalyzed to
H~2SOit in activated carbon
bed which is regenerated
with HaS.
Pritchard SO 2 in gas from incinerator <900 ppmv SOa
absorbed in aqueous ammonia
and converted to ammonium
thiosulfate.
Union Oil All sulfur compounds reduced >98 Percent
to HaS which is then
oxidized to S.
Mineral & S02 in incinerator gas >99.9 Percent
Chemical absorbed by limestone
Resource slurry.
Company
Institut Catalyst promotes oxidation <200 ppmv S
Francais of sulfur compounds to SOa
Cost (Percent
Product of Glaus)
Feed to
Glaus
So or
Feed to
Glaus
Ammonium
Thiosulfate
So
CaS03/
limestone
solids
SO 2
250
Not
Available
75
<200
Not
Available
Not
Available
                                                                        •a
                                                                         en
-------
W. R.  Phillips
With ammonia and sulfide present, the elemental sulfur remains in solution
as polysulfide.   The liquid product from this reaction is then heated above
the melting point of sulfur, mixed with air, and passed over a second catalyst
where any remaining sulfide is oxidized to elemental sulfur.  With no sulfide
remaining to solubilize the sulfur as polysulfide, the sulfur exists as a
separate molten product.

          Tail gas is scrubbed with water to remove ammonia.  Hydrogen
sulfide content in the treated gas is 10 to 100 ppm.  Although it is possible
to design a Sulfox unit which will achieve 1 ppm HaS in the tail gas, the
new source performance standard requires only 250 ppmv S02 or less from a
final oxidizing step.

          Capital costs for a Sulfox system and a Glaus unit are approximately
equivalent, not including the cost of tail gas cleaning.  Sulfox utility costs
are approximately 60 percent of those of a Glaus unit.

Union Carbide UCAP Process41

          This newly publicized integrated Glaus/tail gas treatment process
requires only one Glaus reactor stage to achieve the NSPS requirement of less
than 250 ppmv of SOa.  The process converts HzS to SC-2, absorbs it in tri-
ethanolamine and recycles the S02 to the Glaus unit.  Economics are not yet
available.  The process appears to be a strong candidate for new integrated
plant installations.

Catalyst Regeneration

          SOX emissions may be controlled by flue gas scrubbing systems.
Exxon presently operates four such scrubber systems installed in its coastal
refineries in Texas, Louisiana and New Jersey.

          Exxon's operations have shown that 95 percent of the SOx and 90
percent of the particulates can be removed by a scrubber.^  They believe
that the cost of controlling both particulates and SOX by scrubbing is less
costly both in initial investment and maintenance than a combination of
desulfurization and ESP's.  The space requirements are also less.

          The scrubbers may be the once-through or regenerable type.  A non-
regenerable process has been used for FCCU flue gas, and the resulting spent
scrubbing liquid is handled by conventional wastewater treatment.  It
contains a high concentration of dissolved solids and salts and has a high
chemical oxygen demand (COD).  To date, scrubbers for controlling SOX from
FCC regeneration have been used only where wastewater can be discharged into
the ocean after treatment.  A 50,000 bbl/d FCCU charging a feed with 2^weight
percent sulfur would generate as much as 60-70 tons of sludge per day.
                                     213
-------
W. R. Phillips
Boilers and Process Heaters—SOx Removal

          Emissions of SOX from boilers and process heaters can be minimized
by routing the flue gas to an integrated sulfur removal facility.  Post-
combustion removal of SOx from boiler and heater flue gases using an inte-
grated collection system of course poses serious safety and economic barriers.
Pre-startup firebox purging would be extremely difficult and time consuming.
Cost of ducting would probably be economically prohibitive.  Two such units
are. the IFP-150 and the Aqua-Glaus process described in Table 8.

NOX Removal

          NOX emissions may be reduced in the tail gas by any of three
methods.  These methods are 1) gas scrubbing, 2) catalytic reduction, and
3) thermal reduction with added ammonia.  Post-combustion NOX removal tends
to be more expensive than combustion modification because of the high tem-
perature of the gas, the low NOX concentration, interference from other
pollutants, and high power consumption.  Only thermal reduction appears
economically promising."*"*

          Controlled addition of ammonia and oxygen containing flue gas under
strictly controlled conditions at 1300 to 1900°F can selectively reduce
50 to 70 percent of the NOX remaining after combustion.  This "thermal denox"
process is a balance between two gas-phase reactions:  ammonia reduces NO to
N2 in the presence of the oxygen in the flue gas and ammonia is simultaneously
oxidized to NO.  When conditions are carefully controlled, a major portion of
the NOX can be reduced with little ammonia left over.  This process is more
expensive than combustion modification but can supplement these modifications
should stricter control of NOX be required.

Control Technology from Other Industries

          Control methods developed primarily for other industries can also
be used in the petroleum refining industry with some degree of adaptation.
This is especially true of methods developed by the electric utility industry
for flue gas desulfurization.  Some can be applied to the flue gas from a
Claus incinerator; another with accompanying NOX control can be adapted to
the flue gases from process heaters; still others might be used to control
sulfur emissions from FCC regenerators.

Sulfur Recovery

          Table 9 outlines several sulfur recovery processes from other
industries which might be adapted to Claus tail gas sulfur recovery.  These
sulfur recovery processes are described in greater detail in Radian's refinery
emissions report.  They do not uniformly meet NSPS for S02 emission.
                                      214
-------
        TABLE 9.  POTENTIAL GLAUS TAIL GAS SULFUR RECOVERY PROCESSES FROM OTHER  INDUSTRIES
           Process
                                      Characteristics
H2S or S02
 Removal
   (%)
                                                  Treated Flue Gas
                                                  S02 Concentration
                                                         (ppmv)
                                                                                                             H-
                                                                                                             T)
                                                                                                             cn
Chiyoda Thoroughbred 101
USBM Citrate
            46
Townsend1*7'"*8
Lucas49'50
Takahak
        5 I
Gypsum product.

Elemental sulfur product;
capital = 250 percent of
Glaus cost.  Not
commercialized.

Elemental sulfur product.
Does not remove CS2.   Not
commercialized.

S02 product is recycled.
Capital = 57-80 percent
of Glaus cost.  Semi-
commercial .

Elemental sulfur product.
Allegedly low capital cost.
 97   (S02)

      (S02)
                                                                                         >500
                                                                      (H2S)
                                                                      (S02)
                                                               ^99.9  (H2S)
                          200
                     (+ COS, CS2)
-------
W. R. Phillips
Catalyst Regeneration

          Several FGD methods used by the utility industry have been proposed
for use on FCC regenerators.5   In addition to the ones described below are
some of the regenerable processes touched on earlier, as applicable for
treatment of the Glaus unit tail gas.  One of the processes described below
simultaneously removes SOX and particulates from the flue gas.

The Lime/Limestone Flue Gas Desulfurization Process53—

          Lime or limestone flue gas desulfurization processes are the most
widely used FGD systems.  The systems are very similar; they consume large
quantities of feed material and produce large quantities of waste sludge, but
have relatively low operating costs and are highly reliable.  An S02 removal
efficiency of greater than 90 percent has been demonstrated.

          The economics of large lime/limestone FGD systems has been treated
in great detail.51*

The Dual Alkali Flue Gas Desulfurization Process55—

          The dual alkali (or double alkali) flue gas desulfurization process
can be used to overcome the scaling problem inherent in lime/limestone FGD
systems while retaining the convenience of solid waste disposal.  There are
53 operating dual alkali systems in the United States and Japan; several more
are under construction.

          These systems can achieve SO2 removal efficiencies of greater than
90 percent.  The capacity for more than 99 percent removal of S02 has been
demonstrated.  The dual alkali process itself is capable of greater than
98 percent particle removal.

          Dual alkali systems are economically competitive with lime/limestone
systems; however, a larger disposal area will be required than for a lime/
limestone system because of the higher moisture content of dual alkali sludge.

Boilers and Process Heaters

          The Shell flue gas desulfurization process (SFGD) can be used to
remove SOx and NOx simultaneously from all stack gas in process heaters,
providing they can be collected and sent to one or two stacks.  The SFGD
process can also be used to remove sulfur from the vent of fluid cat crackers,
as well as from Glaus units.56'57'58

          The Shell flue gas treatment process has demonstrated S02 and NOx
removal efficiencies of greater than 90 percent.  The efficiency of the
system is not affected by variations in the SOa or NOX concentration.  Costs
for an integrated SFGD system are not available, because such a system has
not yet been built at a U.S. installation.
                                      216
-------
Wi  R. Phillips
         The processes proposed for control of  SOz  emission from FCCU
regenerators may possibly be applied to flue gases  from boilers and process
heaters as well as from a Glaus unit.

Emission Reduction Through Alternative Operating Practices and Conditions

         Refinery operations are routinely modified to meet product speci-
fication requirements, product marketing  trends, feedstock availability
constraints, and operating cost goals.  The operating choices made include
both deliberate actions concerning processing alternatives (such as which
catalyst to use or which cut point to pick) and  more subtle actions, mainly
in the energy conservation areas (such as attention  to steam leaks or furnace
efficiency).  These choices can also affect the  overall refinery emissions.

         This subsection summarizes the  effects on  emissions of some of
these alternative operating practices.
High Temperature FCCU Catalyst Regeneration
                                           59>60>61>62jG3
         In older FCCU regenerators,  the highly  exothermic oxidation of  CO
to COz is avoided because the resulting high  temperature can damage regenera-
tor equipment, permanently deactivate  the catalyst,  and damage downstream
equipment.  To avoid this oxidation, the flue gas from the regenerator
generally contains little oxygen and large, nearly equal,  amounts  of CO and
C02.

         With high temperature regeneration, coke is  burned from  the catalyst
more efficiently, therefore yield from the FCC unit is increased.   The carbon
monoxide level in the exit gas from the regenerator can be reduced to well
below 500 ppm; in many instances a CO  boiler  is no longer  necessary for
emission control.  Because the catalyst to the FCC unit is hotter, preheat  of
the feed to the unit may not be necessary.   (Five hundred  ppm of CO corre-
sponds to the NSPS limit.)

         Several new catalysts, or promoters, have been introduced in the
last several years to promote the combustion  of CO to  C02.  A promoter may
be chosen to promote complete combustion or partial combustion where
metallurgy cannot withstand the higher temperatures.   Partial combustion  can
also  be used in situations where the CO is needed as fuel.

         One type of noble metal promoter is made part of the catalyst
recipe.  A second type is a liquid injected  into  the regenerator combination
zone.  The third is a solid added to makeup catalyst.

         In situations where partial  combustion  is needed, it is  possible
to combine high temperature regeneration with the CO boiler.  For  this
combined operation, the degree of high-temperature regeneration, and the
final temperature, can be controlled by the amount of  promoter used.
Higher regenerator outlet temperature  partially compensates for the
reduced quantity of CO reaching the CO boiler. Supplemental fuel  to
                                     217
-------
W. R. Phillips
the boiler is still required, but its cost is offset by the  increased product
yields in the FCC unit.

          In 1975, the cost of converting a relatively modern FCCU with
stainless steel cyclones to high-temperature regeneration was $50,000 to
$300,000.  Cost of a CO boiler for the unit was perhaps $2 million to $3
million.

SOX Removal in the FCC Regenerator

          Amoco has developed a catalyst which prevents sulfur from leaving
the regenerator as SOa-  The catalyst holds the sulfur until it is returned
to the reactor, where it is released and converted to H2S.  The H2S leaves
the reactor with the cracked product and is later converted to sulfur in the
Glaus plant; the regenerated catalyst returns to the regenerator.

          Cost for a 60-75 percent reduction in SOx emissions with this
method in a new facility is estimated at $0.03/bbl, compared to $0.22-0.24/
bbl for stack-gas scrubbing and up to $0.27/bbl for feed hydrodesulfurization.
The use of the catalyst for SOX control is also less expensive than other
methods in retrofit applications.

Combustion Modification for Control of N0x61f'65'66

          Of the oxides of nitrogen, only NO and NOa are of environmental
concern.  In combustion sources, NO may be produced either by the fixation
of atmospheric nitrogen in the flame (thermal NOx) or by the oxidation of a
portion of the nitrogen in the fuel (fuel NOX).  N02 from combustion sources
is produced as the NO combines with oxygen in the atmosphere.  Refining
sources of thermal NOX and fuel NOX are given in Table 10.

          A number of specific combustion modifications for NOX control have
been devised.  Those for refinery boilers are summarized in Table 11.  Com-
binations of these methods have been shown to yield a smaller effect than
the sums of the effects from the individual technologies.  The effectiveness
of some of the individual methods and some combinations at different boiler
loads are shown in Table 12.

          The subject of combustion modification is covered  in greater detail
in the refinery emissions report.

CONTROL OF FUGITIVE EMISSIONS

Sources of Fugitive Emissions

Sources Tested in This Study

          The hydrocarbon (HC) emission factors developed in this study  for
valves, flanges, pump seals, compressor seals, drains and relief valves  have
reasonable confidence limits.  Confidence limits for oil-water separator and
                                     218
-------
  W. R.  Phillips
          TABLE 10.   REFINING SOURCES OF THERMAL NOX AND FUEL
             N0>
Classification
High Temperature


Internal
Combustion
Moderate
'•""- 	 	 .-,-,
Source
Power boilers
firing - gas
Power boilers
firing - oil
Power boilers
firing - coal
Engines
Turbines
CO boilers
	 . ^
Thermal
NO
X
Present
Present
Present
Present
Strong
Present
	 ~ ' ~ •
Fuel
NO
X
Possible
Present
Strong
Unlikely
Possible
Present
Temperature
                      Coke and  residual
                      fuels

                      Catalyst
Present
Present
regeneration
Incineration
Process Heating
Gas cracking
Oil cracking
Oil heating
Unlikely
Present
Present
Unlikely
Unlikely
Present
Present
Possible
Possible
Possible
                                  219
-------
   W. R.  Phillips
   TABLE 11.  BOILER COMBUSTION MODIFICATIONS FOR REDUCTION  OF NOx EMISSIONS
        METHOD
      EFFECTS
    GAS FIRED
      UNITS
   OIL FIRED
     UNITS
                         Rating

                         Advantages
 Low excess air
                        Disadvantages
                   Good

                   Improved effi-
                   ciency; less
                   power
                   More complex ducts
                   and controls
                     Good

                     Improved effi-
                     ciency; less
                     power; less
                     chance of cold
                     end deposits

                     More complex
                     ducts and
                     controls
 Flue  gas
 recirculation
 Rating

 Advantages



 Disadvantages
 Excellent

 Very  effective;
 does  not upset
 combustion

 High  initial
 cost; high  operat-
 ing cost; addi-
 tional controls
 Good

 In moderation
 does not  upset
 combustion

 High initial
 cost;  high
 operating cost;
 additional con-
 trols; works
 mainly on
 thermal NO.
Staged combustion

   Two-stage
   combustion with
   over-fire air
   ports
Rating

Advantages



Disadvantages
                                          Excellent

                                          Inexpensive;
                                          very effective
Longer flames;
slight increase
in excess air
Very good

Inexpensive;
moderately
effective

Longer flames;
slight increase
in excess air
                                                        (Continued)
                                     220
-------
     W.  R. Phillips
                              TABLE 11.  Continued
       METHOD
   EFFECTS
   GAS -FIRED
    UNITS
   OIL FIRED
     UNITS
 Staged  combustion
    (Cont'd)

    Off-stoichioroetric
    or Biased firing
 Rating

 Advantages



 Disadvantages
Very  good

No power;
effective
                                           Slightly longer
                                           flames; small
                                           increase in
                                           excess air
                         Disadvantages
                  Reduced unit
                  efficiency;
                  require equipment
Good

No  power;
moderately
effective

Slightly longer
flames; small
increase in
excess air

Direct cooling
Lower preheat or
Water injection
Rating
Advantages
Fair
Simple; no
power
Fair
Simple; no
power
                    Reduced unit
                    efficiency;
                    require equipment
Reduced load or
Oversized fire box
Rating

Advantages


Disadvantages
Very good

Simple; no power
                                           High initial
                                           cost
Very good

Simple; no
power

High initial
cost; more
radiant super-
heater
                         Rating

Burner modifications      Advantages


                         Disadvantages
                  Good

                  Simple; no power


                  None
                    Very good

                    Simple; no
                    power

                    None
                                       221
-------
        TABLE 12.   REDUCTIONS  OF NOV EMISSIONS WITH COMBUSTION MODIFICATIONS AT VARIOUS BOILER  LOADS67
Combustion
Modification
(Percent Full Load)
Fuel Burner
Fired Arrangement
Gas Front Wall
Horizontally
Opposed
Tangential
Average
to
N3
Oil Front Wall
Horizontally
Opposed
Tangential
Average
Percent Reduction in NOX Emissions
Low
85/105
13

17
-
16


27

10
28
19
Excess
60/85
24

15
-
19


20

16
22
19
Air Staging
50/60 85/105 60/85
7 37 30

32 54 35
- -
26 45 31


28 29 20

12 34 34
17
18 30 22
Low Excess Air Flue. Gas Possible Combined
and Staging Recirculation Modifications
50/60 85/105 60/85 50/60 85/105 60/85
30 48 42 36 - -

59 61 48 68 -
- - - 60
52 54 44 52 - 60


20 39 32 21 46 31

47 35 44 42 -
45 10 13
34 38 37 37 28 23
50/60 85/105
43

20 73
66
20 64


50

38
-
47
60/85
42

52
65
51


41

35
59
42
50/60
36

72
-
60


21

55
--
38
                                                                                                                     1-1
Possible combination of modifications on the boilers tested.
-------
W. R. Phillips
cooling tower emission factors were considerably broader; as a result, data
from other sources have been substituted  in Table  13  for  results  from'the
latter two systems.

Relative Importance of Fugitive Emission  Sources

          Table 13 lists emitting sources  in refineries.  All the types of
emission sources listed were field-monitored except for numerically rare
items or those sources otherwise deemed insignificant.  Table 14 ranks the
eight most important emission sources according to the total estimated HC
losses from a hypothetical refinery.68  Emission factors are compared to
current literature values.  Results show  that process valves are typically
the largest fugitive emission source because of their great number in the
refinery.

          The oil-water separator is ranked second in importance in fugitive
hydrocarbon emissions based upon previous  work  (see Table 14 footnotes).
Additional field measurements and/or improved analytical techniques may be
required to obtain satisfactory confidence limits  for separator emission
factors because of the variability of waste oil vapor pressures, composition,
rates, wind effects, etc. from day to day  and from refinery to refinery.
Controlled testing of a covered simulated  separator gave results which would
lower the estimated emission from the AP-42 based  value of 110.5 Ib/hr to
  20.6 Ib/hr.69

          Cooling towers are also temporarily ranked high in HC emissions.
The broad emission factor confidence limits found may be a result of real
differences among cooling towers tested, or may reflect analytical imprecision
resulting from having to analyze water-dissolved hydrocarbons in the 1-5 ppm
range.  Additional field work in this area may be justified.

          Pump and compressor seal emission factors are averages for their
respective arrays, as listed in Table 13.  Types include packed gland,
mechanical face, labyrinth and oil seals,  for both rotary and reciprocating
shaft types, where applicable.  There were an average of about 1.4 seals per
pump in the refineries surveyed, and 2.0  seals per compressor.

          The emission factor for equipment drains is a useful addition to
the literature (0.070 lb/(hr-drain)) because no other factor has been avail-
able except in combination with one for oil-water  separators.

          Pipe flanges constitute the next to smallest fugitive emission
source, even with the largest estimated total number of devices.  The emis-
sion factor is 0.00058 lb/(hr-flange).

          Relief valves contribute the least total HC emission, because of
their relatively small number, but have a  significant emission factor  (0.190
lb/(hr-relief valve).
                                     223
-------
W. R. Phillips
                   TABLE 13.  FUGITIVE EMISSION SOURCES
                                            Number        Estimated Total
                                              of            Emissions, a
     System or Device                       Devices           Ib/hr

Process Piping System
   Valves                                   11,650            289.50
   Flanges                                  46,520             26.98
Safety Relief Valves                           106             20.14
Agitator Seals (Hydraulic, lip,                   Not Measured
   packed, mechanical)
Pump Seals                                     353C            57.63
   Rotaty Shaft
      Mechanical Face
      Packed Gland
   Reciprocating Shaft
      Packed Gland
Compressor Seals                                68             45.46
   Rotary Shaft
      Labyrinth
      Oil
      Mechanical Face
   Reciprocating Shaft
      Packed Gland
Water Systems
Wastewater Systems
   Drains - Process and Storm                  647             45.29
   Primary Treatment - API separator,            1             21e
      CPIe (covered)
   Intermediate Treatment - Air         Air Flot. Units tested; results
      flotation, holding basin            statistically inconclusive
   Secondary Treatment - Biological                 Not Tested
      oxidation processes
   Tertiary Treatment - Carbon
      absorption, filtration, ion
      exchange, reverse osmosis
Cooling Water System - Cooling Towers           V5a           113.3
Solid Waste System Alternatives                     Not Tested
   Land Farming                             1 Site              Og

Total Fugitive HC Emissions, Ib/hr                            618.3
                             (Ib/bbl feed)                   (0.045)

*3
 Estimated from data in Reference 71.
 Basis:   330,000 BCD hypothetical refinery, Reference 72.
 Basis:   1.4 seals (Av.) per pump, Reference 73.
 Basis:   2.0 seals (Av.) per compressor, Reference 74.
p
 Data from Reference 69.
 Based on 6 Ib HC/106 gal H20 (Reference 76) and 0.954 gpm circ. 4- B/D crude
 feed (Reference 77).
^Reference 78.

                                      224
-------


TABLE 14. FUGITIVE AIR EMISSION
Item
(Device or System)
Process Valves
^•1 TT ^ *. Uncovered
Oil-Water Separator ,
Covered
Cooling Towers
Pump Seals
Compressor Seals
Drains
Pipe Flanges
Relief Valves
This Study
Total HC
Emission, Ib/hr
289.5
178
20.6
2.04g'h
57.63
45.46
45.29
26.98
20.14

RANKINGS

Number
of Items
11,650
1
1
5d
353
68
647
46,520
106

- HYPOTHETICAL 330
Emission
This Study
0.0248
12.9 lb/103 bbl1
1.5 lb/103 bbl1
0.408
0.163
0.669
0.070
0.000580f
0.190

,000 B/D REFINERY
Factor, Ib/hr-item
Other Source
0.00625a
252 -lb/103 bblk'C>1
9.6 lb/103 bbl b'c'1
22.7b'd
0.175a
0.354a
N/A
0.006256
o.ioo6
Si
, R. Phillips







 Reference 79.
 Reference 80.
/-*
"Reference 81.
 Reference 82.
""Reference 83.
 The pipe flange emission factor is in units of Ib/hr-flange pair.
"Based upon purge method.
 Equivalent to 0.108 and 12.4 Ib HC/106 gal C.W. circulation and 0.954 gpm circulation of  C.W./
.B/D crude oil feed.
"""Bbl implies  bbl of refinery crude oil feed.
-------
W. R. Phillips
Control Technology - Fugitive Emissions

Process Valves

Existing Levels of Control in Refineries:  Process Valves—
Types of Valves—Process valve technology, per se, will not be covered here;
instead, the focus will be on valve seals.  Excellent recent valve technology
review articles are available; one such is recommended.    With the exception
of the check, plug, and diaphragm valve, valves are generally equipped with
packed stem seals to prevent the working fluid from leaking to the atmosphere.

Packed Stem Seals—Figure 2 is a simplified diagram of the type of packed
seal used for valve stems.  In practice, the stuffing box is filled with
rings of one or more types of compliant packing material.  The packing gland
is gently forced against the packing by tightening the bolts or studs
connecting the packing gland to the stuffing box flange (bolts not shown).
Upon being compressed, the packing material is forced against the stem or
shaft, forming a snug seal face (see figure).  This concentric contact seals
the working fluid from the atmosphere.
                                 -Stuffing
                                    Box
              Working
              Fluid
              End
L-     Spal Far^
                        Stuffing Box
                           Flange
                                                      Packing
                                                      Gland
                                                    Possible
                                                    Leak
                                                    Areas
                                   Packing
                       Figure 2.  Simple packed seal.
                                     226
-------
W. R.  Phillips
Packing Materials—Table 15 shows the diversity of materials used alone or in
combination.85  The following trends in packings have been noted:

               Asbestos packing may continue to be used in high temperature
               service, especially with metal wire core, but is being
               displaced by TFE, TFE-filled glass fibers, etc., up to about
               450-500°F (TFE = "Teflon"®).

