-------
TABLE C2-4. (Continued).
SUALitY CONTROL DATA AND SUMMARY STATISTICS
•~drains**
------ S&x pooled
ID SAMP DATE LfAK RATE X DIFF LEAK RATE X DIFF STAND DEV
32DR
20
BO
50576
0.02U43
192. B
17DR
20
BS
2027B
0.00391
•
17DK
20
BO
20278
0.02720
-150
17DR
20
BO
20278
0.01651
-12H
17DR
20
BC
20278
0 . 00*451
-1H.2
17DR
20
BQ
2037B
0.01285
-107
17DR
20
BO
20376
0.02536
-1<»7
17DR
20
BO
20378
0,03717
-162
17 OR
20
BO
20378
0.03S10
• 160
17DR
20
BO
20378
0.01916
-132
17DR
20
HQ
20378
0.00917
-80. H
17DR
19
BS
20278
O.OOOjq
•
17DR
19
BO
20278
0.00219
-1H7
17DR
19
BO
20378
0.01050
-188
17DK
19
00
20378
0.01173
-189
0.01913 173.9
65.6
0.00619 1BH.I
107
C-55
-------
LtGENOs A = i OOS. B = a OBS. ETC.
0
01
200
150
100
50
.50
-1U0
-150
¦200
A B C D C F C
REFlNtRY
A
A
A
A
A
F
TT
A
B
A
+70%
-70%
FIGURE C2-14. Percent Difference Between Repeat Samples by
Refinery - Valves
-------
legenu; a = 1 oos. n = a obs« etc.
200
0
1
Ul
I
N
L)
I
V
1
u
u
A
L
%
0
1
F
F
150
100
50
-bO
-100
-150
B
D
C
0
B
A
C
+70Z
A
A
B
-A
B
-70Z
-200
REFINERY
Figure C2-15. Percent Difference Between Repeat Samples by
Refinery - Pump Seals
-------
LEGEND: A 3 1 OBS• B » 2 OBS. etc.
0
1
Ln
00
I
N
0
1
V
I
u
u
A
L
X
0
1
F
F
200
150
100
50
-50
¦100
-150
-200
+70Z
-70Z
REFlNtRr
Figure C2-16. Percent Difference for Repeat Samples by
Refinery - Compressor Seals
i
-------
LEGEN05 A = 1 OBS. B = 2 OBS. ETC.
200
160
0
1
(_n
MD
I
N
U
I
V
I
0
0
A
L
X
U
1
F
120
SO
<«0
-40
+705!
• B0
-70X
-120
REFINERY
Figure C2-17. Percent Difference for Repeat Samples by Refinery -
Flanges
-------
LEGtND! A = l OBS» B = 2 OBS» ETC.
+ 70Z
60
<«0
C)
I
as
o
I
N
0
J
V
1
u
u
A
L
S
0
1
F
F
20
-20
-HO
-60
-ao
¦100
A
R -70X
RtFlMtRY
Figure C2-18. Percent Difference for Repeat Samples by Refinery -
Relief Valves
-------
LEGEND: A s 1 OBS. B = 2 OBS. ETC.
?on
150
1
N
0
1
V
I
D
U
A
L
100
50
+70X
0
1
CTn
-50
-100
¦150
A
A
A
C
B
-70X
-200
REFINERY
Figure C2-19. Percent Difference for Repeat Samples by Refinery - Drains
-------
refineries for each of the types of sources bagged. Figures
C2-20 through C2-25 show these same percent differences plotted
versus the average leak rate of the samples from each source.
Control limits of ± 70 percent are included on these plots. A
maximum 70 percent difference between samples was the original
goal for the baggables sampling and analysis procedure. As can
be seen, a significant number of checks were outside these limits.
Leak rates from drains were especially nonrepeatable. The leak
rates appeared to vary considerably with time. In addition, the
sampling technique for drains was not as reliable as that for
other baggable sources. It was difficult to insure that addi-
tional emissions were not induced by the sampling procedure.
Frequency distributions of the percent differences for valves
and pump seals are shown in Figures C2-26 and C2-27.
Figure C2-28 shows a plot of the pooled standard
deviation for the repeat samples versus the average leak rate
of the original and quality control samples. Since the stan-
dard deviation is obviously related to the concentration level,
percent standard deviations were computed. The percent stan-
dard deviations are plotted versus the mean leak rate in Figure
C2-29. The percent standard deviation appears fairly constant
for all levels with a slightly larger percent difference for
leak rates less than 0.01 lbs/hr.
Table C2-5 summarizes the statistical analysis of the
repeat QC samples. The variability for drains is significantly
higher than the other sources while the variability for relief
valves is significantly less. The other sources have a stan-
dard deviation averaging about 40 percent, or a confidence limit
based on a single test of ± 80 percent.
C-62
-------
LEGEND: a r 1 OBS, B = 2 OBS, ETC.
200
ISO
0
1
ON
LO
I
N
0
1
V
I
0
u
A
L
*
0
1
F
F
100
50
¦50
-100
•150
-200
A
A
A
A C AA
tnr
ac a
TT
A a
AA
A A
A
A
A A
A
B A
A AA A AA
A BA
AB a A A
.- — A---- A A—-A — -A —BAAB-B-
C A A B A Q
A A A A A A
AA A AO A AA
A B A
B
— A
A A
A
A A
C A
BA
A
A
+701
-701
C.00
0.01
o.io i.on
NEAN LEAKRATE OF SAMPLE (lb/hr)
10.00
Figure C2-20. Percent Difference for Repeat
Rate - Valves
Samples Versus Mean
Leak
-------
LfGtND! A = 1 OBS, fl = 2 OBS, ETC.
2 00
*
~
i
*
A
n
A
ISO
t
+
A
A
I
N
D
100
* A
t
+
*
t
A
A
A
UA
A
A
A
A
A A
A
A
+70Z
1
V
I
u
0
A
50
o
*
+
t
t A
t
B
A
A
A
0
A
A
A
A
B
A
A
0
A
A
A A
A
A
AA
A
n
A
0
A
A
A
A
AAA
B
A
L
S
U
I
F
-50
*
*
*
+
<
*
A
A
A
A
A AB
A
O A
A
A
AB AA A
A
A
CA A
A A
B
A
A
AO
A
A
B
A
A
-70Z
h
-100
-150
-?00
*
¦f
*
t
*
+
*
*
*
+
A
D
C
A
C
A
B
A
A
o.no
0.01
mean
0.10
LEAKRATC
OF SAMPLE (lb/hr)
1.00
10.00
Figure C2-21. Percent Difference for Repeat Samples Versus Mean Leak
Rate - Pump Seals
-------
200
ISO
100
50
0
-50
100
150
200
LC6ENDI A s 1 OBS, B c 2 OBS, ETC,
A
A
A
S A A A
AA
AAA A A A
A
CA B A A ft
+70X
A A B A
A A A
A
0 A
_fl -70%
A
A A
A
A
0,0 0<1 1.0 10.0 100.0
MEAN LEAKRATE OF SAMPLE (lb/hr)
Figure C2-22. Percent Difference for Repeat Samples Versus Mean
Leak Rate - Compressor Seals
-------
LEGEND! A = 1 oBSf B £ 2 OBS, ETC.
t
200 ~
i
*
A
t
160 ~
*
I
N
0
1
V
1
t
t
120 ~
t
*
*
ttO ~
t
A
+70*
[>
u
A
L
X
u
*
*
HO +
*
*
t
0 ~ -
A
A
A
A
I
F
F
t
*
t
-HO ~
«
1
1
B
A
A
A
-70*
• 60 ~
/
*
-120 ~
n.oo o.oi 0.10 x.oo xo.oo
MEAN LEAKRATE OF SAMPLE (lb/hr)
Figure C2-23. Percent Difference for Repeat Samples Versus Mean
Leak Rate - Flanges
-------
LEGEND! rt = 1 OHS, fl = 2 OBS, ETC,
60
10
+70X
0
1
Ch
--J
I
N
0
1
V
I
u
u
A
L
X
u
I
h
F
20
-20
-40
¦ 60
-60
-100
0.00
0.01
A A
0.10
mean LEAKRATE OF SAMPLE (lb/hr)
A A
1.00
-701
10.00
Figure C2-24 . Percent Difference for Repeat Samples Versus Mean
Leak Rate - Relief Valves
-------
LFGtND: A = 1 OBS, 8=2 OBS, ETC.
0
1
ON
00
200
150
100
50
X
0 -50
-100
¦150
-200
A A
+70Z
-70Z
(1.00
0.01
0 .10
mean leakrate of sample
(lb/hr)
1.00
10.00
Figure C2-25. Percent Difference for Repeat Samples Versus
Mean Leak Rate - Drains
-------
FREOUENrY
"~0
30
0
1
vo
20
in
*****
-175
*****
*****
*****
*****
**~»»
**»»*
*****
*****
*****
-125
¦75 -25 25
PERCENT DIFFERENCE MIDPOINT
75
125
*****
175
Figure C2-26. Frequency Bar Chart for Percent Differences
Between Samples - Valves
-------
frequency bar chart
frequency
30
20
n
i
o
10
*****
*****
**•*•
*****
*****
*****
***••
¦175
¦125
¦75
-25
25
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
75
125
175
PERCENT DIFFERENCE MIDPOINT
Figure C2-27.
Frequency Bar Chart for Percent Differences
Between Samples - Pump Seals
-------
leg*-no: a = i obs. b « 2 obs, etc.
o
s
a
o.
o
O.bH
0.56
0.48
0.40
| 0.32
H
q 0,24
Q
^ °«16
H
t/i
0.08
0.00
A ABA
A AA A A B A AA
BaaBOaCBaB BAB AA CA PA A A AAAA A
A A
A A
0.00
0.01 0.10 1,00
MEAN LEAK RATE OF SAMPLE (lbs/hr)
10.00
Figure C2-28. Standard Deviation of Repeat Samples Versus Mean
Leak Rate - Valves
-------
LEGEND; A a 1 OBst B r 2 OBSi ETC.
0
1
ho
SB
o
w
3
§
a
160
mo
120
luo
00
H
Q 60
3
•~0
20
A
A AA
Pooled standard deviation ¦ 36.61
A A A
AA
A A
A AA
A A
T R
A A
A AA
A A
A A
A A
A AA A
AAA
AA
A
A AA A
.00
0.01
0.10 1.00
MEAN LEAK RATE OF SAMPLE (lbs/hr)
10.00
Figure C2-29. Percent Standard Deviation Between Samples Versus
Mean Leak Rate - Valves
-------
TABLE C2-5. SUMMARY OF BAGGABLE LEAK RATE QUALITY CONTROL SAMPLE
Standard
95 Percent
90 Percent
Number
Deviation of
Reproducibility of
Confidence
of
Total
Average
Sampling/Analysis,2
Sampling/Analysis ,3
Interval About
Source
Sources
QC
Percent
Short-Term
Short-Term
A Sample Test
Type
With QC
Samples
Difference1
Variation
Variation
Result**
Valves
65
137
37.8
36.6
101.4%
± 71.7%
Pump Seals
62
133
44.7
41.9
116.2%
± 82.2%
Compressor
Seals
40
66
39.5
38.1
105.6%
± 74.4%
Flanges
7
12
40.0
39.1
108.2%
± 76.6%
Relief
Valves
16
30
18.5
19.5
54.0%
± 38.2%
Drains
14
33
71.1
59.1
163.7%
±115.8%
OVERALL
204
411
41.9
40.7
112.8%
± 79.8%
1Average % difference - average of pooled percent differences for each soure with QC sample.
Where: % diff= [original - QC leak]/(average of original and QC leak).
2Standard deviation of sampling/analysis short-term variations - estimated standard deviation of the
sampling and analyses procedures for nonmethane hydrocarbons. Estimated from the pool individual
percent differences for each QC sample.
395 percent reproducibility of sampling/analysis short-term variations - quantity that will be
exceeded only about 5 percent of the time by the difference of two test results on a given source
under similar process conditions. The quantity is equal to 2.77 x standard deviation.
**90 percent confidence interval - when taken about a single test result, 95 percent of these intervals
would be expected to include the "actual" leak rate (without bias considerations); the quantity is
equal to 1.96 x standard deviation.
-------
This standard deviation of 40 percent is composed of
variation due to analysis, sampling train components, sampling
team effect, and inherent short-term variability in the leak
rate. In Section 2.2, the standard deviation for the THC
analysis was shown to be about 2.4 percent. In Section 2.3 the
standard deviation for sampling and analysis of standard gases
was shown to be about 17 percent. No significant differences
between sampling teams or sampling carts were found, therefore
a significant portion of the variability in the leak rate
quality control samples is probably due to short-term changes
in the leak rate. These changes can be attributed to varia-
tions in process conditions, environmental changes, and random
variations in the actual leak rate.
Figures C2-30 through C2-32 are examples of the
short-term variation in leak rates for selected sources from
Table C2-4. Significant changes in relatively short periods
of time are obvious from these graphs.
2.4.2 Variance Component Analysis
The variability when measuring the leak rate from a
single source can be put in proper perspective for this program
by comparing this variation due to short-term variation and
sampling/analysis with the total variability of the leak rate
data from all sources. Statistical analyses of variance tech-
niques can be used to separate the total variability of the
measured leak rate into its various components.
Table C2-6 summarizes the estimation of variance com-
ponents for the six baggable source types. The variation of
the logarithm of the leak rate is broken down into four com-
ponents of variation:
C-74
-------
0.028
I 0.024
co
_j
~ 0.020
b-
<
" 0.016 H
<
Hi
a 0.012
5 0.008-
Hi
£
0.004-
0
1/18
ID = 13VA231
*
*
/I
1/27
1/31 2/2 2/3
~
*
ON N £ (71
o o o o o
o o o o o
DATE (OR TIME)
2/6
0.9 -
0.8-
0.7-
0.6-
« 0.5
cc
•o
©
3 0.3-1
O
s
0.2-
0.1.
0
ID = 13VA17
VALVE NO. 13VA17
-r—
5/1
-!—
5/3
T 1 1
'5/9 9:00 11.-00 1:00 3:00
Da I# (or Tim*)
Figure C2-30. Short-Term Variation in Leak Rate
Valves
C-75
-------
0,07 ¦
g 0.06 •
OQ
~ 0.05
01 0.04
*
<
UJ
-1 0.03
0
u
1 0.02
<
UJ
* 0.01
0
ID = 15PU27
*
'»•
V15
¦^v
I I
5/10
DATE
0.0085
0.0075-
0.0065 -
0.0055 -
0.0045-
0.0035 -
0.0025 -
0.0015
0.0005
0
ID = 22PU18
PUMP SEAL NO 22PU18
1 1 1 1 1 1—
7/1 9:00 11:00 1:00 3:00 5:00 7:00
J\r
——1 1 1 1 I 1
7(5 6:00 8:00 1 0tt0 12:00 2:00 4:00 6:00
Oata (orTlma)
Figure C2-31. Short-Term Variation in Leak Rate
Pumps
C-76
-------
€
£ 6.0 •
CC 5.0 -
•J 4.0 •
•o
0> '
w
S 3 0 ¦
(0
©
s
2.0 -
RELIEF VALVE NO. 27RV3
1.0 ¦
4/27
4/27
Date
DRAIN NO. 170R20
=- 0.03 -
0.02S -
"g 0 .02 -
0.01 -
0.005 -
I
V2 1WO 1:00 MO
T
11:00 12:00 1:00
S3
DATE (or Time)
Figure C2-32. Short-Term Variations in Leak Rates -Relief
Valves and Drains
C-77
-------
TABLE C2-6. VARIANCE COMPONENT ANALYSIS - BAGGABLE SOURCES - LN (LEAK RATE)
VALVES PUMP SEALS FLANGES
SOORCE OF
VARIATION
if
Variance
Component
Percent of
Variation
df«
Variance
Component
Percent of
VarlatIon
df*
Variance
Component
Percent of
Variation
Refineries
8
-0.17*
0.0
8
-0.064
0.0
7
-0.141
0.0
Unit/Refinery
43
0.881
14.2
43
0.666
12.0
16
0.973
0.0
Sources/Unit
573
4.952
80.1
326
4.350
78.2
38
4.591
92.1
Saapling/Analyela
Short-tera Variations
116
0.351
5.7
137
0.544
9.8
12
0.396
7.9
TOTAL
7*0
6.184
100X
514
5.560
100Z
73
4.987
100Z
COMPRESSOR SEALS RELIEF VALVES DRAINS
SOORCE or
VARIATION
df*
Variance
Component
Percent of
Verlatlon
df*
Variance
Component
Percent of
Variation
df*
Variance
Conponent
Percent of
Variation
Refinerlea
12
-0.207
0.0
8
0.840
12.7
8
-0.224
0.0
Unit/Refinery
24
3.907
59.0
9
0.822
12.5
23
2.354
37.0
Source/Unit
140
1.927
29.1
40
4.685
71.0
17
2.648
41.6
Saapllng/Analyals
Short-tera Varlatlona
76
0.792
11.9
31
0.250
3.8
34
1.360
21.4
TOTAL
252
6.626
100Z
88
6.598
100Z
82
6.362
100Z
• df ¦ iafiMi of (rwdei
-------
• refineries,
• units within a refinery,
• individual sources within a unit, and
• sampling/analysis, short-term variations.
The estimation technique assumes all the components are random,
i.e., a random selection of refineries, of units within a
refinery, of sources within a unit, and of samples from a
particular source. The degrees of freedom in the table is the
number of independent pieces of data available for estimating
the component of variation.
As can be seen from Table C2-6, the largest percent-
age of the variation in the log leak rate is due to individual
sources except for compressor seals. The variation due to
differences between refineries is negligible for all sources
except relief valves. The percentage variations due to the
sampling and analysis procedures and short-term leak varia-
tions range from 3.8 percent for relief valves to 21.4 percent
for drains. This component for valves is 5.7 percent. The
standard deviation of 40 percent discussed above is not large
when compared to the total variability of the leak rate in the
data base where leak rates span seven orders of magnitude.
Since the emphasis in this program is on overall estimates
rather than estimates of individual leak rates, the variability
of the sampling and analysis process is certainly acceptable
for the program objectives.
C-79
-------
3.0
QUALITY CONTROL FOR HYDROCARBON SCREENING PROCEDURE
The screening of sources during this field sampling
program was accomplished with sensitive portable hydrocarbon
detectors. The principal device used in this study was the
J. W. Bacharach Instrument Company "TLV Sniffer." The Century
Instrument Company Organic Vapor Analyzer (Model OVA-108) was
used for some screening, but not enough data were available for
developing correlations. The instruments were calibrated daily
with standard mixtures of hexane in air. The OVA-108 and TLV
Sniffer give direct readings of hydrocarbon concentrations in
ppm by volume.
When screening valves, pumps or compressors, the
probe of the hydrocarbon detector was normally placed as close
as possible to the intersection of the shaft with the sealing
device (0 cm). The probe was held at this location for a
minimum of five seconds. The detector reading was recorded.
This was repeated at three other points 90 degrees apart around
the shaft. The maximum reading was used as the screening value
for the sampling criterion. This maximum may not be the true
maximum which could be obtained by screening over the whole
360 degrees. Subjective evaluations by the engineers doing
the screening indicated that the difference in maximums would
not be significant.
Flanges were screened by placing the detector probe
at 2-inch intervals all around and right against the outside
perimeter of the flange interface. The maximum detector read-
ing was recorded. Drains were similarly screened. The detector
probe was placed at 2-inch intervals around the perimeter of
the drain. The maximum measured hydrocarbon concentration was
recorded.
C-80
-------
Relief valves were screened by placing the instrument
probe at the valve "horn" exit. The screening value obtained
at that point was used as the sampling criterion.
For evaluation purposes, some readings were also
obtained five centimeters from the source for all source types.
Appendix A discusses further details of the operation
of these hydrocarbon detectors. This section discusses the
various quality assurance activities related to the screening
devices:
• calibration checks,
• repeatability of screening,
• intra- and inter-screener comparisons,
• relationships between TLV and OVA
instruments, and
• screening versus soaping techniques for
leak detection.
3.1 Screening Device Calibration Checks
The TLV and OVA instruments were calibrated each day
they were used. Standards of 500-525 ppmv and 2,000 ppmv
hexane in air were used to get a two point calibration each
day. Before a recalibration was made each day, the values
obtained from the instrument were recorded. This served two
purposes:
• check for instrument damage or
malfunction, and
C-81
-------
• document the stability of the daily
calibration.
The results of these calibration checks at the last
three refineries visited are shown in Figure C3-1 for the lower
standard and in Figure C3-2 for the high standard. Three dif-
ferent TLV instruments were used at these refineries. The
percent differences from the standards are plotted in Figures
C3-3 and C3-4.
Table C3-1 gives a statistical summary of these data.
None of the devices gives any indication of a consistent bias
(or drift) at either the high or low level. The maximum per-
cent differences found were always less than 20 percent of the
known concentration.
Based on these data, it is concluded that the daily
calibration of the screening devices at two levels using stan-
dard gases was adequate for obtaining consistent, unbiased
readings.
3.2 Repeatability and Reproducibility of the Screening
Procedure
Repeatability and reproducibility are estimates of
the variation inherent in multiple screenings of individual
sources. The 90 percent repeatability is the maximum differ-
ence expected between two screenings by the same operator
within a short time period (less than 3 minutes). A difference
of greater than the repeatability statistic would be expected
less than 10 percent of the time. The 90 percent reproduci-
bility is the maximum difference expected between two screen-
ings by different operators within a short time period (less
than three minutes).
C-82
-------
+
4
+
~
+
+
~
+
+
+
4
+
~
+
+
+
+
+
4
4
4-
4-
4-
4-
4
4
4-
4-
4-
4-
4
4-
4
4
4-
Syrabol is code for device:
X TLV 7C7COA
2 TLV 7C7CX6
3 TLV 7ESB55
3*1 11
3
3 2 5 5
5 15
2
5
12 1
1
1 1
1 3
5 I
A
2 3
1 2 3 H 5 6 f e 9 10 11 12 13 14 lb 16 17 ltf 19 20 21 22
ItST
3 OUS "IUULN
Figure C3-1. Calibration Checks for TLV Sniffer - Low Standard
-------
Symbol Is code for device
n
I
00
4>
22U0
T
L
y 21 bO
It
A ^U0
D
I
N .
c 2l)b0
21)00
19bO
8
D l'U0
lUbO
1*00
2 3
1
TLV
7C7C04
2
TLV
7C7C16
3
TLV
7ESB55
1
a l
9 10 11 12 13 m lb 16 1/ 10 19 20 *1 22
TtS I
NOIL :
3 CHS MIUCtN
Figure C3-2. Calibration Checks for TLV Sniffer - High Standard
-------
Symbol is code for Device:
O
CO
Ln
2b
p *0
C
t
0
R
t
0
W
s
T
D
lb
1U
-b
¦10
-lb
1 TLV 7C7C04
2 TLV 7C7C16
3 TI.V 7ESB55
1 3
J 5
3 1
1
6 1
1 2 S 4 5 b t a 9 10 11 12 14 It lb 16 1/ 1? 19 *0 i*l Z*
TCST
Nott : n ims mjuuln
Figure C3-3. Percent Difference for Calibration Checks for TLV
Sniffer - Low Standard
-------
o
00
(Ti
10. 0
' 7.5
C
b.n
D
I
r
r 2.5
T
o 0.0
I -2.5
G
s -b.o
T
-7.5
•10.0
Symbol is Code for Device :
1 TLV 7C7C04
2 TLV 7C7C16
3 TLV 7ESB55
1 2
1
¦i—a-
1 3 1
2 3 1
3
1 1
3 1
1 2 3 H
3
b fa
f a 9 AO 11 12 13 1<4 lb lb l7 1«» 19 20 iil *£.
TLS I
NJtt: 3 cms niuuKu
Figure C3-4.
Percent Differences for Calibration Checks for TLV
Sniffer - High Standard
-------
TABLE C3-1. STATISTICAL SUMMARY OF CALIBRATION CHECKS
Instrument 1
Low Standard (525 ppm)
High Standard (2023 ppm)
Number
of Checks
21
21
Average Average
Difference Percent
(ppm) Difference
- 4.4
-28.0
-0.8
-1.4
Minimum
Percent
Difference
-12.3
-11.0
Maximum
Percent
Difference
19.5
3.8
Standard 95Z Confidence
Deviation Interval for
of X Average Percent
Difference Difference
7.3
4.1
(-4.1
(-3.3
2.5)
0.5)
Instrument 2
Low Standard
High Standard
-16.9
- 4.8
-3.2
-0.2
- 6.1
-14.3
3.8
2.9
5.4
2.9
(-7.7 , 1.3)
(-4.7 , 4.3)
0
1
00
-«J
Instrument 3
Low Standard
High Standard
20
20
- 5.3
-19.8
-1.0
-1.0
-14.3
-11.0
7.9
8.7
5.2
4.1
(-3.4 , 1.4)
(-2.0 , 0.9)
Percent difference = (Measured - Standard) x 100/Standard
-------
The repeatability of the screening process was
investigated by performing repeated screenings on the same
source by the same operators. The issue of repeatability is
again complicated by short term variation in the leak rate.
However, an evaluation of the repeatability of the screening
procedure is important for future use of the screening devices
in standards and maintenance programs. Both the TLV sniffer
and the OVA-108 instruments were used to screen at the sources
and 5 cm from the source. The absolute value of the percent
difference between the duplicate readings is plotted against
the mean of the duplicate readings in Figures C3-5 and C3-6
for maximum reading at the source using a TLV and OVA, respec-
tively. Most percent differences are less than 75 percent for
the TLV and below 40 percent for the OVA. Figures C3-7 and
C3-8 show the same plots of percent difference for TLV and QVA
readings at 5 cm from the source. The percent differences for
the TLV tend to be higher, indicating that the method is not as
repeatable as screening directly at the source.
Quality control studies were run on the TLV sniffer
to determine the reproducibility of the measurement method.
Between one and five sources at selected refineries were
selected with screening values between 200 and 10,000 ppm.
Each day that screening was done, at least one team would
screen each of the sources. Duplicate readings were some-
times performed on each device, both at the source and 5 cm
from the source. Figures C3-9 and C3-10 illustrate typical
results obtained from the repeated screenings using the TLV
sniffer at the source.
Within a day, the screening results from each team
were generally close. A visual comparison of duplicate readings
C-88
-------
legend: a = 1 OBs, B = 2 OBS, ETC.
£00
A
175
ISO
Jzi
A
90% Repeatability » 117%
100
A
75
A A
A
SO
2S
0
+ A
~A
A
AM
A
A
A A A
AAA
A A
A A A
A A AAA A
A A A B A
A A BA A AA A B
AAA B A A
A A
AC AAA
0
LOO
1000 10000 lOOoOO
MEAN T»-V, ppmv
Ftgure C3-5. Percent Difference Between Duplicate TLV
Readings - at the Source
-------
LEGEND: A ¦ 1 OBSt 8 a 2 DBS. ETC.