               Graphite "ribbon" packing, which may be die-formed or formed
               in the stuffing box, is currently used in high temperature,
               high pressure service up to 4000 psi and 1200°F.

Control Valve Packings (High Temperature Steam Service)—One investigator
reported poor results using ribbon graphite packing alone.  The sliding stem
of the control valve, after gland tightening, was roughened, friction was
excessive,  and after as few as 800 cycles, leakage exceeded the target limit
of 0.5 cc/hr for a 0.5 in diameter stem.85  By sandwiching laminated rings
of graphite packing between layers of braided graphite filament packing, the
former acting to control fluid loss parallel to, and along, the stem, and the
latter acting to polish the stem, the life of the packing was extended to
over 47,000 cycles before leakage exceeded 0.5 cc/hr.  This cycle endurance
is roughly equivalent to the mechanical life of many valves, so such a valve
might never require repacking of the stem.87

Frequency of Application:  Valves—

Process Valve Types—Table 16 lists the approximate distribution of Battery
Limits refinery valves within two broad categories (manual and control) and
by valve'configuration.  Radian's battery limits survey shows that 86 percent
of all manual valves are gate valves, and 92 percent of the control valves
are globe-type.  The nine-refinery survey shows that these two categories,
manual gate valves and controlled globe valves, make up 88 percent of all
refinery valves.

Packed Seals—Radian's refinery observations further showed that all the
gate-, globe- and butterfly valves catalogued had packed-gland stem seals.
These packed-stem valves constitute an estimated 94.2 percent of the process
valve population.  On the basis of emission factors found in Table 14, it can
be shown that, on the average, 95.5 percent of the total emission from a
flanged in-line valve will be from the packing gland, and only 4.5 percent
from the two flanges.  Hence, the emphasis in valve emission control for
existing valves must be to select proper monitoring, operating, and mainte-
nance schedules and procedures, as well as suitable packings, for the packed-
stem seals.

Effectiveness of Existing Levels of Control:  Valves—

Factors Affecting Emission Rate—An established plant with a valve emission
problem will seldom in practice solve it by changing the fluid conditions
or the valve type.  The valve originally selected will instead have been
                                    227
-------
                                TABLE 15.  PACKING MATERIALS - PROCESS VALVES
                                                                            68
              Packing Material
        Form
  Used For
 Temperature
OS
      Flexible, all metallic
      Flexible metallic packings
      (aluminum).
       Flexible  metallic  packing  (copper).
       Long-fiber pure asbestos and fine
       lubricating graphite (nonmetallic),

       Closely braided asbestos yarn, top
       jacket reinforced with Inconel
       wire; core: long fiber asbestos.

       Pure asbestos yarn with an Inconel
       wire insert around a resilient
       asbestos core impregnated with
       graphite.

       Twisted long fiber Canadian
       asbestos.
Spiral wrapping.   Thin
ribbons of soft babbit
foil.

Spiral wrapping.   Thin
ribbons of soft annealed
aluminum foil loosely
around a small core of
pure dry asbestos.

Soft annealed copper
foil loosely around a
small  core of pure dry
asbestos.

Graphite  special  long-
fiber  asbestos binder.

Spools,  die-formed
rings.
 Spool form,  die formed.
 Spool form, die formed.
Valve stem
Hot oil valves,
diphenyl valves,
Up to 450°F.
Up to 1000°F,
Hot oil valves,    Up  to  1000'F,
diphenyl valves.
 Extreme
 resilience.
 Up to 7508F.
 High-temperature   Up to 1200eF.
 valves.
 Valve stem for     Stuffing box
 air, water steam   temperature up
 and mineral oil.   to 1200°F,
 Valves handling,
 high and low
 pressure steam.
 Up  to  500'F,
                                                                                                               T)
                                                                                                               cn
                                                                                           (Continued)
-------
                                      TABLE 15.  Continued
        Packing Material
        Form
  Used For
Temperature
Asbestos, graphite and oilproof
binder.

Solid, braided TFE.
Braided asbestos with complete
impregnation of TFE.
 Braided  of  high  quality wire-
 inserted asbestos  over a  loose
 core of  graphite and  asbestos.

 Braided  of  high quality wire-
 inserted asbestos over  a  loose
 core of  graphite.

 Braided of long-fiber Canadian
 asbestos yarn each strand impreg-
 nated with heat-resistant lubricant.

 Long-fiber Canadian asbestos yarn,
 each strand  treated with a synthe-
 tic oilproof binder and  impreg-
 nated with dry  graphite,
Spool form,  die formed,
Coil, spool, ring.
Coil, spool, ring.
Coils, spools,
 Coils,  spools.
 Coils,  spools.
 Coils,  spools,
Shutoff valves.    Up to 550eF.
Valve shaft for    100°F to  500°F.
highly corrosive
service.

Valve stems in     100°F to  600°F.
mild chemical or
solvent service.
Valve stems,
steam, air,
mineral oil.
Up to 1200"F.
 Stainless-steel     Up  to  12008F.
 valve  stems,  air,
 steam, water.

 Valves for  steam,   Up  to  550°F.
 air,  gas  and  mild
 chemicals.

 Refinery  valves.    To  750°F.
                                                                                     (Continued)
-------
                                             TABLE 15.   Continued
to
O
              Packing Material
                                             Form
       Braided/ovetbraided, wire-
       inserted, white asbestos packing
       impregnated with a heat-resistant
       lubricant.

       Braided white asbestos yarn
       impregnated with TFE suspensoid.

       Braided or bleached TFE multi-
       filament  yarn.
Braided TFE multifilament yarn
impregnated with TFE suspensoid.
        Asbestos  jacket,  braided  over  a
        dry-lubricated plastic core of
        asbestos  graphite and elastomers.
                                     Coils, spools.
Coils,  spools.


Spools, coils.



Spools, coils.
                            Used For
                    Temperature
                                      Spools and coils.
Valve stems, for
valves handling
steam, air, gas
cresylic acid.

Valve stems.
                                             Up  to  750°F.
                                                                                   1008F  to  600°F,
Valve stems for     12°F  to  506°F.
highly corrosive
liquids.

Valve stems for     120°F to 600°F.
corrosive  chemi-
cals, solvents,
gases.

Valve  stems,  for    Up to 850"F.
valves  handling
 superheated steam,
 hot gases.
IT)
ff
H-
I—•
t—'
H-
K
-------
         TABLE 16.  APPROXIMATE DISTRIBUTION OF REFINERY PROCESS VALVES3                         ^
                   BY TYPE AND SERVICE                                                     H-
Service
Type Valve
Gate
Globe
Plug
Butterfly
Diaphragm
Total
Manual
64.7
3.8
5.7
0.6
• o.o
74.8
Control
0.0
23.3
0.0
1.8
0.1
25.2
Total
64.7
27.0
5.7
2.5
0.1
100.0
'Check  and sample  system valves  excluded.   No dry-service slide
 valves surveyed  (Radian statistical survey basis).
-------
W. R. Phillips
chosen to fit the fluid flow conditions.   It will further have been based
upon economy of operation and safety in accordance with API code.  The
latter factor of safety, particularly with regard to potential failure in
a fire, cannot be overemphasized.

          Actions most apt to be used to solve or attenuate leakage, listed
in increasing order of cost, are as follows:

          1)   Tighten packing gland;

          2)   Lubricate lantern ring (packed stem valve) or plug (plug
               valve);

          3)   Replace or change type of packing;

          4)   Replace or change type of valve.

Operations/Maintenance Cooperation—Many companies require operating
personnel to check all major equipment once per shift or per day for leaks
for purposes of economy and safety.  Tightening and lubricating valves is
routine.

          Replacing valve packing is considered routine even though not all
maintenance personnel are skilled at it.   A change in type of packing materials
is not generally costly.  The cost breakdown for machinery packing (probably
                                                    " - -	 - ' iT
a pump) outage is:

               Item                                 Percent of Cost

          Packing Material                                  3
          Labor to Pack                                    13
          Fluid Loss                                       21
          Downtime                                         63

          Total                                           100

Statistical Results and Rationale for Results—Table 14 showed the average
valve emission factor for the valve mix in a 330,000 BPCD refinery to be
0.0248 lb/(hr-valve).  Emission rate was interestingly only weakly dependent
on valve size.  Rationale might be:

          1)   Small valves have shallower stuffing boxes than large
               valves, so leak more for their stem size than large
               valves.

          2)   Large valves may get more maintenance and operation
               attention than small valves.

          3)   Large valves may be manufactured to closer tolerances
               than small valves.
                                     232
-------
 .  R. Phillips
Point 1 may be an area for  improvement in valve design.  Aside from retooling
costs, the overall cost of  small  valves should not be increased significantly
by deepening the stuffing box.  Points 2 and 3 are speculative.

Available Control Technology  for  Fugitive Emissions in the Refining Industry:
Valves—

Types of Controls Available for Valves:  "Packless" Seals—The following
types of valve seals are not  apparently used in refineries based upon our
survey:

         •    Diaphragm bonnet seal,

         •    Bellows bonnet seal.

These packless seals when correctly  applied in noncritical, low-stress
conditions of temperature,  pressure  or corrosivity (in the case of the
bellows seal) should approach zero leakage.  The diaphragm material in the
first valve shown limits operation to  about 50 psi pressure differential.90
This type valve has definite  limitations in refinery use; it would fail
catastrophically upon overheating of the elastomer diaphragm, so use would
not include hydrocarbon service where  a fire could be fed by failure.  The
bellows-sealed valve, because of  the corrosion and fatigue failure potential
of the bellows, is limited  in its use  by combined temperature-pressure-
corrosivity stress, which level is best defined by the valve manufacturer.
Back-up stem packing would  appear to be absolutely necessary for these
valves in case of diaphragm or bellows failure.

Valve Maintenance Programs—Valve monitoring and maintenance programs can
be an effective method for  reducing  valve emissions.   These programs and
their effectiveness are discussed in detail in another paper of this
symposium.

Energy Requirements:  Valves—No  primary energy cost would result from
substitution of a very limited number  of packless valves for conventional
packed-stem, bonnet-sealed  valves.  As to a valve maintenance program,
incremental manpower requirement  would probably be necessary if refineries
not already doing so were to  begin comprehensive periodic inspections.

Cost-Available Refinery Technology—Estimates for the substitution of packless
valves for packed-stem valves range  from 150-367 percent91 to 1000-2000
percent92 of packed stem valve cost.  Application of packless valves
(diaphragm-sealed; bellows-sealed) in  critical services is not seen as
probable because of problems  associated with valve failure, so economic
impact is correspondingly nil.

         Because the long-term effects of valve maintenance are not yet
clearly defined, costs of a valve maintenance program have not been developed.
Any valve maintenance program would  probably be more burdensome to the small
                                     233
-------
W- R. Phillips
refiner, because the number of valves inspected/maintained per unit throughput
will be higher for small refineries.

Control Technology from Associated Industries:  Valves—

          Ball valves may possibly find broad use in refining, but with TFE
and TFE-filled fiber seats, are limited to use below about 450°F.  They were
not available in statistically significant numbers in this study, so their
field-tested emission factors are not available.

Development Needs:  Valves—Short-term, there appears to be a need for a
small packed-stem valve with a deeper stuffing box than is currently
available.93

Pump Seals

Existing Levels of Fugitive Emission Control in Refineries:  Pump Seals—

Types of Pump Seals—This survey showed that refinery pump-seal combinations
almost exclusively fall into one of three broad categories:

                                                          Percent of
                                                          Population

          A.   Centrifugal Pump - Mechanical Seal            82.1
          B.   Centrifugal  Pump - Packed Seal               11.5
          C.   Reciprocating Pump - Packed Seal               6.4

                                        Total               100.0

These seals are depicted in rudimentary form in Figure 3.

          The packed seal, Figure 3(A), is used to seal both rotary and
reciprocating shafts against leakage of liquid from the "working fluid" end
of the shafts to the atmosphere.  Compressed packing in the stuffing box
forms a contact seal against the moving drive shaft.  High-speed friction
resulting from this contact requires that either the working fluid be
allowed to leak from the stuffing box housing the packed shaft, or a
supplementary liquid be introduced to remove frictional heat.  A typical
leak rate would be 60 drops per minute (^3 ml per minute).91f

          Mechanical seal application, contrary to the broader applicability
of packed seals to both rotating and reciprocating shafts, is limited to use
on rotary shafts.   Mechanical seals may be used to seal both pump and com-
pressor shafts, but are more universally applied to pumps, specifically
centrifugal pumps.

          Packings - Service limits for selected packings are found in
          Table 17.
                                     234
-------
W.  R.  Phillips
                                                ^-Stuffing box
                                         B. Mechanical packing
                                               ^Stuffing  box
                             Fluid

                         Impeller
                           end
                             Fluid
                                         •b. Mechanical seal
                      Figure 3.   Common pump  seals  - simplified.
                                              235
-------
                                                                                 a
                          TABLE  17.  SERVICE LIMITS FOR SELECTED MECHANICAL PACKINGS '
                                                                                   95
NJ
CO
Break-In.
Leakage
Packing (drops/min)
Asbestos/TFE 120
TFE (lubed) 120
Asbestos/Graphite
Graphite-Fiber
Graphite-Ribbon
Lead
Aluminum
Inconel-Reinforced
Asbestos Over ,
Resilient Core
Running Maximum
Leakage Temperature
(drops/min) (°F)
60
60
60
60
60
60
60


500
500
400
1000 (600)d
1000 (600)d
350
800 (500)d
1200

Pressure at Temperature
Maximum Maximum At Maximum
C C C
Temperature Pressure Pressure
(psig) (paig) (°F)
50 200
50 200
50 250
50 350
50 350
50 400e
50 400e
Unknown

100
100
100
300
.300
100
200


       aBasic data:  2-in shaft, 3550 rpm.  Controlled leakage for 720 h\r  Pumped liquid is water.
        Assumes maximum AT of 100°F  (50°F for flax) due to shaft friction.  Satisfactory results can be
        expected by using these maximum limits and following FSA (Fluid Sealing Assn.) Test Procedure  01.

        Leakage rate:  1 ml/rain a 10 to 20 drops/min,

       cTeraperature is product  temperature; pressure  is stuffing-box pressure.

        Larger number is nonoxidizing environment; smaller number  is oxidizing environment.

       Assumes rings are die-formed.
        For  low-speed shafts only.   (Green, Tweed  and Company).
S3

pa
-------
  . R. Phillips
          Lubricants for packings include  the  following  substances:96

          •     Mica and talc

          •     Graphite

          •     Molybdenum disulfide (MoS2)

               Hydrocarbon type lubricants (greases, tallow, petroleum oils)

          •     Tungsten disulfide

               TFE

          •     Silicone oils.

          Mechanical seals - The mechanical seal in its many forms is the
          predominant pump seal today.  At the time of the Los Angeles
          County,  California study twenty years ago, mechanical seals made
          up only  42 percent of the seals in use there.97  Radian's survey
          revealed that by 1978, approximately 82 percent of the refinery
          pump seals were mechanical type.

          Mechanical seals are prefabricated assemblies which shift the
          point of wear from the drive shaft, as with packed seals, to
          easily-replaced pairs of rings, one of which is attached to the
          pump shaft, and the other to the gland plate or its equivalent.
          Seal faces are perpendicular to the shaft as shown in Figure 4.
          Faces are typically lapped to a flatness of two microns which
          accounts for their typically low leak rate when carefully instailed>
          started  up and flushed properly.

          Double mechanical seals provide a margin of protection against
          seal failure not offered by single mechanical seals.

          If the inner seal should fail, the outer seal prevents escaping
          fluid from reaching the atmosphere; in case of accidental pressure
          loss in  the seal liquid system, however, the pumped liquid will
          contaminate the seal liquid.  If the seal liquid is contained
          within a pressurized "seal pot" system, the problem of contaminated
          seal liquid cleanup is minimized.

Frequency  of Application of Pump Seals—The hypothetical  refinery mentioned
much earlier in  this  paper (Table 14)  was seen to emit an estimated 57.63
Ib/hr from 353  pump  seals for a weighted emission factor  of 0.163 lb/(hr-pump
seal).  This total HC emission rate  places pump seals fourth in importance
among the  process-related fugitive  emissions  studied by Radian.
                                     237
-------
                                                                                                                    Gland

                                                                                                                    plate
UJ
oo
                             i "9 * y^—i


                        KUmped liquid ^
                                         a. Outiide teal
                                                                                               b. Inside
                                Figure 4.   Location of primary ring determines  seal type.
                                                                                                                                        H-
                                                                                                                                        T3
                                                                                                                                        cn
-------
W. R. Phillips
          Table 18 is a comparison of seal distributions and emission factors
taken from the 1958 California Study98 and the current refinery survey."
In 1958,  the percentages of mechanical and packed seals on pumps clearly
favored packed seals:

          Centrifugal Pumps, Packed Seals                  34.7%
          Reciprocating Pumps, Packed Seals                23.1
          Subtotal Packed Seals                            57fQ

          Centrifugal Pumps, Mechanical Seals              42.2
          Total                                           100.0

By 1978 the percentage of mechanical seals used in refineries had almost
doubled; approximately 82 percent of the seals were by then mechanical type.
The Radian survey showed this percentage to be further subdivided into
approximately 67 percent single mechanical seals and 15 percent double
mechanical seals.  No further subdivisions were made.

Effectiveness of Existing Levels of Control in Refineries—Table 18 reveals
that the overall pump seal emission factor has improved slightly over 20
years from 0.17 Ib/hr-seal to 0.16 Ib/hr-seal.

          Pump seal emission factors as shown in Table 18 should not be used
injudiciously without referring to the detailed results and statistical
analyses as found in the Final Report of this study.  To illustrate the
reasoning behind this statement, refer to Table 19.  First, there is little
doubt that pump seals in light liquid service emit hydrocarbons at a higher
rate than those in heavy liquid service (0.256 vs. 0.046).  The two emission
factor confidence intervals do not overlap, adding validity to the estimated
factors (0.17-0.39 vs. 0.02-0.11).  By contrast, the emission factor confi-
dence intervals for the three major seal types do overlap, meaning that
within the limits of certainty  (95 percent), all three classes of pump seals
could have identical emission factors.  These broad confidence limits were
characteristic of the emission factors statistically separated according to
type of seal, regardless of whether the emission factors were analyzed within
stream types (light vs. heavy hydrocarbons) or not.  One factor in the history
of mechanical seal applications should not be overlooked.  Typically, they
were applied first in those services which presented the greatest emission
problems:  especially high pressure, high vapor pressure liquids with little
self-lubrication.

Available Control Technology in the Refining Industry:  Pump Seals—

Types—Application of the types of pump seals already described is relatively
uniform within the refining industry.  This may be the result of a greater
uniformity of feedstocks and products than, say, the chemical industry uses
and produces.  The application of standards published by the American
Petroleum Institute (API) has also undoubtedly led to uniformity among
                                     239
-------
              TABLE 18.   DISTRIBUTION OF TYPES OF,  AND EMISSION FACTORS FOR,  REFINERY PUMP SEALS
N3
-P-
o
Seal Type
Rotary Shaft (Centrifugal)
Mechanical - Single
Mechanical - Double
Subtotal
Packed
Reciprocating Shaft
Packed
Total
(Source) 1958a
Distribution Emission Factor
(%) (Ib/hr-seal)

42.2 0.13
34.7 0.20

23.1 0.22
100.0 0.17
(This
Distribution
(X)
67.1
15.0
82.1
11.5

6.4
100.0
Study) 1978b
Emission Factor
(Ib/hr-seal)
0.19°
0,15°
0.19
0.071

0.14
0.16
       California joint refinery study among federal, state and Los Angeles District agencies.

        Percentages ara baaed upon complete process unit surveys within each of nine refineries, but
        without random selection of unit  types,  Units selected are Isited in Interim Reports
        Emission  factors are average from operational and- standby pumps.

       cEmi8sion  factor confidence limits for the  three basic  types of  seals  (centrifugal-packed,
        centrifugal-mechanical, reciprocating-packed) overlap  to  the extent  that  all three  classes of
        seals  could have identical emission factors.
-------
TABLE 19.
Source
Type


Stream 2J*
Stream 3
Total
Seal Type
Seal Type
Seal Type
Total
Stream 2
Stream 2
Stream 2
Total
Number
Screened





CMC
CPd
RPe

- CM
- CP
- RP


466
290
756
621
87
48
756
404
37
25
Total
Leaking


298
66

312
32
20

264
17
17
PUMP
Percent
Leaking


63.
22.

51.
36.
41.

65.
45.
68.


9
7

0
8
7

3
9
0
SEAL EMISSION FACTORS
97.5 Percent
Confidence Interval


(58.9,
(17.2,

(46.5,
(24.6,
(24.8,

(59.9,
(26.3,
(41.9,


68.9)
28.2)

55.5)
47.8)
60.0)

70.6)
66.6)
87.8)
Estimated
Emission
Factor


0.
0.

0.
0.
0.

0.
0.
0.


256
046

187
071
141

263
082
248
z:
95 Percent "^
Confidence Interval ^
For Emission Factor. £


(0.17,
(0.02,

' (0.13,
(0.02,
(0.057

(0.18,
(0.013
(0.04,
T)
CO
0.39)
0.11)

0.29)
0.22)
, 0.69)

0.41)
, 0.34)
1.2)
  Total
           466
Stream 3
Stream 3
Stream 3
- CM
- CP
- RP
217
50
23
48
15
3
22.1
30.0
13.0
                                                (15.7, 28.4)
                                                (15.6, 47.9)
                                                ( 1.8, 38.5)
                                                              0.044
                                                              0.041
                                                              0.013
(0.02,  0.12)
(0.0083, 0.17)
(10-5,  7.2)
  Total
           290
'Light  liquids
'Heavy  liquids
'CM • Centrifugal pump - mechanical  seal
 CP M Centrifugal pump - packed seal
•RP
Reciprocating pump - packed seal
-------
W. R. Phillips
devices used to control fugitive emissions, not only from pumps, but also
other devices surveyed by Radian.

Effectiveness—All refinery pump seal emission technology believed to be
available today was found among the thirteen refineries surveyed, and the
control effectiveness by seal type was reported in the preceding section.
A detailed account of emission factor statistics is found in the Final Report
of this study.

Energy Requirements and Relative Costs—Industry experience has shown that
mechanical seals lose less frictional energy than packed seals.  One seal
manufacturer100 reports the following relative friction losses:
       Type Pump Seal

Balanced Single Mechanical
Unbalanced Single Mechanical
Packed
                        Average Power
                       Requirement, kW

                            0.333...
                            0.400
                            2.00
Annual Electrical
     Cost, $*

       105
       126
       630
* Cost Basis:
4
-------
W. R. Phillips
satisfactory long-term performance history.  It would probably be necessary
for the API to take a positive position on the inherent design and performance
of this type pump before refiners would be willing to apply them.

Wastewater Systems

Existing Levels of Control in Refineries:  Wastewater Systems—

          Refinery wastewater systems vary tremendously in volume of process
water, storm water, particulates, oil and grease, and other contaminants.
Refinery wastewater systems also vary from one refinery to the next.  About
the only common denominator is an oil and water separator of the API or CPI
type.  As a result of variations in wastewater, reliable data for hydrocarbon
emissions from refinery wastewater systems do not exist.

          Variation in wastewater composition causes corresponding differences
in fugitive emission rates.  This was seen in emission measurements as
reported in Radian's fugitive emissions report.

          Despite the lack of reliable emissions data, control of fugitive
emissions is not complex, because emissions consist primarily of hydrocarbons
released from the collection system and oil-water separator.J Qlf

Characterization of Existing Wastewater Systems—Refinery wastewater systems
have evolved over the years as people have become aware of water pollution
problems, and as various treatment systems have been developed.  The basic
treatment steps may be summarized as follows:105

          •    Primary Separation - The removal of oil by gravity separation.
               Normally an API or CPI type separator is used.

          •    Intermediate Separation - The removal of suspended solids
               and additional oil by chemical sedimentation or air flotation.

          •    Secondary Treatment - The reduction of the biological oxygen
               demand (BOD) with some type of biochemical oxidation.

          •    Tertiary Treatment - Removal of dissolved organics which will
               not degrade with biological treatment methods.  Carbon
               adsorption is the most common form of tertiary treatment.

          Only the collection and primary separation systems will be
discussed.  Losses from intermediate, secondary, and tertiary treatment
systems are small in comparison.

          Since the fugitive emissions from refinery wastewater systems
consist almost exclusively of hydrocarbon losses from the collection system
and the oil and water separator, measurement and control strategies should
be limited to these two areas.   (One additional potential source^of air
pollution is the vent gas from the carbon regeneration system.     This
                                      243
-------
W. R. Phillips
source probably does not produce a significant amount of fugitive emissions,
but should be investigated more fully as the number of refineries using
carbon adsorption increases.)

Estimated Hydrocarbon Losses to the Air—In 1958, hydrocarbon emissions from
wastewater separators in existing refineries in the Los Angeles County area
were estimated to range from 10 lb/1000 bbls refinery capacity to 200 lb/1000
                       i n 7
bbls refinery capacity.

          The third edition of report AP-42 (August 1977) lists the relevant
hydrocarbon emission factors as follows:

          Process Drains

               Uncontrolled                210 lb/1000 bbl wastewater

               Vapor Recovery or             8 lb/1000 bbl wastewater
               Separator Covers

          The relationship between wastewater flow rate and crude oil
throughput has been shown to vary widely among refineries.  Newer or updated
refineries do a better job of segregating process water from storm water.
                                        i n ft
The following ratios have been reported.

          Refinery Classification          Bbl Wastewater/Bbl Crude

                  Older                               6.0
                  Typical                             2.4
                  Newer                               1.2

Another source gives a ratio of 0.8 barrels of wastewater per barrel of
crude.109  The original 1958 Los Angeles factors and the current AP-42
factors are very similar when a ratio of wastewater to crude of slightly over
one is used.

          Table 14 of this report ranks hydrocarbon emissions from oil and
water separators as the second largest source of fugitive emissions from a
refinery-  This ranking is based on the emission factor for uncovered
separators according to Litchfield.ll°  Litchfield's reported emission rates
for covered and uncovered separators were obtained under controlled condi-
tions, and are reasonable bases (with modifications which are to be shown)
for revising the AP-42 emission factors.111

          Using the most recent information available, Arthur D. Little
employed characteristics of raw and treated process wastewater to generate
models of two base case refineries.112  For a 200,000 BPD East/Gulf Coast
refinery, the oil and grease in the API separator effluent was given as 1920
Ib/day, or 9.6 lb/1000 bbls refinery capacity.  For the 330,000 BPD hypo-
thetical refinery in Radian's study, this factor gives 132 Ib/hr oil and
grease in the separator effluent water to the dissolved air flotation unit.
                                     244
-------
W.  R. Phillips
         Separator removal  efficiencies are reported to be 60 to 99
percent,113 and 50 to 87 percent.111*   Using 87 percent as a typical high
efficiency number, oil and grease rate to the separator becomes 1015 Ib/hr,
or 74 lb/103 bbl of crude oil  to  the  refinery.

         In a 1971 study  (using  laboratory data from a simulated API
separator) by D. K. Litchfield of American Oil Company, evaporative losses
of oil  from API separators were found to average 16 volume percent without
covers  and two volume percent  with covers of a cellular glass insulation
(manufactured by Pittsburgh  Corning)  floating directly on the oil.115  The
two volume percent loss with covers is not affected significantly by air
temperature because of the insulating effect of the cover.  The evaporative
oil loss from uncovered separators was found to vary with ambient temperature,
influent temperature, and  the  10  percent true boiling point of the oil.   The
16 percent volume loss reported is for an average ambient temperature of
40.1°F, an average separator water temperature of 140°F, and an average 10
percent true boiling point of  the oil of 300°F.

         The average ambient  temperature of 40.1°F is low for an estimate of
maximum hydrocarbon emissions, so an  average ambient temperature of 80°F
will be used.  Using an evaporation factor of 0.0319 volume percent per °F
increase, the oil evaporation  from an uncovered separator is 17.3 volume
percent and the oil loss with  covers  remains around two volume percent.

         These loss rates give hydrocarbon evaporation rates from the 330,000
BPD refinery of 178 Ib/hr  for  uncovered separators and 20.6 Ib/hr for covered
separators.  If these more recent studies are more representative than the
initial work beginning with  the Los Angeles studies, then the wastewater
system  hydrocarbon emission  factors should be:

         12.9 lb/1000 bbl crude  for  uncovered API separators,

          1.5 lb/1000 bbl crude  for  covered API separators.

Separator covers may therefore be expected to reduce emissions approximately
89 percent.  These factors show that  the fugitive hydrocarbon emissions from
wastewater systems with covered separators should rank well down .on the list
of total emissions, and that API  separator covers produce significant emis-
sions reductions.

Collection System - Process  and Storm Sewers—

         The contribution of  the wastewater collection system to the overall
refinery fugitive hydrocarbon  emissions is shown as 45.29 Ib/hr in Table 15
for the hypothetical 330,000 BPD  refinery.  These emissions result mainly
from allowing oil or oily  water to be exposed to the air in the process
areas or in the drains and sewers.

         In general, available controls for reducing fugitive emissions
from existing process and  storm sewers and collection systems consist of
                                     245
-------
W. R. Phillips
relatively minor modifications such as sealing open  sewer  systems,  altering
pump bases for better drainage, recurbing some process areas  for  separation
of oily water, and improving housekeeping.  These changes  should  be made
wherever applicable.

          Changes which involve substantial capital  outlays  (or which may be
nearly infeasible from a construction standpoint) such as major revisions to
existing underground sewer systems or installation of vapor recovery systems
do not represent best available technology economically achievable.

Primary Treatment - Oil and Water Separator—

          The primary treatment of process water is  the oil and water
separator which is usually of the API or CPI type (corrugated-plate inter-
ceptor).  All U.S. refineries have facilities for gravity  separation of oil
and water.     These separators are effective in removing  free oil  from water.
                                                                    117
but will not separate substances in solution or break up emulsions.

          Covering the oil and water separator is the only effective and
economical means of reducing hydrocarbon emissions from refinery wastewater
treatment systems.  If the separator is operated properly, then hydrocarbon
emissions from the downstream equipment will be negligible, and if  the
separator is covered, then hydrocarbon emissions from that source will be
effectively controlled.

          API separators are covered by floating pontoons or  double-deck type
covers which are sealed against the outer walls of each bay.  A CPI separator
normally will have a fixed roof cover.118

          Using an API separator as an example, the  economic  incentive of
reducing oil losses to the atmosphere by covering the bays will be  examined
here.  Conservative economics here will show the minimum return on  investment
for installing covers.

          Sources indicate separator cover requirements of 0.028 an 0.050 ft2
per BPD wastewater flow.119'120  Corresponding costs escalated by the M&S
equipment cost index (Chemical Engineering Magazine) to third quarter 1979
are $15.84 and $14.85 per ftz, respectively.  A capital cost  of $265,000 is
obtained for a 330,000 BPD refinery using $16.00/ft2 and 0.050 ft2/BPD.

          A typical refinery installing covers would see evaporative oil
losses of approximately 17 percent reduced to less than 2  percent.  For the
330,000 BPD theoretical refinery this means a savings of 157  Ib/hr  or 12
BPD at an assumed specific gravity of 0.87.  At $16  per barrel this savings
is worth $70,000 per year.  If annual maintenance costs of $5000  are incurred
on the covers, then the net savings is $65,000 per year.   This savings yields
a before tax discounted cash flow (DCF) rate of return of  28  percent assuming
a 20-year economic life and no investment tax credits.
                                      246
-------
W. R. Phillips
          The economic incentive to install covers on oil and water separators
combined with the resultant reduction of fugitive hydrocarbon emissions make
a strong case for all separators to be covered.  As of January 1977, 80
percent of the refining capacity was located in states where these covers are
required.J 21

Cooling Towers

Existing Levels of Cooling Tower Control in Refineries—

Types—At the time of the 1958 Los Angeles County California Emissions
Study,122 "atmospheric sections" (splash-cooled heat exchanger tubes) could
still be found in refinery cooling towers, although they were prone to leak
and were difficult to repair.  Chromates and chlorine were used to control
corrosion and biological growth, respectively.  The emission factor for
cooling towers was estimated to be 6.2 lb hydrocarbons/106 gal cooling water
circulation.

          By 1978, wetted "atmospheric" sections had, generally, been phased
out of refineries, organo-phosphates had replaced chromates for corrosion
control, and biological growth was being controlled by combinations of
chlorine and, often, nonoxidizing biocides.

          Today, as in 1958, makeup water ranges from near-pristine snow-
based surface water to sea water.  Some refineries now recycle water from
sour water strippers, which tends to reduce total plant water effluent and
retain phenols in the plant.  If recycled to a cooling tower, the aeration
encourages oxidation of phenols  from the  stripper bottoms water.

          Emission factors determined during this study have been based upon
two analytical methods:  Total Organic Carbon  (TOG) Analysis, and a purge
technique.  These results bracket the 1958 emission factor of 6.2 lb/10
gal cooling water:

                                                 Emission Factor
          Analytical Technique                  lb HC/106 gal C.W.

                 TOG                                  12-4

                 Purge                                 0.108

The purge technique is more precise and accurate than the TOC technique based
upon standardization runs, so one concludes that there is a high probability
that progress has been made in reducing cooling tower hydrocarbon losses.

Cooling Tower Control Technology Available in Refineries and Associated
Industry—

Types and Effectiveness—Forced-draft cooling  towers characterize refineries
and organic chemical plants.  The greater heat release rates of power
                                     247
-------
W. R. Phillips
industry plants make parabolic,  natural-draft towers more economical.  There
is no inherent advantage to either basic type of tower in terms of primary
air emission control.

          Air monitoring and water monitoring instruments for the purpose of
leak detection are common to all the process industries; no analytical
problems of a refinery water system may be considered unique.

Costs—As industries tighten restrictions on water emissions, the likelihood
of having to deal with a broader array of recycle water types increases.
This will probably require progressively more attention to materials of
construction of cooling towers,  heat exchangers and water piping.  Also,
more complex treatment chemicals and application systems may be called for.
Changes in the circulating water may of course affect levels of air emissions
as a result of corrosion-induced leaks.  Air emissions may also increase from
use of cooling towers as bio-oxidation devices, as in the case of recycled
phenols already mentioned.  The overall impact of these forecast changes on
air emission control costs ha's not been addressed, to the best of our
knowledge.
                                     248
-------
 . R. Phillips
REFERENCES

1.    Gary,  J.  H.,  and G.  E.  Handwerk.  Petroleum Refining, Technology and
     Economics.  Marcel Dekker, New York, New York, 1975.

2.    Elkin, H.  G., and R. A. Constable.  Source/Control of Air Emissions.
     Hydrocarbon Proc. 51(10):113, 1972.

3.    Dickerman,  J. C. , T. D. Raye, and J. D. Colley.  The Petroleum Refining
     Industry.   Radian Corporation, Austin, Texas, May 1975.

4.    Bryant, H.  S.  Environment Needs Guide to Refinery Sulfur Recovery.
     Oil Gas J.  71(13):70-76, 1973.

5.    Environmental Protection Agency.  Compilation of Air Pollution Emission
     Factors.   3rd Edition with Supplements 1-8.  AP-42, PB 275 525, Research
     Triangle Park, North Carolina, 1977.

6.    Groenendaal,  W. , and H. C. A. Van Meurs.  Shell Launches HS Glaus Off-
     Gas Desulfurization Process.  Petrol. Petrochem. Int. 12(9):54, 1972.

7.    Radian Corporation.   A Program to Investigate Various Factors in
     Refinery Siting.  Austin, Texas, 1974.

8.    Sussman, U.  H.  Atmospheric Emissions from Catalytic Cracking Unit
     Regenerator Stacks.   Report No. 4.  Joint District, Federal and State
     Project for the Evaluation of Refinery Emissions.  June 1957.

9.    Reference 5.

10.  Radian Corporation.   Control Techniques for Volatile Organic Emissions
     from Stationary Sources.  Austin, Texas, 1978.

11.  Reference 10.

12.  Reference 5.

13.  Laengrich,  A. R. , and W. L. Cameron.  Tail-Gas Cleanup Addition May
     Solve Sulfur  Plant Compliance Problem.  Oil Gas J. 76(13):159-162,
     1978.

14.  Valdez, A.  R.  New Look at Sulfur Plants.  Hydrocarbon Proc. Petrol.
     Refin. 43(3):104-108, 1964.

15.  Reference 1.

16.  Reference 5.

17.  Hydroprocessing Is Lively Topic.  Oil Gas J.  75(28):153, 1977.
                                     249
-------
W. R. Phillips
18.  Barry, C.  B.  Reduce Glaus Sulfur Emission.  Kydrucctibou Proc. 51(4):
     102-106, 1972.

19.  Reference 6.

20.  Reference 17.

21.  Vasalos, I. A., et al.  New Cracking Process Controls FCCU SOX-  Oil
     Gas J. 75(26):141-148, 1977.

22.  American Petroleum Institute, Division of Refining.  Manual on Disposal
     of Refinery Wastes, Volume on Atmospheric Emissions.  API Pub. No. 931,
     Washington, D.C., various dates.

23.  Rheaume, L., et al.  New FCC Catalysts Cut Energy and Increase Activity.
     Oil Gas J. 74(20):103-107, 110, 1976.

24.  Conser, R. E.  Here's a New Way to Clean Process Gases.  Oil Gas J.
     72(13):67-68, 70, 1974.

25.  Barthel, Y., et al.  Sulfur Recovery in Oil Refineries Using IFF
     Processes.  Adv. Chem. Serv. 139:100-110, 1975.

26.  Processes Clean Up Tail Gas.  Oil Gas J. 76(35):160-166, 1978.

27.  Reference 17.

28.  Reference 26.

29.  More Glaus Cleanup Processes Explained.  Oil Gas J. 76(37):88-91, 1978.

30.  Parthasarathy, R., and L. P. Van Brocklin.  Applications of the Phosphate
     Process to Heavy Oil Refining.  Paper No. AM-78-48.  Presented at the
     1978 NPRA Annual Meeting, San Antonio, Texas, 1978.

31.  Reference 26.

32.  Reference 29.

33.  Conser, R. E., and R. F. Anderson.  New Tool Combats S02 Emissions.
     Oil Gas J. 71(44):81-86, 1973.

34.  Dry Process for S02 Removal Due Test.  Oil Gas J. 70(34):67-70, 1972.

35.  Reference 29.

36.  Ball, F. G., et al.  Glaus Process/Gaseous Wastes.  Hydrocarbon Proc.
     51(10):125-127, 1972.

37.  Reference 17.
                                    250
-------
W. R. Phillips
 38.  Reference 17.

 39.  Reference 17.

 40.  Reference 17.

 41.  Atwood, G. R., et al.  New Integrated UCAP Process Treats Low
      Streams, Trims Emissions.  Oil Gas J. 77(35):111, 1979.

 42.  Controlling SOx Emissions from Fluid Catalytic Cracking (FCC) Units.
      Wet Scrubber Newsletter  (48):6, 1978.

 43.  Reference 21.

 44.  Reference 22.

 45.  Reference 26.

 46.  Reference 26.

 47.  Reference 29.

 48.  Ritter, R. E.  Tests Make Case for Coke Free Regenerated FCC Catalysts.
      Oil Gas J. 73(36):41-43, 1975.

 49.  Coke Key to Cleaning Glaus Unit Tail Gas.  Oil Gas J. 74(47): 142-144,
      1976.

 50.  Doerges, A., K. Bratzler, and J. Schlauer.  LUCAS Process Cleans Lean
      H2S Streams.  Hydrocarbon Proc. 55(10):110-111, 1976.

 51.  Reference 17.

 52.  Reference 21.

 53.  Hulman, P. B., and J. M. Burke.  The Lime/Limestone Flue Gas Desulfuri-
      zation Processes.  Radian Corporation, Austin, Texas, 1978.

 54.  McGlamery, G. G. , et al.  Detailed Cost Estimates for Advanced Effluent
      Desulfurization Processes, Final Report.  EPA-600/2-75-006.  TVA, Muscle
      Shoals, Alabama, January 1975.


 55.  Gibson, E. D., T. G. Sipes, and J. C. Lacy.  The Dual Alkali Flue Gas
      Desulfurization Process.  Radian Corporation, Austin, Texas, 1978.

 56.  Arneson, A. D., F. M. Nooy, and J. B. Pohlenz.  The Shell FGD Process
      Pilot Plant Experience at Tampa Electric.  Presented at the Fourth
      Symposium on Flue Gas Desulfurization, Hollywood, Florida, 1977.
                                      251
-------
W. R. Phillips
 57.   Stern,  R.  D. ,  et  al.   Interagency Flue Gas  Desulfurization Evaluation.
      Radian  Corporation, Austin,  Texas 1977.

 58.   Faucett, H.  L.  Private Conversation.   TVA, 26 May 1977.

 59.   Stover, R.  D.   Control of Carbon Monoxide Emissions from FCC Units by
      UltraCat Regeneration.  Ind.  Proc. Des.  Poll.  Control,  Proc. AIChE
      Workshop 6:80-85, 1975.

 60.   Rheaume, L., et al.   Two New Carbon Monoxide Catalysts  Get Commercial
      Tests.   Oil Gas J.  74(21):66-70,  1976.

 61.   Reference  23.

 62.   Davis,  John C.  FCC Units Get Crack Catalysts.  Chem.  Eng. 84(12):
      77-79,  1977.

 63.   American Petroleum Institute, Refining Department.  American Petroleum
      Institute  Refining Department 41st Midyear Meeting, Los Angeles,
      California, May 1976, Proceedings.  Washington, D.C.,  1976.

 64.   Reference  22.

 65.   lya, K. Sridhar.   Reduce NOX in Stack Gases.  Hydrocarbon Proc.  163,
      November 1972.

 66.   Reed, Robert D.  How to Cut Combustion-Produced NOX.  Oil Gas J.  72(3):
      63-64,  1974.

 67.   Reference  22.

 68.   Wetherold, R.  G.   The Distribution of Selected Fugitive Hydrocarbon
      Emission Sources  Among Petroleum Refinery Process Streams, Technical
      Note.  EPA Contract No. 68-02-2665.  Radian Corporation,  Austin,  Texas,
      May 1979.

 69.   Litchfield, D.  K.  Controlling Odors and Vapors from API Separators.
      Oil and Gas Journal 69:60-62, Nov. 1971.

 70.   Reference  68.

 71.   Reference  68,  Table 7.

 72.   Reference  68,  Table 6.

 73.   Reference  68.

 74.   Reference  68.

 75.   Reference  69-
                                     252
-------
W. R. Phillips
76.  Reference 5.

77.  Los Angeles, County of, Air Pollution  Control District,  et al.  Emissions
     to the Atmosphere from Petroleum Refineries  in Los Angeles County,
     Final Report.

78.  Sittig, M.  Petroleum Refining  Industry  -  Energy  Saving  and Environmental
     Control.  Noyes Data Corporation, Park Ridge, New Jersey, 1978.

79.  California, State of, Air Resource Board,  Legal Affairs  and Enforcement
     of Stationary Source Control Divisions.  Emissions from  Leaking Valves,
     Flanges, Pump and Compressor Seals, and  Other Equipment  at Oil Refineries.
     Report No. LE-78-001.  April 1978, p.  1-2.

80.  Reference 5. '

81.  Jones, H. R.  Pollution Control in the Petroleum  Industry.  Pollution
     Technology Review No. 4.  Noyes Data Corporation, Park Ridge, New Jersey,
     1973, p. 144.

82.  Reference 77.

83.  Burklin, C. E.  Revision of Emission Factors for  Petroleum Refining.
     EPA Contract No. 68-02-1889, Task 2, EPA 450/3-77-030.   Radian
     Corporation, Austin, Texas, October 1977.

84.  Pikulik, A.  Selecting and Specifying  Valves for  New  Plants.  Chem. Eng.
     83(19):168, 1976.

85.  Reference 84.

86.  Wilson, R. T.  How to Pack High-Temperature  Valves.   Hydrocarbon Proc.
     57(1):91, 1978.

87.  Reference 86.

88.  Reference 84.

89.  Greene, Tweed, and Co.  Palmetto Packings  for Pumps,  Valves, Hydraulic
     Presses and Rams.

90.  Perry, R. H., and C. H. Chilton.  Chemical Engineer's Handbook, 5th
     Edition.  McGraw-Hill, New York, New York, 1973,  p. 6-55.

91.  Reference 79, p. V-7.

92.  Frazier, W.  Personal Communication.   Crane  Company,  Houston, Texas,
     18 June 1979.
                                      253
-------
W. R. Phillips
93.  Reference 92.

94.  Hoyle, R.  How to Select and Use Mechanical Packings.  Chem. Eng.
     85(22):107, 1978.

95.  Reference 94.

96.  Reference 94, p. 106.

97.  Steigerwald, B, J.  Emission of Hydrocarbons to the Atmosphere from
     Seals on Pumps and Compressors.  Report No. 6.  Joint District and
     State Project for the Evaluation of Refinery Emissions, April 1958,
     p. 29-

98.  Reference 97.

99.  Reference 68.

100. Center for Professional Advancement.  Mechanical Seal Technology for
     the Process Industries.  East Brunswick, New Jersey, March 1978.

101. Chemical Engineering Equipment Buyers' Guide, 1979-1980, p. 448,
     McGraw-Hill.

102. Potter, Charles.  Private Conversation.  Crane-Deming Pumps, Houston,
     Texas, 27 July 1979.

103. Reference 102.

104. Dickerman, J. D., et al.  Industrial Process Profiles for Environmental
     Use, Chapter 3, Petroleum Refining Industry.  EPA Contract No. 68-02-
     1319, Task 34, EPA 600/2-77-0230.  Radian Corporation, Austin, Texas,
     January 1977, p. 82.

105. Beychok, Milton R.  Wastewater Treatment.  Hydrocarbon Proc. 50(12):
     110, 1971.

106. Reference  78, p.  130.

107. Kanter, C. V., et al.  Emissions to the Atmosphere from Eight
     Miscellaneous Sources in Oil Refineries.  Report No. 8, Joint District,
     Federal and State Project for the Evaluation of Refinery Emissions.
     Los Angeles County Air Pollution Control District, Los Angeles,
     California, 1958.

108. Reference 81, p. 255.
                                     254
-------
W. R. Phillips
109. (Arthur D.) Little, Inc.   Environmental  Considerations of  Selected
     Energy Conserving Manufacturing  Process  Options, Final Report.  Volume
     4.  Petroleum Refining  Industry  Report.   EPA Contract No.  68-03-2198,
     EPA 600/7-076-034d.  Cambridge,  Massachusetts, December 1976, p. 28.'

110. Reference 69.

111. Reference 5.

112. Reference 109.

113. Reference 106, p. 202.

114. Azad, H. S., ed.  Industrial Wastewater  Management Handbook.  McGraw-
     Hill, New York, New York,  1976,  pp.  8-27,  8-28.

115. Reference 69.

116. Environmental Protection Agency, Effluent  Guidelines Division.  Interim
     Final Supplement for Pretreatment to the Development Document for the
     Petroleum Refining Industry Existing Point Source Category.  EPA 440/
     1-76-083A.  Washington,  D.C., March  1977,  p.  67.

117. Environmental Protection Agency, Effluent  Guidelines Division.
     Development Document for Effluent Limitations Guidelines and New Source
     Performance Standards for  the Petroleum  Refining Point Source Category,
     Final Report.  EPA 440/l-74-014a, PB 238 612.  Washington, D.C., April
     1974, p. 101.