128
0
1
vO
o
11 i
00
0
1 WH
f
F
*6
3i
907. Repeatability
W
-4"-
0
A
100t>0
100
1000
MEAN OVA. ppmv
100()00
Figure C3-6. Percent Difference Between Duplicate OVA
Readings - at the Source
-------
LECeno: a - i oes« b c 2 obs, trc.
t
*00 ~ A
t
t
A
90 7. Repeatability =¦ 184 %
PS ~
+
*
* A
ISO ~
*
t
t
ItS 4
t
A
A
B
A
t
t
}oo ~
A
1
75 +
* A 8
t
t
50 ~ A
*
t
t A A
29 *
t
*¦ A AA
t
0 <¦ A A
A
>
>
¦>
>
A*
A
A
A A
A B
A
-+¦ +
0 100 XOOO lOOoO lOOoOO
MEAN SCH TlV, ppmv
Figure C3-7. Percent Difference Between Duplicate TLV
Readings - 5 cm from Source
-------
LCGENO; a = 1 OBs. B =2 oas. ETC.
• 0
n
i
vO
to
TO
40
30
D
I HO
F
f
30
20
10
too
907._ Repeatability ^5Z_
1000
MEAN Ova • PP™
lOuOo
1O0000
Figure C3-8. Percent Difference Between Duplicate OVA
Readings - 5 cm from Source
-------
SYMBOL IS CODE FOR SCREENER
lOOOOO
T
L
V 10006
5 800
2
i
1
.2.
100
1 2
note: H 08S HibDEtJ
8
DAV
16
It
ta
13
14
IS
Figure C3-9. TLV Quality Control - Daily Readings at the Source,
Refinery F, Pump Seal 75
-------
SYMBOL IS CODE FOR SCREENER
100000
n
i
T
L
V
A
T
T
H
e
s
0
u
K
C
e
joooo
1060
100
.2 <>-,
.2 2 1,
S 6.7 8
DAr
9 lO 11 |2 13 ^
NOTfc: 3 005 MTOOEN
Figure C3-10. TLV Quality Control - Daily Readings at the
Source, Refinery F, Pump Seal 92
-------
by the same team can also be seen. Figures C3-11 and C3-12
represent the paired TLV readings taken 5 cm from the source
on the same sources. Note that the magnitude of the concen-
tration is different. The magnitude of the difference between
operators is larger, in one case, and about the same in the
other case.
The absolute value of the percent difference between
operators was calculated and is plotted against the mean value
for operators in Figures C3-13 and C3-14 for TLV at the source
and TLV at 5 cm, respectively. For TLV at the source, most
values had less than 60 percent difference. Again, for TLV at
5 cm, the magnitude of the percent differences is larger.
A variance component analysis was run on both TLV
sniffer and OVA-108 data from the reproducibility and repeat-
ability studies on selected devices. The results of this
analysis, broken down by device, are given in Tables C3-2 and
C3-3 for the TLV and OVA. The pooled standard deviation for
all TLV repeat readings at the source (all devices) is 0.50 In
(screening value), yielding a 90 percent repeatability of 117
percent.
The effect of different operators can also be
observed in this analysis. Pooling the data from pumps and
valves, the standard deviation is 56 percent. Ninety percent
reproducibility is then equal to 130 percent. The pooled
standard deviation for all OVA (at the source) repeat readings
is 30.5 percent producing a repeatability of 85 percent. Note
that the repeatability of the OVA instrument appears better
than that of the TLV, and that there are less data available
to evaluate the OVA.
C-95
-------
100000
SYMBOL IS CODE FOR SCREENER
T
L
V
A
T
toooo
looo
0
1
V£>
CTi
c
K
100
2
a
1
3
7 8
DAT
2 2 12
11 1 1
9 10 11 12 13 1H IS
NOTt:
£ 06S hidden
Figure C3-11. TLV Quality Control - Daily Readings - 5 cm from
Source, Refinery F, Pump Seal 75
-------
100000
SYMBOL IS CODE FOR SCREENER
10000
1000
100
2
Z
2.
1
10
ll
12
Day
2
13
1
1
2
2
l **
mote:
1 OBS HIDDEN
Figure C3-12. TLV Quality Control - Daily Readings 5 cm from
Source, Refinery F, Pump Seal 92
-------
Legend; a * i oes. 8 » a oes. £Tt.
907. Reproducibility » 130%
A
A
¦ i
A
A
AA
A A
A A
0 100 1000 10000 100000
HE AN TLV, ppmv
Figure C3-13. Percent Difference in TLV Readings at the Source
by Different Screeners
-------
LEGEND: A = J OBS- 9=2 oBS. FTC.
0
1
VO
VO
160
144
126
*
il2
0
1
F
r 9&
9
c
H 80
64
48
32
90% Reproducibility ~ 247%
100
1000
HEAM SCN TLV . ppmv
10000
-------
100qOQ
Figure C3-14. Percent Difference in TLV Readings 5 cm from
Source by Different Screeners
-------
TABLE C3-2. VARIANCE COMPONENTS FOR TLV MEASURED AT THE SOURCE
Variance Source
TOTAL
INDIVIDUAL VALVES
DAY
OPERATOR
REPEAT
Valves - Ln (screening value)
Degrees of Freedom Variance Component Percent
155 2.847 100
5 1.384 48.6
70 1.134 39.8
39 0.060 2.1
41 0.269 9.5
90% Repeatability = 121%
90% Reproducibility = 134%
Pump Seals - Ln (screening value)
Variance Source
Degrees of Freedom
Variance Component
Percent
TOTAL
46
0.427
100
INDIVIDUAL PUMPS
1
-0.008
0.0
DAY
27
0.192
44.9
OPERATOR
10
0.068
15.9
REPEAT
8
0.167
39.2
90% Repeatability = 95%
90% Reproducibility - 113%
C-100
-------
TABLE C3-3. VARIANCE COMPONENTS FOR OVA MEASURED AT THE SOURCE
Valves Ln (screening value)
Variance Source Degrees of Freedom Variance Component Percent
TOTAL 23 2.342 100
INDIVIDUAL VALVES 2 1.799 76.8
DAY 15 0.401 17.1
REPEAT 6 0.141 6.1
90% Repeatability = 87%
Pump Seals - Ln (screening value)
Variance Source Degrees of Freedom Variance Component Percent
TOTAL 15 0.908 100
INDIVIDUAL PUMPS 1 0.359 39.5
DAY 10 0.528 58.2
REPEAT 4 0.021 2.3
90% Repeatability = 36%
C-101
-------
Table C3-4 contains the results of the variance
components test run on the 5 cm TLV readings. The pooled stan-
dard deviation for repeat readings is 0.79 In (screening value)
and the pooled 90 percent repeatability is 184 percent. This
high repeatability figure shows the 5 cm method to be more
variable than screening at the source. Reproducibility was
also calculated by pooling the variance from both devices that
described the operator effect. The standard deviation is 1.06
and the percent reproducibility is 246 percent, again much
higher than that for screening at the source.
The OVA-108 screening device data from 5 cm was also
checked for repeatability (Table C3-5). The pooled standard
deviation for repeated readings is 0.28 In (screening value)
and the percent repeatability is 65 percent. The repeatability
for the 5 cm OVA readings is slightly better than that for OVA
screened at the source (72 percent) but the difference is not
statistically significant at the 95 percent test level.
3.3 Relationships Between TLV and OVA Readings and Leak
Rates
TLV Versus OVA
The TLV sniffer was the primary screening device used
on this program, but the OVA-108 was also available and used
for some screening. During the duration of this program,
Radian did a study for six San Francisco Bay Area refineries
in which the OVA was used for screening. A special study was
run to relate TLV and OVA readings. This study is reported
here for completeness.
C-102
-------
TABLE C3-4. VARIANCE COMPONENTS FOR 5 CM TLV
Valves - Ln (Screening Value)
Variance Source Degrees of Freedom Variance Component Percent
TOTAL 101 4.613 100
INDIVIDUAL VALVES 4 2.326 50.4
DAY 57 1.327 28.8
OPERATOR 19 0.235 5.1
REPEAT 21 0.725 15.7
90% Repeatability = 198%
90% Reproducibility = 228%
Pump Seals - Ln (Screening Value)
Variance Source
Degrees of Freedom Variance Component
Percent
TOTAL
46
1.484
100
INDIVIDUAL PUMPS
1
-0.046
0.0
DAY
27
0.229
0.0
OPERATOR
10
1,124
75.8
REPEAT
8
Q.360
24.2
90% Repeatability =
139%
90% Reproducibility
- 283%
C-103
-------
TABLE C3-5. VARIANCE COMPONENTS FOR 5 CM OVA
Valves - Ln (Screening Value)
Variance Source Degrees of Freedom
TOTAL
INDIVIDUAL VALVES
DAY
REPEAT
23
2
15
6
Variance Component Percent
4.073 100
2.774 68.1
1.189 29.2
0.110 2.7
90% Repeatability = 77%
Pump Seals - Ln (Screening Value)
Variance Source Degrees of Freedom Variance Component
TOTAL
INDIVIDUAL PUMPS
DAY
. REPEAT
15
1
10
4
0.435
0.056
0.351
0.02
Percent
100
12.9
80.7
6.4
90% Repeatability = 39%
C-104
-------
One hundred and twenty valves with TLV screening
values from 30 ppmv to 100,000+ were each screened in the
following manners :
Device
TLV
TLV
Calibration Gas
hexane
hexane
Distance from Source
0 cm (at the source)
1 cm
OVA
OVA
methane
methane
0 cm
1 cm
For each of the above categories, 120 pairs of screening values
were thus available for analysis. The maximum value was
recorded following the usual screening procedure. Regression
analyses on the logarithm of the maximum screening value were
done and nomographs were developed to enable conversion to
various types of screening and devices. These nomographs are
shown in Figures C3-15 through C3-18. The 90 percent confi-
dence intervals reflect the precision of relationships as
determined from the study in the Bay area. When using these
nomographs, it is important to note that the instrument is
calibrated to the standard gas type listed on the graph.
TLV Versus Leak Rate
sampling program when the source was first located and
rescreening values were obtained nearer to the time that the
source was actually sampled. Correlations and nomographs have
been developed to relate the maximum TLV with leak rates.
These are reported in Appendix B of this report. A number of
summary statistics were evaluated before selecting the maximum
reading. Tables C3-6 and C3-7 report simple correlations
Screening values were obtained during the field
C-105
-------
UPPER LIMIT OF 90S CONFIDENCE
/ INTERVAL FOR MEDIAN
20,000 -
18,000
MEDIAN
16,000 -
LOWER LIMIT OF 90t CONFIDENCE
INTERVAL FOR MEDIAN
UJ
o
oz
14,000
UJ
X
H-
h-
<
w 12,000
<
X
UJ
C3 Z
^ o
§ O 10,000
QC UJ
H-
LEAST SQUARES EQUATION USED TO DEVELOP CHART:
0.90
-j
c
u
STANDARD ERROR ¦ 0.369
8,000
6,000
4,000
2,000 -
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
READINGS FROM OVA (CALIBRATED TO METHANE)
ONE CENTIMETER FROM THE SOURCE
Figure C3-15(a). Nomograph for Relating TLV Sniffer and OVA
Readings - Part I.
C-106
-------
200,000-
180,000
160,000
140,000
o: £
Z UJ
x
« 120,000
5 uj
O ae 2
5 <
•j v
i- o2
z X
3 o 100,000
* I—
UJ
<* a
5"
< «£
5 §§
g r 80,000
60,000
40,000
20,000
UPPER LIMIT OF 90S CONFIDENCE
INTERVAL FOR MEDIAN
MEDIAN
LOWER LIMIT OF 90* CONFIDENCE
/ INTERVAL FOR MEDIAN
LEAST SQUARES EQUATION USED TO DEVELOP CHART:
LOG io(TLVO) - 0.530 + 0.929 L0G,J0VA1)
CORRELATION COEFFICIENT - 0.90
STANDARD ERROR- 0.369
i r
60,000
I !•
80,000 100,000
-i 1 r
20,000 40,000
READINGS FROM OVA (CALIBRATED TO METHANE)
ONE CENTIMETER FROM THE SOURCE
(0VA1)
02-3879-1
Figure C3-15(b). Nomograph for Relating TLV Sniffer and OVA
Readings - Part II.
C-107
-------
42,000
38,000
34,000
30,000
26,000
52 2 22,000
2 « 18,000
p £
14,000
10,000
6,000
UPPER LIMIT OF 90S CONFIDENCE
/ INTERVAL FOR MEDIAN
MEDIAN
LOWER LIMIT OF 90% CONFIDENCE
' INTERVAL FOR MEDIAN
LEAST SQUARES EQUATION USED TO DEV&OP CHART:
LOG 10(TLVO) - 0.854 + 0.929 LOG in(TLV1)
CORRELATION COEFFICIENT » 0.93
STANDARD ERROR - 0.453
2,000
—i i 1 i i 1 i i i i
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
READINGS FROM TLV SNIFFER (CALIBRATED TO HEXANE)
ONE CENTIMETER FROM THE SOURCE
(TLV1)
02-3877-1
Figure C3-16(a). Nomograph for Relating TLV Sniffer Readings
at Zero and One Centimeter - Part I
C-108
-------
420,000
380,000
340,000
300,000
260,000
220,000
180,000-
140,000-
100,000
60,000-
20,000
/UPPER LIMIT OF 90S CONFIDENCE
' INTERVAL FOR MEDIAN
MEDIAN
LOWER LIMIT OF 90t CONFIDENCE
/ INTERVAL FOR MEDIAN
LEAST SQUARES EQUATION USED TO DEVELOP CHART:
LOG 10(TLVO) » 0.854 + 0.929 LOG 10(TLV1)
CORRELATION COEFFICIENT - 0.93
STANDARD ERROR « 0.453
i i 1 i i i r-
20,000 40,000 60,000 80,000 100,000
READINGS FROM TLV SNIFFER (CALIBRATED TO HEXANE)
ONE CENTIMETER FROM THE SOURCE
(TLV1)
02-3876-1
Figure C3-16(b). Nomograph for Relating TLV Sniffer Readings
at Zero and One Centimeter - Part II
C-109
-------
UPPER LIMIT OF 90% CONFIDENCE
INTERVAL FOR MEDIAN
MEDIAN
2000-
UJ
LOWER LIMIT OF 90%
CONFIDENCE INTERVAL
FOR MEDIAN
UJ
X
5
oe
u.
oi
UJ
>—
UJ
s:
1500-
t—
z
UJ
UJ
z
o
UJ
z
<
X
UJ
s:
o
>—
-J
<
<
o
5
a:
u.
500-
o
uj
LEAST SQUARES EQUATION USED TO DEVELOP CHART:
Log10(0VA 1) ¦ -0.424 + 0.0938 Log1Q(0VA 0)
Correlation Coefficient ¦ 0.95
Standard Error » 0.328
<
o
uj
Z
10,000
2000
4000
6000
8000
0
READINGS FROM OVA (CALIBRATED TO METHANE) AT THE SOURCE
(OVA 0)
02-4247-1
Figure C3-17(a). Nomograph for Relating OVA Readings
at Zero and One Centimeter - Part I
C-110
-------
UPPER LIMIT OF 90S
/ CONFIDENCE INTERVAL
FOR MEDIAN
20,000-
UJ
U
ae
MEDIAN
o
UJ
X
§
0£
U_
LOWER LIMIT OF 90%
/ CONFIDENCE INTERVAL
FOR MEDIAN
QC
UJ
>—
£ 15,000-
UJ
X
©
UJ
z
X
o
>—
a
10,000-
UJ
>—
<
ac
CO
•j
<
o
o
5
ac
U_
a
<
5,000-
UJ
ae
LEAST SQUARES EQUATION USED TO DEVELOP CHART:
Log10(0VAl) - -0.424 + 0.938 Log1Q(0VA 0)
Correlation Coefficient ¦ 0.95
Standard Error • 0.328
<
o
80,000
20,000
60,000
100,000
0
40,000
READINGS FROM OVA (CALIBRATED TO HETHANE) AT THE SOURCE
(OVA 0) 02-4248-1
Figure C3-17(b), Nomograph for Relating OVA Readings
at Zero and One Centimeter - Part II
C-lll
-------
5000-»
4000-
3000-
"> 2000 -
1000-
UPPER LIMIT OF 90%
/CONFIDENCE INTERVAL
/ FOR MEDIAN
MEDIAN
LOWER LIMIT OF 90%
~ CONFIDENCE INTERVAL
/ FOR HEDIAN
LEAST SQUARES EQUATION USED TO DEVELOP CHART:
Log10(TLVO) - -0.193 + 0.952 Log10(OVA 0)
Correlation Coefficient » 0.93
Standard Error » 0.384
—I—
2000
—I—
4000
6000
-I—
8000
10,000
READINGS FROM OVA(CALIBRATED TO METHANE) AT THE SOURCE
(OVA 0)
02-4245-1
Figure C3-18(a). Nomograph for Relating OVA Readings
and TLV Sniffer Readings at the
Source - Part I
C-112
-------
UPPER LIMIT OF 90%
/ CONFIDENCE INTERVAL
/ FOR MEDIAN
MEDIAN
LOWER LIMIT OF 90i
/ CONFIDENCE INTERVAL
/ FOR MEDIAN
LEAST SQUARES EQUATION USED TO DEVELOP CHART:
Log10(TLV 0) « -0.193 + 0.952 Log1Q(OVA 0)
Correlation Coefficient ¦ 0.93
Standard Error ¦ 0.384
1 l 1 i 1 1 l 1 1
20,000 40,000 60,000 80,000 100,000
READINGS FROM OVA (CALIBRATED TO METHANE) AT THE SOURCE
(OVA 0)
02-4246-1
Figure C3-18(b). Nomograph for Relating OVA Readings
and TLV Readings at the Source -
Part II
C-113
-------
TABLE C3-6. CORRELATIONS OF SCREENING VARIABLES AND NONMETHANE LEAK RATES
(Lb/hr) - VALVES (All Correlations Based on Log of Variable)
VARIABLE (2) MAX SC (3) MAX RSC (4) AVG RSC (5) 5-CM (6) N. STM (7) N. GL
1. Nonmethane Leak 0.628(584) 0.715(260) 0.739(260) 0.685(246) 0.703(251) 0.511(195)
2. Maximum Screening
Value - 0.745 0.748 0.593 0.677 0.434
3. Maximum Rescreen-
ing Value - 0.978 0.804 0.858 0.633
4. Average Rescreen-
ing Value - 0.837 0.890 0.693
5. Avg. of Maximum
5-cm Reading - 0.733 0.722
6. North Stem Reading - 0.545
7. North Gland Reading
Tabled values are r (m) _ _
Z(Xi-X)(Yi-Y)
r = simple correlation coefficient = / y^2 y ,v ??rz" where X and & are the paired
V MX1 - X; y; variables.
m = number of pairs of data observations used in computing correlation coefficient.
-------
TABLE C3-7. CORRELATIONS OF SCREENING VARIABLES AND NONMETHANE LEAK RATES
(lb/hr) - PUMPS (All Correlations Based on Log of Variable)
VARIABLE (2) MAX SC (3) MAX RSC (4) AVG RSC (5) 5-CM (6) N. STM
1. Nonmethane Leak 0.636(418) 0.578(169) 0.700(169) 0.731(160) 0.716(164)
2. Maximum Screening
Value - 0.766 0.753 0.618 0.765
3. Maximum Rescreen-
ing Value - 0.987 0.825 0.940
4. Average Rescreen-
ing Value - 0.858 0.958
5. Avg of Maximum
5-CM Reading - 0.835
6. North Stem Reading -
Tabled values are r (m)
Z(Xi-X)(Yi-Y)
r = simple correlation coefficient = ; „ —— =ry where X and Y are the paired
V ^(Xi X) (Y± Y; variables.
m = number of pairs of data observations used in computing correlation coefficient.
-------
between leak rates and selected screening statistics (including
individual readings) for valves and pump seals. The maximum
rescreening value at the source was selected because of its
high correlations and simple determination.
3.4 TLV Readings Compared to "Soap Screening"
At one refinery, a short test was made to compare
screening of sources using a soap solution with screening using
a TLV sniffer. No attempt was made to relate the two types of
screening results to leak rates because of the minimal amount
of data. The test gives only a qualitative comparison of the
two screening techniques. Following the usual screening tech-
nique on selected sources, the maximum TLV value was obtained.
Then the source was sprayed with either a "snoop" soap solution
(relatively thin) or a relatively thick solution made from
Ivory liquid soap. Then the "action" or "description" of the
soap solution was recorded.
Table C3-8 summarizes these data. The descriptions
of the soap solution were grouped into one of four classes as
described. The data are plotted in Figure C3-19. As can be
seen, the soap solution formed bubbles for all screening values
greater than 1,000 ppm except for the vertical sources and one
other valve.
C-11C
-------
TABLE C3-8. COMPARISON OF TLV SNIFFER AND SOAP SPRAY SOLUTION
Soap
Source Maxiaum TLV Description of Soap Leak rate Horizontal (H) Snoop Soap (S)
Type Reading (ppsv) Spray Application Class Vertical (V) Ivory Soap (I)
Compressor Seal
11,000
1" Bubble in 2 seconds
2 H
S
fuap Seal
6,000
No Bubbles detected
1 V
S
Valve
260
Slight Bubble formation
barely detectible
2 H
S
Valve
3,400
Very slight bubbles
barely detectible
1 H
S
Valve
10,000
rapidly foraed 1/8" bubbles
3 H
S
Valve
40,000
rapidly formed 1/4" bubbles -
sudsing
3 a
S
Valve
100k+
rapid 1/2" bubbles - spitting
solution
4 H
I
Valve
20,000
no bubbles
1 V
I
Valve
lOOfcf
rapid bubbles - spitting
4 H
I
Valve
100k+
rapid bubbles - spitting
4 H
I
Valve
10,000
rapid 1/2" bubbles
3 H
I
Valve
2,300
no bubbles
1 V
1
Valve
14,000
no bubbles
1 H
I
Valve
100k+
rapid large bubbles - clustering
4 H
I
Valve
80
no bubbles
1 V
1
Valve
lOOkf
rapid 1/4" bubbles
3 H
I
Soap Leak Rate Class 1.
no bubbles - very slight
2* slight bubbles
3.
rapid bubbles
4, rapid, spitting, clustering
-------
100,000 .
I
g: 10,000
1,000
100
10
Q
(?)
V
<3
1—
No
Bubbles
Cl)
w
w
w
V - Valve
C - Compressor Seal
P - Pump Seal
0 - Vertical Service
T-T
T
I
rapid
bubbles
(3)
slight bubbling,
barely detectlble
(2)
Description of Bubbling of Soap Solution
sprayed on source
1
rapid bubbles
clusters, spitting
(A)
Figure C3-19. Relationship of TLV Sniffer Reading and Bubbling from Soap Solution
-------
4.0
QUALITY CONTROL FOR NONBAGGABLE SOURCES
Quality control for nonbaggable sources involved an
evaluation of the accuracy and repeatability of all analytical
procedures. Sampling procedures usually did not lend them-
selves to accuracy evaluations although day-to-day variations
give an upper bound on sampling repeatability.
The procedures discussed in this section include an
evaluation of the quality control for analytical methods used
to measure emissions from cooling towers, wastewater treating
units, and process stacks. The specific sampling and analysis
procedures, including calibration procedures, are discussed in
Appendix A, "Sampling Methodology."
4.1 TOC Analysis for Cooling Tower Evaluation
Total organic carbon (TOC) assays were done during
this program with a Dohrmann DC52D TOC analyzer. This instru-
ment oxidizes organics to carbon dioxide and then reduces the
carbon dioxide to methane. The methane is measured with a
flame-ionization detector.
The instrument was zeroed using a "zero carbon water
standard" which was especially prepared for this analysis by
Radian. The water is deionized, filtered and distilled from
potassium permanganate under helium with a high reflux. This
has been proven superior to commercial standards. The stan-
dard for the analysis is 180 ppm carbon in water available
from Dorhmann. The 180 ppm standard was analyzed on a regular
basis during the time that wastewater samples were being
analyzed in the field and in Radian's laboratory in Austin.
Figure C4-1 shows the results of these 47 analyses. The
following results summarize these analyses:
C-119
-------
220-
210-
200-
190-
180-
170-
CONCENTRATION OF
STANDARD
160 I I I I I I I I I I I I I I I I I I I I I I I I I I
1 2 3 4 5 6 7 8 9 10 12 14 16 18 20 22 24
STANDARDS ANALYZED IN THE FIELD
ttt
26 28
I I I I I I I I
30 32 34 36
I I I I I I 1 I I I
38 40 42 44 46 47
STANDARDS ANALYZED AT RADIAN LAB
FIGURE C4-1. TOC Standards Analysis
-------
Average: 184.2 ppm
Standard Deviation: 13.54 ppm (7.4% of mean)
95 Percent Confidence
interval for average: (180 ppm, 188 ppm)
While these standard analyses do not indicate any bias in the
method, the variability of the standard analyses indicates that
repeat analyses of samples can differ by as much as 21 percent
(95 percent repeatability) at high TOC levels.
Figure C4-2 shows the results of the blank water
analysis done during the analyses of samples at Radian's
laboratory. These analyses averaged about 1.1 ppm with a
standard deviation of 1.8 ppm or 155 percent. These results
indicate that interference and repeatability problems could
occur in samples at low TOC levels (< 10 ppm).
The difference between the TOC concentrations at the
inlet and outlet of the cooling tower (ATOC) is the most
important statistic in calculating cooling tower emissions.
The repeatability of the TOC analysis can be evaluated by
comparing replicate determinations of ATOC on the same sample.
Figure C4-3 shows a plot of the difference between two deter-
minations of ATOC on samples from seven cooling towers.
The average difference for these 48 comparisons was
4.2 ppm. By averaging the squared differences, the standard
deviation for the method can be estimated. For these samples,
the standard deviation is also 4.2 ppm. Note that the average
ATOC's for these towers range from - 2.1 ppm to 12.2 ppm. Six
of the seven towers have an average ATOC of less than the stan-
dard deviation of repeat analysis. From the plot in Figure
C4-3, the repeatability of the analysis appears erratic; for
C-121
-------
• • •
NUMBER OF ANALYSES - 21
AVERAGE ANALYSIS - 1.13 pptn
STANDARD DEVIATION - 1.75 ppm (155%)
957o CONFIDENCE
INTERVAL FOR AVERAGE - (.33 ppm, 1.93ppm)
1—1—i i i i i i i—i—i—i i i i—i—i—i—i—i—i—r-
123456789 1011 1213141516 171819 20 21 22
ANALYSIS OF "ZERO CARBON WATER STANDARD" AT RADIAN
FIGURE C4-2. TOC BLANK ANALYSIS
-------
• •
+T-T+
• • •
• • •
1 I I I I II
2 3 4 5 6 7
TOWER 2
I I I
5 6 7
I I I I I I
1 2 3 4 5 6
TOWER 4
TTT
7 1
TT
2 3
I I II
5 6 7
I I 1 I II
1 2 3 4 5
TTT
1 2 3
t n
5 6
1 2 3 4 5 6
TOWER 1
1 2 3
TOWER 3
TOWER 5
TOWER 6 I TOWER 7
AVERAGE A ppn 1.6
12.2
-2.1
-0.9
3.9
l.S
2.3
FIGURE C4-3. TOC REPEAT ANALYSIS
-------
three of the towers, all differences are less than 5 ppm. This
erratic pattern may be due to the potential interferences dis-
cussed above.
Based on a standard deviation of 4.2 ppm, five sets
of analyses would be required to consistently measure a signifi-
cant change of 4.0 ppm in a sample. Two sets of measurements
averaged would have some negative changes unless the true
change in the sample set were greater than 6.0 ppm. Sampling
and time variations would introduce additional variability
that would further hinder the measurement of emission using
the TOC method.
The magnitude of the sampling and TOC analytical
variations caused some problems in quantifying the low levels
of emissions from the towers. The standard deviation of day-
to-day sampling results using the TOC analyses was close to the
analytical standard deviations when replicate samples are aver-
aged. It appears, then, that most of the variation in the TOC
analytical data is due to the variability of the analytical
technique.
Analytical results obtained by the purge method
(discussed below) were much more precise than those obtained
from TOC analyses.
4.2 Purge Analysis Method
The purge analysis method was used to measure purge-
able volatile orgknics for both cooling tower and wastewater
samples. For cooling tower samples, a 75 ml aliquot was used.
For wastewater aqueous phase sample, a 10 ml sample was used.
For wastewater oil phase samples, a 10 yl aliquot of oil in
C-124
-------
10 ml of water was used. A complete description of the
analytical method is given in Appendix A.
Blind standards were run to determine the percent
recovery obtainable by the purge method. Figure C4-4 graphs
the results of the standards. Almost all the results appear
to fall within two standard deviations of the mean percent
recovery of 85.4 percent. A 95 percent confidence interval
for the average recovery based on the data in Figure C4-4 is
85.4 ± 7.0 percent or 78 to 93 percent, so the recovery appears
to average less than 100 percent.
Multiple analyses of samples were performed on both
cooling tower and wastewater samples. Figure C4-5 shows the
results of the multiple analyses of samples from cooling
towers. The maximum percent difference of the multiple samples
is plotted against the mean sample concentration. The mean
percent difference for all sets of samples was 137 percent with
a standard deviation of 80 percent. This yields a 95 percent
repeatability of 221 percent.
Figure C4-6 shows the same percent difference versus
mean concentration for oil-layer wastewater samples. The mean
percent difference for all samples was 7.0 with a standard
deviation of 2.7 percent. The 95 percent repeatability then
becomes 7.4 percent.
4.3 Gravimetric Method for Evaluating VOC Content of Oil
Total volatile hydrocarbons were also measured in the
wastewater oil samples by gravimetric determination. Samples
were weighed, then stirred for eight hours. A complete
description of the analysis method is given in Appendix A.
C-125
-------
MEAN + 2o - 116.12
MEAN I RECOVERY - 85.41
MEAN - 2a • 54.71
—I 1 1 1 1 1 1 1 1—
3/14 3/15 3/16 3/19 3/20 3/21 3/22 3/23 5/10
DATE
Figure C4-4, Purge Standards Analysis
-------
300-
n
NJ
-j
a
250-
2 200-
s!
2
j- I50H
UJ
ec
e
100-
50
95* REPEATABILITY - 221X
—I i 1 1 1 I I I 1 1 1 I 1 1 1 1—
.02 .04 .06 .08 .10 .12 .14 .16 .18 .20 .22 .24 .26 .28 .30 .32
Ar^
» CI
~l—
.50 .52
MEAN PURGEABLE ORGANICS FROM COOLING TOWER SAMPLES (ppm)
Figure C4-5. Repeatability of Purge Analysis - Cooling Tower
Water Samples
-------
0
1
(-¦
ho
oo
13-
12-
11 -
s! io-
£
2 o J
? 7
bJ 6 -
Ui
C J
t:
u.
M
° 4
£
l-
951 REPEATABILITY - 7.4X
i i I I I I I I l I
600 800 1000 1200 1400 1600 1800 2000 2200 2400
MEAN PURGEABLE ORGANICS FROM WASTEWATER SAMPLES (ppm)
Figure C4-6. Repeatability of Purge Analysis - Wastewater
Samples
-------
Standard mixtures of base oil and isooctane were
prepared and analyzed. Figures C4-7 and C4-8 show the results
of these standards. The average percent recovery was 98.3
percent and recoveries within two standard deviations of the
mean range between 90 and 107 percent. A 95 percent confi-
dence interval for the average recovery is 98.3 ± 3.5 percent,
so there is no evidence that the true mean recovery is differ-
ent than 100 percent.
Duplicate gravimetric determinations were also per-
formed on wastewater samples. Figure C4-9 shows the percent
difference between duplicates plotted against the mean concen-
tration. Note that there are four very high points on the
graphs with percent differences ranging from 68 to 300 percent.
This would seem to indicate that for sample concentrations
below 0.5 ppm the method is unreliable. The mean percent dif-
ference for samples greater than 0.5 ppm concentration is 11.8
percent. The pooled standard deviation for this group is 9.5
percent producing a 95 percent repeatability of 26.4 percent.
4.4 Sampling and Analysis of Process Stack Emissions
The major emphasis on quality control for stack
sampling is on strict calibration of metering and temperature
control devices, leak testing, and laboratory standard
analysis.
Of primary concern in obtaining samples from process
stacks and ducts is that the sampling equipment is in proper
operating condition prior to and during sampling. In order to
achieve this, equipment was inspected and cleaned thoroughly,
monitoring devices checked and calibrated, and volume
C-129
-------
80 -I
70 -
60 "
40 "
o
>
20 -
10 ¦
3
5
9
1
7
8
6
10
4
0
Figure C4-7. Gravimetric Standards Volatile Loss "During
Eight Hours
C-13C
-------
0
1
t-1
u>
106-
104
102
« 100-
98-
o
u
oc
M
96-
94-
92-
90-
MEAN +2a = 106.8%
MEAN PERCENT RECOVERY = 98.3%
MEAN - 2o - 89.8%
T
T
10.05 25.5 50.0
STANDARD LEVEL (ppm)
Figure C4-8. Standards for Gravimetric Determination
-------
50
46
42
38
34
30
26
22
18
14
10
6
2
(300)
> <283)
(150)
• (68)
95% REPEATABILITY = 26.4%
t 1—i 1 1 r-*-i 1 1 1 1 1 1 1 1 1—r"
.2 .4 .6 .8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4
MEAN VOLATILE HYDROCARBON CONCENTRATION (ppm)
Figure C4-9. Repeatability of Gravimetric Determination
-------
measurement devices calibrated prior to sampling. During
sampling, equipment was monitored continuously for proper
operation.
One of the most important methods involved to assure
that a proper sample is obtained is the calibration of sample
volume metering devices. In this program the devices used are
dry gas meters and rotameters. The dry gas meter used in the
EPA Method 5 train was calibrated generally in conformance to
EPA's publication Maintenance, Calibration, and Operation of
Isokinetic Source Sampling Equipment.6 Instead of using a wet
test meter as a standard, a Hastings Model AHL-5 flowmeter was
used. This instrument traveled much better than a wet test
meter and calibration of the unit remained constant. Calibra-
tion was performed over a similar range as suggested in the
above publication and calibration correction factors for the
gas meter and orifice meter in the sampling unit were calcu-
lated according to suggested procedures. Similarly, the gas
meter in the SASS train was calibrated but only at one flow
rate, close to the expected constant sampling rate. Meter
correction factors were calculated. The rotameter in the grab
sampling train was checked rather than calibrated as any differ-
ences in the measured range of rates between the rotameter and
the standard were negligible.
Proper operation of equipment was monitored continu-
ously during all sampling activities. Equipment was shut down,
inspected, and repaired if required, before continuing sampling.
Upon completion of sampling, great care was taken with sampling
containers in order to eliminate contamination of the samples
prior to analysis in the laboratory trailer.
C-133
-------
The primary quality control procedure during the
analysis of the stack samples was the analysis of standards.
The various methods used in the chemical analysis of these
samples is described in detail in Appendix A, "Sampling
Methodologies." Blind standards were analyzed for aldehydes,
sulfur gases, and N0X. Table C4-1 contains the results of
these standard analyses. Figures C4-10 and C4-11 show the
percent difference from the standard plotted versus the stan-
dard concentration for the aldehyde and sulfur species
standards.
The percent differences for the 28 aldehyde stan-
dard analyses average 0.8 percent with a standard deviation
of 5.2 percent. The variability appears greater at the lower
concentration levels (about ± 12 percent) than for the higher
concentration standards (about ± 6 percent). The aldehyde
analysis procedure is concluded to be unbiased with a precision
averaging about ± 10 percent.
The percent differences for the 18 sulfur analyses
averaged 0.5 percent with a standard deviation of 14.6 percent.
Only two standards above 100 ppm were tested. The percent dif-
ferences ranged from - 39 to + 20 percent, but only 3 of the 18
analyses were low. The overall accuracy (including both bias
and precision) of the method for concentrations below 100 ppm
is about ± 30 percent.
The three standard analyses for N0x ranged from 21 to
73 percent low, indicating potential inaccuracies in the
analytical method utilizing potassium dichromate-aqueous
sulfuric acid solution. A chromotropic acid method of
--134
-------
TABLE C4-1. BLIND STANDARDS FOR STACK SAMPLE ANALYSIS
ALDEHYDE STANDARDS
SULFUR SPECIKS STANDARDS
0
1
I—1
LO
Ul
KNOWN
"LAS
DI»"F
PDIFF.
*06.0
447.6
-<~0
8
-10.029
H06.8
•~•+0. t>
-33
7
-8.284
OC6.U
6/1.0
-65
0
-8.063
406.a
<~«: 6.3
-19
b
-*,71*
•068.0
(<210.0
-142
0
-3.491
"~06.0
4i2.1
-5
3
-1.303
4068.0
tlil.U
-53
0
-1.303
4Ui8.0
<~103.0
-35
0
-0.860
40b8.«
4103.U
-35
0
-0.860
HOfO.U
<~103.0
-35
0
-0.060
40feU.U
<~04)5 . 0
-17
0
-0.418
406.»
<~05.0
1
8
0.442
Hte.a
<~05.0
1
8
0.442
406.8
405.0
1
8
0.4<*2
•~ue-a.y
<~030.0
18
0
0.442
4068.0
<~050.0
16
0
0.442
4060.0
<~032.0
36
0
0.885
MU61.U
<~032.0
36
U
0.883
4C6.8
3*7 . 9
8
9
2.168
•»ofce.«
39bl.O
107
0
2.630
4060.U
3943.0
125
0
3.073
O06.u
7 f2. 0
34
0
4.218
-------
12
-6
¦12
* A
Legend; A =
1 OBS, B = 2 OBS, ETC.
* A
* A
A
B
t >
A
0
* A
* C
D
U
t A
A
*
A
t A
* A
A
* A
+ ---
-----~
n 400
BOO 1200
1600 200u
2400 2600 3200 3600 4000 4400
4800
MEASURED CONCENTRATION, ppa
Figure C4-10. Blind Standards Analysis - Aldehydes
-------
0
1
M
LO
p
E 2H
R
C
E 16
N
T
D 8
I
I °
E
R
E
N
C
E "16
F
R -2,
M
K -32
N
0
W "to
N
• 8
A A
A
AB A
Legend: A = 1 OBS, B = 2 OBS, ETC.
0 2S 50 7b 100 125 150 175 200 225 250 27S 400 325 350 A/5
MEASURED CONCENTRATION, ppm
FIGURE C4-11. Blind Standards Analysis - Sulfur Gases
-------
analysis was tried. This method was finally used exclusively
because it appeared to be more reliable and accurate than the
potassium dichromate method.
C-13S
-------
5.0
DATA QUALITY CONTROL AND VALIDATION
A comprehensive data base management system was
required to handle the data generated in this program. The
total data obtained during the program resulted in more than
one-half million records. Initially data was loaded into
sequential files on a UNIVAC 1108 system. As requirements
for statistical analysis, reporting, editing, and other data
manipulations increased, it became necessary to obtain a data
management/analysis software package.
The SAS system on Boeing Computer Corporation's IBM
370-168 TSO system was used to manage and analyze the data
after six refineries were completed. SAS offered extensive
statistical procedures, as well as data manipulation capabili-
ties. The data base was kept on on-line disk storage with
weekly backups on tapes.
5.1 Data Collection and Coding
Formatted data sheets were provided for field and
laboratory use in recording all data. These forms were
designed so that keypunching could be done directly from
the original sheets, thus eliminating transcription errors.
Data sheets were collected daily in the field and reviewed
by the field supervisor for completion and general accuracy.
At the end of sampling at each refinery, the completed forms
were hand-carried to Radian, sorted, and immediately bound
in notebooks. Then a data analyst reviewed each form and
completed the coding of certain variables. Next the data
sheets were keypunched and verified.
C-13S
-------
5.2
Computer Files
Sequential files of data were developed for each of
the source types studied. The data were validated during the
loading process by the following means:
• Each variable was checked for appropriate
numeric or alphanumeric characters.
• An acceptable "range of values" was
checked for each variable. These were
established by knowledge of possible
codes as well as engineering judgment
on process variables.
• All leak rates were calculated by com-
puter code in the load program. The
computed leak rates were printed out
and compared to previously hand calcu-
lated values.
• Each variable in a data record was
assigned weights describing the importance
of the variable toward using the data
record (e.g., unit and ID were considered
essential variables). The weights were
used to produce an error report after each
data file was loaded. The records with a
large error code were checked and correc-
tions made by referring to the original
data sheets.
C-140
-------
5.3
Quality Control Checks
After a new data file was created, a number of
reports were prepared for both reporting and quality control
purposes. The entire data set was printed in an easily
readable format for future reference. Selected key variables
were printed out in a report. This report was checked against
the original field data, in some cases laboratory notebooks,
by a data analyst. A formatted report of process' variables
and leak rate data was prepared, reviewed by Radian engineers,
and sent to the refinery from which the data were collected
for further revision and verification.
A number of statistical reports were run to further
validate the data. A report generated the mean, standard
deviation, minimum, and maximum for each variable in each
refinery. Another report generated sample correlation sta-
tistics for all combinations of continuous variables. Two-way
plots and frequency tables were produced for each process
variable and the leak rate. Control charts were used to dis-
play laboratory measurements made over time. All of these
plots and reports were studied for potential outliers and for
possible trends and relationships among the variables.
5.4 Data Analysis Files
After the data from each refinery had been validated,
an "analysis" data base was developed. This data base was used
to produce summary reports, develop correlations, develop emis-
sion factors, and to do statistical analyses to summarize the
progress of the project. A number of new variables such as the
process stream classifications were created as the analysis
data base was produced.
C-141
-------
Other operations performed on the data sets
included:
• Leak rates from any quality control samples
on each source were averaged with the origi-
nal sample result to get one average leak
rate for each individual source.
• A process stream classification variable
was created from individual process stream
codes.
• Logarithms of both leak rate and screening
values were produced.
• Data sets with critical missing parameters
were omitted from the files.
• Summary statistics of screening values
such as the average and maximum value
were developed and included in the data
base.
• The files were sorted by units and leak
rates within units. This consolidated
the data from each refinery into summary
files.
Additional specific data bases were created to handle
particular aspects of the program. These include:
C-142
-------
• a short-term maintenance file for valves
selected in the last four refineries (and
at Refinery "F") for maintenance study,
• a file of leak rate quality control data, and
• special study data files.
These files are currently kept on disk storage for quick analy-
sis, reporting, and editing. Each file is backed up on a
weekly basis on tape. Editing is done either by updating the
central file and recreating the data bases, or by updating the
individual data bases as well as the central file.
C-143
-------
6.0
STATISTICAL PROCEDURES FOR ANALYZING EMISSIONS DATA
A number of statistical analysis procedures were used
in analyzing the emissions data developed during this program.
The discussions in the previous sections have shown that hydro-
carbon measurements were not very precise (precision was
usually greater than ± 50 percent) and that variability of leak
rates from different sources spanned several orders of magni-
tude. This extreme variability made the use of properly
selected statistical models and techniques very important in
extrapolating the data collected during this program to the
population of fugitive emissions.
The estimation of emission factors was one important
objective of this program. Because of the high degree of
skewness in the distribution of nonmethane leaks rates from
baggable sources, conventional statistics were inadequate for
efficient estimation of emission factors and their variances.
In addition to the skewness, a large percentage of the sources
studied were considered "nonleaking" (i.e., had a screening
value < 200 ppm) . These sources affect the emission factor
and therefore had to be considered in developing estimates for
these factors. Another statistical problem which had to be
addressed in developing the emission factors was the estimation
of leak rates for sources which screened greater than or equal
to 200 ppmv but were not sampled for economic reasons.
The population to which the data from this study can
be extrapolated is the total number of sources from all United
States refineries. The selection procedure for sources was
described in Appendix A of this report. For this analysis, it
is assumed that a random selection was made for refineries,
units within a refinery, and sources within a specific choice
C-144
-------
variable category within a unit. The "true value," e.g., of
an emission factor, is an abstract concept. Essentially, this
"true value" is that number which would be obtained if at a
given point in time all sources of a particular type in the
population could be sampled, analyzed and averaged.
This section discusses the statistical procedures
used in the following areas:
• estimating emissions from nonsampled
sources,
• statistical models for leak rate
distributions,
• development of confidence intervals, and
• development of nomographs.
6.1 Estimating Emissions for Nonsampled Sources
Due to time and equipment constraints, it was not
always possible to sample all sources that screened greater
than 200 ppmv. At the fifth refinery, a sampling strategy was
developed to reduce the sampling workload. All sources screen-
ing greater than 10,000 ppmv were sampled, but only one-fourth
of the valves and pumps with screening values between 200 and
10,000 ppmv were sampled. In order not to bias the distribu-.
tion of leaking sources, it was necessary to develop estimated
values for all sources screening greater than 200 ppmv and not
sampled. The number of sources sampled and estimated for each
source type is shown in the following table:
C-145
-------
Baggable Source
Type
Total Sources
Sampled or Screened
> 200 ppmv
Sources
Sampled
Sources to
be Estimated
Valves
627
474
153
Pump Seals
382
281
101
Compressor Seals
Hydrocarbon Service
102
83
19
Hydrogen Service
69
60
9
Flanges
62
43
19
Drains
49
28
21
Relief Valves
58
31
27
Least-squares regression analyses were done for each
device type, regressing the logarithm of the nonmethane leak
rate on the logarithm of the maximum screening reading. Both
the original screening value and rescreening values (taken
closer to the time of sampling for leak rate) were evaluated
and a "best" equation was selected for each device as summa-
rized in Table C6-1.
Using the equations in Table C6-1, predicted log-
nonmethane leak rates were computed for each source not sampled
with a screening value greater than or equal to 200 ppmv. Leak
rates (lb/hr) were then computed using
leak rate = expio [log leak + z (standard error of estimate)],
the number of sources estimated, where z is a random number
from a standard-normal distribution. The use of the random
number is an attempt to yield a predicted distribution of leak
rates which would approximate the distribution if all sources
C-146
-------
TABLE C6-1. PREDICTION EQUATIONS FOR NONMETHANE LEAK RATES BASED ON MAXIMUM TLV
SCREENING OR RESCREENING VALUES
SOURCK TYPE
LEAST - SQUARES EQUATION
NUMBER CORRELATION STANDARD
OF DATA ^COEFFICIENT ERROR OF
PAIRS (r) ESTIMATE
Valves
LOG(NMLK)- -5.41
+
0.88
LOG (MXTLV-RS)
177
0.78
0.736
Pump Seals
LOG(NMLK)- -4.64
+
0.89
LOG(MXTLV-RS)
171
0.68
0.820
Compressor Seals:
48
0.58
0.791
Hydrocarbon Service
LOG(NMLK)- -4.77
+
0.92
LOG(MXTLV-RS)
Hydrogen Service
LOG(NMLK)- -3.66
+
0.44
LOG(MXTLV- S)
44
0.36
0.884
Flanges
LOG(NMLK)- -5.11
+
0.84
LOG(MXTLV- S)
47
0.74
0.535
Drains
LOG(NMLK)- -5.02
+
1.16
LOG(MXTLV- S)
60
0.72
0.807
Relief Valves
LOG(NMLK)- -4.47
+
0.87
LOG(MXTLV-RS)
53
0.78
0.637
NMLK - Nonmethane leak rate (lb/hr)
MXTLV- S - Maximum value - original screening (ppmv)
MXTLV-RS - Maximum value - rescreening (ppmv)
LOG - Logarithm, base 10
-------
were sampled. No bias correction factor is needed in
converting from the log to linear scale since the mean leak
rate is not being predicted. The predicted leak rates were
used in further analyses and development of emission factors.
Because the true leak rate/screening relationship is
unknown, there is a potential bias introduced when these pre-
dicted leak rates are used in developing emission factors. The
potential bias is proportional to the standard error of the
estimates adjusted for number of data pairs used to develop the
equation. The impact of the bias on emission factors depends
on the percent of sources leaking. The potential bias for each
source type estimated was approximated by weighting the 95
percent confidence limits of the predicted values according to
the percent of estimated sources. These values were expressed
as a percent of the emission factor estimate. The potential
biases were found
to be
as follows:
POTENTIAL PERCENT
BIAS
IN EMISSION FACTORS
SOURCES
FROM ESTIMATING
Source Type
- Percent Bias
+ Percent Bias
Valves
0.9
2.4
Pump Seals
2.7
2.9
Compressor Seals
Hydrocarbon Service
11.8
14.1
Hydrogen Service
1.6
3.2
Flanges
0.3
0.3
Drains
3.5
A.2
Relief Valves
5.1
6.0
These potential biases were taken into consideration
in developing confidence intervals as discussed in Section 6.3.
C-14S
-------
6.2 Statistical Distribution Models for Leak Rates
A lognormal distribution was used to model the
distribution of leaking sources. This distribution has the
property that when the original data are transformed by taking
natural logarithms, the transformed data will follow a normal
distribution. The lognormal distribution is often appropriate
when the standard error of an individual value is proportional
to the magnitude of the value. The form of the lognormal dis-
tribution is as follows:
exp
|1 (&n x - y)2 j
xa V 2tt
f (x) = '
for 0 > x > 00
for x < 0
Mean = exp
K]
Variance = exp[2y + 2a2] - exp[2y + a2]
In order to develop estimates for emission factors,
the nonleaking sources (leak rate assumed equal to zero) also
had to be modeled. A mixed distribution, specifically a
lognormal distribution with a discrete probability mass at
zero, was used for this purpose. Letting p equal the fraction
of nonleaking sources in the population, this mixed-lognormal
distribution has the following form:
C-149
-------
r = number of sources screened < 200 ppm or
with measured leak < 10"5 lbs/hr
m = n - r = number of "leaking" sources
g(t) = infinite series
_ (m- l)t (m- l)3t2 (to - 1) 5t3
m m 2! (m+1) m33! (m + 1) (m+ 3)
x = average of the logarithm of leak sources
n-r
= 2 £n (nonmethane leaks)/(n - r)
1
s2 = variance of the logarithm of leaking sources
n-r
= 2 [£n (nonmethane leaks) - x]2/ (n - r - 1) .
1
The mean and variance formulas hold whenever there is
more than one leaking source (n-r > 1). When only one leaking
source is identified, the following estimates are appropriate:
x x?
mean = — and variance = — ,
n n
where x is the single measured leak. If no leaks are found
(r=n), then the best estimate for both the mean and variance
is zero.
C-15]
-------
Computer programs were developed for these estimators
and the estimator for the mean was used for all emission factors
presented in this publication. Finney3 (1941) showed that this
estimator is more than twice as efficient as the arithmetic
mean for data distributed similarly to the leak rates from
baggable sources.
Since data distributed lognormally can be trans-
formed to a normal distribution by taking natural logarithms
of the data, the distribution assumption for the leaking
sources can be tested by examining distributions of the log
leak rates. Histograms displaying these distributions were
constructed for all important source type and process stream
classifications and are shown in Figures C6-1 through C6-12.
The data for most sources appear to adequately approximate a
normal distribution. The compressor seal data from hydro-
carbon service and the heavy stream data for pump seals both
appeared skewed to the left. Compressor seals with sampled
leak rates less than 10"3 lb/hr were considered as negligible
(zero) to minimize this skewness.
To statistically test the assumption of a normal
distribution for the log-leak rates, skewness and kurtosis
statistics were computed for each data group and tested for
departures from their expected value of zero in a normal
distribution. Table C6-2 summarizes these statistics,
Only three of the twelve cases indicate significant
lack of normality, confirming the conclusions from the
histograms.
C-152
-------
FKEUUEMCY
CI
I
-------
frequency
n
Cn
*
t
*
*
***
***
* * *
*** *****
50
+
*** ***** *****
£ «
*** ***** *****
*** ***** ***** *
***
v** ***** ***** *
***
*
*** ***** ***** *
** *
<~0
~
***** *
*** ***** ***** *
***
t
***** *
*** ***** ***** *
***
*
***** *
*** ***** ***** *
***
*
>
***** *
*** ***** ***** *
***
t
***** *
*** ***** ***** *
***
30
+
***** *
*** ***** ***** *
***
*****
***** *
*** ***** ***** *
***
t "
*****
***** *
*** ***** ***** *
***
t
*****
***** *
*** ***** ***** *
***
t
*****
***** *
*** ***** ***** *
***
20
~
*****
***** *
*** ***** ***** *
***
*****
t
*****
***** *
*** ***** ***** *
***
*****
t
*****
***** *
*** ***** ***** *
***
*****
i
*****
*****
***** *
*** ***** ***** *
***
*****
t
*****
*****
***** *
*** ***** ***** *
***
*****
*****
10
~
*****
*****
*****
***** *
*** ***** ***** *
***
*****
*****
t
*****
*****
*****
***** *
*** ***** ***** *
***
*****
*****
t
*****
*****
*****
***** *
*** ***** ***** *
* * *
*****
*****
* *****
*****
*****
*****
***** *
*** ***** v**** *
***
*****
*****
t * * * * *
*****
*****
*****
***** *
*** ***** ***** *
** *
*****
*****
---------
-------
1
1
i
1
1
i
1
1
i
1
1
1
1
1
t
t
1
t
1
1
1
-----
-------
------
•11
-10
-a
-7
-6 -S -H
-i:
-1
MIUPU1NT
OF ln up nun meihane
leak
RATt.