118. Hustvedt, K. C. , and R.  A.  Quaney.   Control of Refinery Vacuum Producing
     Systems, Wastewater Separators and Process Unit Turnarounds.  EPA
     450/2-77-025, OAQPS No.  1.2-081.  Environmental Protection Agency,
     Office of Air Quality Planning and Standards, Office of Air and Water
     Management, Research Triangle Park,  North  Carolina, October 1977, p. 3-2.

119. Air Pollution Control Association, ed.   Emission Factors and Inventories,
     Anaheim, California, November 1978.   Proceedings of the Specialty
     Conference, Pittsburgh,  Pennsylvania, 1978.

120. Reference 118, p. 4-20.

121. Reference 118, p. 5-3.

122. Reference 77.

123. James,  E. W., W.  F. McGuire,  and W.  L. Harpel.   Using  Waste Water  as
     Cooling  System  Makeup  Water.   Chem.  Eng. 83(18):95,  1976.
                                     255
-------
J. A. Mullins
                                  REVIEW

                                    by

                               J.  A. Mullins
                             Shell Oil Company
                              Houston, Texas


                                    on

                 REFINERY AIR EMISSION CONTROL TECHNOLOGY
                                  RESUME

         Mr. Mullins is employed as Staff Environmental Engineer in the
Environmental Affairs Department of Shell Oil Company, Head Office in
Houston, Texas.  He received his B.S. degree in Chemical Engineering from
the University of Colorado in 1952.  After military service, he started
with Shell in Chemicals-Manufacturing.  Other assignments have included
design and pilot plant operations.  For the past eight years he has been
involved in environmental engineering, design and regulatory analysis.  He
also serves on several trade organization committees and task groups related
to the environmental area.
                                    256
-------
J. A.  Mullins
                                  REVIEW

                                    by

                               J.  A. Mullins
                             Shell Oil Company
                              Houston, Texas

                                    on


                 REFINERY AIR EMISSION CONTROL TECHNOLOGY
SUMMARY

         This paper had as its objectives to review  the state-of-the-art
of selected process and fugitive emission controls and to discuss available
control technology.  In order to meet these objectives, three selected
process emission sources are described and the emissions are at least
qualitatively discussed.  Because of the vast differences which exist
between refineries and because of their complexity,  it is recognized that
it is difficult to present a single simplified description or evaluation
of emissions.  In general, while there has been presented a lengthy listing
of conceivable controls, we do not believe that there has been a proper
evaluation of those technologies nor their applicability to petroleum
refineries.

         The classification system chosen for control technologies is
misleading to the reader since it leaves one with the impression that many
technologies can be readily applied to reduce atmospheric emissions when
that is not  the case.

         We  would readily agree that the controls listed as "existing" have
been applied in petroleum refineries and are in usage at many locations.
This fact, however, should not be interpreted that such controls are
economic at  all refiners or even necessary to meet ambient air quality
standards.

         The second listed grouping of control technology is incorrectly
titled as "available."  We believe this to be a poor grouping of possible
processes which contains control technologies that have actually been
applied in a limited number of refineries and those which have been con-
sidered but  never applied.  The use of the word "available" implies that
                                    257
-------
J. A. Mullins
all the processes listed could be used at least to some extent.  Although
there are general statements as to cost and sometimes safety, we believe
that the review and discussion given these systems is far too shallow to
justify the classification of "available."

         The third classification of process emission control technology—
namely technology transfer—again implies that the process discussed can be
used in the refining industry.  We disagree with the presumption that
limited success and usage in some application classifies a process for
transfer to another industry.

         The portion of the paper dealing with fugitive emissions and the
techniques for controlling them appears to have more technical backup than
the process'emission section and does a reasonable job of defining the
potential leakage rates of various sources.  The paper, in no way, however,
attempts to place in perspective the relative importance of fugitive
emissions to the overall hydrocarbon emission potential.  As pointed out
in the report, the second and third highest fugitive emission sources in
the hypothetical refinery are based upon old or questionable testing data.
Since these two sources alone would be judged to account for over 35 percent
of the total fugitive emissions, additional studies are certainly indicated
before any major conclusions can be made.

DETAIL COMMENTS

         I would like to point out some of the specific areas where we
believe clarification or correction may be warranted.

Process Emissions

         Although the discussion section of the Glaus process indicates the
presence of hydrocarbon in the unit feed gas, Table 2 fails to indicate that
a typical feed gas contains 1 ± 0.5% hydrocarbon.  This can cause control
problems with a Glaus unit and sometimes results in excessive temperatures
in a tail gas incineration device.

         The discussion of catalytic cracking emissions would indicate that
no new emission factors were determined in this study, and that, in the
previous study, "uncontrolled" emission factors excluded control devices
such as cyclone separators.  Justification is presented to review these old
emission factors because of changes in control technology.  This presents a
confusing picture to the reader and should be clarified.  A second reason
for review of the 1956 emission factors seems to be -a concern that one of
the six tested units may have been atypical, and that, if this unit were
excluded, an average emission factor of under 50 pounds per 1,000 barrels
of feed would result.  Since the excluded factor is 181, we do not under-
stand how the published factor 242 was arrived at.  This entire question
requires further review and clarification.
                                    258
-------
J. A. Mullins
         Because of vast  differences  in complexity of refineries, we  do not
believe that average steam  and  fuel usages  are meaningful.   As  energy con-
servation projects are  implemented these factors will also  change.  Emis-
sions from the boilers  and  heaters are strongly dependent upon  the mix of
fuels used and this also  varies significantly from refinery to  refinery.
The emissions factors shown in  Table  4 are  not in  agreement with those in
AP-42 and the reference for these  factors was not  indicated.

Existing Control Technology

         In the sulfur  recovery discussion,  the inclusion of Glaus tail gas
incineration as a clean-up  device  is  misleading since it only converts the
sulfur compounds to a different form.   In the discussion and accompanying
table for tail gas incineration there are discrepancies which should  be
corrected.  Incineration  temperature  is generally  950 - 1150°F rather  than
the 1200°F stated in the  discussion or the  752°F shown in Table 6.  Tempera-
tures below about 900°F will not convert the H2S and  temperatures above about
1100°F are unnecessary  and  consume large quantities of fuel.  Excess  air is
also more likely to be  100  percent or greater since most of these devices are
natural draft with manual control.

         Although hydrodesulfurization of feedstock to catalytic crackers
is not generally practiced  for  purposes of  SOX contol, it does  occur  in
about 20 percent of the cracking feed in the United States.   We believe that
an estimate of the degree of control  achieved by this technology should be
stated.  It is implied  in the discussion that the  degree of desulfurization
used today is 92 to 95  percent.  This level  is commercially uneconomical
and rarely, if ever, achieved.   More  commonly levels  of 60  to 70 percent are
practiced.

         In the discussion  of FCCU particulate removal, a collection
efficiency for ESP's of 99.5 percent  is stated to  be  a common occurrence.
Again, we believe this  to be rarely,  if ever, achieved in practice.   The
more likely long-term efficiencies are in the range of 90 to 95 percent.

         Control of CO  emissions by CO boilers is, as stated, a fairly
common practice; however, the required combustion  temperature is usually
1800 to 2000°F rather than  the  1300°F indicated.

Available Control Technology

         As discussed earlier,  we  disagree  with the  use of the term
"available" since it implies a  greater degree of assurance  that the technol-
ogy can be used than is justified.  Examples of this  are the UOP Sulfox
Process and the Union Carbide UCAP Process which are  reported as only being
in the testing stage.   The  use  of  many of the alternative methods of  Glaus
tail gas clean-up shown in  Table 8 are also  questionable.   The  Catalytic
Incineration process does not by itself reduce sulfur compound  emissions
and should not be listed  as a clean-up device.
                                     259
-------
J. A. Mullins
         The system of SOX removal discussed for boilers and process heaters
is not realistic.  The concept of ducting all furnace and boiler flue gases
to an integrated sulfur removal facility is purely theoretical and cannot be
justified either technically or economically.  The safety barriers pointed
out in the discussion would only be the "tip of the iceberg" if this
practice were to be implemented.  Even if such a centralized treatment
system were feasible, the suggested processes for SOX removal would not be
the likely choice.

         For NOX removal (or reduction) from boiler and furnace flue gases,
three add-on techniques are suggested with the conclusion that only the
thermal process appears promising.  We question not including combustion
modifications in this discussion.  In fact, we believe it would be more
appropriate if all of the techniques included in the section devoted to
changes in operating practices were discussed and classified as "existing,"
or "available" technology.  We see no reason to consider these pollution
reduction methods in a different light than the add-on techniques described
in the earlier sections.  Many of them may be more cost-effective than the
proposed add-on devices and should be considered.

         Concerning the "thermal denox" process, we believe the stated 50
to 70 percent reduction is optimistic and would, on a long-term basis, be
in the range of 40 to 50 percent.  Since some of the possible combustion
modifications appear capable of achieving near the same reduction, the use
of add-on techniques may not be justified at this time.  Also, the tempera-
ture range cited for the thermal denox process is incorrect by a factor of
10.  The proper temperature range is 1300 to 1900°F.

         In the discussion of technologies that could be transferred from
other industry for SOX and/or particulate removal from boiler and furnace
flue gases, several processes that are either in commercial use or test for
flue gas scrubbing are suggested.  I would not disagree that this may be
possible from a technical point of view, but I question the economics and
reliability of such schemes.  It is stated that the lime/limestone SOX
removal process is highly reliable.  Considering the recent comments of the
utility industry relative to the SOX scrubber requirements under the
revised NSPS for utility boilers, I would question this conclusion.  It is
especially important when you consider that refinery catalytic crackers are
expected to operate with very high stream factors.

Alternative Operating Practices^

         As discussed earlier, I believe that the various techniques
described in this section should be placed, as appropriate, in the sections
of the paper where technology is described as "existing" or "available."
I see no reason to attempt to segregate things like catalyst changes from
add-on scrubbing when considering the potential to reduce emissions and
evaluating the economics of such changes.
                                    260
-------
J.  A. Mullins
         In the discussion of high  temperatures  FCCU  catalyst  regeneration,
it is stated that the CO level  can  be  reduced  to well below  500 ppm.  In our
experience the 500 ppm level is  about  the  lower  limit.  We are not aware of
any catalyst supplier or vessel  designer that  would guarantee  CO  levels of
less than 500 ppm.  A recent Federal Register  notice  also states  that 500
ppm CO is the lower limit for high  temperature regneration.

         The newly developed catalysts which prevent  sulfur  from  leaving the
regenerator offer great promise  for SOX emission reduction.  We would, how-
ever, question the costs cited  as realistic.   The  estimate of  $0.03 per
barrel of feed appears to be taken  from an Amoco study in early 1977.
Escalation of these costs should be made as well as some allowances made
for increased size and operating costs of  the  Glaus unit required to handle
the additional load that will be generated.

FUGITIVE EMISSIONS

         As discussed initially, much  of the emission data for fugitive
emissions appears to be on a sounder basis.  A significant effort by Radian
to quantify the emissions was made  and is  the  subject of another paper at
this conference.  It does concern me,  however, that the second and third
highest fugitive emission losses in the hypothetical  refinery  are stated to
be the two categories of sources that  had  the  lowest  confidence limits of
the study and were therefore estimated on  the  basis of  previous work.  The
recommendation that these two potential sources  be investigated further is
certainly justified.

         Even though a separate  paper  is being presented on  the subject of
fugitive emissions, I believe it should be made  clear in this  paper that
the emission factors given for  a particular type of source represent average
leakage or loss, and that by far the majority  of fugitive emissions are the
result of a relatively few leaking  sources.  It  is this fact which makes
some type of leak detection and  maintenance program a viable alternative of
control.  As was discussed, the  use of "leak-proof" equipment  such as
diaphram values does not offer  a reasonable or even achievable solution to
the problem.

         Of significant interest was the finding that,  despite a  great
increase in the usage of mechanical pump seals in  the past 20  years, the
average emission factor for all  pumps  has  essentially not changed.  This
would indicate that the proposals to require mechanical seals  in  certain
processes would not result in emission reductions  commensurate with the
costs involved.  Double mechanical  seals of the  tandem type  are described.
The other type of double seal operates with the  seal  fluid pressure higher
than the pump suction and leakage is therefore into pump.  Leakage across
the outer seal to the atmosphere will  occur; however,  in limited  cases this
could be a nonhydrocarbon.

         "Canned" pumps for horsepowers up to  250, heads to  3,000 feet, flow
to 50,000 GPM and temperatures  to 340°F are suggested.  We are not aware of
                                    261
-------
J. M. Mullins
any such applications and believe that the few that are used in refineries
are near 30 HP, 200 feet of head, 100 GPM and temperatures of 100°F.  It
may be that confusion exists between the sealless "canned" pump and the
vertical can pump which is sealed by mechanical seals.  We would agree,
however, that general refinery usage of sealless pumps will not occur until
these pumps have met API  standards.
                                    262
-------
W. R. Phillips
                           QUESTIONS AND ANSWERS


COMMENT/Rosebrook - I would like to make a  comment  at  this  time, because I
think something has happened which makes a  point.   Too often  the only infor-
mation available to consulting firms such as  ourselves,  to  enforcement
agencies, to many other types of firms, is  information which we glean from
the literature, and attempt to appraise based on sound engineering principles.
But it is people like Jim Mullins and Shell Chemical,  and Mobil and Exxon
and others, who tell you that it is not 1300°,  it is 1800.  It really makes
a difference in the economics of that control technology and  the effective-
ness of the control technology, if indeed as  opposed to  the literature,
practical experience, day-to-day running in the field,  shows  that instead of
giving 90 percent efficiency they give 60 percent or 70  percent.

Q.  Joseph Zabago/Mobil Oil - Perhaps you can elucidate  Table 13 and 14,
where you are making an attempt to prioritize sources  with  specific refer-
ence again to Mullins' commentary on the oil/water  separators and the cool-
ing towers.  I am not as interested in your answer  to  the question as I am
in applying the question to a paper that will be given tomorrow, where
people will be talking about dispersion analysis for an  entire refinery.
Table 13 and 14 are the first place that I have seen in  the document where
one has taken the whole A. D. Little hypothetical refinery  and turned out a
number.  We got a number based on all those numbers that Lloyd told us about
this morning, and I like them, in terms of what he  has already done, but then
we have wastewater separators and cooling towers and they just blow Table 13
and 14 out of the water.  It is my way of saying that  I  don't think the
tables are consistent, and I think they should be strongly  qualified as one
is based on a study and one is a paper study.   One  is  based on empirical
work and the other is totally a paper study.

A._- First of all, on the oil/water separators, as  I originally said, I tried
to put these emission factors and total emissions from a hypothetical
refinery in some sort of pecking order.  Well,  for  the oil/water separator,
you can see it only fits into the table if you will look at the uncovered
or uncontrolled oil/water separator.  The emission  in  pounds  per hour is
178, but if you look at a covered separator,  which  I am beating the drum
for here, it drops clear to the bottom of the table.   So, it  is there in
                                     263
-------
W. R. Phillips
position number 2 only by virtue of what it will do if it is uncovered.  Now,
as to the data that went into that, those who did the separator simulation,
actually took a body of water and an oil layer and covered it and uncovered
it and took their temperatures and samples and so forth.  We feel like that
is probably better than the original work that went into the AP-42 numbers
represented here as 252 pounds per 1,000 barrels.  Well, anyway, not only
are those numbers lower than the original AP-42 numbers, but I think they
were done in a controlled manner.  As to cooling towers, the values which I
show you here for Table 14 include both the values that we got using our
purge method and our TOG method.  I don't believe it appears in your copy.
The original order here was established by the values given by the TOC
results for cooling towers of 234.  It was only after going through the data,
going through the procedures, looking at both the precision and the accuracy
of the two methods that we concluded that the lower values based upon the
purge method were what we should go with.  Now, obviously in that case that
would drop cooling towers way down in the table.  So, your point is well
taken.  Those two, if they are covered, in the first case, and if we use the
purge method in the second, which we feel is the more reliable of the two
numbers, then certainly those two sources of emissions are out of the running
for being important.  Incidentally, we did take a look at rough economics of
covering the API separator.  It is a runaway first choice for doing what was
mentioned earlier today, namely holding onto some of those good hydrocarbons
that you really don't want in the air, that you would like to recycle into
your refinery.  There is a high return I won't mention it because someone
will hang his hat on it.

COMMENT / Joseph Zabago/Mobil Oil - Thank you!  The 234 is not included in
our copies.  That clarifies my total confusion in figuring out why number 3
was the lowest, I now understand that and why a number of cooling towers
should be, say 2.04, put at the bottom of the priority list, and I appreciate
your clarifying the business on the oil/water separators.  The commentary
about the economics is still premature pending the data that we determined
from studies that were talked about earlier.
                                     264
-------
R.  L. Honerkamp
                          CORRELATION OF FUGITIVE

                        EMISSION RATES FROM BAGGABLE

                           SOURCES WITH REFINERY

                             PROCESS VARIABLES

                              R. L. Honerkamp
                             Radian Corporation
                               Austin, Texas


                                  ABSTRACT

         The  effects  which refinery process variables have on fugitive emis-
sion rates  are  discussed in this paper.   Correlations are presented for both
continuous  and  discrete process variables.   Continuous variables include
temperature,  pressure,  size,  age, etc.   Discrete variables may relate to
function,  type, manufacturer, etc.   The only dominating correlation observed
was for stream  composition;  components containing higher volatility process
fluids tend to  have higher emission rates.   Factors which affect the observed
correlations  or lack  of correlation are also presented.
                                   RESUME

          Russell  Honerkamp is a Staff Chemical Engineer at Radian Corporation.
He received his  B.S.  degree in Chemical Engineering from the University of
Texas at Austin  in 1975.   Since 1976, he has worked on several projects at
Radian pertaining  to  atmospheric emissions of VOC from industrial processes.
He is currently  working on a contract with EPA to support and develop New
Source Performance Standards for fugitive emission sources in the Synthetic
Organic Chemical Manufacturing Industry.
                                      265
-------
R. L. Honerkamp
             CORRELATION  OF FUGITIVE  EMISSION  RATES  FROM BAGGABLE

                   SOURCES WITH REFINERY PROCESS VARIABLES
  INTRODUCTION AND  OBJECTIVES

            Refinery  process variables  can be  categorized as  continuous  or
  discrete  variables.   Continuous  variables exist  over a range with common
  unit(s) of measure  which  describe  the range.   Continuous variables include
  pressure,  temperature,  size,  age,  capacity,  etc.   Discrete  variables include
  function,  type, manufacturer,  configuration,  etc.   Discrete variables  must
  be  compared individually  and  no  interpolation between levels of the variables
  is  possible.   Some  variables,  such as stream composition, can be treated as
  either  continuous or  discrete depending  on how they are defined.  The  objec-
  tive of this paper  is to  present and  explain the correlations observed in
  this study between  process variables  and fugitive  leak rates.

           There are  two primary reasons for investigating the effects of
 process  variables  on fugitive emission rates.  The more important reason is
 to provide useful  information for developing  fugitive emission control strat-
 egies.  If certain classes of emission sources do not leak due to the effects
 of process variables,  these sources can be excluded from emission reduction
 maintenance programs.   If  changes in certain  process variables increase
 emission rates  significantly,  more  intensive  maintenance programs might be
 indicated  for  the  higher emission rate sources.  Alternatively, changes in
 design,  construction,  or operation  of these source types could be applied to
 counteract the  emission rate increase caused  by the process variables.   The
 second reason  for  examining these correlations is that numerous correlation
 theories were  proposed at  the outset of the sampling program.  Many people
 felt that  these effects would be borne out by sampling data "without a doubt".
 Thus far,  the  surprises have outnumbered the  expected <_= t-r-v. L s .


           One  constraint which complicates the effort to relate emission
 rates and  process  variables is the  accuracy of recorded  process variables.
 A large  number  of  process  variables were collected for each source screened
 or sampled. (Figure  1).   The amount of time and effort that could be expended
 in collecting  these  variables  was limited by  the schedule and budget, and
 sometimes  the  information  simply was not available.  Age is a good example
 of this  problem.   The  "best"  age to record for a source would be the age
 since maintenance  or installation.   But often the only estimate of age
 available  was  the  age  of the process unit.  FUJI a large number of diverse
 sources  (especially  valves)  It was  impossible to obtain accurate process
 variable information.
                                      266
-------
R.  L. Honerkamp
                                        DATA SHEET  -  PUMP  SEAL
                                                                                          500
        1. Radian ID//	
                     I   23   45   6  "7   8~
        3. Refinery ID#
                    9
                          I - Inboard
                          0 - Outboard
                                                 Note:  Space 5  should be seal identification
                                                        letter (A,B,C etc.). Use sane ID on
                                                        sampling  sheet.
                                                       In-Service/Out  of Service
                                                        10
        VARIABLES:
        4. Discharge pressure, psig

        5. Temperature, °F

        6. Puajp/seal type

        7. RPM or strokes PM
                              11  12 13  14
        8. Stroke length (Recip, in)
                                    26  27 28  29
        9. Capacity; GPM
                .    IF - product leakage   "I
       10. Seal/lube w-v.cer
                    JH - bydrccarbcra lubricaatl
                                 30 31  32  33  34
                                                35
        PROCESS FLUID DESCRIPTION:
       18. Name
                                               1 N - no quench gland]
                                 11.  Gland type f - mi q-^nch
                                               |^V - water quench  _j
                                 12.  Single or double  (S, D)

                                 13.  Shaft diameter,  in 	

                                 14.  Age,  yrs

                                 15.  Manufacturer
                                              16.  Ktls  of  constr
                                              17.  Horizontal  or vertical (H, V)
                47  48 49.' 50  51  52  53 54  55  56
        SCREENING DATA:
       19. Date of screening

       21. Max TLV

       23. TLV data
                                    -1	L_
                                              20.  Screening  team
                       57  58 59  60   61  62
                                         |1 22.  Liquid  leak?  (Y, N)
                                                                      63 64
                       65  66  67 68  69  70
                                                                         71
                        DATA-
                1C
                    '   '	1	1	'
Eadian IH'J	
           'T  £   3.   4  5   6   T
Screening Concentration,  ppm
                                                 Screening
                                                      Date
                                                              i    '    '	I	L
                                                           9   10   11. 12  13  14
                                                              |Screening
                                                              I     Team
                                                                                           15   16
                                                                                i   t   '	I	L.
         5 Cm
17 13 19 20 21  22

  ~.   .   .   .  I
                                      23 *24 *25 26  *27 28  29 30 31 32  33  34 35  36 37 33 39 40
                       >  t7 68
                                                                                      [SJ
                                                                                      60
               Figure  1   Process  Variable  Data Collection  Sheet
                                                 267
-------
R. L. Honerkamp
 CONCLUSIONS

           The only dominating effect on emission rate that has been observed
 is related to stream composition.   This effect became evident in the early
 stages of the program.   One problem with this effect was related to the units
 used to describe stream composition.  Refinery streams are primarily multi-
 component hydrocarbon mixtures with wide variations of vapor pressure, molecular
 weight, chemical class (paraffins,  aromatics), viscosity,  etc.  Since stream
 composition changes numerous times  within a single process unit, it was im-
 possible to determine complete stream composition variables for each emission
 source screened or sampled.  The most available type of stream identification
 available was the "stream name", such as atmospheric overhead, debutanizer
 bottoms, reformer reactor outlet,  etc.   The actual composition of streams
 with the same name may vary considerably between refineries depending on crude
 composition,  desired products, operating conditions, and other factors.
 Radian developed a stream identification code system to categorize streams
 based on the most volatile class present in >_ 20 weight percent.  (Figure 2).
 Analysis of emission rates showed three distinct "stream classifications".
 Highest emission rates were observed for sources containing gases or vapors.
 Lower emission rates were observed  for sources containing  light liquid or two
 phase streams,  and sources in heavy liquid service had the lowest emission
 rates.  The split between "light" and "heavy" liquids is approximately between
 heavy naphtha and kerosene.  This  corresponds to a vapor pressure of about
 0.1 psia @ 100°F.  Examination of  other process variables  was performed after
 separating stream categories in order to separate the stream composition effect
 from effects  of other variables.