*****
* ~ * * *
*****
Figure C6-2. Histogram of Ln of Nonmethane Leak Rate Valves
Light Liquids/Two-Phase Streams
i
-------
FREQUENCY
9
+
*
*****
*****
8
+
*
*****
*****
7
+
*
*****
*****
*****
*****
6
~ '
*
*****
*****
*****
*****
5
+
t
*****
* * * * *
***** *****
***** *****
H
+
*
*****
*****
AC * A ± A A A A A AC *
***** *****
3
+
*****
*****
*****
***** *****
*****
t
*****
*****
*****
***** *****
*** **
2
+
*****
*****
*****
***** ***** *****
* * * **
*****
*****
*****
***** ***** *****
*****
1
*****
*****
*****
***** ***** *****
*****
*
*****
*****
*****
***** ***** *****
** ** *
-9
-a
-7
MIDPOINT OF
-6 -t>
LN UF NUN MElHANt LLAK KATt
-3
Figure C6-3. Histogram of Ln of Nonmethane Leak Rate Valves
Heavy Liquid Streams
-------
FNEUUtNCY
O
Ln
lif
11
10
«#
s
7
*****
*****
*****
*****
* * * * *
*****
* * * * *
* ****
*****
t
* * * * *
*****
*****
*****
*****
*****
*****
*****
* * * * *
*****
*****
* * * * *
*****
*****
*****
*****
*****
*****
* * * * *
*****
** * * *
* * * * *
*****
* * * * *
*****
*****
* * * * *
*****
*****
*****
* * * * *
*****
** * * *
*****
*****
*****
*****
*****
*****
*****
* * ** *
-10
-9
¦6 -7 -6 -5 -<+
MIDPOINT op ln up nun methane llak KAJL
-3
-2
Figure C6-4. Histogram of Ln of Nonmethane Leak Rate Valves
Hydrogen Streams
-------
FKEUUENtY
7 +
*****
*
t
*****
*****
t
6 +
t
*
*****
*****
*****
t
5 +
** *****
*****
*****
*****
*
*~
* * *****
*****
*** ** %
t
~ ~
** *****
*****
*****
*
»»
** *****
*****
*****
+
~ *
** *****
*****
*****
*
**
** *****
*****
*****
*
**
** *****
*****
*****
t
#• *~
** *****
*****
*****
3 ~
, " **
** *****
*****
*****
* * * * *
**
** *****
*****
*****
*** * *
t
**
** *****
*****
*** **
* * * v *
t
**
** *****
*****
*****
*****
2 +
**
** *****
*****
*****
*****
*****
* * V * *
t
**
** *****
***4*
*****
*****
*****
*****
*
**
** *****
*****
*****
*****
* > * * *
*****
**
** *****
*****
*****
*****
*****
* * * * *
1 +
***** **
** *****
*****
*****
*****
*****
*****
*
***** **
** *****
*****
*****
*****
*****
*****
*
***** **
** *****
*****
*****
*****
* * * * *
*
***** **
** *****
*****
*****
*****
*****
*****
-y -8 -7 -6 -b -5 -a
MI13P01NT OF LN OF NON METHANE LEAK KATt
Figure C6-5. Histogram of L'ri of Nonmethane Leak Rate
Open-Ended Valves
-------
FKEUUEfoCY
0
1
t—1
t-n
00
50
+
****
t
* ***
* ***
t
****
****
*
*****
****
****
«*0
+ t
*****
****
****
*
*****
****
****
t
*****
** *w
* * * *
*
**** *****
****
*** *
**** *****
* ***
****
30
+
**** *****
****
****
t
**** *****
****
****
*¦
**** *****
****
****
t
**** *****
****
****
*
**** *****
****
****
20
^ »»
**** *****
****
****
V
**** *****
****
****
*
**** *****
** **
****
t
*****
**** *****
****
****
*
*****
*****
**** *****
****
****
10
•f
*****
*****
**** *****
****
** * *
*****
*****
**** *****
****
****
*****
*****
**** *****
****
****
£ *****
*****
*****
**** *****
****
****
£ *****
*****
*****
**** *****
****
* ** *
-------
--------
- —
„ — — —
-9
-a
-7
-6 -5
*****
* #***
** ** *
*****
*****
**** *
** ***
*****
*****
%
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
* * * * *
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
-d -1 0
MIDPOINT OF UN UF NUN MCTHANE LLAK KATt
Figure C6-6.
Histogram of Ln of Nonmethane Leak Rate Pump Seals
Light Liquid Streams
-------
fkewuency
It)
10
0
1
M
U1
vO
*****
-10
*****
*****
*****
*
***
•
* v***
*
***
*****
*****
*
***
*****
*****
*****
*****
*
***
*****
*****
*****
*****
*
***
*****
*****
*****
*****
~
* * *
*****
*****
*****
*****
*
***
*****
*****
*****
*****
*
***
-7
-b
-t>
-4
-4
MIDPOINT OF
LN UF NUN
Mf. THANE
LEAK
KAT*-
***
**.*
* * *
** *
* * *
* * *
***
*» *
***
*****
*****
*****
*****
*****
*****
* * ** *
-y
-8
-2
-1
Figure C6-7. Histogram of Ln of Nonraethane Leak Rate Pump Seals
Heavy Liquid Streams
-------
FKEQUEfjCT
ON
O
25
20
15
10
t
*
t
+
t
*
*
*¦
*
t
*
t
*
~
t
*
t
*
***** *****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
V****
*****
*****
*****
* * * * *
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
* * * V*
* * V¦ * *
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
-9 -8 -7 -6 -5 -4 -2 -1 0
MIDPOINT OF LN UF NUN METHANE LE.AK KATt
Figure C6-8. Histogram of Lri of Nonmethane Leak Rate Compressor Seals
Hydrocarbon Service
-------
frequency
0
1
M
o>
16
~
*****
*****
*****
*****
12
+
*****
*****
t
*****
*****
11
+
*****
*****
*
*****
*****
10
+
*****
*****
*****
t
\
*****
*****
*****
9
~
*****
*****
„ * * **
#
*****
*****
*****
8
~
*****
*****
*****
*****
*****
*
*****
*****
*****
*****
* * * * * *
7
+
***** *****
*****
*****
*****
*****
t
***** *****
*****
*****
*****
*****
6
+
***** *****
*****
*****
*****
*****
*
***** *****
*****
*****
*****
*****
5
~
***** *****
*****
*****
*****
*****
t
* • ~
***** *****
*****
*****
*****
*****
~
***** *****
*****
*****
*****
*****
*****
*
***** *****
*****
*****
*****
*****
*****
3
+
*****
***** *****
*****
*****
*****
*****
*****
** ***
***** *****
*****
*****
*****
*****
*****
2
+
*****
***** *****
*****
*****
*****
*****
*****
t
*****
***** *****
*****
*****
*****
*****
*****
1
~
*****
***** *****
***** *****
*****
*****
*****
*****
*****
t
*****
***** *****
***** *****
*****
*****
*****
*****
U * * * *
*» + **
-10
-9
-0
-7 -b -5 -i* -4
MIUPOINT OF LN UF NUN ME IHANE LEW* KAJL
-2
-1
Figure C6—9. Histogram of Lri of Nonmethane Leak Rate Compressor Seals
Hydrogen Service
-------
FKEOULNCY
li>
1U
0
1
M
(T>
N>
.*»***
-11
*****
-10
*****
-9
*****
*****
*****
*****
*****
*****
*****
***** v
*** **
*****
*****
*****
*****
*****
*»*
**
*****
*****
*****
***
**
*****
*****
*****
»* *
**
*****
*****
*****
»**
**
* * * * *
*****
-a - ( -6 -5 -n
MIUHIHNT OF LN Up NUN METHANE LLAK KATL
-Z
-1
Figure C6-10. Histogram of Ln of Nonmethane Leak Rate Flanges - All Streams
-------
FREUUENCY
9
+
*****
t
*****
6
+
***** * * * * *
*
***** *****
7
~
***** *****
t
***** *****
6
~
*
*****
***** ***** *****
t
*****
***** ***** *****
5
~
*****
***** ***** *****
*****
*****
t
*****
***** ***** *****
*****
*****
-------
FREUUENCT
n
I
f—¦
13
+
*****
*****
12
+
*****
t
*****
11
+
*****
t
*****
10
+
*****
*****
*****
*
*****
*****
*****
9
*****
*****
*****
*
*****
*****
*****
8
+
*****
*****
*****
*****
*****
*****
7
+
*****
*****
*****
t
*****
*****
*****
b
+
*****
*****
*****
*****
*
*****
*****
*****
*****
5
+
*****
*****
*****
*****
*****
*****
*****
*
*****
*****
*****
*****
*****
*****
* * * * *
«~
+
*****
*****
*****
*****
*****
*****
*****
*
*****
*****
*****
*****
*****
*****
*****
3
+
*****
*****
*****
*****
*****
*****
*****
*
*****
*****
*****
*****
*****
*****
** ** *
2
+
*****
*****
*****
*****
*****
*****
*****
*
*****
*****
*****
*****
*****
*****
*****
1
+
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
-9
-a
-7
-5
-3
u
1
HIUPOANT
OF UN UF
NUN MEIHANE
LfcAH KATt
Figure C6-12. Histogram of Ln of Nonmethane Leak Rate Relief Valves - All Streams
-------
TABLE C6-2. SKEWNESS AND KURTOSIS STATISTICS
Number of
Source Type/Stream Group Leaking Sources Skewness Kurtosis
Valves1
Gas/Vapor Streams 154 0.19 - 0.33
Light Liquids/Two-Phase 330 - 0.16 - 0.18
Heavy Liquids 32 0.28 - 0.88
Hydrogen Streams. 59 - 0.18 - 1.09*
Open-ended Valves 30 - 0.01 - 0.98
Pump Seals1
Light Liquids 296 0.03 - 0.36
Heavy Liquids 66 - 0.77* 0.06
Compressor Seals
Hydrocarbon Se
Hydrogen Service 69 - 0.29 0.69
Hydrocarbon Service 102 - 0.99* 1.16*
Flanges 62 0.39 0.20
Drains 49 - 0.04 - 0.47
Relief Valves 58 - 0.05 - 0.21
Probability < 0.05 given a normal distribution.
XA11 data needed to classify sources into stream type were not available
for all pump seals and valves. Those particular sources are not included
in this analysis.
C-165
-------
The other assumption made in using the mixed-lognormal
model was that the sources with screening values less than 200
ppmv (calibrated to hexane) had insignificant leak rates which
could be assumed equal to zero. A number of sources with TLV's
less than 200 were sampled during the program in order to evalu-
ate this assumption. Table C6-3 summarizes the leak rate data
for these sources. The median leak rate is a conservative esti-
mate (high) of the typical concentration for sources screening
less than 200 ppmv since most of these sources had TLV values
of zero. A "worst-case" impact of this zero-emission assump-
tion on emission factor estimates can be evaluated by comparing
the median value times the percent of sources screening < 200
ppmv that were used in computing the emission factor. Table
C6-4 summarizes this comparison.
Only for flanges does the zero assumption appear to
have a potential impact on the emission factor estimate. For
flanges, the median leak rate for the 5 sources screening < 200
ppmv was approximately equal to the emission factor. Setting
all sources that were considered zero to 0.00054 lbs/hr would
almost double the emission factor. This potential bias was
accounted for in developing confidence intervals for the emis-
sion factor estimate for flanges (see Section 6.3). Since
there is only a potential bias from this assumption, as opposed
to a measured bias, adjustments to the emission factors are not
appropriate.
6.3 Confidence Intervals for Percent Sources Leaking and
for Emission Factors
Confidence intervals for the percent of leaking
sources were computed using the Binomial Distribution. The
Binomial is used to model data when'a random sample is selected
C-166
-------
TABLE C6-3. LEAK RATES FOR SOURCES SCREENING LESS THAN 200 PPM.
Source
Type
Compressor Seals
Drains
Flanges
Pump Seals
Relief Valves
Valves
// Sampled
<200ppmv
3
5
5
12
8
30
MAX TLV (ppmv)
Minimum Maximum
0
0
0
8
40
0
Leak Rates (lb/hr)
Minimum Median Maximum
% of Total Sources
Screened <200
140
0.00086
0.00416
0.1058
23.5
120
0.00056
0.00197
0.1078
81.8
110
0.00007
0.00056
0.0047
96.9
180
0.00006
0.00137
0.0052
51.5
180
0.00037
0.00132
0.0765
60.8
190
0.00001
0.00042
0.0383
67.2
TABLE C6-4. IMPACT OF "ZERO LEAK RATE" ASSUMPTION ON EMISSION FACTOR
Source
Compressor Seals
Drains
Flanges
Pump Seals
Relief Valves
Valves
Approximate Emission
Factor Estimate (lb/hr)
0.8
0.07
0.00058
0.17
0.19
0.023
Median leak rate for
<200 sources lb/hr
0.00098
0.00161
0.00054
0.00071
0.00080
0.00029
Median Times % of Sources
<200 Expressed as Percent
of Emission Factor
0.1
2.3
92.8
0.4
0.4
1.2
-------
and each item is classified into one of two categories (leaking
or nonleaking here). Exact confidence limits (level 1 - a) for
the estimate of percent leaking can be obtained by iteration
solving for P^ in
E QpJ a-p,)"-1-!
for the lower limit and for Pu in
i=0 V i /
for the upper limit, where
n = number of sources screened
k = number of leaking sources.
Tables of these solutions, available for most cases, were used
to develop 95 percent confidence intervals for reporting and
for computing 97.5 percent confidence intervals which were used
in developing confidence intervals for emission factors. The
97.5 percent was selected so that 95 percent confidence inter-
vals for emission factors would result when the estimated per-
cent leaking was combined with the estimated mean leak rate
(0.975 x 0.975 « 0.95) .
Patterson** (1966) described how confidence intervals
for the mean from a lognormal distribution can be computed
using estimators developed by Finney3 (1941). The 97.5 percent
confidence intervals were computed for the average, y, of the
C-168
-------
transformed data, y = £.n (leak) , using
C0 = lower limit = y - 2.24 [s2/(n-r)]
1/2
and
= upper limit = y + 2.24 [s2/(n - r)j17 2
where
s2 = the variance of the transformed data
n-r = the number of leaking sources.
Then, following Patterson's arguments, confidence intervals for
the mean leak rate can be computed using:
and
= lower limit = expfC^] g(s2/2)
= upper limit = expfC^] g(s /2)
where
g(t) is the series given in Section 6.2
To obtain 95 percent confidence limits for the emis-
sion factors, the confidence limits for the percent leaking
and for the mean leak rate were combined as follows:
C-169
-------
lower 95% limit for emission factor = P^ (Cj£)
upper 95% limit for emission factor = P (C)
r u u
These confidence intervals are conservative in the sense that
95 percent is a lower bound for the confidence coefficient for .
the intervals. The confidence intervals should be interpreted
as follows:
When we state that the true emission factor falls
within the limits computed as described above, we expect
to be correct at least 95 percent of the time.
These confidence intervals consider random sampling
variation and random test error, with no adjustments for
potential bias in the sampling and analytical methods. The
potential sources for bias have been discussed in previous
sections of the appendix:
• recoveries from sampling,
• analytical inaccuracies,
• biases in estimating leak rates from
nonsampled sources, and
• bias in assuming sources screening
< 200 could be considered as zero leak
rates.
Sections 2.1 and 2.3 of this appendix discussed
potential biases from sampling and analysis. It was concluded
that there was no evidence of bias in the final leak rate data
except for Refinery "F" where a low bias of about 15 percent was
indicated. Based on the percent of sampled sources from
C-17C
-------
Refinery "F," the estimated total effect of this bias was as
follows:
Estimated Bias (%)
Source Type In Emission Factor
Valves - 1.8
Pump Seals - 2.0
Flanges - 0.7
Compressor Seals - 0.3
Relief Valves - 1.4
Drains - 2.1
These biases were considered as systematic errors and there-
fore the emission factors and confidence limits were adjusted
upward by the appropriate percentage as given above.
Potential biases from estimating leak rates from
screening values were discussed in Section 6.1. Estimates of
negative and positive potential biases were given. To account
for this potential bias, lower confidence limits were adjusted
downward by the appropriate percentage for positive bias and
upper confidence limits were adjusted upward by the negative
potential bias estimate.
Only for flanges did the assumption of zero leak rates
for sources with screening values less than 200 ppmv appear to
have a potential impact on the emission factor estimate. This
was discussed in Section 6.2. To account for this potential
bias, the upper confidence limit for the flange emission factor
estimate was increased by 93 percent.
C-171
-------
The systematic errors discussed above were considered
independent so the net effect of combining all types of system-
atic errors was used in adjusting the emission factors and con-
fidence limits. The following table summarizes these net
systematic adjustments made to emission factors and confidence
intervals:
Total Systematic Adjustments (%)
Source Type
Lower Confidence
Limit
Upper Confidence
Limit
Emission Factor
Estimate
Valves
- 0.6
+ 2.7
+ 1.8
Pump Seals
- 0.9
+ 4.7
+ 2.0
Flanges
+ 0.4
+ 93.6
+ 0.7
Compressor Seals:
Hydrocarbon
- 13.8
+ 12.1
+ 0.3
Hydrogen
- 2.9
+ 1.9
+ 0.3
Relief Valves
- 4.6
+ 6.5
+ 1.4
Drains
- 2.1
+ 5.6
+ 2.1
6.4 Development of Nomographs
Three types of nomographs were developed as part of
the statistical analyses for this project:
• predicting mean leak rate from screening
values,
• cumulative distribution of sources by
screening values, and
C-172
-------
• cumulative distribution of total emissions
by screening values.
This section describes how these nomographs were constructed.
6.4.1 Predicting Mean Leak Rate from Screening Values
Section 6.1 describes least-square linear regression
equations developed for predicting leaks from nonsampled
sources in the data base with screening values greater than
200 ppmv. For prediction purposes outside the data base, a
statistical analysis of covariance was done to determine if
different equations were required for the various source types
and stream groupings. The data used to develop these relation-
ships and the resulting nomograph are given in Appendix B,
"Detailed Results." Although the equations were developed on
a logarithmic scale, the nomographs are shown on an arithmetic
scale for ease in reading and interpolation. Predicting the
arithmetic mean leak rate for a given screening value is simi-
lar to predicting the mean from a lognormal distribution as
discussed in Section 6.2. The mean value for a given screening
value on the nomograph was computed as follows:
mean = expi0[B0 + Bi logi0 (screening)] g(SE2^n/2)
B B
= (10) 0(screening value) 1 (scale bias correction factor)
where
B0 = log regression intercept
Bi = log regression slope
C-173
-------
SE^n = standard error of estimate in natural log scale
g(t) = series described in Section 6.2.
The 90 percent confidence intervals for the predicted mean leak
for a given screening value were computed in a similar manner
to the confidence intervals for the mean leak rate as described
in Section 6.2.
6.4.2 Cumulative Distribution of Sources by Screening
Values
Another set of nomographs included in Appendix B con-
tains the estimated cumulative distribution of log screening
values. The nomographs show 100 percent minus the cumulative
percent, or the estimated percent of sources which would have
screening values greater than any particular screening value.
These cumulative distribution functions were estimated by
fitting a lognormal distribution, as described in Section 6.2,
to the screening data and then generating the cumulative
distribution.
There was some difficulty in fitting the lognormal
distribution to the screening values. Figure C6-13 shows a
typical histogram of log screening values for valves in gas
service. The histogram appears to approximate a normal dis-
tribution adequately up to 10,000 ppm (4.0 on log10 scale).
The spike at 10,000 ppm was due to the inability of the screen-
ing device to measure beyond 10,000 without a dilution probe.
The dilution probe was used in only a few cases in the screen-
ing process during this program. The dilution probe was used
for rescreening in all refineries except the first three
C-174
-------
0.125
0.375
0.625
0.875
1.125
1.375
1.625 -
1.875 -
2.125 -
2.375 -
2.625 -
2.875 -
3.125 -
3.375 -
3.625 -
3.875 -
4.125 -
*.375 -
*.625 -
*.875 -
- 0.374
- 0.624
- 0.874
- 1.124
- 1.374
- 1.624
- 1.874
- 2.124
- 2.374
- 2.624
- 2.874
- 3.124
3.374
3.624
3.874
4.124
4.374
4.624
4.874
5.124
10
20
30
40
50
60
70
80
90
No. Valves - Gas/Vapor Streams
FIGURE C6-13. Distribution of Log10 (MAX Screening Value)
-------
visited. However, the original screening values were required
for the analysis.
To overcome the bias caused by this spike, only log
screening values less than 4.0 were used to estimate the param-
eters of this distribution. Formulas from "censored" normal
distribution theory (Cohen,2 1959) were then used to arrive at
unbiased estimates of the entire distribution. These estimates
were used to generate the cumulative distribution function for
each source type/process stream grouping.
Confidence intervals for these cumulative functions
were obtained using the Binomial Distribution as in Section
6.3. The 95 percent confidence interval for individual proba-
bilities were approximated using
p ± 1.96 [p(l - p)/n]1/2
where p is the estimated cumulative percent and n is the number
of screening values for each particular source type and stream
group.
Assuming that the sources screened approximate a
random sample from the population of a particular source type,
these confidence intervals can be interpreted as follows:
When we state that the true percent of sources in
the population which have screening values greater than
any selected screening value lies within the confidence
bounds, we expect to be correct about 95 percent of the
time.
C-176
-------
Note that these limits apply to the entire population for a
source type and are not necessarily applicable when addressing
a particular situation concerning a small number (less than
100) of sources.
The estimated cumulative distribution functions were
compared with the sample distribution function and appeared to
fit the data for each case except compressor seals. Figures
C6-14 through C6-22 show the estimated and sample distribution
for the source types and important stream classifications.
Discrepancies were found at the 10,000 ppmv screening value
(4.0 log screening value) in almost all cases, but this was
to be expected since the sample function had a big jump at
this point.
For compressor seals, censoring the data at 10,000
ppmv eliminated 64 percent of the observations, so the log-
normal parameters were reestimated using all the data as
recorded. These estimates resulted in a "better" agreement
between the sample and estimates distribution function, and
were therefore used to generate the cumulative distribution
function for compressor seals in both types of service.
6.4.3 Cumulative Distribution of Total Emissions by
Screening Values
A third set of nomographs given in Appendix B,
"Detailed Results," contains a function estimating the cumula-
tive percent of the total emissions attributable to each
particular source type/stream group as a function of increasing
log screening values. As before, 100 percent minus the cumula-
tive function is shown so that the percent of total emissions
C-177
-------
0
1
M
CO
<0
0
1—I
n)
>
"O
a)
•U
a
M
a
-------
VO
0)
3
iH
>
TJ
0)
4J
O
0)
iH
(1)
4)
3
iH
(fl
>
00
a
•H
£
20
to
-------
a)
3
i-H
>
T)
72
a)
4J
a
a)
i-H
0)
C/5
60
A
4)
3
i-H
CO
>
48
00
a
•H
a
0)
a)
u
34
o
CA
J2
4J
U
3
24
n
a>
o
M
3
o
m
12
U-l
o
4J
a
a)
o
0
-------
at
n
i
i—*
00
> 40 +
a)
•u
(U 33
CO
01
30 +*
nj
25
60
-------
96 +
. AA
« »
AA
AABA» ....
. . ABABA» . . .
....B*»* ..
.ABCA* . . .
. . AABA-*. . .
. . C**. .
. ACA* . .
..CC» ..
.ADA» . .
. ACA». .
.CF» . .
.CB* . .
. AB3B. .
. . AA«. .
. BCC .
. . . CC. .
. . *DB. .
. . *BF. .
. . . «»«. .
. . . *»*
* - Estimated Inverse Cumulative Distribu-
tion Function
o - Upper & Lower 95% Confidence Intervals
A,B,etc. ~ Points from Sample Function
A
A
AAAACB
AB
—+—
0. 0
0.3 1.0 1.3 2.0 2.3 3.0 3.3 4.0 4.3
LOGio (Screening Value, ppmv)
B
__+_
3. 0
3. 3
6. 0
6. 3
Figure C6-18. Inverse Cumulative Distribution Function for Pump Seals - Light Liquid Service
-------
72 +
Q>
3
rH
cd
>
T3
0)
4J
O
0)
£ 60
0>
3
To 48
>
00
a
a
Q>
-------
0
1
CO
•a
3
rH
<0
>
60
a
u
o
w
0]
-------
0)
3
r-j 36
flj
>
TJ
0)
«48
0)
»H
0)
CO
40
0)
3
nj
>
00
c
•H
d
0)
0)
u
o
KS)
J3
4J
m
4)
CJ
U
3
o
VI
IH
o
32
24
16
* * * *
** **A
»»##
* - Estimated Inverse Cumulative Distribu-
tion Function
o - Upper & Lower 95% Confidence Intervals
A,B,C - Points from Sample Function
****
A ***
A A ***
A *•** . . .
... B *** . . .
.... A A ***
... BA ***
AAAA»*
AA**»*
... A A ***
... AA ***
AAA ****
AAAA***
AAAA***
.... AB BA*
A*AB
*AC»*
•**A*
A*
A.
a
m
a
u
-------
a)
0
96
i—i
aJ
>
T)
00
NV
3
a
48
a)
a)
u
o
Crt
36
4-1
•H
>
CO
24
-------
attributable to sources with screening values above any
selected screening value can be determined.
This cumulative function was estimated by integrating
the leak/screening regression relationship over a lognormal
distribution of screening values. This function has the
following form:
S0 = selected upper screening value for
integration,
C = log/arithmetic scale bias correction
factor,
B0 = log 10 regression intercept term,
Bi = log10 regression slope term,
u = mean of the logg (screening values),
a2 = variance of the loge (screening values),
x = screening values over which the integra-
tion is being done, and
CF = cumulative function described above
in lbs/hr.
The form of the cumulative function can be simpli-
fied by algebraic reduction and change of variables to obtain:
So
where
C-187
-------
CF = C(10)B° exp[- " (^Bl02)]^[£n (So) ~ U ~ BlCj2 j
where $ is the cumulative function of a standard normal
distribution.
This function was used in developing the cumulative
emissions function shown on the nomographs. The censored dis-
tribution parameter estimates described in Section 6.4.2 were
used for the lognormal distribution parameters in each case
except compressor seals. The log/log least-squares regression
estimates described in Section 6.4.1 were used for the scale
bias correction factor and for B0 and Bx.
The scale for the above cumulative function is in
lbs/hr. To obtain a cumulative percent function, the number
obtained in lbs/hr at each screening value was divided by the
value of the function at a screening value of one million ppmv.
This forced the cumulative function to 1.0 at one million ppmv.
These scaled values were then subtracted from 1.0 and multi-
plied by 100.0 to obtain the functions shown on the nomograph.
The estimated cumulative emissions functions were
compared with the sample functions and found to adequately
approximate the data in most cases. Figures C6-23 through
C6-31 show the estimated and sampled functions for the source
types and important stream classifications. Again, the biggest
discrepancies were near the 10,000 ppmv screening value where
the sample function has a big jump. This area is more critical
for this function than the cumulative distribution function
since most of the emissions are attributable to sources with
screening values greater than 10,000 ppmv. It is important to
C-188
-------
0
1
M
00
vo
43
128
U
ri
9
a>
CO
3
3
o
CO
a)
o
4-1
a
96
u
a)
r-f
0)
a)
r-f
cn
X*
cd
a)
80
u
2
¦u
43
•H
a
U
U
cd
43
64
u
4-1
<
M
CO
a)
a
4-1
48
o
cd
•H
O
H
00
d
16
U-l
•H
o
d
-------
A
u
n
i
h-1
vO
O
112
to
0)
u
u
d
o
CO
o
4J
-s
4J
3
5
u
4J
4J
d
rH
cd
>
13
0)
4J
a
0)
rH
a>
CO
a)
4J
&
4J
n
a)
4J
cd
a>
M
O
0)
(V
d
rH
cd
>
00
fl
•H
s
a)
u
o
en
96
BO
64
48
32
16
* » « »A «AA»AA»AAAAABABBBABAABBDBBDC3PB»CAGBBBBBBCBBCA*»»«**
CBA *
A
* - Estimated Cumulative Function
A,B,C - Points from Sample Function
A A A AA
A
B
-16 +
-0. 9 0. 0 0. 9
1.0 1.5 2. 0 2. 5 3. 0 3. 5 4. 0
LOGio (Screening Value, ppmv)
4. 9
9. 0
—+—
S. 9
6. 0
+
6. 9
Figure C6-24. Cumulative Distribution of Total Emissions by
Screening Values for Valves, Gas/Vapor Streams
-------
T6T-D
3
fD
0
On
1
NJ
U1
w o
° 6
1 B
ID C
(D M
c
H* rt
o
3 H*
o
00 <
H
(D
o
<3
(d a
/"N
M H*
W
C cn
r>
(1) rt
it
CO i-i
(i)
H*
(D
Ml O*
a
o c
H*
i-l rt
a
H*
OQ
< o
(D 3
<
M
(u
< O
M
(D Mi
c
cn
(1)
- H
«•
O
f rt
H. (13
T3
00 M
3
3*
rt M
3
r* h*
H- cn
.a cn
C H»
H* O
a. 3
\ 01
«? o*
o ^
1
TJ
3*
flJ
cn
(1)
Percent of Total Emissions Attributable to Sources with
Screening Values Greater than the Selected Value
o
i
o i +
• +
CI
u
M
*
GO
*
A
CD
O
-0
o
10
CD
CI
w
u
u
o
(J
CI
CI
CI +
m ^
rt td
o -
. o
I
I
TJ M
O CO
H* rt
3 C"
S g
rt
f-h fD
Q.
O
0 s
w B
0) c
B M
T3 0)
rt
ft H*
fD
O
P
O
9
>
«
>
>
>
ta
>
G3 00
o o
> 0
~
~
~
»
~ m
> o
~
*
-------
* - Estimated Cumulative Function
A,B,C - Points from Sample Function
» A ~ »»A*A*A»AA»A»»*BAAAAAA
A DA ABB«BA AAABABBBA
«»» BABA
#»# AAAA
*** A
#»
•» AAA
#
##
»
**
*#
#
~
*
* A
* AAA
* AA
*
#
* A A
+ + + + + + + + + + ~ + + ~
0 0.3 0. & 0.9 1. S 1.9 1.8 2. 1 2.4 2.7 3.0 3.3 3.6 3.9 4.2
LOGio (Screening Value, ppmv)
Figure C6-26. Cumulative Distribution of Total Emissions by
Screening Values for Valves, Heavy Liquids
-------
0
1
M
vO
CO
.C
4-1
•H
0)
a)
u
u
s>
o
¦s
4J
a
^5
0)
a
o
•H
0)
CO
¦a
w
o
EH
o
u
a>
PM
0)
a
H
128 +
a)
>
•a
a>
u
112 +
o
a>
H
a>
en
96 +
a>
rfl
u
d
aj
80 +
rfl
u
u
u
o
0)
48 +
00
32 +
e
•H
a
0)
a>
u
16 +
o
in
* - Estimated Cumulative Function
A,B,C - Points from Sample Function
etc.
~ ~ AA AAAABBBABABA3CBABADCACDDBA»»
BDFCCBBCA##
BCCCC**
DBBF»
«*
¦»*#
**
##
A AA
AACBAB
+
—_ +—
0. 0 0. s
A
A
1.0 1.3 2.0 2.3 3.0 3.3 4.0 4.3
LOGio (Screening Value, ppmv)
3. 0
3. 3
6. 0
6. 3
~"
7. 0
Figure C6-27. Cumulative Distribution of Total Emissions by
Screening Values for Pump Seals, Light Liquid
Service
-------
0
1
I-1
vO
0)
128 +
0)
3
a)
iH
C
a
u
>
C
3
C
o
co
a)
*j
H2 +
r
o
a
L
u
a)
iH
c
a)
a)
r
iH
CO
96 +
X*
r
<0
a)
L
U
c
3
4J
•o
r
•H
g
80 +
M
cd
4J ^3
c
4J
*j
c
<
c
0)
a
a)
4J
64 +
o
C
•H
a)
r
(0
L
(0
o
c
CO
48 +
W
a>
3
C
iH
iH
c
r
4J
>
32 +
o
H
c
C
M-l
~H
c
O
S
c
U
a
a
CO
c
u
r
a)
L
*
o +
» » «A» AA#A»A»A*AA-»BA»AAABAAADABC BCCA3 A
*»«•*•»**** BA
**»#* ABB
*»»• AAB
### A
»#~
~#
~~A
*A
* - Estimated Cumulative Function
A,B,C - Points from Sample Function
etc.
~AA
•~A
A
AB
AA
#A
» A
•A
A
A
-+—
-0. 8
-0.4 0.0 0.4 0. B 1.2 1.6 2.0 2.4 2. B
LOG10 (Screening Value, ppmv)
3. 3
3. & 4. 0
4. 4 4. 8
Figure C6-28. Cumulative Distribution of Total Emissions by Screening
Values for Pump Seals, Heavy Liquid Service
-------
112 +
n
t
t-1
VO
Ut
0)
CO a
HI i-I
U n)
Vj >
a
O 73
I/)
o
H 60
C
>W -H
° s
4J a)
a ^
-------
n
i
t—¦
vo
as
•fl
4J
•H
s
a)
W
3
0)
T—J
a
u
>
o
CO
a>
4J
o
o
4J
a)
rH
o
H
00
a
U4
•H
o
a
0)
4J
0)
a
u
0)
a
a
CO
u
0)
112 +
96
80
64
48
32
16
* # **#»#* #»A****»AA#AAB»*A#A#BAAAABB AABAAAAA
*#*•»»#*¦»•**• BAAA
*#*«•#* AB BAA A
##*#
#»# A
~~~ A A
### A
~#B A
##
* A
*
*
* - Estimated Cumulative Function
A,B,C - Points from Sample Function
-16 +
c
——
-0. 8
-0.4 0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8
LOG10 (Screening Value, ppmv)
3. 2
3. 6
4. 0
4. 4
~
4. 8
Figure C6-30. Cumulative Distribution of Total Emissions by
Screening Values for Drains
-------
A* A A »BAC AA»AAAAA*»*»
A AA»»AAAA A
*»*AA
* - Estimated Cumulative Function
A,B,C - Points from Sample Function
##*
•»**
***
*#~
AA AB ~~
~~
#*
#*
#*
AA ABA
0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0 4.4 4.8 3.2 3.
LOG10 (Screening Value, ppmv)
Figure C6-31. Cumulative Distribution of Total Emissions by
Screening Values for Compressor Seals
-------
note that very little screening data are available with
screening values greater than 10,000 ppmv. Thus, this portion
of the curve is based on extrapolations using models developed
from screening values less than 10,000 ppmv.
This cumulative function is a very complex nonlinear
function of five sample statistics:
• the intercept and slope from the
regression of log leak rate on log
screening value,
• a bias correction factor used when
converting the logarithmic to the
linear scale, and
• the mean and variance of the natural
logarithm of the screening values.
Due to the complexity of this function, it was not possible to
derive a closed-form analytical expression for the confidence
intervals. Thus, a Monte-Carlo computer method was used to
generate the confidence intervals.
This method involved regenerating the cumulative
function 400 times. Each time, the data collected in the
project (the number of sources with screening values greater
than zero) were regenerated, except with an independent set
of random variations. The distributional properties of the
leak rate and screening data were used in computing the
required random numbers.
C-198
-------
For each of the 400 trials, sample estimates of the
five parameters required to compute the cumulative function
were computed. Then these estimates were used to generate a
new cumulative function. The five percent lower result and
the five percent upper result from the 400 trials for any
given screening value were then selected as the 90 percent
confidence limits for the population cumulative function.
These approximated 90 percent confidence limits can be
interpreted as follows:
When we state that the true percent of total emissions
for the population of sources, attributable to source
with screening values greater than a selected value, is
within the confidence bounds, we expect to be correct
about 90 percent of the time.
Since these confidence limits address the uncertainty in the
cumulative function for the entire sampled population of a
particular souce type, they are not necessarily applicable to
a finite sample of sources in a particular situation. The
variation of this function depends on the number of sources
in a complex manner, so it is not possible to draw a general
conclusion for the effect of sample size. Monte-Carlo simu-
lation techniques can be used to approximate intervals for a
finite random sample of a particular source type.
As an example, Figure C6-32 shows the confidence
intervals for the cumulative percent of emissions functions
for valves in light/two-phase service. Intervals are shown
which are applicable to a random sample of 100 valves and a
random sample of 1,000 valves. Also included are the confi-
dence intervals for the entire population. As can be seen,
C-199
-------
100
90
80
I 70
M
«*
*S 60
w
M
1 50
3 40
a
- 30
«
e
20
10
0
i ¦
s>
\\
ror rorc«m or cjauiions v *
fro* Total Population of \ \
Valves (n • •) \ \
• Estimated Percent of \ \ 0\
Total Mass Emissions » s N \
90S Confidence Interval
for Percent of Emissions
------90S Confidence Interval
for Percent of Emissions
In a Random Sample of
1000 Valves
•— - — 90% Confidence Interval for
Percent of Emissions 1n a
Random Sample of 100 Valves
1 2 3 45 10 50 100 1000 10,000 100.000 1,000,000
Screening Value (ppev) (Log1Q Scale)
Figure C6-32. Cumulative Distribution of Total Emissions
Screening Values - Valves - Light Liquid/Two
Phase Streams - Comparison of Confidence
Intervals
C-200
-------
the intervals applicable to a finite number of sources are
significantly wider than those for the population.
These intervals for finite populations were also
developed using simulation techniques. Four hundred Monte-
Carlo Trials generating 1,000 sets of data and then 100 sets
of data were run. In each of the trials, the generated sample
was ranked according to screening values and a sample cumula-
tive leak rate•function computed. Each sample function was
scaled by dividing by the total leak rate generated. Then
the five percent lower result and the five percent upper
result from the 400 trials for any given screening value were
selected as the 90 percent confidence bounds.
These confidence intervals can be interpreted as
follows where "m" is the number of randomly selected sources
in a particular situation:
When we state that the cumulative percent of total
emission function, which would be generated from a
random sample of "m" sources, will fall within the
confidence bounds, we expect to be correct about 90
percent of the time.
C-201
-------
7.0
CONVERSION FACTORS
To Convert From
Btu
bbl
gal
ton
lbs
cm
ft3
psi
g/gal
Btu/bbl
kWh/bbl
lb/bbl
lb/106 Btu
grain/ft3
gal/106ft3
gpm
lb/1000 gal
To Multiply By
kcal
0.252
Z
159.0
Z
3.785
kg
907.2
kg
0.454
in
0.394
m3
0.0283
kg/cm2
14.223
g/A
0.264
kcal/i.
0.0016
kWh/Z
0.0063
kg/Z
0.0285
g/Mcal
18.0
g/m3
2.29
Jl/106m3
133.7
m3 /hr
0.227
mb/i.
119.8
C-202
-------
REFERENCES
1. Aitchison, J. On the Distribution of a Positive Random
Variable Having a Discrete Probability Mass at the Origin.
American Statistical Association Journal, 50 (9):901-908,
1955.
2. Cohen, A. C., Jr. Simplified Estimators for the Normal
Distribution When Samples are Singly Censored or Truncated.
Technometrics 1:217-237, 1959.
3. Finney, D. J. On the Distribution of a Variate Whose
Logarithm is Normally Distributed. Journal of the Royal
Statistical Society, Series B, 7:155-161, 1941.
4. Patterson, R. L. Difficulties Involved in the Estimation
of a Population Mean Using Transformed Sample Data.
Technometrics 8 (3):535-537 , 1966.
5. Wetherold, R. G., and L. P. Provost. Emission Factors and
Frequency of Leak Occurrence for Fittings in Refinery
Process Units. EPA-600/2-79-044, EPA Industrial Environ-
mental Research Laboratory, Research Triangle Park, 1979.
6. Rom, J. J. Maintenance, Calibration, and Operation of
Isokinetic Source-Sampling Equipment. Environmental
Protection Agency, Office of Air Programs, Research
Triangle Park, N.C., March 1972.
C-203
-------
APPENDIX D:
DETAILED ENVIRONMENTAL ASSESSMENT
D-i
-------
APPENDIX D: DETAILED ENVIRONMENTAL ASSESSMENT
TABLE OF CONTENTS
Section Page
1.0 INTRODUCTION D-l
2.0 DEFINITION OF THE HYPOTHETICAL REFINERY MODEL . . D-4
2.1 Refinery Process Configuration D-5
2.2 Refinery Layout D-7
3.0 EMISSION CALCULATIONS D-10
3.1 Emission Factors and Fitting Counts D-10
3.2 Emissions of Criteria Pollutants and
Total Hydrocarbons D-15
3.3 Emissions of Selected Hydrocarbon
Components D-l 7
3.4 Summary of Hydrocarbon Species Emissions . . D-30
4.0 ATMOSPHERIC DISPERSION MODELING D-33
4.1 Choice of the Dispersion Model D-33
4.2 Application of the Dispersion Model to
the Hypothetical Refinery D-34
5.0 IMPACTS ON AMBIENT AIR QUALITY D-40
5.1 Criteria Pollutants and Total Hydrocarbons . D-40
5.2 Selected Hydrocarbon Components D-44
5.3 Discussion of Results D-54
D-ii
-------
TABLE OF CONTENTS (Continued)
Section Page
6.0 EFFECTS OF EXISTING AND POTENTIAL
REGULATIONS AND POLICIES D-60
6.1 State Regulations D-60
6.1.1 Particulate and Visible Emissions . . D-61
6.1.2 Sulfur Emissions D-62
6.1.3 NOx Emissions D-63
6.1.4 Carbon Monoxide Emissions D-63
6.1.5 Hydrocarbon Emissions D-64
6.1.6 Effects of State Regulations on
the Environmental Impacts of
Refineries D-65
6.2 Federal Regulations and Policies D-66
6.2.1 New Source Performance Standards. . . D-66
6.2.2 New Source Review D-68
6.3 Potential Regulations and Policies D-72
6.3.1 State Regulations D-72
6.3.2 Federal Standards D-74
6.3.3 Effects of Potential Regulations and
Policies on the Environmental Impacts
of Refineries D-74
7.0 EFFECTS OF NEW AND DEVELOPING TECHNOLOGY D-76
7.1 Process Technology D-76
7.2 Emission Control Technology D-77
8.0 CONVERSION FACTORS D-78
REFERENCES D-79
D- iii
-------
LIST OF TABLES
Table Title Page
D2-1 Large Capacity Existing Refinery Module Key ... D-9
D3-1 Process Sources and Emission Factors D-ll
D3-2 Fugitive Sources and Emission Factors D-12
D3-3 Estimated Number of Individual Emission Sources
in 15 Specific Refinery Process Units D-14
D3-4 Summary of Emissions from the Model Refinery. . . D-16
D3-5 Distribution of Unit Fugitive Emissions
by Stream D-18
D3-6 Example Stream Component Analysis - Cracked
Naphtha D-21
D3-7 Summary of Stream Quality Data (PPMW) D-23
D3-8 Fluid Catalytic Cracking - Fugitive Emission
Characterization D-27
D3-9 Relief Valve Distribution D-28
D3-10 Relief Valve Summary Fugitive Emission
Characterization u-29
D3-11 Estimated Composition of Inlet Oil, Hydrocarbon
Vapor, and Outlet Oil Streams Around an API
Separator - D-31
D3-12 Summary of Hydrocarbon Species Emissions
From Fugitive Sources D-32
D4-1 Relationship Between Stability Categories
and Surface Meteorological Conditions D-36
D5-1 Source Severity Factors for Criteria
Pollutants D-41
D5-2 Hydrocarbon Species Ambient Concentration at
the Point of Maximum Total Hydrocarbon
Concentration D-45
D-iv
-------
LIST OF TABLES (Continued)
Table Title Page
D5-3 Maximum Ambient Concentration of Selected
Hydrocarbon Species ..... D-47
D5-4 Summary of "F" Values D-50
D5-5 Source Severity Factors for Selected
Hydrocarbon Species D-52
D-v
-------
LIST OF FIGURES
Figure Title Page
D2-1 Block Flow Diagram of Model Refinery D-6
D2-2 Model Refinery Layout D-8
D5-1 Hydrocarbon Isopleth D-43
D6-1 PSD Applicability Chart D-69
D-vi
-------
1.0 INTRODUCTION
The environmental assessment presented here is a
method of examining the potential effects of refinery emissions
on the surrounding atmosphere. It will utilize the large volumes
of emission rate data generated in this program to estimate
ambient pollutant levels. It also attempts to examine the
environmental and public health effects of the predicted pollu-
tant concentrations. Finally, a brief survey of the effects
of existing and potential regulatory policies and developing
technology is presented.
The primary objective of the environmental assessment
is to provide guidance in identifying potential problem areas.
For instance, it can provide insight into which sources and
which pollutants appear to pose potential hazards. The results
are semi-quantitative in nature, which allows a relative ranking
of such problem areas. This can help to focus attention on
those areas needing further research. The environmental assess-
ment is only a tool to aid in the relative evaluation of
potential environmental impacts, not a method for making precise
and accurate predictions of such impacts. The results should
not be regarded as an absolute value which can be used to pre-
dict violations of standards, public health hazards, require-
ments for additional pollution control technology, or regulatory
requirements.
This type of analysis is particularly important for
refinery fugitive emissions, where hydrocarbons are the only
significant pollutant species. The rationale behind controlling
hydrocarbon emissions is based on two diverse effects: the
formation of photochemical oxidants and the toxic effects of
D-l
-------
some hydrocarbon species. Only through atmospheric modeling
(or the even more expensive ambient monitoring) can the latter
effects be assessed.
The approach to the environmental assessment follows
these steps:
• Define a hypothetical refinery.
• Estimate its emissions.
• Estimate ground level concentrations by
atmospheric dispersion modeling.
• Compare those ground level concentrations
to some acceptable concentration.
The parameter which is used to quantify environmental
impacts is called source severity. This concept was developed
by Monsanto Research Corporation under contract to the EPA.1
A source severity factor is defined as the ratio of the maxi-
mum ground level concentration of a pollutant in a "standard
receiving atmosphere" to the "acceptable pollutant concentration,"
as shown below:
where:
S
\iax = the maximum ground level concentration of
the pollutant, and
F = the acceptable pollutant concentration.
X
c, _ max
b F
= the source severity factor,
D-2
-------
This acceptable concentration is derived from either National
Ambient Air Quality Standards (NAAQS) or from Threshold Limit
Values (TLV's). If the resulting ratio is greater than 1.0,
then further emission reduction is probably needed. If the
ratio is below about 0.01, then further control is probably
not needed. Intermediate values are in a gray area where
further emission reduction may or may not be needed.
D-3
-------
2.0
DEFINITION OF THE HYPOTHETICAL REFINERY MODEL
The first step in the environmental assessment is the
selection and definition of the model refinery. The require-
ments of this model refinery are much broader than most. Not
only must the refinery processing be characterized, but also
its physical configuration. There is ample documentation of
the difficulties involved in trying to synthesize a "typical,
representative refinery." Refineries are very diverse, and
only a very rough approximation can be achieved with a single
model. When size and layout are added to the model, the task
is complicated further. Therefore, it should be noted through-
out this discussion that this is not a model that attempts to
represent the total industry, but rather a model of one hypo-
thetical refinery that reflects the "real world" as much as
possible.
The source for the model refinery is an EPA report
prepared by Pacific Environmental Services (PES)2 in which
detailed descriptions of the processing and physical layouts
of several types of refineries are provided. The large exist-
ing refinery was chosen as the model for this study because it
appears to represent the worst case. It is difficult to deter-
mine exactly which model would pose the worst case, because of
the diversity of effects between stack emission sources, pro-
cess area fugitive sources, and wastewater related fugitive
sources. A complex refinery will carry its processing further,
resulting in the production of higher volumes of aromatic com-
pounds which will often be separated into relatively pure
streams. This increases the likelihood of the occurrence of
localized high concentrations of some of the more hazardous
materials. An existing refinery was chosen because a stage-wise
growth pattern over many years is likely to result in a less
D-4
-------
organized layout than a new grass roots refinery. This could
aggravate the impact of emissions by placing a large source
near the boundary line rather than in a well-planned central
processing area buffered by surrounding greenbelt. Thus, if the
environmental assessment indicates minimal impacts for a large
existing refinery, then smaller, less complex, and more effi-
cient grass roots refineries would probably create a lesser
impact.
2.1 Refinery Process Configuration
Figure D2-1 shows the basic processing configuration
of the model refinery. This refinery processes 350,000 barrels
per stream day (BPSD) of mixed crudes to produce a full range
of low sulfur fuels and specialty products. All of the normal
refinery unit operations are represented, including:
• Atmospheric and vacuum crude distillation.
• Extensive hydrotreating of all ranges of
product streams (such as naphtha, middle
distillate, gas oils, and resid).
• Catalytic reforming.
• Aromatics extraction and separation of BTX.
• Hydrogen manufacturing.
• Fluid catalytic cracking with electrostatic
precipitator and CO boiler.
• Sulfuric acid alkylation.
D-5
-------
I
o\
LPG and Gas
S3
U ti
• *
X H
S-d
< Q
Naphtha
Naphtha
Hydro-
treating
Middle Distillates
Heavy At«. Gas Oil
Hydro-
Treating
Vac.
Gas Oil
Hi
Plant
TT
o
a
Vac. Cas Oil
Res Id
Hydro-
Treating
Delayed
Coking
Fuel Caa and LPG
Catnlytlc
Reforming
Gas
Sulfgr
Tail "
Gas
Treating
Treating
Recovery
BTX
H
Catalytic
Cracking
Aromatlcs
Extraction
llyd ro-
treating
Reformate
Gasoline
Cycle Gas Oils
To
Hydrotreaters
Hi
Hydro-
treating
C?/C„*
JCi_
Asphalt
Plant
Alkylation
Alkylate
Gasoline, Naphtha, Middle Distillates ^
Sulfur
P
R
0
D
U
C
T
S
Gasoline
Solvents
Aviation
Fuels
Diesels
Heating Oils
Aromatlcs
Petrochemical
Feedstocks
Asphalts
Industrial
Fuels
Refinery
Fuel Oil
**2 *-
Figure D2-1. Block flow diagram of model refinery.
-------
• Claus sulfur recovery, Wellman-Lord tail
gas treating, and a sulfuric acid plant.
• Gas processing (oil absorption/stripping
and distillation).
• Delayed coker.
• Rerun stills for recovered oils.
• Many miscellaneous treating, brightening,
etc., types of processing.
Again, it should be stressed that this configuration is not
intended to represent the total industry. But is is a
reasonable example of a modern refinery supplying low sulfur
fuels and specialty products.
2.2 Refinery Layout
The plot plan of the refinery (shown in Figure D2-2)
will give evidence of the detail which was presented in the PES
report. The functions of the various refinery modules are
detailed in Table D2-2. The process areas tend to form two
clusters, probably the result of a stage-wise expansion. Con-
siderable detail has been included in the physical model. All
of the appropriate vital functions have been accounted for and
distributed in a realistic manner. These are critical points
in achieving meaningful results from the atmospheric dispersion
model.
D-7
-------
LARGE CAPACITY EXISTING REFINERY
W = 1865 m
23
24
25
26
27
25
26
30
31
32
22
I I I I
<5
46
47
46
58
SS
50
60
68
51
61
33
34
48
52
62
S3
10
35
36
37
39
38
>40
41
42
53
64
69
70
71
72
65
66
67
73
£4
11
13
14
15
12
16
17
18
19
20
21
•43
44
55
56
57
74
75
76
126
Figure D2-2. Model refinery layout2
D-8
North
-------
TABLE D2-1. LARGE CAPACITY EXISTING REFINERY MODULE KEY 2
Module No.
Description
Module Ho.
Description
LI
Buffer Zone
L36
Catalytic Reformer
L2
Feedstock Storage
L37
Aromatics Extraction
13
Crude Oil Storage
L38
Catalytic Cracking
L4
Feedstock Storage
L39
Para-Xylene Plant
L5
Feedstock Storage
LAO
Delayed Coker
L6
Crude Oil Storage
L41
Barrel Storage
L7
Feedstock and Product
LA 2
Barrel Reconditioning
Storage
L43
Feedstock Storage
L8
Crude, Feedstock, and
L44
Storm Water Impound
Froduct Storage
Basin
19
Crude, Feedstock, and
L65
Warehouse '
Product Storage
L46
Gas Holder/Blcwcovn
L10
Oil-Water Separator
Stack
Lll
Product Storage
L47
Cas Holder/Blovdown
L12
Product Storage
Stack
L13
Distillation ahd Gas
L48
Fire Prevention Train-
Recovery Unit
ing Facility
L14
Jet Hydrofiner/Catalytlc
L49
Oil-Water Separator
Reformer
LSO
Asphalt Plant
L51
Solvent Treating Plant/
LIS
Naphtha Hydrotreater
Boiler House
L16
Hydrotreater (Lt Cycle
L52
SO? Treating Plant/
Oil)
Tanks
LI 7
Hydrogen Manufacturing
L53
Lube Oil Packaging
L18
Partial Oxidation Unit
L54
Coke Storage
L19
Future Expansion
L55
Crude Oil Storage
L20
Cooling Tower
LS6
Feedstock Storage
L21
Flares
L57
Tanks/Impound Basin
L22
Feedstock and Product
L58
Administration
Storage
L59
Oil-Water Separator
L23
Naphtha Hydrotreater
L60
Gasoline Sweetener/
L24
Vacuus Cas Oil Unit
Crude Distillation
L25
Benzene Fractionation
L61
Crude Distillation/
L26
Steam Rerun Stills
Crude Desaicer
L27
Future Expansion
L62
Specialty Crude
L28
Crude Distillation
Distillation
L29
Catalytic Reformer
L63
Speciality Crude Dis-
L30
Vacuus Residua De-
tillation/Condenser
sulfurlzer
Box
L31
Hydrogen Manufacturing
L64
Gasoline Fractionating
L32
Alkylation
Unit
L33
Distillate Hydrodesul-
L65
Tank Loading/Truck
furlxation (Hvy Gas
Loading/Vapor Re
Oil)
covery
L34
Sulfur Recovery
L66
Buildings
L35
Tanks/Cooling Towers
L67
LPG Storage And Blending
L63
Vapor Recovery/Gasoline
Rectifier/Tanks
L69
Main Pump House
L70
Product Storage
L71
Wastewater Treatment
172
Building
L73
Product Storage
L74
Shops and Warehouse
L75 Crude Oil Storage
L76 Crude, Feedstock, and
Product Storage
The oil/water separator in Module L10 treats aqueous discharge from
Modules L1-L21.