           For all other process variables, no major significance was observed.
 In some cases a statistically significant correlation coefficient was ob-
 served, but no dominating effects,  other than stream composition, were seen.
 The lack of significant correlations may be due to the dominating effect of
 stream composition, the inaccuracy  of measuring process variables,  the
 variability of  leak rates and measurement techniques,  or combinations of all
 of these.   The  only conclusion that can be made is that stream composition
 is the only dominating  variable. Other correlations either don't exist, or
 the data base isn't accurate enough to  identify them.

 PRESENTATION  OF RESULTS

 COMPLICATING  FACTORS

           There are several peculiarities of fugitive emissions and/or this
 particular data base which make it  difficult to draw conclusions about ob-
 served correlations.   These factors include:

              •   The extreme skewness  of the leak rate data with a large
                   percentage of most source types not leaking

               •   The variability of leaks and leak measurement techniques
                                     268
-------
R.  L.  Honerkamp
  STREAM GROUP1

  Gas/Vapor
  Light Liquids/Two-Phase
  Heavy Liquids
HYDROCARBON STREAM DESCRIPTION2

Ci-Ca Hydrocarbons
Ca-Ci* Hydrocarbons
Cg-Cg Hydrocarbons
Ci o+ Hydrocarbons
Mixed Molecular Weight Hydrocarbon
      Streams
Aromatic Hydrocarbons
Miscellaneous Organic Compounds
Hydrocarbon Streams Containing H2 ,
      and H20
C2 Hydrocarbons
Cs-Ci* Hydrocarbons
Cs-Cs Hydrocarbons
C?-C9 Hydrocarbons
Naphtha
Light Distillate
Aromatic Hydrocarbons (low molecular
      weight)
Miscellaneous Streams

Kerosene, Diesel, Heating Oil
Gas Oils
Atmospheric Resid/Vacuum Gas Oil
Vacuum Res id/ Asphalt
Aromatics/Polymers
Mixed Molecular Weight Streams
Non-distillate Solvents
Miscellaneous Organic Streams
 'Stream group is determined by the stream conditions within the process lines
 The most volatile stream component present at a concentration of 20% or  m
 determines the stream classification.
              FIGURE 2,  PROCESS STREAM CLASSIFICATION BY GROUP
                                   more
                                      269
-------
R. L. Honerkamp
             •   The dominating effect of stream composition on
                 leak rates

             •   The inaccuracy or unavailability of process
                 variable data

The effects of the first two factors listed above were minimized by  trans-
forming the data to logio.  This normalized the data and gave homogeneous
variability for all levels of logio leak rates.

          Any discussion of the effect of process variables is complicated by
the confounding between variables in the data base.  This  confounding is due
to the lack of independence between process variables as they naturally occur
and the fact that all combinations of levels of many variables could not be
obtained in the study.  A fractional factorial experimental design was fol-
lowed in selecting sources with selection based on key process variables.
This design allowed the estimation of the main effects of  important  variables,
but not all variable interaction effects could be estimated.  Most second
order interactions (such as stream type by line size, by source type) and
higher order interactions are either confounded or there are not enough
replicate data to quantify by their effects with any precision.  This means
that it is difficult to break sources down by more than two variables at a
time to determine emission factors or effects.

         A good example of the difficulty introduced by the distribution of
leak rates is the effect of line size on flange emission rates.  The percent
of flanges leaking, mean leak rate, and emission factor estimate are shown as
a function of line size in Figure 3.  Although there are significant differ-
ences in percent leaking and a significant effect of line  size on leak rate,
the confidence intervals for the five emission factors all overlap.  Since
this effect of overlapping confidence intervals occurs for many other source
and variable interactions, comparison of emission factors  is not a good way to
determine significance of process variable effects.   For continuous variables,
the simple correlation coefficient "r" is an indicator of statistical sig-
nificance of the correlation.   Discrete variables can be compared visually
by preparing "box and whisker" diagrams.

CONTINUOUS VARIABLES

          At the beginning of the sampling program, several trends were ex-
pected to be present.   The process variables which were expected to  show
greatest significance were:

              •   Pressure

              •   Temperature

              •   Size

              •   Age
                                     270
-------
R. L.  Honerkamp
           15.0
           10.0
           5.0
           0.0
      ^ t/l   J A
      — uj  -4.0
        is
      Ji
           -5.0
           -6.0
                    I
                                                       2.96 Overall
                                                       Percent
                                                       Leaking
In (Leak Rate) • -6.69 + 0.103 -(Line Size)

Correlaflon Coefficient (r) « 0.34
Standard Error of Estimate « 0.87 In(leak),
                                                       x - estimate

                                                       I - 35% confidence
                                                          interval
                  2468

 Njnber Screened   2471 1189  |   340
                  LINE SIZE (INCHES)
             10  12  14  16  18  20  22   24  26  28  30
                    86
                                      118
     Figure  3   Effect of  line  size on  emissions
                    from flanges
                                            271
-------
R.  L. Honerkamp
 Increasing pressure might provide increasing  driving  force  for  emissions
 through the sealing element.  Temperature extremes might  adversely affect
 the degradation rate of sealing elements.  Larger sizes would be  expected  to
 have larger potential emission area, and therefore greater  emission rates.
 Age or time since last maintenance was also expected  to result  in increased
 degradation of the sealing element.  None of  these expected  results has been
 conclusively determined from the data base.

          The inaccuracies in determining some of the process variables reduce
 the sensitivity of the correlation analysis.  For instance,  the variable "age"
 recorded was usually the age of the unit.  A more useful  age determination
 would have been the years in service of each  individual source, or  possibly
 the time since last maintenance was done, but it was  impractical  to obtain
 this information for the large number of sources studied.  Therefore, the con-
 clusions concerning process variables pertain to the variables as measured
 or determined in this study.

          Table 1 lists the simple correlation coefficients  between the log
 leak rate and the appropriate continuous process variables  for  each source
 type and stream classification.  Correlations significantly  different than
 zero are noted.  The simple correlation coefficient is a  statistical measure
 of  the linear relationship between  two variables.  The correlation between "X"
 and  "Y" is  computed as:
                             (X.-X)2 Z(Y.-Y):
and is bounded:     -1 < r   < 1.
                          XY


          The value of r2 indicates the approximate percentage of the total
variation in the log leak rate that is accounted for by the relationship of
the leak rate with the correlating variable.  For instance if r = 0.50, then
r2 = 0.25 and about 25 percent of the variation in the leak rate is attri-
butable to the relationship with the process variable.  The remaining 75 per-
cent of the variation is due to other variables and random variation.

          The sampling distribution of values of r is highly dependent on
the sample size.  Small values of r (0.1-0.2) may be statistically significant
for large sample sizes while large values of r (0.4-0.7) may not be signifi-
cant for small sample sizes.  Statistically significant refers to a statis-
tical test of the hypothesis that the correlation is equal to zero , i.e.,
no relationship between the variables.  A significant correlation therefore
does not imply a large value of r, since values of r < 0.2 may be significant
for large sample sizes.
                                     272
-------
to
§
TABLE 1 CORRELATIONS BETWEEN CONTINUOUS VARIABLES AND LOG.n LEAK RATE £
10 03
— i
Pressure
Valves
Raa/Vnpor Streams .230*
Liglit Liquid Ktrcamr, .103*
Heavy Liquid Streams -.351*
Hydrogen Service -.088
Open-Ended .236
Pump Senls
Light Liquid Service .0(18
Heavy Liquid Service .097
Flanges .072
Compressor Seals
Hydrocarbon Service .346*
Hydrogen Service .398*
Drains -
Relief Valves .045
Temperature Ap,n

.077 .263*
.051 .096
.144 .220
.129 -.531*
.242 .230

-.012 .062
-.098 .237
.021 -.180

.218* .105
.312* .052
-.408*
.096
LI HP Slroko
Size DJnmoter Aren RTH Cnpnc.Jty Lond Lcnp.tli

.150* - - - • -
.143* ______
.046 -
.288* ______
-.078 ______

.021 - -.064 -
.128 - -.182 -
.336* ______

.278* - -.143a -.138 -.087 -.012
.343* - -.034 .218 -.099 -.074
-.039 -.191 _
-.075 ______
                 * Correlation Coefficient statistically different from zero (P >  .90).
                   Log ifl  RPM was correlated with login  leak rate.
-------
R. L. Honerkamp
          The correlation coefficient, r, can  sometimes be misleading  for the
following reasons:

          •  r does not describe how much Y changes  for
             a given change in X, what the shape of  the curve
             connecting Y and X is, or how accurately Y can be
             predicted from X.

          •  A correlation between X and Y may be due to  their
             common relation to other variables.

          •  Outliers and highly skewed data can distort  the
             frequency distribution of r.

          •  Selecting values of X at which Y  is measured can
             distort the frequency distribution of r.

          •  r may be unduly high because of sampling from two
             different populations instead of  one.

          In order to examine the actual data  used in calculating the cor-
relations presented here, scatter plots of the log leak rate data (in pounds
per hour) and the process variables were developed.  Several of these plots
have been selected to illustrate the "best" correlations observed.  (Figures
4 through 18).  Each plot selected shows a correlation that was considered to
be statistically significant, and although the correlations are statistically
significant, the data show a lot of scatter throughout the range.  Each plot
shows a line representing the mean value of the correlation.  A one order of
magnitude change in leak rate is indicated by  two solid dots on the line.  In
all cases, the variation of leak rates at given values for the process vari-
able is at least one order of magnitude, while the variation in mean leak rate
across the entire range of the process variable is often less than one order
of magnitude.  These plots represented the process variables which showed the
most significant correlations.  All other continuous variables for all other
source types showed more scattering and less statistical correlation.
Several examples of this scattering are shown  in Figures 19 through 25.
These figures show how the expected correlations were not observed in the
data base.

DISCRETE VARIABLES

          Unlike continuous variables, correlation coefficients are not
easily interpreted for discrete variables versus leak rate.  A visual method
for comparing the relationships between levels of the variable and leak rate
is the schematic plot.  On each plot, the level of the variable is represented
by a "box and whisker" figure that identifies  the mean, median, upper  and
lower quartile and range of values.  Because of small sample sizes and over-
lapping values, most of the correlations with  discrete variables  are  not con-
sidered to be significant.
                                     274
-------
                                                 A = 1 UO.S, H r: 2 UllS.  LT<-.
                                         Correlation Coefficient (r) = 0.230*
                                         Number of Data Pairs       = 157
      t
      t
      t
L   i) «•
U     /
G     /
1     /
it  -1 »•

L     t-
t     f
A  -2 +
K     X
   t
-.5 »
   /
   /
   /
-•» +
                                   AA
                           AA
A<- A
A«- A
       A
  A    A   A
   A      A
      A     A     UAA U
A   A A       A       A
 «  AA A  AA      AHA
  (1A	
            UAA
i\   tl       A A   U   A
 A 0 C     A     A   A
 U  I)     A       A
A A »A      A  A
 U  A    A  A  A
  />      A          A
     A    A
    AH                 A
                                            AA
        A       A
            A
          A
        A  A
                                                                       A
                                                                       A
                                       A   A
                                       A
-f.
     . - > - •
      -V»r
                .4	i	|	
                (I    b I    1UU
                                   200   
-------
L
0
ti
1
0
H
A
t
C
    •f »
   -1
L     /
L     t
A  -2  +
K     /
-3 *•
   *
   *
   t
-14 +
   X
   *
   t
-5 t
   #
   /
   t
"6 »
             «AA
               Mi
                                          LEKLNO: A = i  ousi 0=2 uus» trt.
                                          Correlation Coefficient (r) « 0.150*
                                          Number of Data Pairs      =156
                                                                                                       TO
                                                                                                       t-t
                                                                                                       7?
                        A
                        A
                        A
A
A
            0     3     h     9   1?    Ib    10
                                              ---4..
                                               21
              .4	4.
              >•»    ?7
	+	4
  36    3V

             Figure  5  Leak  rate  vs.  line  size - valves, gas/vapor streams.
-------
                                                   A = 1  (|US, II =
                                           Correlation CoefflclRnt (r) = 0.263*
                                           Number of Data Pairs      = 82
                                                                                                              EC
                                                                                                              O
                                                                                                                 7?
                                                                                                                 P
L
0
G
L
t
A
K

H
A
1
E
 0 +
   t
   i
   t
-1 »
   t
   t
   t
-? *
   t
   *
   t
-.} *
   -i) »
   -b +
      #
      t
      i
   -6 *
•t --- + --- + --- 4 --- f ---
 3   5   7   ?   11  1
                                  --- 4 --- 4 --- 4---4 --- 4 --- 4 --- 4 --- 4
                                    15   17   19  21  i?3   *5  i!7  ^V

                                                   (YEARS)
                                                                       --- 4---4 --- > --- 4-
                                                                        JJ  3S  37  iV
                   Figure 6   Leak rate vs.  age  -  valves,  gas/vapor streams
-------
-4

00
1 »
                                      LElitNO:  A  = 1 UWS,  H = 2 OuS, LTC.

                                     Correlation Coefficient (r) = 0.103*

                                     Number of Data Pairs      = 331
                                                                                                                  o
                                                                                                                  3
                                                                                                                  fp
                                             'IP (I
                                                                               ---+-•
                                                                                1320
                                                     PHtSSUML (HSI6»
                Figure  7  Leak rate vs.  pressure - valves,  light liquid/two phase  streams
-------







L
0
G
1
0
L
t
A
K

K
A
T
C




t
I +
t
0 +
_ t
t
*
-1 »
/
i
t.
-2 +
t
#
-3 +
t
t
t
-H +
<
'
-5 +
<
-b *
*
-7  +
                                            A = I
                                    Correlation Coefficient (r)
                                    Number of Data Pairs
0.143*
3Z6
                                                                                                   CD
                                                                                                   "O
                                                 10
                                                         .- + .
                                                          12
                                                                  m
                                                                           it.
                                                                                    10
 Figure  8   Leak  rate  vs. line size  - valves,  light  liquid/two phase  streams
-------
   -O'.f,
   -1.2
              A
              A «
                                                                A = 1 OHSt  H =  if UnSi
                                                        Correlation Coefficient (r)  =-0.351*
                                                        Number of Data Pairs       » 32
                                                                                                          I
                                                                                                          CD
N>
oo
o
L  -1.0  4
0       t
G       y
1       /
n  -;>.n  4
             L
             L
             A
             K
t
t
   -3-0
   -M.O 4
        »
        /
        y
   - 5. 'I +
             ,f	^	4..
             II     fcSO    r)0ll
                                             750   1000   12t>0   IbOU   1 ^5U   ?OUO   2y-i«   ?300   27SO   300U
           Figure 9   Leak rate vs. pressure  - valves,  heavy  liquid  streams.
-------
L
0
G
1
0

L
E
fl
K

H
A
T
t
                                                   «  =  1 ()IJ8i t) = 2 
                                                                                                                     i-i
                                                                                                                     7^
                                                                                                                     1
         l.nn
'I. SI)
O.UO
---+-.
 9.'5
11.50   13. ?
lb.00
• -_ +	
16.75
18.50   20,25  22.00  23.75
             Figure 10   Leak  rate  vs.  line size  -  valves,  hydrogen streams.
-------
NS
00
NJ
L
0
G
1
0

L
t
A
K

H
A
T
                  ii.n
                 -l.fe
-3.2 4
     1
     t
     1
-'l.O •»
     1
     1
     t
-'».»> +
     1
     t
     1
-3.6 +
                 -fp.'t
                                     A
                                     H
                                                            LE.bt.riU! A = 1 ObSt H = 2 OnS«
                                                            Correlation Coefficient (r) • -0.531*
                                                            Number of Data Pairs      * 33
                                                                                                                               (D
                                                                                                                               1-1
                                                                                                                               7?

                                                                                                                               1
                                     2    'I    f,    0   10   12   1't   16   10  20  22  2'l  *f>  20  -id  32  3«l
                                                                     L (YEARS)
                               Figure  11   Leak  rate vs.  age -  valves,  hydrogen  streams
-------
                 1.6
                                                            :  A  = l.«»S«  » = 2 UPS,
                                                      Correlation  Coefficient (r) • 0.346
                                                      Number of Data Pairs      « 10Z
                                                                                                     I
N3
00
             L
             0
             (4
             1
             a

             L
             t
             A
             K
   u.n 4
   o.o 4
     •  1
       i
       t
   -u«8 +
   -1.6
n  -«

l
L
                                 rtA
                                                                                                                   I
                -'••0  4
                     i
                     i
                     t
                -'(.«  f
                     r
                                    eo
                                                                               '([10   '110
                                                                            •-4 -•
                                                                            •too
                                                                                                    bbU
                Figure  12  Leak rate vs.  pressure  - compressor seals,  hydrocarbon service
-------
                                                        A = I  (IMS, M = 2
                                                  Correlation Coefficient (r) = 0.218
                                                  Number of Data Pairs     « 102
oo
          t.
          0
          o
          1
          II

          L
          L
          l\
          K

                                                                                                             o
                                                                                                             0
                                                                                                             (D
                                                                                                             i
            Figure  13  Leak rate  vs. temperature - compressor seals,  hydrocarbon service,
-------
                                                     Correlation Coefficient (r) = O.Z70

                                                     Number of Data Pairs      • 00
IS3

OO
                                                                                                                  Pi
                                                                                                                  o
                                                                                                                  I
3.0
                                                         A . b       i« , (I




                                                          ILK (1NCMKS)
'•.5
                                   S.O
5.5
t>.0
                Figure  14  Leak rate vs.  diameter - compressor seals, hydrocarbon  service.
-------
to
oo
L
U
0
1
0

L
t
A
K

H
A
I
L
                  n.o 4
                     y
                 - U • <> 4
                  1.2 +
                     f
                     t
                     i
                  1»R 4
                     *
                  2.M 4
                 -3.0 4
                 -3.6 4
                                                              ft a  \ 0»Si  U = 2
                                                       Correlation Coefficient (r) » 0.398*
                                                       Humber of Data Pairs      = 62
                                                                                                                en
                                                                                                                o
                                                                                                                0
                                                                                                                n>
                                    H

                                    A
                                                                                       souo
                                                                                   ..-.-+--
                                                                                     3bUU
                  Figure  15  Leak rate vs.  pressure  - compressor seals,  hydrogen service.
-------
                 1
             il. 0 *
                                          Lt.bt.NU:  A  = 1 (»MS«  H  = 2 OHS»  t|t,
                                          Correlation  Coorricicnt (r) • 0.311
                                          Number of Data Pairs      = 59
to
            -o
       *
   -1.2 +
L      /
0  '    /
G    ,  1
1  -l.B *
0    .  t

L      t
t  -2.4 »
A      X
K    -  *
       t
K  -J.O »
A      *
 r      /
t      /
                                                                                                                      -I
                                                 0

                                                 A
                            .«!    00   96
                                                                            4-00
                                                                                       2MO
                                                      I tMPtl
-------
            -1.2
         L  '
         0    "
         G
         1  -JUH
         0
L
E
A
K
            -2. "I  *
1-0
00
00
M
A
T
c
   -3-0
            -3.6
            -H.2 4
                                                          I! A =  )  UHS, M = 2 U|1S«  ML,
                                                    Correlation Coefficient (r) • 0.343
                                                    Number of Data Pairs      « 27
                                                                                                                       8s
                                                                                                                       0
                                                                                                                       0>
                                                                                                                       f-t
            -ll.f.  (
                                                                                                              I
            -'1.0 +
                1

                    V. . I) It
                                      3.00
                                                   3.50
--«•--	4--
'I.UO    1.
  4.	_.-4.__- — -4.-.
»»,50   H.75   5.«0
               Figure 17  Leak  rate vs.  diameter  -  compressor  seals,  hydrogen service.
-------
                                                      Lf. GLNU: A = 1  (JUS. b = ? U»S'  tfC-
                                                      Correlation Coefficient (r) « 0.336*
                                                      Number of Data Pairs      = 60
00
L
0
G

1
0

L
E
A
K
R
A
T
E
                                                                                                                          TO
                                                                                                                          H


                                                                                                                          I
                                                       LlNt.
                                                                 (1MCMLS)
                                  Figure 18   Leak rate  vs.  line  size -  flanges
-------
NJ
*£>
O
/
1 +
t
1
1
0 ^
t
t
t
L -1 *
0 t
G 1-
1 t
0 --2 +
*
L /
L /
A -5 •»
K t
f
U /
A -'« +
T /
t /
1
-5 +
/
t
t
-t, *
f
-7 *




A
A
U
t.
• 	 U-
A 0
C
L
(\
A






n






A
A
A
t
U
l\

-U—
A
A
11
A
II


1)









Jl


A
II
t
n
L
—A- 	
U U
A
n
C «













5 7
uurreiai ion v
Number of Dat
A
A A
A A
t A
AW U
A A n A t
U L U
— 	 	 	 TI 	 0 A
U 13 A
ti
U
A A
tt A
U A
A



n






V 11 U 10 17 19 *i
.Uf 1 1 H. ! Ulll \I / " V.VJV
:a Pairs = 181
A
A
A
A
U B
ABA
A A D
A U A A
ti A 	 — 	 6 	 — 	 ~ 	 ~ 	
A A
AD C
A A
A B
A B
A U A
B
A
A









26 Kb i!7 i!V 31 33 35 37 iV MI
                                                                                                   1
                                                    AOL (YEARS)
                Figure 19  Leak  rate vs.  age - valves,  light  liquid/two phase  streams
-------
to
                        A
              A
             II AH
                A
             A   M
             MC  A H     A
                     A A   A
                                      A

                                    A A
2 t
      1
    1  4
      t

      t    A  A A
    0  4 A   A ft
L     *   A  A
0     /A   AAUA
G     /A CTtA AA CUHAAA U  A
1  -1  4      nnCAB  u&n AUAU
0     «AA A A MA AOHt C A A 0
      /AA A tAA A  OUAA  AA U   A
L     fAHAl'ACAH AtlAAHUAA AAA A
1-24    AMtAA A t HA  OA
A     t    A  CHA «AA ARCO AA
K     JAMAA  1) AA  A A     AA

K  -3 +      HA A   ft         A
A     t         A  UCA       A
T     t      A  A    AAA
L     t  A AA
   -M 4                    A
                                              A
                                              A A
                                             A A
                                                  A
                                               A AB
                                                A
                                                A A
                                                 B
                                                                A = 1 OUSi » =  2  U|)S,
                                                        Correlation Coefficient (r) »  -0.012
                                                        Number of Data Pairs       =  291
                                                                                                                         (0
                                                                                                                         I-J
                                                                                                                         7?
                                                                                                                         I
      <
   -S 4    A A
      t
      t    A
      t
   -6 +
      *
      -4	1	
       o      jf,u
                  Zd     MOU
                                                faMO     000
                                                               V6U    1120
                                                                              1
-------
N3
                                                   iLfctwo: A = i  o»s. a s v <>,,s, LT*-,

                                                   Correlation Coefficient (r) • -0,012









L
0
b
1
0

L
t
A
K

H
A
t
L











t
* *
t
t
1 +
t
i
t
0 *
/.
t
t
-I +
*
t
t
-2 *
t
t
t
-3 +
t
/
*
-i| +
i
t
t
-5 +
t
•t
t
-6 +
t

v v * i u i a v ' un t*u u i i i ^. i T; 1 1 1. \i/ ~ w ( v 1 1
Number of Data Pairs • 294

A
n
A
A HC A
AAA A
AA IIAA A A A
A AAA H A A A A A
A A CtU AAA A A
A CA UtILU AAA8A A A A
A A OG A C A AAA AAA
A A DUO AA A A A AA AA
UCAUAtAA A AA A BA «
A l\ CCfl UHA AA A AAUA A A A
A A AAU'lUB A A A AAA A
A A 0 C UAU A A 0 A . A
AU tAA A A A A A
A f I-AU A A A
A If A A A A
AAUDA A A
OAA A
II A
A