The separator located in Module L59 treats aqueous streams fron Modules
L58-L60, L70, L71, and L7J-L76.
The wastewater separator la Module LM treats discharges froa the rsaaIc-
ing aodules.
D-9
-------
3.0 EMISSION CALCULATIONS
Once the model refinery is defined, the next step is
to estimate its emissions. For point source emissions, avail-
able emission factors are used. For fugitive emissions, both
emission factors and fitting counts are required. Finally, the
total hydrocarbon emissions are broken down into emissions of
selected hydrocarbon components.
3.1 Emission Factors and Fitting Counts
Emission factors and the corresponding unit capacities
for the model refinery are presented in Table D3-1. All of
these emission factors were taken from AP-423 with the exception
of the non-methane hydrocarbon factor from the FCC unit CO boiler
and the S0x factors from the Sulfur Recovery Complex.
Table D3-2 presents estimated fitting counts and
emission factors for fugitive sources in the model refinery.
All of these emission factors were taken from Radian testing
except for those for the oil/water separator.
The estimate of the population of each type of fitting
is as important as the emission factor in determining total
emissions. The PES model contained estimates of fitting popula-
tions, but they were not broken down into the service categories
corresponding to the emission factors. Radian data on fitting
counts were generated during the field testing phase (see
Table D3-3), but these unit configurations would not necessarily
match those from the model refinery. The following procedure
was developed to generate fitting counts compatible with emission
factor service classes and to represent the model refinery as
closely as possible:
D-10
-------
TABLE D3-1. PROCESS SOURCES AND EMISSION FACTORS
EMISSION FACTORS
SOURCE
CAPACITY
PARTICULATES
SO,
NO,
CO
NM HYDROCARBONS
Process Heaters/Boilers
- oil fired1 36.7 * 10' gal/hr 6 lb/10' gal 47.7 lb/10* gal 60 lb/101 gal 5 lb/10J gal 1 lb/101 gal
- gaa -fired*
2.27 * 10* ft'/hr 5 lb/10* ft3
0.6 lb/10s ft' 120 lb/10* ft1 17 lb/10* ft' 3 lb/10* ft'
Fluid Catalytic
Cracker CO Boiler*
2.086 x 10'Bbl/hr 45 lb/10'Bbl. 493 lb/105Bbl 71 lb/10'Bbl . Negligible 13.3 lb/lO'Bbl.
Sulfur Recovery Coaplex*
- Claua plants plua
Wei Inn-Lord Tall
Gas Treating Unit
- Sulfuric Add Plant 179 LTPD
408 long tons/day
(LTPD)
3.6 lb/LT
14.6 lb/LT
Flarea*
350 x 10'Bbl/day Negligible
26.9 lb/10'Bbl 18.9 lb/10'Bbl 4.3 lb/10JBbl 0.8 lb/10*Bbl
1 Based on AF-42* factors for Ho. 6 Fuel Oil with 0.3 wt. X sulfur.
* Based on AP-42* factors for nstursl gas.
* Baaed on AP-42* factora for a Fluid Catalytic Cracker with an Electrostatic Precipitator and a CO Boiler,
except the NMHC factor waa taken froa Radian testing.
* Based on sulfur recovery efficiencies taken from Hydrocarbon Processing2-stage Claus • 92X, Wellman-Lord
TCTU - 99%, Sulfuric Acid Plant - 99X
* Based on AP 42s factors for blowdown systens with vapor recovery and vents to flares.
-------
TABLE D3-2. FUGITIVE SOURCES AND EMISSION FACTORS
ESTIMATED POPULATION SERVICE HON-METHANE HYDROCARBON
SOURCE (OR CAPACITY) CATEGORY (NMHC) EMISSION FACTORS
Pump Seals1
313
340
Light Liquid
Heavy Liquid
0.25 lb/hr
0.046
./source
Valves1
1714
4198
7422
8442
Hydrogen
HC Gas
Light Liquid
Heavy Liquid
0.018
0.059
0.024
0.0005
Compressor Seals1
82
48
Hydrogen
HC Gas
0.11
1.40
Flanges1
84346
NA
0.00056
Relief Valves1
171
NA
0.19
Process Drains1
1105
NA
0.070
¦
Cooling Towers1
(10,668 x 103 gal/hr)
NA
0.006 lb/103 gal.
Oil/Water Separators2 (160.3 x 103 gal/hr)
(1719 x 103 gal/hr)
Uncontrolled
Controlled
• 5.0 lb/103 gal.
0.2 lb/103 gal.
Dissolved Air
Flotation1
(220.5 x 103 gal/hr)
NA
0.01 lb/103 gal.
'Emission factors
^Emission factors
based on'Radian testing,
based on AP-42.
D-12
-------
1) Use PES pump count times 1.4 to estimate
the total number of pump seals. Use pump
service descriptions to determine the
service class, with naphthas and lighter
being light liquids, all others being
heavy liquids.
2) Use PES compressor count times 2 to estimate
the total number of compressor seals. Use
compressor service descriptions to determine
the percentage in hydrocarbon and/or hydrogen
service.
3) Use the pump seal count times 41 to estimate
total valves.
4) Use the percentage in gas service from the
technical memorandum, and split the total
valves in gas service into 29 percent hydro-
gen and 71 percent hydrocarbon service.
5) Break down valves in liquid service into
light liquid and heavy liquid based on the
breakdown of pumps for that unit.
6) Use the valve count times 4 to estimate the
total number of flanges.
7) Use the process drain count from Table D3-3.
If none is available for a given unit, use
the pump seal count times 1.9.
D-13
-------
TABLE D3-3. ESTIMATED NUMBER OF INDIVIDUAL EMISSION SOURCES
IN 15 SPECIFIC REFINERY PROCESS UNITS
Estimated Number of Sources Within Battery Limits of Process Units
Relief
Process Unit Valves Flanges Pumps® Compressors11 Drains Valves
Atmospheric Distillation
890
3540
31
1
69
6
Vacuum Distillation1
500
2700
16
O1
35
6
Fuel Gas/Light Ends Processing
180
760
3
2
11
6
Catalytic Hydroprocessing
650
2600
10
3
24
6
Catalytic Cracking
1310
5200
30
3
65
6
Hydrocracking
930
3760
22
3
58
6
Catalytic Reforming
690
2760
14
3
49
6
Aromatics Extraction1
600
2400
181
O1
41
6
Alkylation
680
2280
11
0
41
6
Delayed Coking1
300
1240
91
O1
28
6
Fluid Coking
300
1240
9
4
28
6
Hydrodealkylation*
690
3760
141
31
58
6
Treating/Dewaxing
600
2290
18
1
44
6
Hydrogen Production
180
640
5
3
17
4
Sulfur Recovery1
200
800
61
O1
20
4
'Sources were not counted in process units of this type. The number of sources was estimated.
*Only those sources In hydrocarbon (or organic compound) service.
'Number of pump seals - 1.4 x number of pumps.
*Nusber of compressor seals - 2.0 x number of compressors.
-------
8) Estimate the number of relief valves as an
average of 6 per unit. This average was
developed from field source counts.
The figures for capacity of various process units and wastewater
treating facilities were taken directly from the PES report.
Although emissions from storage tanks were not within
the scope of this study, they were estimated to provide a basis
of comparison to other hydrocarbon emission sources. The PES
report contained a detailed breakdown of the storage facilities,
their service, capacities, and annual turnover. Emission factors
were taken from AP-423 and applied to these facilities to estimate
total emissions. The PES data indicated the use of floating
roofs to control emissions on all tanks containing liquids with
Reid vapor pressures greater than 0.5 psia. This stringent
level of control probably explains the low estimate of storage
emissions as compared to other hydrocarbon sources in the model
refinery (see Table D3-4).
3.2 Emissions of Criteria Pollutants and Total
Hydrocarbons
By applying the appropriate emission factors, unit
capacities, and fitting counts documented in Section 3.1, a
slate of emissions from the model refinery can be estimated.
Table D3-4 is a summary of those emissions by pollutant type
and by source.
D-15
-------
TABLE D3-4. SUMMARY OF EMISSIONS FROM THE MODEL REFINERY
Emissions in Tons/Year
Pollutant Point Sources1 Fugitives2
Particulates 1,425
SO 14,650
X '
CO 1,247
NOx 12,693
Non-Methane
Hydrocarbons 364 8'767
Storage
Total
3,308
1,425
14,650
1,247
12,693
12,439
1 Includes combustion sources, fluid catalytic cracker, CO
boiler, sulfur recovery complex, and flares.
2 Includes process fittings (pumps, valves, 'flanges, compressors,
drains, and relief valves), cooling towers, oil/water
separators, and other wastewater treating units.
D-16
-------
3.3 Emissions of Selected Hydrocarbon Components
The emissions estimates in Table D3-4 are sufficient
inputs to the model to estimate the ambient concentrations of
criteria pollutants, but a species breakdown is necessary to
evaluate specific hydrocarbon concentrations. Analyses of the
components in various process streams were made in this program
and supplemented by literature sources. The application of
these stream analyses is not straightforward, however, since
the emissions were estimated on a unit basis. The necessary
approach involves three steps:
• Distribute total unit emissions among
the appropriate streams.
• Apply stream analyses to get component
emissions for each stream.
• Sum the stream component emissions to
get unit component emissions.
First the unit emissions must be broken down into
the streams characteristic of that unit. As an example, Table
D3-5 shows the breakdown for the fugitive emissions from the
fluid catalytic cracker. It's characteristic streams are:
• Atmospheric gas oil (feedstock).
• FCC Make-gas.
• Olefinic LPG.
• Cracked naphtha.
D-17
-------
TABLE D3-5. DISTRIBUTION OF UNIT FUGITIVE EMISSIONS BY STREAM
Stream
Example.- Fluid Catalytic Cracking Unit
Percent of
Fittings for
the Stream
®
Weighted Emission
Factor for the
Stream
®
lb/hr/source
Relative Percent of
Emission Unit Fugitive
Rate for Emissions for
the Stream the Stream
® x ®
Atmospheric Gas Oil
15
0.0016
0.024
1
FCC Make-gas
10
0.059
0.59
30
Olefinic LPG
15
0.030
0.45
23
Cracked Naphtha
30
0.030
0.9
45
Lt. Cycle Gas Oil
20
0.0016
0.032
1
Hvy. Cycle Gas Oil
10
0.0
0.0
0
Totals
100
N. A.
1.996
100
-------
Light cycle gas oil.
• Heavy cycle gas oil.
Although this does not include every possible product or inter-
mediate stream, it is detailed enough to allow a reasonably
good characterization.
The next step is to estimate the percentage of total
fittings in each stream service. These are engineering esti-
mates based on familiarity with the unit operations. The next
important variable is the weighted emission factor for each
stream. This is determined by first classifying the stream
as gas phase, light liquid, heavy liquid, or hydrogen. For
gas phase and hydrogen service, the valve emission factor was
used as the weighted average indicator. For light and heavy
liquids, a ratio of 41 valves per pump was used to estimate
the weighted average indicator. Since valves and pumps
contribute greater than 70 percent of the process fugitive
emissions, it is reasonable to base the weighted average emis-
sion factor on them alone. It should also be noted that this
factor is being used only to compare the relative emissions
contribution of each stream, not as an absolute value to
calculate emissions. When these two factors are multiplied
together, the resulting product is in proportion to each
stream's tendency to cause fugitive emissions. By summing
these products and determining each product as a percentage of
the sum, the total unit emissions can be allocated to each
stream by that percentage.
Next the component analyses can be applied to these
stream emissions. The component analyses come primarily from
GC-MS work done on samples collected in the refineries during
fugitive emission sampling. This was supplemented where
D-19
-------
necessary with data from a previous Radian literature survey,3
an API medical research report, and engineering estimates.
Tables D3-6 and D3-7 are examples of a GC-MS data sheet and the
stream quality summary, respectively. All of the GC-MS data
sheets are included in Appendix B, Section 5.
It was necessary to consolidate these component
analyses to minimize calculations and to yield reasonable data.
This consolidation was done on the basis of the availability of
both discrete concentration data and quantifiable toxicity data
for any given component. If both were available, then the
component was treated individually. If either was missing, the
component was lumped into a family of components such as "other
alkylbenzenes." This resulted in a list of discrete components
which included:
• Benzene.
• Toluene.
• Ethylbenzene.
• Mixed xylenes.
• Naphthalene.
• Anthracene.
• Biphenyl.
• Hexane
D-20
-------
TABLE D3-6. EXAMPLE STREAM COMPONENT ANALYSIS-CRACKED NAPHTHA
Bulk
Vapor on Vapor on
Peak
Compounds
Liquid
XAD
Tenax
Number
(In Retention Order)
(ppm)
(ug)
(yg)
1
Benzene
6,600
260
0.72
(IS)
d6-Benzene
—
—
(0.035)
2
Toluene
47,700
8,100
25.3
3
Ethylbenzene
10,600
4,400
4.0
4
m-+p-Xylene
57,200
8,000
21.3
5
o-Xylene
21,300
7,500
8.7
6
Isopropylbenzene
—
130
0.21
7
n-Propylbenzene
3,000
850
8
3- + 4-Ethyltoluene
32,500
7,100
f 19.8
9
1,3,5-Trimethylbenzene
15,100
2,800
10
2-Ethyltoluene
7,100
1,280
11
1,2,4-Trimethylbenzene
46,000
6,150
13.3
12
1,2,3-Trimethylbenzene
9,600
880
3.2
13
C<,-Alkylbenzene
—
72
0.33
14
Indan
4,000
250
1.2
15
C<,-Alkylbenzene
17,200
1,000
16
C<,-Alkylbenzene
19,600
960
7.8
17
C<,-Alkylbenzene
2,400
210
18
C t, - Alky lb en z ene
13,200
520
4 1
19
C^-Alkylbenzene
13,600
480
20
2- Methylindan
2,500
85
0.41
21
C^-Alkylbenzene
2,000
32
0.74
22
C^-Alkylbenzene
19,600
340
2.3
23
Methylindan
2,500
10
0.22
24
Methylindan
2,800
30
0.24
25
C <,-A1 ky lb en z ene
2,800
—
0.49
26
C s-Alkylbenzene
27,000
—
1.2
27
Cs-Alkylbenzene
2,700
—
0.44
28
Naphthalene
15,600
66
0.03
29
Cfc-Alkylbenzene
1,200
—
—
30
Ca-Alkylindan
2,400
—
0.11
31
Cfc-Alkylbenzene
600
—
—
32
C s-Alkylbenzene
4,000
—
1.2
33
C s-Alkylbenzene
1,700
—
0.46
34
Ca-Alkylindan
400
—
«
Continued
D-21
-------
TABLE D3-6. CONTINUED
Bulk Vapor on Vapor on
Peak Compounds Liquid XAD Tenax
Number (In Retention Order) (ppm) (yg) (ug)
35
C2-Alkylindan
600
—
—
36
Ca-Alkylindan
400
—
—
37
Cj-Alkylbenzene
100
—
—
38
C5-Alkylbenzene
1,000
—
—
39
2-Methylnaphthalene
8,700
—
0.03
40
1-Methylnaphthalene
3,600
—
0.01
(IS)
dxo-Anthracene (IS)
(100)
(1,000)
—
D-22
-------
TABLE D3-7. SUMMARY OF STREAM QUALITY DATA (PPMW)
Straight
Conpound or Crude Run Middle Atmospheric Vacuum H? Recycle Desulfurlzed
Functional Family Oil Naphtha Distillate Gas Oil Gas Oil Reformnte Gas Naphtha
Benzene
60
253
0
0
0
5,400
0
253
Toluene
680
2,621
5
8
5
77,700
0
2,621
Ethylbenzene
220
887
9
6
6
33,500
0
R87
Xylenes
880
1,623
52
16
22
170,900
0
1,623
Other AlkyIbenzenea
3.739
16,578
835
61
368
324,400
0
16,578
Naphthalene
1,463
100
4
28
7,400
0
1,463
Anthracene
UO
5
56
3
12
0
0
5
B1phenyl
320
628
0
0
9
0
0
628
Other PNA's
7.8R0
14,983
5,507
220
663
700
0
14,983
n-Hexane
18,000
38,838
0
0
0
24,000
0
38,838
Other Alkanes
877,210
499,613
842,536
949,673
948,887
356,000
650,000
499,613
Olefins
0
0
0
0
0
0
0
0
Cycloalkanes
58,300
422,508
100,000
50,000
50,000
0
0
422,508
Other Compounds
tndlcated Present
Carbonyl
- 500 ppm
Thiols
•* 25,000 ppm
SulfIdrs
- 6,000 ppm
Qulnollneo
~ 200 ppn
PyrIdlnes
Thiols
Sulfides
Pyridines
Thiols
Sulfides
~ 51,000 ppm
Qulnollne8
Pyridines
Thiols
Sulfides
Qulnolines
^ 9 ppm
Pyridines
Thiols
SulfIdes
Quinolines
Hj - 350,000
Pyridine*
-------
TABLE D3-7. CONTINUED
Cmpound or
Functional Paally
Hydrotreated
Middle
UlRt I Hate
Refinery
Fuel
Cas
Liquefied
Petroleum
Gns (I.PG)
Raff inate
Aromatlea
Extract
Benzene
Toluene
Xylenes
Benzene
0
0
0
50
17,840
993,000
1,000
0
Toluene
5
0
0
750
256,700
2,000
992,800
1,000
Ethylbcnzene
9
0
0
300
110,670
0
4,000
162,420
Xylenes
52
0
0
1,500
564,590
0
1,000
828,580
Other Alkylbenzene
-------
TABLE D3-7. CONTINUED
Compound or
Functional Family
LFG
Olefins
Alkylate
Cracked
Naphtha
FCC
Light Cycle
Cae OU
FCC
Heavy Cycle
Gas Oil
Heavy
Aromatlea
Extract
(SO* Plant)
Asphalt
API
Separator
Skim Oil
Vacuum
Reald
¦enzene
0
0.1
2,880
0
740
0
0
87
0
Toluene
0
0.3
89,780
40
10,000
0
0
1,713
0
Ithylbenzene
0
0.1
21,430
0
1,200
0
0
661
0
Xylenes
0
1.1
171,450
610
11,800
0
0
2,510
0
Other Alkylbenzenes
0
3.3
243,470
26,670
38,200
750,000
i
12,751
i
naphthalene
0
0.3
10,950
59,000
14,000
0
0
990
0
Anthracene
0
0
0
10,270
0
0
2
457
2
¦lphenyl
0
0
0
10,180
0
0
0
2,351
0
Other PNA's
0
2.2
6,480
624.480
22,500
200,000
200
29,700
200
a-Rexane
0
96
11,830
0
0
0
0
i '
0
Other Alkanes
400.000
998,956
204,110
190,800
701,560
45,000
999.798
948,780
999,798
Oleflna
600,000
930
170,740
36,750
50,000
0
i
i
Cycloalkanes
0
11
68,880
41,200
150,000
5,000
i
i
Other Compounds
Indicated Present
Thiol*
Pyridines
Thiols
SulfIdes
Qulnollnes
Phenols
Carbonyls
Pyridines
Thiols
Sulfides
Qulnollnes
Pyridines
Carbonyls
Thiols
SulfIdes
Qulnollnes
Mote: The •yabol i. «eana that the component hat bean indicated to be present, but the exact concentration la unknown.
-------
The general family groups included:
• Other alkylbenzenes.
• Other polynuclear aromatics.
• Other alkanes.
• Olefins.
• Cycloalkanes.
It is not really meaningful to talk about the toxicity of such
a broad group as olefins or cycloalkanes. These were included
to assure a consistent analysis based on a closed mass balance
to be synthesized from several diverse sources.
The dispersion model inputs emissions on a unit basis,
so the next step is to combine the stream breakdown with the
stream analyses to get a component analysis of unit emissions.
An example of this process is shown for fugitive emissions from
the FCC in Table D3-8.
A similar operation was performed separately on relief
valves, since they are not distributed uniformly across the
streams. Relief valves are usually placed at the top of a
fractionating column or reactor vessel, and thus are exposed
primarily to lighter streams. Table D3-9.shows the allocation
of relief valves for the Aromatics Fractionation Unit. The
number of relief valves in each stream service was then totaled,
and the stream analyses were applied to the emissions, as shown
in Table D3-10. According to observations in the refineries tested,
few relief valves in existing refineries are vented to a control
D-26
-------
TABLE D3-8. FLUID CATALYTIC CRACKING - FUGITIVE EMISSION CHARACTERIZATION
Weighted Contribution of each Conponent to Unit Emissions, In ppnv
Streaa
Aittos. Gas Oil
Fuel Gas
LI*C Olefins
Cracked Naphtha
Lt. Cycle Gaa Oil
ttvy. Cycle Gas Oil
Total
30
23
45
1
0
Eaisa.
Rate
lb/hr
0 0
0 0
0 0
1296 40401
0 0
1296
40401
0
0
0
9644
0
9644
0
0
0
0
77153
6
77159
0
0
109562
267
109830
0 0
0 0
4928
590 103
5518
103
102
102
2 0
0 0
0 0
2916 5324
6245 0
9163
5324
9495
276000
92000
91850
1906
471251
0
18000
138000
76833
368
233201
500
'31008
0
0 6000
0 0
30096 0
412 0
6000
59.8
.078
2.42
.577
4.61
6.57
.33
.006
.006
.548
.318
28.18
13.95
1.85
.359
-------
TABLE D3-9. RELIEF VALVE DISTRIBUTION
Example: Aromatics Fractionation Unit
Total Relief Valves = 6
Stream No. of Relief Valves
Benzene 4
Toluene 2
Xylenes 0
D-28
-------
TABLE D3-10. RELIEF VALVE SUMMARY FUGITIVE EMISSION CHARACTERIZATION
0
1
NJ
vO
Weighted Contribution of each Component to RV Emlaalona, In pptw
8treaa
(42)
H, Recycle Gaa
22.5
0
0
0
0
0
0
0
0
0
0
146250
0
0
787V)
(12)
Fuel Caa
6.5
0
0
0
0
0
0
0
0
0
0
59800
3900
0
1300
(58)
IPG
31.1
0
0
0
0
0
0
0
0
0
0
311000
0
0
0
(4)
LFG Olefin*
2.2
0
0
0
0
0
0
0
0
0
0
8800
13200
0
0
(33)
S.R. Naphtha
17.6
45
461
156
286
2918
257
1
111
2637
6835
87932
0
74361
0
(13)
Cracked Naphtha
7.0
202
6285
1500
12002
17043
767
0
0
454
828
14288
11952
4682
0
(8)
Reforaate
4.3
232
3341
1441
7349
13949
318
0
0
30
1032
15308
0
0
0
(2)
Extract
1.1
196
2824
1217
6210
528
1
0
0
1
1
21
0
0
0
. (2)
Rafflnata
1.1
1
8
3
17
25
1
0
0
1
693
10252
0
0
0
(4)
Benrene
2.2
21846
44
0
0
0
0
0
0
0
no
0
0
0
0
(2)
Toluene
1.1
11
10921
44
11
0
0
0
0
0
0
13
0
0
0
(4)
S0t
2.2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
(2)
Middle
Distillate
1.1
0
0
0
1
9
!
0
0
61
0
9828
0
ltOO
0
TOTALS
100.0
22533
23884
4361
25876
34472
1345
1
111
3184
9499
663492
29052
80143
80050
Normalised
Total
23040
24421
4459
26458
35247
1345
1
113
3256
9713
678416
29705
81946
81850
-------
device. All relief valves in the model refinery are assumed to
vent to the atmosphere.
Still a different procedure was required to charac-
terize the hydrocarbons emitted from the API separators.
Analyses were available of the inlet oil to the separator and
the recovered oil. A hydrocarbon material balance was then made
to determine the composition of the evaporative emissions from
the separator, as shown in Table D3-11. The available analyses
showed only the aromatic components, so the balance of the oil
was assumed to fall in the alkane family.
This material balance approach assumes that any com-
ponent lost from the oil phase is lost as evaporative emissions.
This neglects the slight solubility of certain components which
could result in mass transfer to the water phase (or eventually
even the sludge phase). This results in another conservatively
high, worst case assumption.
3.4 Summary of Hydrocarbon Species Emissions
The emissions of selected hydrocarbon species were
calculated by the above methods. The results are summarized in
Table D3-12. These figures represent only fugitive hydrocarbon
emissions and not the point source emissions from process
heaters, fluid catalytic cracking regeneration, etc. The stack
hydrocarbon emissions did not significantly affect any of the
critical points for hydrocarbon species because of the effects
of height of release and plume rise.
D-30
-------
TABLE D3-11. ESTIMATED COMPOSITION OF INLET OIL,
HYDROCARBON VAPOR, AND OUTLET OIL
STREAMS AROUND AN API SEPARATOR
Estimated Composition of Streams Wt %
Vapor Outlet
Inlet
From
(Skimmed)
Component
Oil
Separator
Oil
Benzene
0.03
0.07
0.01
Toluene
0.22
0.22
0.22
Ethylbenzene
0.06
0.06
0.06
Xylenes
0.21
0.21
0.21
Alkylbenzenes
0.80
0.80
0.80
Naphthalene
0.29
0.29
0.29
Anthracene
0.04
0. OA
0. OA
Biphenyl
0.18
0.18
0.18
Polynuclear Aromatics (PNA's)
2.04
0.15
2.98
Alkanes
96.13
97.98
95.21
100.00
100.00
100.00
Skimmed Oil Rate = 667 lb./lOOO lb. inlet oil
Vapor Lost from Separator = 333 lb./lOOO lb. inlet oil
D-3]
-------
TABLE D3-12. SUMMARY OF HYDROCARBON SPECIES EMISSIONS FROM
FUGITIVE SOURCES
Source
Component
V, P, C, F, D, CT*
Relief Valves
API Separators
n-Hexane
Other Alkanes
Olefins
Cycloalkanes
Hydrogen
TOTALS
16,000
654,000
6.3
255.9
46,000 18.1
135,000 52.9
31,000 12.3
9,700 0.2
678,000 11.3
30,000 0.5
82,000 1.4
82,000 1.4
I** I.
980,000 502.4
391.2
16.8
512.8
Totals
ppmw
Vg/hr
ppmw
Vg/hr
ppmw
kg/hr
ppmw
kg/hr
Benzene
7,200
2.8
23,000
0.4
700
0.4
3,900
3.6
Toluene
21,000
8.2
24,000
0.4
2,200
1.1
11,000
9.7
Ethylbenzene
5,600
2.2
4,500
0.1
590
0.3
2,800
2.6
Xylenes
31,000
12.1
26,000
0.4
2,100
1.1
15,000
13.6
Other Alkylbenzenes
42,000
16.6
35,000
0.6
7,900
4.1
23,000
21.3
Naphthalene
1,700
0.7
1,400
0.02
2,900
1.5
2,400
2.2
0
1
Anthracene
20
0.01
1
0.0
390
0.2
220
0.2
CO
to
Biphenyl
230
0.1
110
0.0
1,800
0.9
1,100
1.0
Other PHA's
7,700
3.0
3,300
0.05
1,500
0.8
4,200
3.9
7,100 6.5
840,000 769.6
20,000 18.6
59,000 54.3
15,000 13.7
920.8
Fugitive emissions from valves, pumps, compressors, flanges, drains, and cooling towers.
^Components marked with "i" are indicated present, but no quantifiable concentration data vera available.
-------
4.0 ATMOSPHERIC DISPERSION MODELING
With the hypothetical refinery and its emissions
defined, the next step is to predict ground level concentrations
of the various pollutants. This can be- done by using any one of
a large variety of computer modeling techniques. This section
will discuss the rationale behind the choice of the model used
and will give the details of its application to the hypothetical
refinery.
4.1 Choice of the Dispersion Model
There are a variety of atmospheric dispersion models
available, each with distinct advantages in certain situations.
Many of the more sophisticated models have been developed to
account for unusual conditions of site topography, meteorological
conditions, etc. These models would add little to this analysis
of a hypothetical refinery because no hard data for such variables
are available.
There are several guidelines to be used in choosing a
model. First it should be an established, well accepted model.
It should have the capacity to handle a large number of both
point sources and area sources. It must be able to give not only
the total concentration of the pollutant at any given point, but
also the relative contribution of each source to that total.
Although not alone in satisfying these requirements,
the EPA guideline model RAM is certainly the most well-known.
There are two versions of RAM, the rural and urban verions. The
urban version has slightly higher dispersion coefficients to
account for the numerous heat sources typical of an urban environ-
ment. As with other unconstrained choices, the worst case was
D-33
-------
chosen, which means the rural version of RAM. The rationale
behind always choosing worst case conditions is the desire to
generalize the results of the hypothetical refinery analysis.
If the worst case analysis shows little or no impact, then it
can be said with some confidence that refineries in general have
little or no impact. The worst case scenario is softened some-
what by the omission of storage tank emissions from the modeling,
but the effects of this will be discussed in the conclusions
section.
The choice between the urban and rural versions of RAM
is not straightforward. A case could be made for the use of the
urban model with larger dispersion coefficients because of the
"heat island" effect of the refinery itself. While this approach
might better simulate the "real world" effects of the refinery,
it would make it impossible to draw any conclusions from the
results of the environmental assessment. It should also be
noted that the primary objective of this assessment is not the
prediction of absolute values of various pollutant concentrations
outside the hypothetical refinery, for that would not be possible
given the general nature of the refinery model and the state-of-
the-art for both emissions estimating and atmospheric modeling.
The results of this assessment are intended to show the relative
impacts of the many sources in a refinery and to draw rough con-
clusions as to the potential health effects of refinery emissions
With these objectives in mind, the use of the RAM rural model was
justified.
4.2 Application of the Dispersion Model to the
Hypothetical Refinery
The air quality impacts of the model refinery were
predicted with RAM, an EPA guideline model.8 It is capable of
predicting a 1- to 24-hour average concentration of relatively
D-34
-------
unreactive pollutants. A maximum of 250 point and 100 area
sources can be modeled. Concentrations are predicted at a
maximum of 150 selected locations (receptors).
RAM uses Gaussian steady-state dispersion algorithms
for areas where one wind vector for each hour is a good approxi-
mation. Concentrations are calculated hour by hour as if the
atmosphere had achieved a steady-state condition.
One set of inputs to the model are hourly meteorologi-
cal data consisting of:
Variable
Assigned Value
Wind speed
Wind direction
Temperature
Stability class
Mixing height
4.5 m/sec
Worst case - Due West
25°C
500 m
These meteorological parameters are set by the "standard
receiving atmosphere" as defined by Monsanto in their source
severity work.1 It is recognized that a persistence of these
conditions for a 24-hour period is quite improbable, and that
this again results in a worst case approximation of "real world"
conditions. The source severity methodology is specific in
requiring that these conditions be used, and no provisions were
given to incorporate variations when modeling for longer averaging
times.
D-35
-------
The dispersion coefficients (the Gaussian o's in the
vertical and horizontal directions) are empirically determined
as functions of atmospheric turbulence, distance from the source
and the concentration averaging time. .Thus, the spread of the
plume is dependent on these three factors. The atmospheric
turbulence is defined by stability classes. These classes,
which range from very unstable to neutral to very stable atmos-
pheres, are determined by wind speed and insolation during the
day, or wind speed and cloud cover during the night. This rela-
tionship is presented in Table D4-1.9 The most unstable class
is A with F the most stable. The C stability class used here
is considered neutral.
TABLE D4-1. RELATIONSHIP BETWEEN STABILITY CATEGORIES
AND SURFACE METEOROLOGICAL CONDITIONS9
Day Night
Solar Radiation Thinly Ovei
or
m/sec Strong Moderate Slight ^4/8 Low Cloud Cloud
Surface Wind Incoming Solar Radiation Thinly Overcast < •?/«
Speed (at 10 m), or
<2 A A - B C
2-3 A-B B C E F
3-5 BB-CC D E
5-6 CC-DD D D
>6 C C C D D
The mixing height was set at 500 meters. This mixing
height allows the plumes to rise to their maximum height without
causing them to be trapped above or reflected off the upper
boundary of the mixing zone.
D-36
-------
The worst-case wind direction is dependent on source
geometry and emission parameters. The worst-case direction was
determined by modeling with the wind coming from 16 directions
for one hour each and comparing the predicted concentrations.
All the meteorological conditions, except wind direction, are
constant for the 24 hours. The alternating wind directions are
repeated in sequence every 3 hours (5 degrees either side of and
including the worst-case wind direction). Thus, the predicted
24-hour concentration would be the same as the 8-hour and 3-hour
concentrations.
Annual concentrations (for comparison with NO2 NAAQS)
can be predicted with Larsen statistics.10 Using empirically
determined ratios, the maximum annual concentration can be
determined from mean concentrations for shorter averaging times.
These ratios are functions of the standard geometric deviation
of the shorter averaging times. Using data collected in 1977
by the Texas Air Control Board, a typical standard geometric
deviation of 1.85 per 24-hour N0X concentration was determined.
This figure was determined by averaging the standard geometric
deviations reported for about 50 noncontinuous NO2 monitors
located in TACB Region VII (which is located on the upper Texas
Gulf Coast). Based on a standard geometric deviation of 1.85,
Larsen's model estimates that the ratio of the average 24-hour
N0X concentration to the expected maximum annual concentration
would be 4.85.
RAM can accept both point sources and area source
inputs. The data required to characterize a point source
includes:
• Source coordinate.
• Emission rate.
D-37
-------
• Physical height.
• Stack diameter.
• Stack gas exit velocity.
• Stack gas temperature.
Area source parameters consist of the:
• Coordinates of the southwest corner.
• Side length.
• Total area emission rate.
• Effective height.
Stacks, flares, etc., were modeled as point sources.
Fugitive emissions were modeled by three different methods.
• As a single point source originating
in the center of the process unit plot.
• As a pseudo-area source (where the single
point source was divided into three point
sources distributed across the unit in a
plane perpendicular to the worst-case wind
direction).
• As area sources.
It was hoped that the point source approximation would not
significantly affect the results, since this type of calculation
D-38
-------
requires much less computer time. This approach gave very
unrealistic boundary line conditions, however, with quite
large concentration peaks directly downwind of the unit center-
lines and very low concentrations elsewhere. The pseudo-area
approach had some smoothing effect, but' only the rigorous area
source approach gave satisfactory results.
Concentrations from the point sources are a function
of the distance downwind and crosswind from the source to the
receptor. Concentrations due to area sources are calculated
using the narrow plume approximation. This neglects diffusion
in the crosswind direction and assumes that an area source con-
sists of many narrow plumed point sources. As a result, any
receptor that has no area sources directly upwind receives no
contribution to its predicted concentration from area sources.
A more detailed discussion of the narrow plume approximation is
beyond the scope of this assessment, but the theory and valida-
tion of this approach are discussed at length in the
references.11'12'13 The five degree variation in wind persis-
tence did add some dispersion outside the worst-case wind
direction streamline.
The locations of a series of permanent receptor sites
were also input to the model. The locations consisted of a
grid placed in the area of greatest impact as predicted by the
worst-case wind direction. The model then calculated the
24-hour average concentration at each receptor. From these data,
maximum concentrations were determined. Also, isopleths (lines
of equal concentration) were plotted. Not only were the total
ambient concentrations displayed for each receptor, but these
concentrations were broken up into their component contributions
from each of the sources.
D-39
-------
5.0
IMPACTS ON AMBIENT AIR QUALITY
The outputs from the atmospheric dispersion model
consist of estimated maximum ground level concentrations for
criteria pollutants and nonmethane hydrocarbons. In this
section, these outputs will be presented and compared to National
Ambient Air Quality Standards (NAAQS). In addition, the total
hydrocarbon concentrations will be broken down into selected
components for comparison to toxicity data. The sensitivity of
these results will then be examined and their significance
discussed.
5.1 Criteria Pollutants and Total Hydrocarbons
The modeling results for criteria pollutants and non-
methane hydrocarbons are summarized in Table D5-1. Note that
the four criteria pollutants do not exceed the NAAQS, those
being particulates, oxides of sulfur, oxides of nitrogen, and
carbon monoxide. The estimated maximum concentration for non-
methane hydrocarbons does exceed the listed standard, but it
should be noted that this is only a guideline standard based on
the potential contribution of nonmethane hydrocarbons to oxidants.
The maximum ground level concentration of particulates
was 68 yg/m3 as compared to the NAAQS of 260 yg/m3. It should be
noted that this considers only process particulates which result
primarily from the FCC and oil fired heaters, and does not include
fugitive dust from unpaved roads, construction activities, or
coke handling. The point of maximum concentration occurred due
west of the refinery center at a distance of 1.5 kilometers from
the fence line.
The maximum concentration of S0y was found to be
288 yg/m3 as compared to the NAAQS of 365 yg/m3. The maximum
D-40
-------
TABLE D5-1. SOURCE SEVERITY FACTORS FOR CRITERIA
POLLUTANTS
Pollutant
X t
max
yg/m3
Fft
yg/m3
sftt
Particulates
68
260
0.26
SO
X
288
365
0.78
CO
16
10,000
0.0016
N02
55
100
0.55
Nonmethane Hydrocarbons*
9644
160
60.3
t
tt
X is the maximum ground level concentration,
max °
is the acceptable pollutant concentration (which is the NAAQS for
criteria pollutants).
+++,
is the source severity, with the following decision levels,
if S ^ 1: Additional Emission Reduction Probably Required
if 0.1** < S < 1.0: May or May Not Require Additional Emission
if S < 0.1**:
Reduction
Additional Emission Reduction Probably Not Required
* The nonmethane hydrocarbon standard is a guideline standard based
on the estimated contribution of hydrocarbons to oxidant formation.
** The lower critical value may need to be as low as 0.01 where large
uncertainties are involved.
D-41
-------
point was due west of the fluid catalytic cracking unit and
occurred at one-half kilometer from the refinery boundary.
The maximum 8-hour concentration of CO was predicted
to be 17 yg/m3 as compared to an NAAQS of 10,000 yg/m3. The
maximum point occurred due west of the refinery center and at a
distance of 1.25 kilometers from the boundary line.
The maximum 24-hour average N0y concentration was
estimated to be 269 yg/m3. By applying the Larsen statistics
discussed in Section A, the annual average concentration of N0y
was estimated to be 55 yg/m3. This figure is well below the
NAAQS value of 100 yg/m3 as a maximum annual average. Since
this analysis used the conservative assumption that all N0y was
emitted as NO2, the actual NO2 levels should be even lower.
An analysis of the source severity factors indicates
that none of the criteria pollutants has a high probability of
causing a public hazard (as indicated by S > 1). On the other
hand, only CO has a source severity factor low enough to have
strong confidence that it does not create a hazard. The others
are in the area of uncertainty where no clear decision can be
made, but the worst case scenario used would make the probability
of any hazard resulting very low.
The total nonmethane hydrocarbons were found to be in
excess of the 160 yg/m3 guideline standard, with a maximum con-
centration of 96AA yg/m3. This point was located on the refinery
boundary and due west of the main processing area. Although the
concentrations fell off rapidly from the maximum, the 160 yg/m3
isopleth extends about 3.5 kilometers downwind and encompasses
about four square kilometers, as shown in Figure D5-1. The
source severity factor for total hydrocarbons is quite high, but
D-42
-------
9644 pg/m
CO
Pollutant: HC
NAAQS : 160 ug/tn3
Area in Excess of
East
4.05 km2 (1.57 mi2)
South
Figure D5-1. Hydrocarbon isopleth
-------
this must be interpreted carefully. The NAAQS guideline for
hydrocarbons is based on the prevention of the formation of photo-
chemical oxidants rather than on primary toxicity data. While
the health effects and aesthetics of smog are important, more
emphasis will be placed on the source severity factors for hydro-
carbon components for which toxicity data is available.
5.2 Selected Hydrocarbon Components
The ambient concentration of any given hydrocarbon
species can be determined by the following relationship:
Xg = (XT) (PPMWs) (10-6)
where
Xg = the hydrocarbon species ambient concentra-
tion contributed by a given source,
= the total hydrocarbon ambient concentration
contributed by a given source, and
PPMW = the concentration in weight parts per million
O
of the subject species in that source.
This is based on the dispersion model assumption that all
species will disperse at the same rate; or in other words, that
atmospheric turbulence far outweighs any differences in molecu-
lar diffusion between species.
The first point of interest is the receptor showing
the largest total hydrocarbon concentration. Table D5-2 shows
the component breakdown at that point. This maximum point is
D-44
-------
TABLE D5-2. HYDROCARBON SPECIES AMBIENT CONCENTRATION AT THE
POINT OF MAXIMUM TOTAL HYDROCARBON CONCENTRATION
Location: On the west boundary line at a point 1650 meters from the north-
west corner; directly downwind of source L49 (an API separator).
Component Concentration, yg/m3 Concentration, ppmv
Benzene 6.6 0.0019
Toluene 21.2 0.0051
Ethylbenzene 5.7 0.0012
Xylenes 19.8 0.004
Other Alkylbenzenes 102.2 0.017
Naphthalene 27.5 0.0047
Anthracene 3.6 0.0005
Biphenyl 16.5 0.0025
Other Polynuclear Aromatics 22.7 0.0030
n-Hexane 2.8 0.0007
Other Alkanes 9380.0 1.876
Olefins 0.0 0.0
Cycloalkanes 33.7 0.009
H2 1.8 0.020
Total Nonmethane Hydrocarbons 9644.0 yg/m3 1.95 ppmv
D-45
-------
located directly downwind of the API separator (Source L-49),
and 97.8 percent of the hydrocarbon species at that point came
from the separator. The bulk of the hydrocarbons are from the
alkane family (9380 yg/m3 or 1.9 PPMV), but both the aromatics
and polynuclear aromatics species are present at the part per
billion level (PPB).
It is also desirable to find the point of maximum
concentration for each potentially hazardous component, if that
should prove to be different from the point of maximum total
hydrocarbons. A limited search was carried out to find these
species maximum points by finding the maximum points for units
with high concentrations of the subject species. For example
in the case of benzene, the maximum point for each catalytic
reformer, the aromatics extraction unit, the aromatics frac-
tionation unit, and the fluid catalytic cracking unit were
determined. A complete component breakdown was calculated at
each point to detect unit interactions, and the point of maxi-
mum benzene concentration was selected. A similar procedure
was carried out for each component, and the resulting maximum
concentrations are summarized in Table D5-3.
It is interesting to note that all of the species
maximum concentrations came from the two highest ranked points
for total hydrocarbons. Five species (including benzene,
naphthalene, anthracene, biphenyl, and the general polynuclear
aromatics family) have their maximums adjacent to the API
separator. The other species maximum values were found at a
receptor on the west boundary about 1380 meters from the north-
east corner. The largest contributor to this point was the
crude distillation unit (L28-1). Other significant contributing
units included:
D-46
-------
TABLE D5-3. MAXIMUM AMBIENT CONCENTRATION OF SELECTED HYDROCARBON SPECIES
Location
Ambient Concentration
Component
yg/nr
ppmv
On the West Boundary,
XXXX meters from the
Northeast Corner
Benzene
Toluene
Ethylbenzene
Xylenes
Other Alkylbenzenes
Naphthalene
Anthracene
Biphenyl
Other Polynuclear
Aromatics
n-Hexane
Olefins
Cycloalkanes
6.6
26.3
10.7
53.6
105.5
27.5
3.6
16.5
22.7
58.5
37.6
365.8
0.0019
0.0063
0.0022
0.0092
0.0179
0.0047
0.0005
0.0025
0.0030
0.0152
0.010
0.099
1650
1380
1380
1380
1380
1650
1650
1650
1650
1380
1380
1380
-------
• Two catalytic reformers (L36-1 and L29-1).
• Aromatics extraction (L37-1).
• Alkylation (L32-1).
• Fluid catalytic cracker (L38-1).
• Delayed coker (L40-1).
• Hydrogen plant (L31-1).
• Resid HDS (L30-1).
The largest concentration for any single component examined was
found to be hexane at a concentration of 15 PPBV.
To assess the impact of a given concentration of a
pollutant species, quantifiable toxicity data must be available.
The Monsanto approach uses the term "acceptable pollutant con-
centration" as the level at which there is a very low proba-
bility of adverse impacts on the general public. For criteria
pollutants, the Primary Ambient Air Quality Standards were used
as the acceptable pollutant concentrations. For other species,
the acceptable concentration can be estimated from the Threshold
Limit Value (TLV) as shown below:
F « TLV(G),
where
G = (8/24) (1/100) - 1/300,
so
D-48
-------
F = TLV/300.
The factor "G" is defined as a conversion factor to change a
TLV into an "equivalent PAAQS" and is calculated from two factors:
• The ratio (8/24) converts the TLV from an
8-hour per day basis to a 24-hour basis.
• The factor (1/100) is a safety factor to
account for the fact that the general public
is more susceptible to illness than the
industrial work force (for whom the TLV was
set).
Table D5-4 shows a summary of the acceptable pollutant
concentrations that result from this operation. Note that some
of the values are in parentheses. These are values arbitrarily
assigned to a family of chemicals, some of whose members have
TLV's that average out to the assigned value. These values
are interesting, but they should be used with caution. Not all
of the members of such a family are equally toxic, nor is it
certain that their effects would be additive. If the source
severity factors based on these values are low, then it can be
said with some confidence that no damage will be done by those
compounds. If the values are high, however, no conclusion can
be drawn.
The whole process of basing an acceptable pollutant
concentration on TLV's should be critically appraised. These
values were set for use in industrial hygiene studies within
plant boundaries, and the American Conference of Governmental
Industrial Hygienists (ACGIH) specifically warns against their
use:
D-49
-------
TABLE D5-4. SUMMARY OF "F" VALUES
Pollutant
F yg/nr
Based on
Benzene
Toluene
Ethylbenzene
Xylenes
Other Alkylbenzenes
114
1,388
1,586
1,586
(488)
TLV = 10 PPM
TLV = 100 PPM
TLV = 100 PPM
TLV = 100 PPM
TLV = (25 PPM)
Naphthalene
Anthracene
Biphenyl
Other Polynuclear Aromatics
194
0.66
4.4
(25)
TLV =10 PPM
TLV = 200 yg/m3*
TLV =0.2 PPM
TLV = (1 PPM)
n-Hexane
Other Alkanes
1,281
(16,665)
TLV = 100 PPM
TLV = (1,000 PPM)
Olefins
(12,344)
TLV = (1,000 PPM)
Cyloalkanes
(4,937)
TLV = (400 PPM)
*Based on "Coal Tar Pitch Volatiles" of which anthracene is a
major component.
D-50
-------
• as a relative index of hazard or toxicity,
• in the evaluation of community air pollution, or
• in estimating the toxic potential of continuous,
uninterrupted exposure.
While such usage is discouraged, the fact remains that no other
source of quantifiable toxicity levels is available. Therefore,
the use of TLV's to estimate acceptable pollutant concentrations
is used here in accordance with the source severity methodology.
It is felt that this comparison, however tenuous it might be, is
better than ignoring the problem. This is especially true as
long as the proper qualifications and limitations on the results
are explicitly stated.
Taking the maximum ambient concentrations presented
in Table D5-3 and the acceptable pollutant concentrations shown
in Table D5-4, it is straightforward to calculate source severity
factors for each component. These factors are shown in Table D5-5.
Using Monsanto's recommended decision levels, it can
be said that there is a significant probability that anthracene
and biphenyl need further application of control technology.
Several things should be noted in that context:
• The high concentrations of these species were
contributed by a covered API separator.
• That separator was located right on the plant
boundary line, which is quite unusual.
D-51
-------
TABLE D5-5. SOURCE SEVERITY FACTORS FOR SELECTED
HYDROCARBON SPECIES
Component
max
ug/m3
ug/m;
Benzene
Toluene
Ethylbenzene
Xylenes
Other Alkylbenzenes
6.6
26.3
10.7
53.6
105.5
114
1388
1586
1586
(488)*
0.06
0.02
0.007
0.03
(0.22)
Naphthalene
Anthracene
Biphenyl
Other Polynuclear
Aromatics
27.5
3.6
16.5
22.7
194
(0.66)
4.4
(25)
0.14
(5.5)
3.8
(0.9)
n-Hexane
58.5
1281
0.05
Olefins
37.6
(12344)
(0.003)
Cycloalkanes
365.8
(4937)
(0.07)
If
Values in parentheses are an average of the F values for
several selected members of the family group, and are not
true F values for the entire family.
"7
D-52
-------
• There is a great deal of uncertainty in the
emission factor for separators. No conclusive
results were obtained from limited testing of
separators on this program, so AP-42 factors were
used in this analysis. The EPA has since begun a
program to improve these factors.
• The emissions from an API separator are
highly variable in component breakdown (much more
so than process unit emissions), and the species
breakdown for that unit is based on several grab
samples which may well not be reflective of
"typical" operation.
• These component emissions (calculated by an oil-
phase volatile hydrocarbon balance) may be over-
stated due to solubility effects.
It can also be stated that there is a very low prob-
ability of the need for further control of ethylbenzene and the
olefin family. All other species fall into the range where no
clear decision can be made. The uncertainties involved in the
calculation of these source severity factors make it impossible
to make clean cut decisions for the range from 0.01 to 0.99.
It should also be noted here that all of the quoted
hydrocarbon species maximum points occurred on the refinery
boundary. Because they are released close to the ground and
with little velocity or thermal buoyancy, the vapors tend to
stay at ground level. Dispersion does proceed at a relatively
rapid pace when moving downwind. This establishes two inter-
esting points:
D-53
-------
• The sphere of influence for hydrocarbon
species that were noted as potential problems
at the boundary line does not extend more than
a few hundred meters.
• This further suggests that buffering areas
with a high potential for fugitive emissions
could be effective in reducing or eliminating
high source severities.
5.3 Discussion of Results
The detailed results of the environmental impacts of
emissions from the hypothetical refinery were presented in the
last section. This section will examine the significance of
those results and draw conclusions where possible.
The first step is an examination of all the input
variables to see how sensitive the results are to changes in the
assumptions that were used. The most significant variables to
consider are:
• Refinery processing configuration.
• Refinery layout.
• Calculated emissions.
• Atmospheric dispersion model choice.
• Meteorological conditions.
• Hydrocarbon component breakdown.
D-54
-------
• Basic toxicity data.
• Modified toxicity data.
Several of these can be considered in a group. A
change in the calculated emission rates will produce a propor-
tional change in the predicted maximum concentrations. These
emissions will vary with a change in refinery processing con-
figuration, emission factors, or fitting counts. This in
itself is enough to prevent a complete generalization of these
results to the refining industry. A simple topping refinery,
for instance, will have lower emissions and quite different com-
ponent breakdowns, probably resulting in lower source severity
factors.
Another point of uncertainty is the potential contri-
bution of storage emissions to the impacts predicted for the
refinery process area. As described in Section 3, total storage
tank emissions were estimated to be 3308 tons per year or about
27 percent of all nonmethane hydrocarbon emissions. These
emissions were estimated for each of the storage modules listed
in Table D2-1. Since no layout information within the module was
available, each storage module was treated as an area source.
The specific emission rates (lb/hr/ft2) were calculated for
each module and compared to the API separator and the process
area which contributed to the two highest points of nonmethane
hydrocarbon concentration. The following conclusions can be
drawn:
• The worst storage tank module (L-22) had a
specific emission rate of 85.4 lb/hr/106 ft2.
This figure is much lower than either the covered
API separator (L-49) with 1169 lb/hr/106 ft2 or
D-55
-------
the worst process area (near L-28) with
320 lb/hr/106 ft2.
• Since all of these sources are directly adjacent
to the west boundary line, the predicted impacts
should be roughly proportional to the specific
emission rates. This is actually somewhat con-
servative, since the height of release of the
storage tank emissions would be considerably
higher than for a separator or for process
fugitives. Therefore, it can be concluded that
the specific impact of the worst storage module
would be significantly less severe than the two
worst points cited in this assessment.
• By examining the relative contributions of
adjacent process sources to the predicted
maximum and applying similar ratios to adjacent
storage modules, it was determined that the
inclusion of storage emissions in the modeling
would not have significantly increased the
estimated maximum concentration. It would,
however, have greatly increased the area of
impact in which relatively high concentrations
of nonmethane hydrocarbons would occur.