A A

A



-f.u i) bo i?u jnn z'io 300 4&o 'tyu MHO SMO 600 f-^o 720 /ou
                                                                                                                o
                                                                                                                0
                                                                                                                n>
                                                                                                                I
                                                    TtMI'LKATUKt I ^ I
                 Figure  21  Leak rate vs.  temperature - pump seals, light liquid  service,
-------
OJ
                                                           A = i cms. u =. a unsi tn-

                                                    Correlation Coefficient (r) • 0.062
t
? +
t
t
1
1 *
*
/
0 *
u t
o /
b *
i -i *
o t
f
L #
e. -2 *
A t
K *
/
» -3 «•
A t
T t
E /
-'i t
*
*
-0 +
*
*
*
-fc +
*

Number of Data Pairs = 140


A

A A
« U
A A A
A A A ti
n A A A A B A A
AAAA AA AD H
t H B A A B A
A 11 A A BA n A U
A t) A A H B A A
AAAl) BA tAt
U A A A A A
AC AA B A A AH
A A A B B U
rt A B U A A
A A U A
A AAAA
A A




A











A

A "
A

A A


A A


B

A


A









	 + _, 	 4. 	 . 	 -_-«.
                                                                                                                     o
                                                                                                                     3
                                                                                                                     n>
                       O.'j   3.5   fc.b   9.5  12.5  lb.5  l«.5  21.b  ?M.5  27.5  3".5  53.5  36.5  3V.5



                                                           AGL (YEARS)
                       Figure  22   Leak  rate  vs.  age -  pump  seals, light  liquid service
-------



L
0
(,
I
I)

L
L
A
K

H
A.
T
E









*
2 +
t
t
I +
*
1
0 +
*
t
t
-1 *
it\
1
to
-2 «•
t
*A
*A
-3 +
/
t
t
-t +
t
-5 +
1
t
t
-f, *
1
0.8
A
A
l\
U
U
I-
L>
^
i
^
j
u
n
L
1
A
U
t
IV


«






A



o
K
t
n u
F
C
U
C
A
A












1.4
feV 1 1 C 1 <
Number
A
u

c
u
u
G
c
u
t
J
e
i
F
G
H
t
A
A








2.0
ni IUH \AJC i i i u ten t \ i
of Data Pairs
A
A
A
u
A
C
L
U
U
1)
C
c

A
B
U

A








2.6
f — V * UC. t
= 295
A
A
A A
A
A
t) B
C
U A B
n A
H A
A
C A
U
A
A A
A

A

A

A

A



3.2 3.8
                             UlAdtTtH
Figure 23  Leak rate vs.  diameter - pump seals, light liquid service.
-------
                                                  A =
                                           Correlation Coefficient (r) » 0.105
                                           Number of Data Pairs      = BO
                                                                                            PC
                                                                                            o
                                                                                            n>
L
0
        t
        1
    0.0  4
i  -u.n  t
o       i
        t
L       X
t  -1.6  4
«       *
K       •/
        *
K  -a.<»  4
A       /
I       X
C       *
   -3.2  4
c «
c
u
A
(t
  U
A «
                               A
                               l\
                               A
                                t
                                i\
                                n
                                i\
    II
  A
  A
    t
A A
    H
  A «
A A «
  U A
    A
C
                                                                        C
                                                                        A
                                                                        H
                                                                        H
A A


A
                                                A A
                                                B
                                                A
                                                                                                          I
        t
   -M.n 4
        y
        1
        t
   -4.0 4
             .-4--_>-__4	4	+	4	4	4	4	4.	+	4	4---4	4.--4-	4---4	4--- +	4	•

               1   3   5   7   9  11   13  15  17   19  21  4? 3  25  VI  *9   31  33  35   37  3V  «ij.


                                                  AWL (YEARS)
        Figure 24  Leak  rate  vs.  age  -  compressor seals,  hydrocarbon  service
-------
                                                  : A = t OHS.  B  =  * Of|S.
                                            Correlation Coefficient (r)  - 0.045
    !•* *                                    Number of Data Pairs       -47

        :.

        •/
    U.O 4-
        *                  "
        <                      A A
        /                                  A        A
        ^
L   0,0 *                                     A
0       1
(if                  ft
I       /                      A
0       *      l\       A                                         A
   -o.o 4           n c      A
i.       -t             W
t       r     A       n        A              A
A       /             A        -A                                  A
K       /          A
   -1.6 +
H       i             fl  A                                  A
A       r     A    A   ft                       A
T       /             A                           A
L       *     «       rt               » A
   -2.«4 +
        1
        t
        t             A          A
        /                                              A
   -.1.2 4                                          A
        r                      A
        /
        /             A
        f
   -'1.0 4
        	,	^	+	+ _	f	|	f	_. +	+ -
           ')       bU       1UU      lt>0     200      i!bO      4»U     4SO      MOO
                    Figure 25   Leak  rate vs.  pressure  - relief valves
-------
R.  L.
          Several trends were expected to exist for discrete variables.
 Control valves were expected to  leak more then block valves because of a
 higher frequency of operation.   Vibration was expected to be directly pro-
 portional to leak rate.  Packed,  single mechanical, and double mechanical
 seals were expected to represent the range of highest to lowest emission rates.
 It was also speculated that sources  made by different manufacturers would
 exhibit different emission rates.  Several examples of discrete variable
 correlations are shown in Figures  26 through 34.

 CONCLUDING REMARKS

          Even  the  "most  significant"  correlations observed showed a lot of
 scattered data  and  the correlations  were not dramatic.  The only exception to
 this is the effect  observed  for  stream composition.  These results do not
 necessarily indicate  that correlations between emission rates and process
 variables are nonexistent.   There were several factors discussed previously
 which probably  explain why no  significant correlations were observed, and
 therefore no  conclusions  can be  drawn  from the observed results.  As fugitive
 emissions become  subject  to  regulatory constraints, additional data may be
 collected by  the  regulated industries.  This expanded data base'may eventually
 reveal significant  correlations  between process variables and fugitive emis-
 sion rates.
                                      297
-------
i.nn »,__..
t
1
t
t
t
t
t
/
-O.J3? *
1
1
t
t
1
t
.1
t
"1.67 +
'C *
1 :
~* *
S '
- ;
3 -.1.00 »
i ;
S t
S *
^ o i
00 ST !
3 .„.„ t
*
f
t
t
t
t
2
*
-5.67 *
*
t
t
1
t
t
t


V


0 0
0
f
*
0
*
*
t
t
t
t t
i *
»---* *
* * » 	 1 t 	 +


t t 	 * t «
* * < *
* f « 	 1
»---+ * *
*
t
t
*
*
tf
0 0
D
N-108 o
0
0
C •
*
N-50
N-26B

*





Gas/Vapor Light
Block Control Block






0
t
t
t
*
* «
* 0 *
1 t
* * *
* » ». — »
t 	 + « * »

> « » * *
* t +.--« t t t
*-*„• *---> 
-------
    1.90
  -0..133
S  -1.67
3
a  -3.00
   -1.33
   -5.67
   -7.00
*
*
*
V
*
>
/
1
4
*
*
f
r
i
t
t
t
4
/
f
r
<
'
^
4
1
1
1
1
t
i
1
i
4
*
i
i
t.
t
t
1
i
4
*
t
1
t
1
1
t
1
II

0
*
/
t
t
t
t
1
1
t
4---4
* <
t
*
-4-»
*
*
4 — — "• 4
(
*
*
t
t
1
*
t
t

 «"«
0 *
0 N»92
N-292 °
0 0




















*
.--



...
^
*
# +

» ^
+ — »" +
*
*
*
«
/
*
*
<
*
t

N'J






*





t
*
*
*
t
t
*
*
+
<
*
*
*
*
t
*
*
4
t
*
*
*
*
*
*
t
4
*
*
*
/
*
*
*
<
4
*
*
*
*
*
*
t
t
4-
*
«
<
*
*
*
*
*
                                                                                                                      8
                                                                                                                      H
                                                                                    KEY
                                                                                             Detached value (1 In 700)
Upper qu«rt11e
Mean
Median

Lower quartlle
                                                                                           0  Outside v.ilur (I in
                                                                                           0
                                         Manufacturer

                                      It       5
            Figure  27   Schematic plot  for valves  by  manufacturer variable.
-------
                   ?.»«
OJ
o
O
               _  -0.667
               I

               3
               a
                  -2.00
£
o
f
                  -J.33
                  •n.*T
                  -6.00










t
1
1
*
\
1
t
t
t
t
t
t
f
t
t
t
1
+
*
f
*
<
$
1
1
t
0
0
*
*
*
* — +
*— -» t t *
t t »_.-» « .--» »-—»
* * * — *
* * ... — «
* * 4 * 	
• • + - t •»»- ^
t t t f * *
tt » * * ---+
* . * * * *
« X
0 «. — *

N-15 H-3

0
H-17 N-ie
0 0
N*262 "

H'4B


•
*

*



                                               Light Liquids
                                           CM       CP        DP
    Heavy Liquids
CM        CP
                                                                                              Rp
                                                                                                            KEY
                                                                                                                    * Detached value (1 In ZOO)
                                          Upper quartlle
                                          Mean
                                          Median

                                          Lower quartlle
                                       0  Outside value (1 In 70)
                                       o
                                   CM • Centrifugal Mechanical
                                   CP • Centrifugal Packed
                                   RP • Reciprocating Packed
                                                                                                                                                  §•
                                                                                                                                                  0
                                                                                                                                                  (D
                                                                                                                                                  1-1
                                                                                                                                                  9?
                                                                                                                                                  §
                                                                                                                                                  •O
                                                                 Pump Type
                                 Figure  28   Schematic plot  for  pumps  by  pump type  variable.
-------
?.nn <
i
t
t
i
i

»
0.6A7 *
/
1
t
i
r
*
' *
»
-0.667 4
t
1
T '
f 5
^r i
A
***
3 -a, oo
|Q
oe
M
3
»
c
5
*j
5 -3.33
o
ST
_j


-t|.67
&£ ft rt
6 • W V




0
0
0
*
*
*
t
t
t t
t t
t t
t «
* « t
4---4 .--» <
t t 0 4 — .4
f it
t ---»

* -4- ' 4 *
t t
t * t
«-. -4 * 4 	 	 4
* * 4 f
t • 	 *
t *
> t
4 	 4 t
t 0
t
t
n
N-9

H-38 0
•""' N-54

*
*
t
1
Light Liquids Heavy Liquids
Double Single Double Single
Seal Seal seal Seal
*
t











KEV
* Detached value (1 In 200)



-., Upper quartlle
4 Mean

	 Median


-r-J Lower quartlle

1
0 Outside value (1 In 20)
0











                                                                           SJ
                                                                           fD
                                                                           t-i
Figure 29  Schematic plot for pumps by  seal variable
-------
z «ou







O.«.ft7







J -0.667
>i»
a
"^
s
i*
i
8- -._
»?«00
1
i
i
o
f

-3.33




-It. 67





-«. nn
1
t
t
9
t
* 9
t 0
' V
t *
* t
1-1 t
t t t
t t
t t 0
t t t
* * *
t t t
4 4-4 4-4 4-4 t t t


t t t t - * t t -* */**4***»
* * t ft ***-• «-« t
t 4- t t 4- -4 4-« * * * *
f - * *4- * t t t a
\ ** ; T :: «•"
; M H-S v •
* . *
* *
^ #
» K-29 »-3' «•«* "•" 0 ""'.i,
t m o c "•'*
; * K-89
J N-24
*
4
*
» ' •
* *
*
f »












4



*
«
4-*
* * 4«»
*-* •»»
* * *4*
* *
4-4
*
t
t
*
*
t

n-is
*
*





















KEY
* Detached value (1 In 200)

1
r-1--) Upper quartlle
t Mean
	 Median
. - • — J Lower Qusrti Ic

i
0 Outside value (1 In 20)
0














                                                                                        o
                                                                                        3
                                                                                        (C
    Light Liquids





2315
     Manufacturers


678     J
                                      Heavy Liquids




                                 23156
Figure 30   Schematic plot for  pumps by manufacturer variable.
-------








-0.667




— -1.73
e
a
3
^
AC
*t
3
^ -*.*<>
2
N
1
UJ
S |
-3.07



-0.33










0







t

»••
-*-
' t
t f
\ * \ *
t •— t * *
h — » J *
: i j H-8
r *4« * H«
r N*4 t"
t 1
t t »---
» +•--» t
* * *
* * *
t t
t IMS
4
t
fit
*
»
t
/
» •
»
»
<
/
t
1
t
1
t
\
t
t
' «Y
* Detached value (1 In 200)


r- — t Upper quart) 1e
4 Mean
-— • Median

- -J Lower quart Me


0 Outside value (1 In 20)
0













                                                                                   05
                                                                                   I
           High
                    Medium
Slight
                                       None
Figure  31   Schematic  plot of  flanges by vibration variable.
-------
    ?.nn
   o.njs
~. -0.333
   -l.SO

   •2.67
   •J.B3
                                        4...

                                        *
                              +-..»
                                 i
                                 t
                                 *
                              *—»
                               IM
II-78
                               Hydrocarbon Stream

                              Double       Single
                     *
                     *
                    ».--•»
                    *   *
                    *.+-»
                    *   »
                    *   >
                    *.--«
                     >
                     *
                     *
                                                             H-?1
4--.»

  *

  *
                                                                        N-22
                      Hydrogen Streams

                    Double      Single
                                                                                      Ml
                                                                                             * OeUched value (1 In 200)
                                                       Upper quartlle

                                                       Mean

                                                       Median


                                                       Lower quartlle
                                                                                             0 Outside vglue (1 In 20)
                                                                                             0
                                                                                                                          to
                                                                                 r
                                                                                 (D
                                                                                 H
                                                                                 ?^

                                                                                 I
  Figure 32   Schematic  plot  for  compressor  seals by  single/double seal variable.
-------
*
/
t
t
1
t
1
0.03; »
t
t
t
t t
t t
t t
t, t
T * '
« -0.333 * +— t
a t t »-.-(
C. i t t
X > t t
*'••'•
1 » ' *
" / •--- «
« 1 t » t
• -liSO » * •+-
5 t t
I !
8 | I...
u. 3 ,
1
-7.67 #
# +-*-
* 1
t 1
t 1
t t H-7
J N-19
-3.C3
































f
»•


























*
*
>---






4





-"


















t 0

* » 0
» t t
f *--- *
t t » r
* * * *
» * * — •
t • r <
* < /
* « ---»
t * * * *
* * * /

N-11 J
#
^
^

0

N-16

0
0 *
0
0
N-16
»
t
t
i









\ KEY
* * Detached value (1 In 200)




... .- Upper quartlle
* Mean
	 Median


- -J Lower quartlle


0 Outside value (1 In ?0)
0












                          Manufacturer
Figure 33   Schematic plot for compressor seals by manufacturer variable.
                                                                                       33
                                                                                       O
                                                                                       I
-------
  -O.J1T

   -1.45
f
   -J.H9    «•
                                                             t
                                                             4--
                                            N-33
                                                                N«16
                                               Visible Vapor Emissions
                                            No                 Tes
                                                                                          *  Cetaehtd value (1 In 200)
                                                                                             Upper qutrttle
                                                                                             Hem
                                                                                             Htdlan

                                                                                             lower quartUe
                                                                                          0 Outside value (1 In ?0)
                                                                                          0
                                                                                                                     sa
                                                                                                                     o
                                                                                                                     I
              Figure  34   Schematic plot of drains  by  visible  vapor  variable.
-------
A. F. Pope
                                   REVIEW

                                     by

                                A.  F.  Pope
                      ARCO Petroleum Products Company
                               A Division of
                        Atlantic  Richfield  Company
                          Los Angeles, California
                                     on
           CORRELATION OF FUGITIVE  EMISSION  RATES FROM  BAGGABLE
                  SOURCES WITH REFINERY PROCESS VARIABLES
                                   RESUME

         Arthur F. Pope  is  the Manager, Environmental Policy and Planning,
 for ARCO Petroleum Products  Company,  a division of Atlantic Richfield
 Company.  He received his B.S. degree in Mechanical Engineering from the
 University of Detroit in 1969.   Since 1974,  he has worked  for Atlantic
 Richfield as a Project Engineer, Manager, Air and Water  Conservation, and
 Manager, Environmental and  Energy  Conservation, at the Watson refinery,
 Carson, California.  He  has  participated in  the several  projects conducted
 by EPA and the California Air Resources Board which attempted to ascertain
 atmospheric emissions of VOC from  refinery sources, including valves, during
 his assignments at the Watson refinery.
                                     307
-------
A. F. Pope
                                  REVIEW

                                    by

                                A. F. Pope
                      ARCO Petroleum Products  Company
                               A Division  of
                        Atlantic Richfield Company
                          Los  Angeles,  California

                                    on

           CORRELATION OF FUGITIVE EMISSION RATES FROM BAGGABLE
                  SOURCES  WITH REFINERY PROCESS  VARIABLES

INTRODUCTION

         Refinery fugitive VOC emissions have been studied a number of times,
starting in 1957 with the Joint Study1 of the then Los Angeles County Air
Pollution Control District and culminating most recently with the completion
of the EPA/RADIAN program under discussion at this Symposium.  The Atlantic
Richfield Watson refinery has been involved in each of these studies except
the recent RADIAN effort.  Therefore, the observations made in these previous
studies should be useful in comparing the results from our experiences with
the results from facilities examined by RADIAN.  I will limit my discussion
to the subject of R. L. Honerkamp's paper, "Correlation of Fugitive Emission
Rates From Baggable Sources With Refinery Process Variables."

         The data I will present corroberates what Honerkamp found—the
dominating correlation is stream composition.  I will also suggest that, for
refinery valves at least, process unit designations may be more suitable and
are certainly more practical than specific stream composition in designing
inspection and maintenance programs for refinery fugitive VOC sources.
                                    308
-------
A.  F.  Pope
DISCUSSION

         As reported by J. H. Nakagama2 at  a previous  EPA/RADIAN workshop
on this subject at Jekyll Island, Georgia,  in  1978, Atlantic  Richfield
Company undertook a study of all the valves and  fittings  in one of  its  crude
oil distillation units.  The objective of this study was  to assess  the  costs
that would be incurred .for the inspection of all valves and fittings  for the
detection of any leaks, using a soap-solution  method.  Of the more  than
11,000 potential leak sources checked in this  38,000 B/D  crude unit,  38
leaks were discovered.  Thirty-one leaks were  repaired with the unit
on-stream.

         Subsequent to this study, the California Air  Resources Board,  using
enforcement personnel, conducted a two-week study of refinery valves  and
flanges in the South Coast Air Basin including Watson.  Upon  reinspection of
this same crude oil distillation unit, using soap-solution methods, no  leaks
were found.  Approximately eight months had elapsed between studies.

         Some conclusions that can be reached  are:  (1) once  repaired,  crude
oil distillation unit valves and fittings will not begin  to leak for  at
least eight months, and (2) crude oil distillation unit process stream  con-
ditions of pressure, temperature, stream composition,  etc., collectively
yield a low probability that a component will  be found to be  leaking, when-
ever it is inspected.

         In each of the studies mentioned,  virtually all  valves and flanges
were examined.  As a result, it would be more  accurate to say that  these
studies had characterized the emission factor  for the  valves  for that speci-
fic crude oil distillation unit, rather than the emission factor for  the
category of valves in general.  Indeed, the data ought not be considered in
statistical analysis, since the valves were not  selected  on a random  basis.
Therefore, it is not possible to compare these results directly with  results
obtained by RADIAN.

         However, these data and that which I  am about to describe  generally
support the finding that stream composition is an appropriate variable  to
consider when developing fugitive emission  control strategies and that, as
noted in the crude oil distillation unit case, the composite  effect of  the
process variables associated with a process unit may be more  practical  in
identifying those refinery sources which have  a  higher probability  that an
individual component within that process unit  will be  found to be leaking.
                                     309
-------
A. F. Pope
         While Atlantic Richfield Company disagreed3 with the inspection
methodology and the conclusions reached by the California Air Resources
Board enforcement personnel, some insight on the composite effect of process
variables can be gleaned from the data obtained during this effort.  The
process units inspected at Watson included the crude oil distillation unit
mentioned previously, two additional crude oil distillation units, an
alkylation unit, a super fractionation and isomerization (SFIA) unit, and
LPG storage vessels and associated transfer piping.

         The refinery obtained duplicate gas samples of each component which
was bagged by GARB and analyzed the sampes for VOC.  Using the GARB leak rate
measurement and the mass spectrometer results from the refinery laboratory,
the following estimated emission factors were developed for the valves and
flanges associated with each specific process unit:

                           VOC EMISSION FACTOR
      UNIT TYPE                LB/DAY/VALVE           LEAKING MATERIAL

      Crude                     0                     N/A

      Alkylation                0.0299                Butanes

      SFIA                      0.0362                Light Gasoline
                                                        Components

      LPG                       0.0596                LPG
         These emission factors incorporate all the variables possible for
each process unit without the need to determine their individual influence.
In general terms, these factors also indicate that the stream composition
plays a significant part in the unit emission factor, i.e., the lighter and
higher pressure streams exhibited higher leak rates.

         A crude oil distillation unit will have many different streams at
different temperatures, compositions, and pressures.  This type of unit was
found not to have any leaks.  At the other end of the spectrum, the LPG
facilities handle a single stream at relatively constant temperature and
pressure and was found to leak most.  If one were to consider all streams
within a process unit and determine an "average" stream, the stream compo-
sition effect is easily seen.
                                     310
-------
A. F. Pope
CONCLUSIONS

         Several conclusions  concerning  fugitive VOC  emissions result from
the preceding discussion:

         1.  Process unit VOC emission factors  for valves correlate
             well with "average"  stream  composition.

         2.  Process unit designation offers  a  more practical
             method for evaluating  fugitive emission  control
             strategies than  stream type.

         3.  Process unit designation minimizes the problem of
             attempting to  correlate a multiplicity of process
             variables, including component age, line size,
             pressure, temperature,  and  stream  composition.

         4.  The data discussed support  the conclusion of
             R. L. Honerkamp  that the dominating correlation
             is stream composition.

REFERENCES

1.  "Joint District, Federal, and State  Project for the Evaluation of
    Refinery Emissions," Los  Angeles Air Pollution Control District,1957.

2.  "Inspection and Monitoring Concepts  for Refinery  Fugitive Emissions,"
    J. H. Nakagama, 1978.

3.  Letter, N. E. Pennels  (Atlantic Richfield Company) to J. J. Morgester
    (California Air Resources Board), May 5,  1978.
                                     311
-------
R. L. Honerkamp
                           QUESTIONS  AND ANSWERS

Q.  James Stone/Louisiana Air Control Commission - The first question,
maybe frivolous.  Did you assess operator motivation as a variable?  In
other words, in an HF alkylation unit, he is very likely to see to it that
any leak is repaired immediately, while in some other units, like a coker
nothing dripping is going to hurt him, so he wouldn't go after the leak
quite as fast.

A. - No!  We did not attempt to assess that particular variable.  The results
that I presented here were not really a documentation of any existing main-
tenance practices, but based on the entire data base collected from the
thirteen refineries nation-wide.  All sources in the stream composition
service shown pose the effect of the process variables on the emission rate.
But, perhaps looking at it on a unit basis, as Art Pope suggested, might show
differences that could be attributable to that factor.  But, I don't have any
data that I could present that would show how that effect could be quantified.
Art, do you have anything else?

COMMENT/A. F. Pope/ARCO Petroleum Products Company - One of the units in the
study that I did not present any information on was that in our plant we
happened to have a benzene unit and there were no high VOC emissions from
that unit.  There were some hydrogen leaks, but there were no VOC leaks.
So, that might tend to answer your question.

Q.  James Stone/Louisiana Air Control Commission - The second question is
that you have been putting a lot of effort into your statistical evaluation
of the data.  Do you think that a statistical approach would have validity
in regulations which we might write to cover maintenance of this type?

A._ - I think it certainly might, although I don't think that any of the
correlations presented here this morning other than stream compositions are
significant enough to warrant such attention.  All the variations that we
saw were primarily about an order of magnitude variation in leak rate
throughout the entire range of the variable.  Although there were much
greater variations than that at individual values.  So, based on that I
don't think we can conclude that any particular temperature range or pressure
                                     312
-------
R. L.  Honerkamp
range should have more or less frequent inspection  of maintenance.  Although
the effect does seem to be there as far as stream composition.

Q.  James Stone/Louisiana Air Control Commission -  What  I  really mean  is
that we have a regulation in Louisiana that says that best practical main-
tenance shall be used at all times in a refinery.   And,  as we inspect  the
place, if we pick a statistical number of valves rather  than try to inspect
everything on paper, and if we find that below a certain number is leaking,
they are in compliance, and above a certain number,  they would be considered
to be in violation of that regulation.

A.   (By Rosebrook) - Lloyd Provost has a very definite view on that.   The
line of reasoning is that he can, given a sufficient data  base for the
enforcement people to work with, he can set confidence intervals based on
the  time since the refinery last did their inspection and  maintenance  in a
particular unit, he feels very strongly a set of numbers could be developed
which would allow you to come in and for example screen  100 valves or  200
valves, rather than screening the entire unit.  If  you find a certain  number
leaking then you could call it a violation with some kind  of confidence, if
you  find some leaking but not the limiting number then it  should not be a
violation.  I can't go through that explanation near as  well as Mr. Provost,
and  perhaps we can at least have you get together with him to answer that
question.

A.   (By Honerkamp) - One thing that such a scheme does require is that
several variables be understood.  Those variables will be  discussed in
greater detail in the final paper that EPA will present  this afternoon,
those include:  what level of repairs are achievable; and, how long after
repair do the leaks reoccur.  You must know, with confidence, what those
factors are before you can attempt to develop the statistical enforcement
program.

COMMENT/Person unknown - I encourage the regulators not  to write that  number
as zero.  The data that you have seen would tell you that  on any given day,
in any given plant, there is going to be at least one component that is
going to leak to some degree or another.  And, what I am trying to tell you
is that if you have a nonattainment area and you have a  source requesting a
permit, and that source has to certify that he is in compliance with all
federal, state, and local regulations, under penalty of  perjury, how can he
do that if you have a regulation that requires all  discrete components in a
plant, not to leak.  That is a no growth regulation.
                                     313
-------
R. L. Honerkamp
Q.  Nancy A. Kilbourn/PEDCo Environmental, Inc. - My first question is how
much did it cost Atlantic Richfield to monitor all the valves?  And, my
second question is how much do you think that you will save after you
establish an I and M (inspection and maintenance) program?

A.  (By Pope) - We are subject to California regulation now and have been
for almost a year to conduct I and M in the Southern California area at our
Watson refinery, and we have utilized the resources of a contractor to do
that inspection and minor maintenance work.  The costs this year are going
to run on the order of $100,000 to do the inspection and perhaps another
$50,000 to do some special repairs on-line, to avoid shutting down unit to
make that repair.  The State of California regulation for your information,
has a maintenance provision in it.  It allows two working days from the
discovery of a leak to repair it to a no-leak condition.  And, a leak in
this case is 10,000 ppm at 1 cm.  A repair is 1,000 ppm at 1 cm.

Q.  R. L. Honerkamp/Radian - Do those numbers include regular maintenance or
is that just special maintenance?

A.  (By Pope) - No, that is just special maintenance for this program.
There would be additional costs to the number that I gave you for the
contractors costs to us, the internal cost of our own people.  I'm really
not sure what they are at this point.  It would be more.

Q.  Nancy A. Kilbourn/PEDCo Environmental, Inc. - How much do you think you
would save then in recovering your volatile organic compounds?

A.  (By Pope) - I don't have a number for you.

Q.  James Stone/Louisiana Air Control Commission - Just going to follow-up
on that same question.  When your contractor does that sampling, do they
sample all valves or flanges or do they use some percentage to work by?

A.  (By Pope) - No, the regulation we are subject to, requires a comprehen-
sive inspection of all valves.  Flanges are not included.

Q.  James Stone/Louisiana Air Control Commission - How many is that?  How
many valves for your facility?

A.  (By Pope) - I think we estimated, since we don't really have a complete
count with a high degree of accuracy, something on the order of 100,000 to
130,000 components.  They are $1.60 a piece to check if you want.
                                     314
-------
R. L. Honerkamp
Q.  K. C. Hustvedt/US-EPA-RTP - On  the whisker  diagrams,  looking  at  the
difference between the block valves and  the  control  valves,  the emission
factors before you crossed the lines  looked  like  the same  for all three
process stream variables; the gas,  the light liquid,  and  the heavy liquid.
I think that it has biased the diagrams  by excluding the  nonleakers  from
them.  You are showing emission factors  that are  not the  real emission
factors.

A_._ - But I am showing leak rates  for  the ones that did  leak.  As  I pointed
out in the beginning, the combination of that effect  plus  the effect on the
percent that leaks, i.e. how many that don't leak at  all,  is what results in
the emission factor.  But rather  than look at emission  factors for my pre-
sentation I looked at just the effect on the ones that were leaking.  Since
there were no dramatic effects that were going  to be  shown at all, I felt
that it would be more interesting to  look at the  effect on the ones  that
did leak, rather than whether or  not  they leaked  at  all.   That effect has
been split out in the stream category designation.

Q.  K. C. Hustvedt/US-EPA-RTP - To  completely compare the  two between block
and control valves you have to know what your total  population looked like,
not just some arbitrary subset, say over 200 ppm, to be able to compare the
effects of those if 90 percent of the block  valves did not leak and  50
percent of the control valves did not leak.   Just comparing their average
emission factors for leakers doesn't  tell you on  the  average if block valves
or control valves leak more or less.  You have  to look at  the total  popula-
tion, I would think, to see what  a  real  true comparison between what the
effect of those two is, not a subset  based on an  arbitrary cutoff.

A_._ - We weren't looking at the effect on emission factor,  but the effect on
the ones that leaked.  It is true that if such  an effect did exist that
emission factors were significantly different for block and controlled valves,
that would have been the incorrect  presentation to look at.  I don't believe
such an effect, on emission factors does exist, as far  as  block and  control
valves.  That is, I think the confidence intervals probably do overlap
significantly, if you look at emission factors.   But  it is true,  that what
we looked at were just the ones that  were leaking.

Q.  A. F. Pope/ARCO Petroleum Products Company  -  I would  like to  encourage
you not to think just of the components, as  I mentioned in my discussion.  I
think you will be better served in  looking at this whole  spectrum of things
that need to be done on a practical level.   The people  out there  like myself
have to implement something.  If  you  can focus  on getting  to where you want
                                    315
-------
R. L. Honerkamp
to go in a practical manner, some of these discrete things you might like
to evaluate would be good to know, and perhaps will be helpful in
redesigning components for minimizing losses from those components.  But
in terms of I & M programs, I don't think that they are really worthy of
the significant effort to characterize them in a very discrete manner.
                                     316
-------
R. G.  Wetherold/S. L. Preston
          THE EFFECT OF MAINTENANCE PROCEDURES ON THE REDUCTION OF

                FUGITIVE HYDROCARBON EMISSIONS FROM VALVES IN

                             PETROLEUM REFINERIES

                      R. G. Wetherold and S. L. Preston
                              Radian Corporation
                                Austin, Texas


                                   ABSTRACT

           Regulations  pertaining to the screening and maintenance  of  refinery
  process valves are being  proposed by regulatory agencies.   Under an EPA
  contract, Radian has studied the effect of simple maintenance  practices  on
  the  reduction of hydrocarbon emissions from refinery valves.   The  results
  are  presented here.  Included in the study were block and  control  valves in
  the  major types of refinery process stream services.  The  reduction in
  hydrocarbon emissions  after maintenance was determined for  valves  having
  initial leak rates ranging from large to small.   The merits of hydrocarbon
  monitoring during the  performance of maintenance were evaluated.   Finally,
  the  effectiveness of valve maintenance over short (one week) and long
  (six months) time periods was investigated.

                                    RESUME
                               R. G. Wetherold
           Robert G. Wetherold is a Senior  Staff Engineer at Radian Corpora-
  tion in Austin, Texas.  He is currently Project Director for several
  programs associated with  the study of fugitive emissions from  petroleum
  refineries and chemical plants.   He received his B.S. in chemical  engineer-
  ing  from Texas A&I University,  a M.S.  in chemical engineering  from Texas A&M
  University, and his Ph.D. from the University of Texas at  Austin.  Before
  coming to Radian, he was  employed as an Associate Engineer in  the  Process
  Development Division of Mobil Chemical Company.   He also worked in the
  Process Design Division of Chevron Research Company.  He is a  member  of
  AIChE.

                              Sheryl L. Preston

           Sheryl Preston  is a Data Management Specialist at Radian Corpora-
  tion.  She has a B.S.  from the University of Arizona, and  is currently  in  a
  Masters program at the University of Texas.  She is a member of ASQC.  For
  the  past 3 years she has  managed the Fugitive Emissions from Petroleum
  Refining data base at  Radian and contributed in data analysis.
                                      317
-------
R. G. Wetherold/S. L. Preston
          THE EFFECT OF MAINTENANCE PROCEDURES ON THE REDUCTION OF

                FUGITIVE  HYDROCARBON EMISSIONS FROM VALVES IN

                            PETROLEUM REFINERIES
          The effects of relatively simple maintenance procedures on the
reduction of fugitive emissions must be defined to evaluate the cost
effectiveness of inspection and maintenance procedures.  The reduction of
fugitive hydrocarbon emissions from valves as a result of maintenance has
been studied as part of the EPA's program for the environmental assessment of
petroleum refineries.  The emission reduction study is described in this
paper.  The results of the study are presented.


OBJECTIVES

          The objectives of this maintenance study are given below:

               •   To select for the program those fugitive hydro-
                   carbon emission sources which could be studied
                   in the most cost-effective manner.

               •   To define a group of the selected emission sources
                   which would provide a representative sample for the
                   maintenance study.

               •   To determine the immediate effects of directed and
                   undirected maintenance activities on the reduction
                   of hydrocarbon emissions.

               •   To define the short and long-term effects of main-
                   tenance procedures on the reduction of fugitive
                   hydrocarbon emissions.


MAINTENANCE PROCEDURES

          The six "baggable" sources were all considered for maintenance
studies in this program.  These sources include valves, flanges, pump seals,
compressor seals, relief valves, and drains.  The types of maintenance con-
sidered for each source are described below.  They are listed in order of
increasing difficulty, complexity, and generally, cost.
                                      318
-------
R. G.  Wetherold/S. L. Preston
          The majority  of  valves  in refineries  are  gate  valves  (on/off) and
 globe valves  (control).  Plug valves are  also present.   The  types of main-
 tenance considered  for  valves include:

               •    Tightening packing gland nuts (gate and  globe valves)

               •    Adding  grease  (plug valve)

               •    Replacing  the  valve packing

               •    Injecting  sealant into the packing area

               •    Replacing  the  entire valve

          Pump seals  are either of  the packed or mechanical types.  The main-
 tenance procedures  applicable to  these seals are:

               •    Tightening of  the packing gland nuts  (packed seal)

               •    Replacement of the packing (packed seal)

               •    Replacement of the mechanical seal

          The maintenance  of  compressor seals takes the  same form as that of
 pump seals.   Included are:

               •    Tightening of  the packing gland nuts  (packed seal)

               •    Replacement of the packing

               •    Replacement of the mechanical seal

          Repair of flange  leaks  can generally be accomplished by one of
 these procedures:

               •    Tightening of  the flange bolts

               •    Replacing  the  flange gasket

               •    Replacement of the flange or flange face

          Those pressure relief valves  venting to the atmosphere were also
 considered for maintenance  studies.   If these devices are leaking through
 the valve seat to the atmosphere, mechanical repairs are generally required.
 If a dual relief or manual  bypass system  is available, the more simple
 mechanical repairs  might be made  with the valve in the line but blocked out
 of service.   In many  cases, however,  the  entire valve must be removed and
 repaired.
                                      319
-------
R. G. Wetherold/S. L. Preston
          There is generally no simple maintenance procedure that can be
used to reduce emissions from open drains.  The drainage system must be
revamped to include items such as covers and traps.


EXPERIMENTAL DESIGN OF THE MAINTENANCE STUDY

          It was not possible to study the effectiveness of various types
of maintenance on all six baggable source types.  The study was limited by
time, funds, and practical considerations in an operating refinery.  As an
aid in the definition of a useful but limited program, the total emissions
from the six source types in the major process units of a refinery were
estimated.  The Gulf Coast Cluster Model Refinery developed by A. D. Little,
Inc.1 was used as a basis to these estimates.  The number of sources in each
unit were developed from source counts made during the course of the sampling
program or from source counts of relatively similar process units.

          The estimated total number of baggable source types in each major
process unit is shown in Table I.  In Table II the estimated percentage of
leaking source types and the relative emission contribution of these sources
are presented.  The most numerous source types are valves and flanges.  How-
ever, the percentage of flanges that leak is quite low, and their contribution
to the total emissions is small.  A considerable amount of time would be
required to locate a sufficient number of leaking flanges for a satisfactory
sample size.

          Valves are quite numerous, over a third of them leak, and they
contribute 60 percent of the baggable source emissions.  Preliminary studies
indicated that a significant reduction in valve emissions could be achieved
through simple maintenance.

          Nearly half of all inspected pump seals leaked to some degree.
With the exception of valves, pump seals contribute more emissions than any
other source type.  In the case of packed seals, simple maintenance consists
of tightening the packing gland.  This can be done while the pump is in
service.  Leaking mechanical seals must be replaced.  The pump must be taken
out of service to make this replacement.  Refineries generally have spare
pumps which can be quickly placed in service in place of many of the more
important pumps.  Thus, in many cases, a pump can be taken out of service
for maintenance without disrupting the process.

          The majority of compressor seals leak.  Because there are relatively
few of them, however, their emissions are only 9 percent of the total baggable
source emissions.  Additionally, maintenance of packed seals and replacement  of
mechanical seals can be a major procedure.  In many cases, the process would
have to be shut down to repair or replace the compressor seals.

          Maintenance of relief valves is also not an insignificant viffort.
While over a third of the inspected relief valves leaked, the emissions only
make up 4 percent of the total baggable source emissions.
                                     320
-------
          TABLE I.  ESTIMATED DISTRIBUTION  OF FUGITIVE EMISSION SOURCE TYPES IN MAJOR PROCESS  UNITS OF A
                    HYPOTHETICAL REFINERY3
OJ
N)
Process Unit
Atmospheric Distillation:
Unit 1
Unit 2
Vacuum Distillation
Light Ends /Gas processing
HDS - Reformer Feed:
Unit 1
Unit 2
HDS -Light Gas Oil
HDS - Light Cycle Oil
HDS - Vacuum Gas Oil
HDS - Coker Naphtha
Fluid Catalytic Cracking
Hydrocracking
Catalytic Reformer:
Unit 1
Unit 2
Aromatics Extraction
Allaylation
Coking
Hydrogen Production

Valves

890
890
500
190

650
650
650
650
650
650
1300
940

690
690
600
570
310
180
11650
Flanges

3560
3560
2000
760

2600
2600
2600
2600
2600
2600
5200
3760

2760
2760
2400
2280
1240
640
46520
Pump
Seals

43
43
21
4

14
14
14
14
14
14
42
31

20
20
17
15
13
3
456
Compressor
Seals

2
2
-
4

6
6
6
6
6
6
6
6

6
6
0
0
0
6
74
Relief
Valves

6
6
6
6

6
6
6
6
6
6
6
6

6
6
6
6
6
4
106
Drains

69
69
35
11

22
22
22
22
22
22
65
58

49
49
41
41
28
8
655
                                                                                                                    O
                                                                                                                    rt
                                                                                                                    i-j
                                                                                                                    o
                                                                                                                    Tl
                                                                                                                    fD
                                                                                                                    CD
      Hypothetical  refinery  units  taken from Arthur D.  Little Gulf Coast Cluster Model Refinery with a
       capacity of 330,000  BPD.
-------
R. G. Wetherold/S. L. Preston
  TABLE II.  ESTIMATED DISTRIBUTION OF FUGITIVE HYDROCARBON EMISSIONS FROM
             SIX SOURCE TYPES IN THE MAJOR PROCESS UNITS OF A HYPOTHETICAL
             REFINERY3
Source Type
Valves
Flanges
Pump Seals
Compressor Seals
Relief Valves
Drains
Estimated
Percent
Leaking
27
3
48
76
39
19
Emissions Contributed
By Each Source Type,
Percent of Total
60
6
12
9
4
9
100
 aArthur D. Little:  Gulf Coast Cluster Model Refinery- 330,000 EPS.
                                      322
-------
R. G. Wetherold/S. L. Preston
           It was decided that the most  cost effective sources to study were
 valves and pump seals.  Furthermore, the maintenance study would concentrate
 on valves, since they represent the greatest emission source of the baggable
 SOU1TC6 typGS •

           The maintenance to be performed on valves consisted of:

             •   Simple adjustment/tightening of the packing gland,  or

             •   Injection of grease into the fittings of plug valves.

 Additionally, some valves were to be monitored for extended time periods to
 determine the effectiveness of valve maintenance over an extended period of
 time.
           The number of valves required to make the above evaluations  was
 limited through selective experimental design.  The wide variation in  leak
 rates between valves was circumvented by using paired measurement schemes
 for maintenance evaluations.  Only valves with particular selected leak rates
 were studied.

           The factors that were considered in selecting valves for the main-
 tenance study were:

             •   Process stream group (gas/vapor streams, light
                 and two-phase streams, and heavy liquid
                 streams.

             •   Valve type  (block/gate, block/other, control/globe,
                 control/other).

             •   Leak rate or screening value range (500-5000 ppm
                 screening value, 5001-50,000 ppm screening value,
                 and > 50,000 ppm screening value).

 In addition, data were collected on all of the parameters normally included
 in the program.

           A total of 28 valves were proposed for study at each refinery.
 The distribution of these valves is shown in Table III.

           Pump seals were to be selected for the maintenance study in  a
 manner similar to that for valves.  The factors that were considered in the
 selection included:
                                      323
-------
R. G. Wetherold/S. L. Preston
      TABLE III.  DISTRIBUTION OF VALVES TO BE STUDIED IN EACH REFINERY
Process Stream Valve
Group Type
I Gas -Vapor
Streams


II Light Liquid &
Gas-Liquid
Streams

III Kerosine
& Heavier
Streams

BG
BO
CG
CO
BG
BO
CG
CO
BG
BO
CG
CO
Low
(500-
5,000
ppm)
XO
0
X
0
XO
0
x D
0
X
0
X
0
Medium
(5001-
50,000
ppm)
xoQ
X
X
X
XO
X
X
x n
X
X
x n

High
(> 50,000
ppm)
XO
X
x n
X
xoD
X
X
X
X
0
X
0
Total
X's

10



10



8


 Total X's
12
10
28
  Determined by maximum "TLV Sniffer" reading.
  BG = Block, gate; BO = Block, any type other than gate;
  CG = control valve, globe; CO = control valve, other than globe.
  [  I = control point, i.e. select a valve but do no maintenance.
  X  = select a valve here if possible; 0 = secondary choice for valve
       selection.
                                     324
-------
R. G.  Wetherold/S.  L.  Preston
             •   Leak rate category (medium leak = 0.5 - 1.0 Ibs/hr,
                 high leak rate = > 1.0 Ib/hr).

             •   Pump type (centrifugal-packed, centrifugal-single
                 mechanical seal, centrifugal-double mechanical seal,
                 centrifugal-packed, etc.).

           It was hoped that 10 leaking pump seals suitable for a maintenance
 study could be found in each of four refineries.

 PROCEDURE

           The steps below were generally followed during the maintenance
 studies:

             •   screening to locate potential sources

             •   rescreening of selected sources

             •   sampling of sources

             •   performance of maintenance

             •   resampling of sources

             •   additional short and long-term screening

             •   application of quality control procedures

           A Bacharach "TLV Sniffer," a sensitive hydrocardon detector,  was
 used to locate and select sources for study.   With a dilution probe, the
 range of this instrument is 0 - 100,000 ppmv.   The TLV Sniffer is calibrated
 with hexane.   For source selection the TLV Sniffer probe was placed as  close
 as possible to the points of potential leakage (valve stem and gland, pump
 seal).  Readings were taken at eight different points around valve stems and
 glands and at four points around pump seals.   The maximum reading was taken
 as the basis for estimating the leak rates.   Leaking valves which fit into
 the desired distribution (Table III)  were tagged for further consideration.
 Selected pump seals were similarly tagged.

           When all the required valves and pump seals were located, prepara-
 tions were made for measuring their leak rate.  Each selected source was
 rescreened immediately prior to sampling.   All data were recorded.   The
 leaking source was then enclosed in plastic.   A sampling train was  attached
 to the enclosure and the leak rate from the source was determined.

           After the initial leak rate was measured, maintenance was performed
 on the leaking source.   This maintenance was defined as either "directed" or
 "undirected."  Directed maintenance involves simultaneous maintenance and
                                     325
-------
R. G. Wetherold/S. L. Preston
 screening of the source with a hydrocarbon detector.  Maintenance activities
 are continued until no further reduction in hydrocarbon  concentration can be
 achieved.  Undirected maintenance consisted of the normal maintenance pro-
 cedures without any hydrocarbon concentration monitoring during the
 activity.

           When the maintenance procedures were completed, the maintained
 source was again screened and sampled.  The leak rate immediately after
 maintenance was thus determined.

           Whenever possible each maintained source was rescreened several
 times during a period of one to two weeks immediately following the main-
 tenance.  The purpose of this activity was to get an indication of the
 short-term effectiveness of directed and undirected maintenance.

           Arrangements were made at some refineries to obtain some data
 regarding the long-term effects of maintenance on the reduction of emis-
 sions.  In these cases, refinery personnel agreed to monitor selected
 maintained valves at intervals of one week to one month  for a period of six
 months.

           As part of the experimental study, quality control procedures were
 implemented.  These generally consisted of replicate and multiple source
 screening, replicate source sampling, accuracy testing of the sampling
 train, frequent calibration checks, and frequent analysis of standard gases
 in the laboratory.

 RESULTS

           A total of 120 valves were included in the maintenance study.
 Eighty-six of these actually underwent maintenance.  The remaining 34 valves
 were not maintained.  They were screened, however, and were also, in some
 cases, sampled.  The unmaintained group provided data on the variability of
 screening values and the change in leak rate as a function of time.

           Twenty-seven valves underwent directed maintenance.  Fifty-nine
 valves were subjected to undirected maintenance procedures.

           No maintenance studies were performed on pump  seals.  Difficulties
 were encountered in locating leaking pump seals in the proper leak rate
 categories.  In addition some pumps that were found to be leaking could not
 be adequately isolated for seal replacement.  In some cases, there were no
 spare pumps available to replace the leaking pump.  In other cases, it was
 felt that the time required and the cost incurred for seal replacement was
 not justified by the size of the leak.

           The effect of maintenance procedures on leak rates can be expressed
 as a percentage reduction in leak rate.  The percentage  reduction can be
 calculated from Equation 1.
                                     326
-------
R. G. Wetherold/S. L. Preston
                       LB  ~  LA
                   R = —	 x 100
                           B

 where

           R  = leak rate  reduction, %

           L  = leak rate  before maintenance,  Ib/hr

           L  = leak rate  after maintenance,  Ib/hr.


           Negative percentage reductions  in  leak rates  can be obtained if
 the leak rate is higher after maintenance than  it was before maintenance.
 The highest achievable positive reduction is  100%.   It  is possible, however,
 to get negative percentage  reductions that are  much  greater than 100 percent.
 Thus is particularly true if  the original leak  rate,  LB,  is very low.

           The effect of maintenance procedures  on the leak rates of the
 individual valves involved  in this  study  are  tabulated  in Tables IV and V.
 The data are plotted in Figures 1 and 2 where the effect of maintenance can
 be seen more clearly.  In these figures the  leak rate of the individual
 valves  after maintenance is  plotted as a function of the valve leak rate
 before maintenance.  This is  done for both undirected and directed mainten-
 ance procedures.  The valves  exhibiting a reduction  in  leak rate from main-
 tenance activities are indicated by those points that fall below the
 diagonal line drawn in each figure.   Those valves whose leak rate increased
 after maintenance are represented by the points  which fall above the diagonal
 line.  It can be seen that  the points in  Figure 2 generally fall further
 below the diagonal and closer to the horizontal axis than those plotted in
 Figure 1.  It appears from  these figures,  then,  that directed maintenance
 procedures are generally  more effective than  undirected maintenance activ-
 ities in reducing valve emissions.   Also, a  smaller fraction of valves
 exhibit an increase in emission rate after directed  maintenance than after
 undirected maintenance.

           The data are plotted in the form of histograms in Figure 3.  The
 results of the directed and undirected maintenance studies are shown.  The
 greater effectiveness of  the  directed maintenance procedures is clearly
 shown in this figure.

           The effects of  the  valve  maintenance  studies  are summarized in
 Table VI.  The results are  shown for both the directed  and the undirected
 maintenance programs, and are grouped according to the  level of emission
 rates.  Two results are noteworthy.   It is evident that the average percent-
 age leak reduction for those  valves that  were subjected to directed mainten-
 ance ±8 considerably greater  than that of the valves which underwent
 undirected maintenance.
                                      327
-------
R. G. Wetherold/S.  L.  Preston
  TABLE IV.  THE EFFECT OF UNDIRECTED MAINTENANCE PROCEDURES  ON LEAK RATES
             FROM INDIVIDUAL VALVES
Valve
ID
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
, 21
22
23
24
25
26
27
28
29
30
31
48
49
50
51
52
53
54
Valve
Function
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Control
Control
Control
Control
Control
Control
Control
Measured Leak
of Nonmethane
Before
Maint.
0.0320
0.0437
0.0158
0.1476
0.6572
0.5801
0.0018
0.0327
0.0871
0.1963
0.1071
0.0026
0.0109
0.1673
0.0019
0.0449
0.0381
0.0295
0.1256
0.0023
0.1995
0.1019
0.0264
0.1761
0.0015
0.0614
0.00049
0.0034
0.0083
0.0182
0.0293
0.2703
0.6235
0.2253
0.0923
0.0227
0.0286
0.5863
Rate, Ib/hr
Hydrocarbons
After
Maint .
0.00001
0.00002
0.00028
0.0051
0.0231
0.0481
0.00015
0.0031
0.0094
0.0288
0.0168
0.00045
0.00191
0.0365
0.00047
0.0174
0.0192
0.0157
0.0767
0.0015
0.1354
0.0714
0.0198
0.1328
0.0012
0.0508
0.00054
0.0044
0.0174
0.0462
0.1398
0.0009
0.0045
0.0017
0.0012
0.0018
0.0023
0.0553
Reduction
After
Maintenance,
Percent
100
100
98
97
96
92
92
90
89
85
84
83
83
78
76
61
50
47
39
34
32
30
25
25
20;^
17
- 10
- 29
-110
-153
-377
100
99
99
99
92
92
91
                                                            Continued
                                     328
-------
R. G.  Wetherold/S.  L.  Preston
                            TABLE IV.  Continued
Measured Leak Rate, Ib/hr
of Nonmethane Hydrocarbons
Valve
ID
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
Valve
Function
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Control
Before
Maint .
0.0058
0.0054
0.0161
0.0063
0.0514
0.0039
0.1641
0.0276
0.00037
0.0009
0.0055
0.00063
0.0127
0.0234
0.0119
0.0027
0.0015
0.0011
0.00031
0.00013
0.0019
After
Maint.
0.0008
0.0007
0.0029
0.0013
0.0202
0.0018
0.0758
0.0141
0.00026
0.00065
0.0040
0.00049
0.0115
0.0244
0.0133
0.0035
0.0024
0.0027
0.00078
0.00085
0.1673
Reduction
After
Maintenance,
Percent
86
87
82
79
61
54
54
49
29
28
27
22
9
4
• - 12
- 30
- 60
- 145
- 152
- 550
-8745
                                      329
-------
R. G. Wetherold/S. L. Preston
 TABLE V.  THE EFFECT OF DIRECTED MAINTENANCE PROCEDURES ON LEAK RATES FROM
           INDIVIDUAL VALVES
Measured Leak Rate, Ib/hr
of Nonmethane Hydrocarbons
Valve
ID
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98 •
99
100
101
109
110
111
112
113
114
115
116
117
Valve
Function
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Block
Control
Control
Control
Control
Control
Control
Control
Control
Control
Before
Maint .
0.0011
0.0111
0.0891
0.1396
0.0075
0.0383
0.0126
0.0115
0.0307
0.0032
0.0045
0.0800
0.00066
0.0014
0.00078
0.00197
0.00055
0.0053
0.0095
0.0181
0.0065
0.0173
0.0126
0.0025
0.0021
0.0005
0.0016
After
Maint .
0.0000
0.0002
0.0017
0.0028
0.0002
0.0017
0.0009
0.0008
0.0025
0.0004
0.0007
0.0124
0.00013
0.00039
0.00032
0.00106
0.00090
0.0130
0.0000
0.0004
0.0003
0.0010
0.0011
0.0003
0.0005
0.0003
0.0035
Reduction
After
Maintenance,
Percent
100
98
98
98
97
96
93
93
92
87
84
85
80
72
59
46
- 63
-145
100
98
95
94
91
88
76
40
-119
                                      330
-------
                                                                                                                                         o
LO
U)
   i.uoo +
         7
         t
         J-
         t
         t
         t
             L
             t
             A
             K   U.10U
             It
             A
             r
             E
A
F
1
t
rt

M
A
I
N
T
                U.U1U
                U.U01  +
                       t
                       4-
                       1
                                                                    :  A = 1 OUS« H  =  a OUSi LIC,
Leak rate before =
  Leak rate after
  maintenance
                                                                              A  A
                                              U.UU1
                                                         u.uiu

                                               Ct«l\ HATt. Ut-HOKL  ilAilMT.,  (LB/HR)
                                                                                                                                         rt

                                                                                                                                         (D
                                                                                                                                         P-!
                                                                                                                                         O
                                                                                                                            hd
                                                                                                                            f-i
                                                                                                                            n>
                                                                                                                            cn
                                                                                                                            It
                                                                                                                            o
                          Figure 1.   The Effect of  Undirected Maintenance on the Leak  Rate From Valves
-------
                                                                                                                       o
                                                     : A = 1 UbSt  H = U OHS«  ttt.
L
L
A
K

K
A
T
t

A
F
T
M
A
1
                                         Leak rate before «
                                           Leak rate after
                                           maintenance
                                                         t\
                                0.001
          o.uiu

LEAK KATL btHUHt MA1NT»> LB/HR
                                                                                                    1.000
                                                                                                                       ro
                                                                                                                       i
                                                                                                                       o
                                                                          i-t
                                                                          (D
                                                                          en
                Figure 2.  The  Effect of  Directed Maintenance on the  Leak Rate  From Valves
-------
R. G.  Wetherold/S.  L.  Preston
10


 9


 8'H


 7
      u
      c
      01
         2-
         -I
         3-
                Undirected Maintenance
             g
          ~. I  >  I I
                        j
                        /y-^
                       I
                  i
%m
                                           B!
           MS


                                  oooooooo
                              Percent Reduction
LO-]
9-
3-
7-
6"
5-
4-
3-
2-
1-





^

-£
Directed Maintenance




^1





\
y/
//.





/•/
//





SJ




I

6
//

\
i
\
i
^


\
\



//







          OcTiar--i£iLn»rmcNi—   -^c>jro»runu3r-a3cr>o

          "T	   Percent deduction
       Figure  3.  Histograms for Percent Reduction in Leak Rate

                  From Directed and Undirected Maintenance
                                     333
-------
                           TABLE VI.   SUMMARY OF MAINTENANCE REDUCTION BY LEAK RATE LEVEL
bJ
U>

Original Leak Rate

Level Range (Ib/hr)
1



2



3



4



n =

P -
pw =
pm =
< 0.001



0.001 - 0.01



0.01 - 0.1



> 0.1



Number of valves maintained
Average percent reduction =
Weight percent reduction = •
Median percent reduction
n
P"
pw
pm
n
P~
pw
pm
n
F
pw
pm
n
P"
pw
pm

EPi/n, where

Directed Maintenance
4
30.7
35.2
52.6
12
48.7
56.9
86.2
10
93.8
93.0
93.8
1
98.0
98.0
98.0

„ _ (leakage before - leakage

Undirected Maintenance
6
- 105.5
- 26.3
5.6
16
- 530.0
- 276.4
30.4
22
31.7
45.1
60.9
15
73.4
83.5
85.4

after maintenance) „ , nn
1 leakage before maintenance
Eleakage before maintenance - Zleakage after


Eleakage before maintenance

maintenance ..
——————— x J.UU

                                                                                                                        o
                                                                                                                        i-l
                                                                                                                        O
                                                                                                                        CO
                                                                                                                        •

                                                                                                                        r1
                                                                                                                        i-t
                                                                                                                        fD
                                                                                                                        cn
                                                                                                                        rt
                                                                                                                        O
-------
R. G. Wetherold/S. L. Preston
           It  is also  apparent  that  the level of the initial  leak rate  has  a
 marked effect on the  percentage  reduction in emission rate  for both  directed
 and undirected maintenance.  The percentage reduction achieved by mainten-
 ance  is lower for  the initially  small leak rates.   In the low initial  leak
 ranges, < 0.01 pounds per hour,  the average and weight percent reduction
 in emissions was actually negative  for undirected maintenance.

           It  should be noted that as the magnitude  of the leak rate  becomes
 smaller, both the  average percent reduction and weight percent reduction
 decrease rapidly.  Both of these parameters are dependent on the magnitude
 of the initial leak rate and are highly influenced  by extremes within  the
 leak  rate range.   The median percent reduction, however,  is  a more definitive
 measure of  central tendency and  cannot be affected  by the very large nega-
 tive  values of percent reduction encountered at low leak  rates with  undirected
 maintenance.

           The median  percent reduction does show the same pattern as the
 average and weight percent reductions.  The comparison between the median
 percent reductions for the two types of maintenance indicates that directed
 maintenance yields a  higher reduction in leak rate.   Undirected maintenance
 appears to  be even less reliable at low leak rate levels  (<  0.001 Ib/hr).
 This  type of maintenance appears to have a greater  potential for producing
 an  increase in valve  emissions  after  maintenance.

           The percent reduction  from the two maintenance  methods was plotted
 against the original  screening value in Figures 4 and 5.  The positive
 percent reductions with directed maintenance generally appear to be  higher
 than  the reductions achieved with undirected maintenance.  Also,  a greater
 percentage  of the  valves undergoing undirected maintenance appears to have
 increased in  leak  rate (compared to those subjected to directed maintenance)
 after being maintained.   Table VII  bears out these  observations.  The median
 percent reduction  with directed  maintenance (91.2 percent) is significantly
 higher than that with undirected maintenance (53.8  percent).

           The valves  are grouped by function (block or control)  in Table VII.
 Control valves which  had directed maintenance had a slightly higher  median
 percent reduction  in  leak rate than block valves which had the same  type of
 maintenance.  However, the opposite is true for valves which underwent
 undirected  maintenance.   Again,  even within the block/control groupings,
 directed maintenance  appears to  yield a higher percent reduction in  leak
 rate  than undirected  maintenance.

           It  should be noted that leaking valves in some,  categories  of the
 original experimental design were not found.  Very  few valves in the heavy
 liquid stream classification were found to be leaking, particularly  at the
 higher rates.  Valves in some  categories were found in some  refineries,  but
 not in others.  In many cases, substitutions from other categories were  made
 to provide  an adequately sized data base.
                                      335
-------
                                                ft =  1  OUSi b = ? UUS I  LTC.








p
L
C
j_

N
i
K
t
L)
U
C
r
i
0
N





t
t
f
J.MI +
t
1
t
1 U 0 t ;\
t
t
t
bO +
y

/ /t H


^
4
-tjl) f
r
y
/
- 1 U U 4
;
-I'JO +
^
-r'UO i
; (-550)
(1
(0
rt
cr
S
o
i — '
1 -
A /\ A A A C •
H A t A A E ^
ft A A A r
A A Htf
AA A D H
b A P
ff)
A A A A A rt
A A §
A A
A A


A



A
A A A


(-8745) (-377)
	 + 	 •" 	 + «•- - — 	 	 +--
1UU 1C11U iUUUO 1UUQUU
                                          MAX SCREENING VALUE. PPMV
Note:  3 values were out of range
           Figure  4.   Undirected Maintenance - Percent Reduction  in Emissions  as  a Function
                       of Initial Screening Values
-------
UJ
                                                                                                                                o
                                                                                                                                •


                                                                                     LTL.
                                                                                                                                n>
                                                                                                                                i-j
                                                                                                                                o
                                                                                                                                i->
                                                                                                                                &.

                                                                                                                                CO
liltl <
1 U 0 «
I1 '
L /
K r
t Ml t
t. /
N /
t ,' -
K i
t. ^
U /
U -UJ() I
t ^
1 /
1 t
0 -UU 4
N
r
-1'jO <
^
<
-Jo II t
/\ ft tt A A A AC
« A « A A A ' H
A A A
A A
A
A
A
                                                                luiiu                   10 nun


                                                          MAX SCREENING VALUE,  PPMV
                          Figure 5.   Directed Maintenance - Percent Reduction in Emissions as a Function

                                      of  Initial Screening Values
                                                                                                                                it
                                                                                                                                O
-------
                    TABLE  VII.
STATISTICAL SUMMARY  OF  MAINTENANCE  DATA  -  PERCENT  REDUCTION
OJ
00
                Screening
                                    Block Valves
                             DJ-rected Mrtlntrnimce

                                      Control Valves
Range
(ppov)

<5K


5K-50K


>50K

G/V
Stri-nm
2 58.8
56.5
58.8
2 76.1
90.7
76.1
3 93.8
97.8
98.0
LL
Strrnm
5 63.1
90.5
93.1
4 89.8
89.0
90.1
2 -26.4
56.7
-26.4
HL
Strrnm
0


0


0





Total*
Block
7 61.8
86.5
87.3
6 85.2
89.1
88.7
5 45.7
92.3
91.7







G/V
Strnnm
0
1


1


18 64.2 (32,96)
91.0 (82,99)
86.2 (75,97)
45.7
'.5.7
45.7
77.2
77.2
77.2
LL
Rl ronm
4 39.5
84.9
89.8
1 95.0
95.0
95.0
2 97.2
96.4
97.2
HL
Stream
0


0


0





Total All
Control Vnlvos
4 39.5
84.9
89.8
2 70.4
91.5
70.4
3 90.5
95.0
94.5







11 53.74
85.6
88.4
8 81.5
89.2
8R.7
8 62.5
92.6
93.1
9 66.8 (12,100)
89.7 (79,99)
91.2 (9.3,98)
(3.7,100)
(72,99)
(18,98)
(65,98)
(69,100)
(-55,96)
(-7.9,100)
(81,100)
(-33,99)







27 64.6 (38,91)
90.7 (83,98)
91.2 (79.95)
                *Numhers In parentlieaes Indicate an approximate 95X confidence Interval for the average  reduction Tor the three different estimations.
                                                                                                                                     (ContinimoT
o
•


n>
rt

ro

o
H
&

W3
                                                                                                                                                             T)
                                                                                                                                                             i-l
                                                                                                                                                             n>
                                                                                                                                                             en
                                                                                                                                                             rt
                                                                                                                                                             O
                                                                                                                                                             0
                  Code for

                  Each Cell

                  In Table
     1 = Number of valvea maintained

                           .   .    .              .    .     100 x (leak before - leak after maintenance)
     2 - Average of percent reduction where percent reduction > 	Leak before maintenance	
                                                 Weight percent reduction •« —


                                                 Median percent reduction
                                £_JL_ea_k rateM>eforc maintenance - Llcak rate after maintenance

                                             J^leak rate before maintenance
-------
                                                     TABLE  VII.
Continued
u>
                                                                Undirected Maintenance
Screening
Value
Range
(ppmv)

<5K


5K-50K


>50K

Block Valves
C/V
Stream
6 54.0
52,2
65.2
4 69.8
47.8
B2.6
3 75.3
88.4
84.3
I,L
St rciin
6 42.6
58.9
76.9
4 -64.9
- 9.0
28.2
4 81.3
93.0
90.9
ML
Stream
4 -26.1
-43.4
7.37
0


0





Total*
Block
16 29.7
48.5
33.1
8 2.4
20.2
50.1
7 78.7
91.1
85.4







Control Valvefl
U/V
Stream
7 -1321)
- 717
-58.4
2 54.2
53.8
54.2
8 29.4
81.3
19.3
LI.
SL roam
5 5.2
91.1
26.56
4 87.8
96.9
95.6
1 90.6
90.6
90.6
111,
Stream
0


1 82.1
82.1
82.1
0


31 33.7 (-1.8,69)
68.7 (48,89)
61.1 (31,85)
Total*
Total All
Control Valves
12 - 769
-50.5
24.1
7 77.4
90.2
82.1
9 36.2
87.0
29.5







28 -312
33.0
28.9
15 37.4
67.4
82.1
16 54.8
89.6
67.0
28 298 (-940,100)
81.0 (64,98)
51.4 (13,85)
(-950,100)
(-39,100)
(-0.5,79)
(-28.100)
(34,100)
(42,88)
(31.78)
(81.98)
(21.92)







59 -]24 (-410.100)
73.9 (69.88)
53.8 (29.82)
'Numbers In p
Code for
Each Cell
In Table
aronthcscu Indicate
1 2
3
4
1
2
3
an approximate 952 confidence
m Number of valves maintained
• Average of percent reductlo
- Weight percent reduction -
interval for the average percent reduction for the three different estimations.
. . 100 x (leak before - leak after maintenance)
wiidc Ki-*..i-.n- m.t-uMi.i.^vi. Leak before maintenance
F.lenk rule before maintenance. - Elenk rate after maintenance *QQ
^leak rate before maintenance
•s>
•

o
                                                                                                                                             fD
                                                                                                                                             It
                                                                                                                                             rr
                                                                                                                                             m
                                                                                                                                             i-i
                                                                                                                                             o
                                                                                                                                             i-1
                                                                                                                                             on
                                                                                                                                             i-i
                                                                                                                                             fD
                                                                                                                                             CO
                                                                                                                                             rt
                                                                                                                                             O
                                     4 - Median percent, reduction
-------
R. G. Wetherold/S. L. Preston
           A comparison of emission reduction by range of screening value can
 also be made.  For directed maintenance, the median percent reduction stays
 approximately constant across the screening value range.  However, for the
 undirected maintenance group the median percent reduction increases
 dramatically with increasing screening values.  The median percent reduction
 is very low, only 28.9 percent, for those valves having low screening values.
 This may indicate that undirected maintenance at this screening level is not
 effective at all.  For the middle screening value range, the median percent
 leak reduction for valves which underwent directed maintenance increases to
 82.1 percent, almost as high as the reduction with directed maintenance
 (88.7 percent).  However, the median percent leak reduction with directed
 maintenance is somewhat lower (67 percent) for the valves in the high screen-
 ing value range.  The effectiveness of the maintenance program appears to be
 much more consistent when the directed maintenance method is used rather than
 the undirected method.

           The differences in percent reduction discussed above should be con-
 sidered as trends.  Confidence intervals were calculated for the key values
 and these are presented in Table VII.  Differences in the percent emission
 reduction cannot be considered statistically significant if confidence limits
 for the estimates overlap.

           A graphical representation of the differences between the effect
 of maintenance on block and control valves is shown in the next several
 figures.  The leak rates before and after maintenance are plotted for block
 and control valves in Figures 6 and 7.  The percent reduction in leak rate
 for each valve is plotted against the original screening value for block and
 control valves in Figures 8 and 9.

           Finally, Figures 10 and 11 are histograms of percent reduction for
 block and control valves for directed and undirected maintenance.  While no
 large differences between valve function are obvious, the differences between
 the percent reduction in emissions for valves undergoing directed and
 undirected maintenance can be seen.  The advantages of directed maintenance
 are apparent.


 THE SHORT AND LONG TERM EFFECTS OF VALVE MAINTENANCE

           A number of the valves which underwent maintenance were screened
 several times during a one week period following the maintenance.  The
 results are summarized in Table VIII.  The advantage of directed maintenance
 can be clearly seen.  Fifty percent of those valves with initial screening
 values > 10,000 ppmv still had screening values in excess of 10,000 ppmv
 immediately after undirected maintenance.  By the end of one week,  60 percent
 of these valves had exhibited screening values  above 10,000 ppmv.

           By contrast, only 2 of the 10 valves subjected to directed main-
 tenance had screening values in excess of 10,000 ppmv.  One additional valve
 developed a screening value above 10,000 ppmv by the end of the week.
                                     340
-------



L
t.
A
K

K
A
T
t

A
h
1
L
K

M
A
1
N
T
t

L
B
/
H
R
i
J . uiiij •»
*'
I
/
/
•J.I 00 «
t
t
t
t
/
f
t
U.Oll) •*
1
*
t
t
f
7
*
U . u U 1 +
f
J
t
4
t
/
f
U , U 0 U +
                U = Undirected

                D = Directed
                                                          u   u
                            u.uui
                                                  U.U1U                  U.1QU



                                         LtAt\ KAfL Utt-OKt  MAlNTt. LB/HR
                                                                                                              o
                                                                                                              fD
                                                                                                              n
                                                                                                              o
                                                                                                               fD
                                                                                                               en
        i  t.Mts  muut.i'1
Figure  6.   Directed and  Undirected Maintenance - Leak Rate  After Maintenance as a Function of

            the Leak Rate Before Maintenance - Block Valves.
-------
             1 , 0 0 0 4
to
          L
          L
          A
          K

          K
          A
          T
          E

          A
          F
          T
M
A
I
N
T
          L
          B
          /
          H
          R
   o.ioo
/
t
<
t
t
/
4
t
t
1
t
t
t
                              U
                              D
                       Undirected
                       Directed
ft)
o
i-1

en
*
r1
                                                                                                                 (0
                                                                                                                 en
                                                                                                                 rt
                                                                                                                 O
                   U . l IIU
                                          . uui
                                                               0 . 01 U
                                           Ll ftH HI* It  I1LHIKL
                                                                                     u.iou
                                                                           ,  LB/HR
              Figure 7.  Directed and Undirected  Maintenance - Leak Rate After Maintenance as  a  Function of
                          the Leak Rate Before Maintenance - Control Valves
-------
I1
E
C
t
N
T
K
E
0
U
C
T
S o
LO N

i
t U • Undirected
150 * D = Directed
/
/
lull t i' u U D U U D
/ II III) U U U (I I) U
x URDU
, U U
bO * D U
* U U
t u U U U
* u
/ u
/ o
Jt
-t>0 +
/ u
*
-iuo +
t- U
*
>
-1'jO 4 U
^
^
I (-377)
r IUO 100M 1UUOO
1)
n
It
n

                                     MAX SCREENING VALUE, PPHV
                                                                                                           o
                                                                                                           0)
                                                                                                           II
                                                                                                           o
                                                                                                           I-1
                                                                                                           A-
                                                                                                           cn
                                                                                                           rt
                                                                                                           O
                                                    U UIIS II1UULIM
Figure  8.   Directed and Undirected Maintenance - Percent Reduction in Emission Rate as a

            Function of the  Screening Value - Block Valves
-------
                                                                                                       CD
r


* U = Undirected
, D « Directed
^
/
*UO + U U U U U
P * U U U U
E * U
R t u
C bO + 0 UU
E X
N * U u
J j

R *

E *
0 t
U -1)0 +
C t U
T /
1 /
U -100 +
N 1
* I)
-1'JU -f U O
-2UO <
r (-550) (-8745)

Si
0>
rt
ro
i-i
o
i — i
e^
DM c/?
u
D IT1
•
U "T3
11