The refinery layout may be even more critical than the
complexity and the resulting overall emission rate, especially
from the hydrocarbon species. Fugitive emissions are released
near ground level, and thus are subject to much less dispersion
than stack emissions. A refinery layout with process units
right on the boundary line (such as the model used here) will
show much higher hydrocarbon concentrations at the boundary than
one with a buffer zone around the processing area.
D-56
-------
The situation is further complicated when looking at
individual species. For instance, a gas processing facility
near the fence line would result in high concentrations of
total hydrocarbons, but it would probably not cause any large
source severity factors. On the other hand, a complex con-
sisting of a reformer, an aromatics extraction unit, and a BTX
fractionation unit would result in moderate to low total hydro-
carbons, but they would probably result in high source severities
for the aromatics components.
The choice of which type of dispersion model to use
could affect the predicted pollutant concentrations signifi-
cantly. None of the available models is perfect, and predicted
maximum concentrations may vary from half to ten times (or more)
the actual concentrations measured from a source. The use of
the rural version of RAM to model a refinery is a conservative
choice since the heat island effect of a refinery will tend to
increase atmospheric diffusion. This should definitely be
considered in interpreting these results.
The predicted impacts will also vary with meteoro-
logical conditions, as illustrated by sensitivity runs on the
model. A 22 percent decrease in wind speed resulted in a 28
percent increase in predicted maximum ground level concentra-
tions for total hydrocarbons. Use of the more stable atmo-
spheric stability class D resulted in 3 percent higher predicted
concentrations. While this sensitivity is interesting, the
meteorological conditions used in the base case were firmly
fixed by the "standard receiving atmosphere" for source severity
calculations. The appropriateness of these conditions for pre-
dicting "real world" concentrations, however, is highly
questionable.
D-57
-------
The hydrocarbon component breakdown is also quite
critical. Individual component source severity factors will
vary in direct proportion with the predicted concentration of
that component in the emitting source. There are fairly wide
confidence intervals that should be applied to those component
breakdowns. Component concentrations will vary from day to
day with changes in feedstocks, operating conditions, etc.
This effect is amplified by several orders of magnitude when
discussing API separator emissions. The small number of
samples on which the component analyses are based is insuffi-
cient to confidently average out such process variations.
There is a shortage of quantifiable toxicity data
for the many organic compounds present in refinery processing.
This makes it difficult to prepare a comprehensive environmental
assessment. The accuracy of existing toxicity data is also
questionable, and this effect is compounded by the transforma-
tion from TLV to an "acceptable pollutant concentration."
Dividing the TLV by three to account for the difference between
eight hours per day and continuous exposure assumes that the
toxic effects are cumulative. For some compounds that is
certainly true, but others require some critical concentration
to be harmful and are easily assimilated by the body below that
concentration. The safety factor of one hundred to account for
the greater susceptibility of the general public is obviously
arbitrary and therefore questionable. Any change in the accept-
able pollutant concentration will produce an inversely pro-
portional change in the calculated source severity.
Recognizing the high degree of uncertainty in the
results, the following conclusions can be drawn:
D-58
-------
There is no certainty of public hazard
resulting from the emissions of this
hypothetical refinery.
Conversely, there is no certainty that
it does not create a hazard.
If any hazard exists due to hydrocarbon
species, the most likely species to cause
problems would be the polynuclear aromatics.
This approach to an environmental assessment
of a generalized source is of limited value
in providing specific information on whether
steps need to be taken to further reduce
emissions of a given pollutant.
The results can be useful in indicating the
relative impacts of various emission sources
and species. For instance, API separators
appear to pose more of a potential hazard
than fluid catalytic crackers; polynuclear
aromatics emissions appear to be more
troublesome than benzene. Such relative
ranking of emission sources and species can
be useful in directing emphasis towards
potential problem areas.
If this approach were used to assess the
impact of a specific plant, it might yield
more useful results. The range of uncertainty
would be much narrower because the input fac-
tors could be more firmly defined.
15-59
-------
6.0
EFFECTS OF EXISTING AND POTENTIAL REGULATIONS
AND POLICIES
This section will examine the effects of environmental
regulations and permit policies on the .emissions from petroleum
refining. The first two subsections will deal with state and
Federal regulations, respectively. The final section will
address the effects of potential new regulations.
6.1 State Regulations
Existing facilities are regulated by the states,
rather than by Federal standards. State regulations were
obtained from the Environmental Reporter. 1 ** Standards for
the South Coast and Bay Area regions of California are considered
here with the state standards. Though some state standards were
amended as late as 1979, most were adopted in the early 1970's.
There is a lot of disparity among the standards in
some categories; the general trends which could be discerned are
presented here along with notable exceptions. No attempt is made
to describe the standards of the individual states per se.
All states are included even if they presently have no
refineries. This should not be interpreted, however, to mean
that all states have specific standards for all the categories
included here. Some states have standards only for specific
existing facilities; others have no standards except for support
of federal standards. Some of the summaries presented below
represent the standards of only a few states.
D-60
-------
6.1.1 Particulate and Visible Emissions
Most states have specific standards for the maximum
opacity and darkness of emissions. The strictest standard, and
by far the most common, calls for a maximum opacity of 20
percent and a maximum darkness of No. 1 on the Ringlemann Chart.
In some states these stricter standards apply only to new
sources while existing sources are allowed an opacity of 40
percent and a darkness of No. 2 Ringlemann. In other states
these more lenient standards apply to new and existing sources.
One state allows 40 percent opacity for new sources and 60
percent for existing sources.
Some state standards specify either opacity or dark-
ness, but not both. Exception to the above standards is some-
times allowed for the flue gases from catalytic cracking
catalyst regeneration and fluid coking: these gases may be
allowed 25 to 40 percent opacity where other gases are limited
to 20 percent. In all states with visibility standards, pro-
vision is made for varying amounts of upset time.
Particulates are generally regulated by source. For
process emissions in general, many state standards have a chart
with pounds per hour allowable emissions versus tons per hour
process weight, with all stacks considered collectively.
Again, catalytic cracking catalyst regeneration is
sometimes considered separately. Although no exact comparison
of the various standards can be made because of widely varying
formats, the one pound per ton of coke bum-off standard of two
states appears to be the most stringent. When a CO boiler is
installed on the regenerator, an allowance is usually made for
the added emissions from fuel-burning.
D-61
-------
Particulate emissions from fuel-burning are also often
considered separately. The stipulation is generally made that
all fuel-burning at the facility is considered collectively.
Standards range from 0.1 to 2.5 pounds of particulates per
million Btu of heat input; a number of .the standards stipulate
a maximum of 0.6 lb/106 Btu or less. Some standards have varying
maximums for different size units. One state standard specifies
that afterburners must be used.
6.1.2 Sulfur Emissions
Several states limit SO2 emissions from any source in
a refinery to 500 ppmv; a common maximum is 2000 ppmv. One
state limits total SO2 emissions from the refinery to 10 percent
of the sulfur in the crude; another limits total S02 emissions
to 0.3 pounds per barrel of oil processed. Many states, however,
consider separately the sulfur emissions from fuel burning and
sulfur recovery. One state limits emissions of mercaptans
specifically to 0.25 pounds per hour.
Most standards for S02 emissions from fuel burning are
of two types. Some states limit the sulfur content of the fuel
burned while others specify a maximum amount of SO2 that may be
emitted per million Btu of heat input. When the sulfur content
of the fuel is limited, allowance is usually made for equiva-
lent alternate means of SO2 emission control.
Where the sulfur content of the fuel is limited, state
standards stipulate maximums of up to 2.5 weight percent sulfur.
Most maximums are 1.0 weight percent sulfur or less. Sulfur
content of gaseous fuels (often specifically fuel gas) is
expressed in grains of H2S per dry standard cubic foot (gr/dscf).
In this case the common maximum is 0.1 gr/dscf.
D-62
-------
Allowable SO2 emissions from sulfur recovery units are
sometimes expressed in pounds of SO2 per pound of sulfur processed.
These allowances range from 0.004 to 0.12 lb S02/lb S. Several
state standards contain a chart of allowable emissions versus
sulfur input. One state allows up to 1-000 pounds of SO2 per
hour. Limits of 500 to 2000 ppmv SO2 are in some instances set
specifically for sulfur recovery units.
Hydrogen sulfide emissions from sulfur recovery units
are addressed by a few states. One state allows 0.3 pounds of
H2S per hour. One state allows 0.1 ppm H2S and two others
allow 10 ppm H2S.
6.1.3 N0x Emissions
State standards for the control of N0x emissions from
fuel burning are quite consistent. These standards, which
normally apply only to units larger than 250 million Btu, allow
gas-fueled units to emit 0.20 pounds per million Btu and liquid-
fueled units to emit 0.3 pounds per million Btu. Solid-fueled
units are allowed 0.7 pounds per million Btu. When different
fuels are burned simultaneously, the applicable standard is
determined by proration.
6.1.4 Carbon Monoxide Emissions
One state limits carbon monoxide to 200 ppmv in fuel-
burning units larger than 107 Btu. All other CO standards are
for the catalytic cracking catalyst regeneration and for fluid
coking. Some states limit CO emissions from these sources to
500 ppmv; in some instances this limit applies only to new
sources. One state allows existing sources to emit 20,000 ppmv
CO, another allows the emission of 5 tons of CO per year.
D-63
-------
r. r f
..j5.6 Literature Cited
American Cyanamid Company. 1959. The chemistry of acrylonicrlie. 2r.d
Edition. pp.11-66. Petrochenicals Department, American Cyar.a^id
Company, New York. Hop,
Altshuller, A.P., I.R. Cohen, S.F. Sieva, and S.L. Kopczynski. 1962. Air
pollution: photooxidation of aromatic hydrocarbons. Science
138(3538):442-443.
Bowman, R.M., T.R. Chamberlain, C.W. Huang, and J.J. McCullough. 1974.
Medium effects and quantum yields in the photoaddicion of naphthalene
and acrylonitrile. J. Am. Chem. Soc. 96(3):692—7C*C .
Broderius, S.J. 1977. Personal communication concerning the face of
cyanides in the aquatic environment and EPA Grant R8C5291, Dec. 5, 1977.
Univ. of Minnesota, St. Paul.
Bruson, H.A. 1949. Cyanoethvlation. Organic Reactions 5:79-135.
Gut, I., J. Nerudova, J. Kcpeckv, and V. Holecek. 1975. Acryloni;rlie
biotransformation in rats, mice and Chinese hamsters as influenced by the
route of administration and by phenobarbital, SKF 525—A, cysteine,
dimercaprol, or thiosulfate. Arch. Toxicol. 33:151-161.
Laity, J.L., I.G. Burstain, and B.R. Appel. 1973. Photochemical smog ar.d
the atmospheric reactions of solvents. Chap. 7, pp.95-112. Solvents
Theory and Practice. R.W. Tess (ed.). Advances in Chemistry Series 124.
American Chemical Society, Washington, D.C.
Lank, J.C., Jr. and A.T. Wallace. 1970. Effect of acrylonitrile cn an-
aerobic digestion of domestic sludge. Eng. Bull. Purdue Univ., Eng. Ext.
Ser. 137(Part 2 ):518—527 . (Abstract cnlv). CA 1972 . 76:131165z.
Leo, A., C. Hansch and D. Elkins. 1971. Partition coefficients and their
uses. Chem. Rev. 71:525-612.
Sanchez, A., A. Hidalgo, and J.M. Serratosa. 1972. Adsorption of nitriles
on moncmorillonites. Proc. Int. Clay Conf., Madrid, 1972. J.M.
Serratosa (ed.). 617-626. (Abstract only), CA 1974. 81:96688b.
Shackelford, W.K. and L.H. Keith. 1976. Frequency of organic compounds
identified in water. U.S. Environmental Protection Agency, (ERL),
Athens, Ga. 617p. (EPA 600/4-76-062).
105-7
-------
equivalent methane or 8 pounds of hydrocarbons per hour before
controls must be applied; another state allows 5 tons of hydro-
carbons per year.
Other sources of hydrocarbon.emissions are mentioned
infrequently in state standards. Several standards specify that
hydrocarbon emissions from condensers, hot wells and accumulators
be incinerated, compressed, or equivalently controlled. One
standard allows no hydrocarbon emissions from fuel burning;
another specified 95 percent control of hydrocarbons from
vacuum systems and from process unit turnarounds. One standard
states that relief valves in pipes over one inch in diameter
must be vented to vapor recovery or disposal, be protected by a
rupture disc, or be maintained by an approved inspection system.
In one standard, emissions from air blowing must be incinerated
at 1400°F for 0.3 second or equivalently controlled.
6.1.6 Effects of State Regulations on the Environmental
Impacts of RefinerieF
It is quite difficult to assess the effects of state
regulations because of the great variety in standards. There is
no'doubt that significant emission reductions have been achieved
over the last ten to fifteen years by virtue of these regulations.
The model refinery used in this environmental assessment, however,
already reflects the control technology required by the consensus
of regulations for existing sources. Some reduction of the impacts
could be expected if the refinery were located in one of the
stricter states.
D-65
-------
6.2
Federal Regulations and Policies
Federal regulations apply primarily to new or modified
sources. These take the form of New Source Performance Stan-
dards (NSPS) and New Source Reviews required for permitting.
6.2.1 New Source Performance Standards
NSPS specific to refineries are contained in 40 CFR
Part 60, Subpart J. These standards apply to fluid catalytic
cracking unit regenerators, fluid cokers, sulfur recovery units,
and fuel sulfur levels. Subpart D contains standards for fossil-
fuel fired steam generators with a heat input greater than 250
million Btu. Subpart K includes standards for storage vessels
containing petroleum liquids, but these are outside the scope
of this study.
6.2.1.1 Particulate and Visible Emissions
Federal standards state that gases from fossil-fuel
fired steam generators may not exhibit more than 20 percent
opacity except for one 20-minute period per hour of not more
than 27 percent opacity. These gases also may not contain more
than 0.1 pound of particulate matter per million Btu of heat
input from the fossil-fuel.
Gases from fluid catalytic cracking catalyst regen-
eration may not exhibit more than 30 percent opacity, except for
one six-minute average reading per hour. These gases also may
not contain more than 1.0 pound of particulate matter per 1000
pounds of coke burn off.
D-66
-------
If the gases from the regenerator pass through an
incinerator or waste heat boiler in which auxiliary or supple-
mental liquid or solid fuel is burned, excess particulate
emissions may be allowed. These excess emissions may be 0.1
pound or less per million Btu of heat input attributable to the
added fuel.
6.2.1.2 Sulfur Emissions
When liquid fuels are used for steam generation,
sulfur dioxide emissions must be no more than 0.8 pounds per
million Btu of heat input. Any fuel gas which is burned in a
combustion device must contain no more than 0.10 grain of H2S
per dry standard cubic foot or the sulfur dioxide emissions
from the combustion device must be controlled in an equivalent
manner. Flares for the combustion of process upset gas or fuel
gas from relief valve leakage are exempt from this standard.
Sulfur dioxide emissions from Claus plants must be limited to
0.025 percent (250 ppm) by volume on a dry basis at zero percent
oxygen if emissions are controlled by an oxidation system (one
which converts emissions to sulfur dioxide) or by a reduction
system (one which converts emissions to hydrogen sulfide) followed
by incineration. If emissions are controlled by a reduction system
not followed by incineration, emissions from the unit may be 0.030
percent (300 ppm) reduced sulfur compounds and 0.0010 percent
(10 ppm) hydrogen sulfide calculated as sulfur dioxide at zero
percent oxygen on a dry basis. Reduced sulfur compounds include
hydrogen sulfide, carbonyl sulfide, and carbon disulfide.
6.2.1.3 Carbon Monoxide Emissions
The standards for carbon monoxide states simply that
no gases which contain more than 0.050 percent by volume (500
D-67
-------
ppmv) carbon monoxide may be discharged to the atmosphere from
a fluid catalytic cracking catalyst regenerator.
6.2.1.4 NO^ Emissions
Allowable N0X emissions from fossil-fueled steam
generators vary with the type of fuel used. When gaseous fuel
is used, emissions are limited to 0.2 pounds per million Btu of
heat input. For liquid fuels the limit is 0.3 lb/106 Btu and
for solid fuels 0.7 lb/106 Btu. When different fuels are burned
simultaneously, the applicable standard is determined by
proration.
6.2.2 New Source Review
The 1977 Amendments to the Clean Air Act emphasize the
control of atmospheric pollutants from new or modified facilities
by establishing a New Source Review (NSR) process. This is
essentially a Federal permit to construct any major emission
source. The review process can take one of two paths depending
on whether or not the source is to be built in an area in attain-
ment of the National Ambient Air Quality Standards (NAAQS). If
so, the Prevention of Significant Deterioration (PSD) regulations
apply. If not, then nonattainment regulations apply. Frequently,
both paths must be followed, since attainment is judged on a
pollutant-by-pollutant basis.
6.2.2.1 Prevention of Significant Deterioration
The PSD review process is a multilevel examination
of the emission levels and air impacts of the new source. The
overall process can best be illustrated by the flowchart shown
in Figure D6-1. It would not be pertinent here to examine in
detail the many applicability criteria which determine the level
D-68
-------
ModHied SiationMy Sources In PSD Areas
According to Alabama Power Decision
Mod Weal ton
Determination
Significant Nat Increaae
Oattrmlnalion
0
1
o>
vQ
Itp
TI
> KOtenefr
tetany pelHrtant?
\
TCS
)• rrt
lor any ponutam?
WotridaMttncraaao
In amltatona o<
MES pollutant wmK,
including any
bMratovrc* ollMti
and eumulatt** amJaMon
htct«BMt tine* Vtm.
PSOdoot
net apply
mafe* moAlieatlon
PSD appfict unfaaa
othamlt* nempt
WouM incfitu in PTC e< any
regutatad pollutant eic***)
denitnlmlt amotion
BACT ravivw lot
that pollutant
da m
rtd increaaa aieeed
mhfria impact Irai?
1
NO
mpMl (««4m to*
thai poJtut ant
Modi He atton «fetM
f*ntt In significant
nMlncnm BACT
and amtolant revtaw required
to* each pollutant unl««t
otharwtsa aiampt
ABBREVIATION CODE:
PtO« PumillaaofaigrtHcawtdatertaraUun
•ACT • Bait wallrtN fcawtrolinhnalf |
NAAOB- WaWenel urtlim at* NAAOS'
Would sourca impact
m+t Class I araa"*
» Ak quality review
~ Monitoring
» Vrsibtfiry.soMt. vegetation n
•0} nNr
-------
of review required. Suffice it to say that if a new or modified
refinery (which is one of the 28 major industry categories) has
the potential to emit more than 100 tons per year of any given
atmospheric pollutant, and that represents a net increase in
emissions since 8/7/77, and the net increase is greater than
the "de minimis" level, the new or modified section must demon-
strate the use of Best Available Control Technology (BACT).
BACT is defined as the level of emission control which
gives the lowest emissions while taking into consideration the
cost of control, energy efficiency, and technical feasibility.
BACT must, therefore, be determined on a case-by-case basis to
evaluate these effects. When an NSPS is available, this usually
forms the minimum criteria for BACT. When no NSPS exists, then
all possible methods of emission reduction must be catalogued.-
When one of these methods has been proposed as BACT for the new
source, all methods giving lower emissions must be shown to be
inappropriate in terms of cost, energy impact, or technical
feasibility.
6.2.2.2 Nonattainment Requirements
The requirements for permitting a source which will
emit significant levels of a pollutant for which the area is
not in attainment of the NAAQS are quite stringent. First, the
source must use the Lowest Achievable Emission Rate (LAER). It
must then offset the resulting emissions by reducing emissions
from another source in the area by a more than equivalent amount.
There are additional requirements relating to other sources owned
by the applicant and to assuring a net positive air quality
improvement, but these are not pertinent to this discussion.
LAER is defined as the strictest control technology
required for this type of source by any State Implementation
D-70
-------
Plan (SIP), or the lowest emissions achieved by any operating
source of the same type, whichever is more stringent, unless
the owner or operator of the proposed source demonstrates that
such limitations are not achievable. This does not take cost
or any other side effects into account." It also recognizes the
transfer of control technology from one type of source to another,
if technically feasible.
The resulting emissions after LAER must be offset on a
pollutant-by-pollutant basis by reducing emissions from other
sources in the area. For nonreactive pollutants the offset must
be from another source in the immediate vicinity. For N0X and
hydrocarbons, however, offsets can be obtained over a broader
area. The offsetting emission reduction must be greater than the
emissions from the new source, thus causing a new positive air
quality improvement.
6.2.2.3 Effects of New Source Reviews on the Environmental
Impacts of Refineries"
The effects of the New Source Review process on the
environmental impacts of refineries should be significant. Any
new refinery permitted under this system should have much lower
emissions than existing refineries. This would be particularly
true in the area of hydrocarbon emissions, but it would also
occur for N0X and S02.
The NSR process will also discourage expansion in non-
attainment areas, where the combined impacts of a heavily
industrialized area have already caused a deterioration in air
quality. If an expansion were to be made in such an area, it
could only be done by achieving a greater than equivalent offset.
D-71
-------
Thus, the impetus to build new facilities can provide the impetus
to clean up older facilities. The net effect of these policies
should be an improvement in existing air quality. In attainment
areas, refinery construction and expansion will be limited by
the constraints of PSD increments. Unlike the situation in non-
attainment areas, few existing sources will be available to
provide emission offsets which will be needed to allow growth
once the PSD increment is consumed.
6.3 Potential Regulations and Policies
There are many standards and regulations currently
under consideration that would have a significant impact on
refinery emissions. It would be quite difficult to document all
of these since many have not even been published as proposals
at this time. Several examples will be discussed in this section
to illustrate regulatory trends. Caution shoiuld be used in
interpreting or applying these regulations since they are only
proposed at this point, and they may be significantly modified
before being adopted.
6.3.1 State Regulations
Only the new and developing standards for the Bay
Area (San Francisco) and the South Coast (Los Angeles) regions
of California are summarized here. Other regulatory agencies
may be similarly updating their standards. Most of these new
and proposed standards are concerned with the emission of hydro-
carbons from refineries.
Two levels of control for S0X emissions from catalytic
cracker catalyst regeneration are being considered by the South
Coast region, one of which is expected to become a standard by
1982. One proposed standard calls for replacement of the
D-72
-------
conventional catalyst with a newly developed catalyst which can
reduce S0X emissions by 90 percent without additional controls;
the other proposes the addition of alkaline scrubbers for 90
percent control of S0x. The South Coast region also proposes
that the allowable sulfur contents of fuels be halved by 1982.
Many of the proposed standards are concerned with
fugitive emissions, an area not emphasized by present standards.
The South Coast region proposes that by 1980, leak rates, main-
tenance schedules, etc. for random hydrocarbon emissions be
established. Pumps and compressors within three miles of the
control center would be inspected every eight hours, all others
every 24 hours.
The South Coast region also proposes that natural
gas-fired control devices such as afterburners must have a
stand-by fuel system for use during natural gas curtailment.
By 1980, all relief valves would be vented to vapor recovery or
disposal. By 1982, combustion modification and/or ammonia injec
tion for control of N0X would be required on heaters and boilers
A Bay Area region standard which went into effect in
December 1979, limits valve leakage to 10,000 ppmv VOC measured
one centimeter from the leak. It is proposed that this standard
be applied also to flanges.
By March 1980, emissions from condensers or vacuum-
producing systems must be incinerated,compressed and added to
fuel gas, or controlled equivalently. It is proposed that
emissions from steam ejectors be similarly controlled.
D-73
-------
Also, since March 1980, hot wells and/or accumulators
associated with contact (barometric) condensers must be covered
and the organic vapors either incinerated or contained and
treated. It is proposed that this standard apply to the hot
wells and/or accumulators associated with all condensers.
Emissions from process vessel depressurizing must,
by 1980, be passed through a knockout pot to remove condensable
hydrocarbons, then incinerated, flared, or compressed and added
to the fuel gas. It is proposed that emissions from process
vessel purging be similarly controlled.
Another Bay Area region standard effective March 1980,
is similar to those in several other states: oil/water separators
must have a solid cover, a floating pontoon or double-deck cover,
90 percent effective vapor recovery, or other approved control
equipment.
6.3.2 Federal Standards
Petroleum refineries are among those industries for
which New Source Performance Standards (NSPS) will be formulated
in the near future. It is expected that within two to three
years additional standards will be added to Subpart J of 40 CFR
Part 60 and parts of the existing Subpart J may be revised.
Additional standards may concern such emissions as S0x from cata-
lytic cracking catalyst regeneration and fugitive emissions.
6.3.3 Effects of Potential Regulations and Policies on the
Environmental Impacts of Refineries
These proposed regulations will tend to bring more
uniformity to the determination of BACT. They do not generally
increase the stringency of measures already required through the
D-74
-------
New Source Review process. If other states follow California's
lead in upgrading their SIP's, however, the allowable emissions
from existing refineries could be greatly reduced. Such strin-
gency in state regulations may or may not be warranted, depending
on the magnitude of any air quality problems in the specific
state.
D-75
-------
7.0 EFFECTS OF NEW AND DEVELOPING TECHNOLOGY
Although petroleum refining is considered a mature
technology area, it is constantly undergoing a process of improve-
ment. The environmental impacts of refineries will be reduced
in the future through new developments in both process technology
and emission control technology.
7.1 Process Technology
Process technology in petroleum refining has continually
evolved to meet the demands of the end-use sector. Some of the
current evolutionary trends in refining include the shift to
produce lead-free gasoline, increased use of hydrodesulfurization
to achieve lower fuel sulfur contents, and a push for greater
energy efficiency. Some of these trends will tend to aggravate
emissions while others will reduce them.
The production of lead-free gasoline requires signifi-
cantly more processing in units like catalytic reformers, alkyla-
tion units, and isomerization units. While the units are not
major emitters, they do contribute to fugitive emissions and
emissions from combustion sources. Since there is a decrease in
gasoline range yields with this type of processing, more crude
must be charged to maintain the same gasoline production. This
will cause slight increases in emissions across the board. Much
of this effect is now behind us, but a phase-down of gasoline
pool lead content will cause continued emissions increases.
Sulfur levels in many fuels are being regulated down-
ward. This will require an increased use of hydrodesulfurization
to achieve these low sulfur levels, which will, in turn, increase
the load on Claus plants and tail gas treating units. The hydrogen
D-76
-------
demand will also begin to exceed that provided by catalytic
reforming, and thus require construction of hydrogen plants.
Both of these effects will tend to cause an increase in refinery
emissions unless countered by more effective control technology.
The trend toward greater use of energy conservation
will tend to reduce the emissions from combustion sources. The
recovery of process heat and the use of intrinsically more effi-
cient processes will reduce the heat required from process heaters
and steam boilers. Since the emissions from combustion equipment
are proportional to the fuel burning rate, this should result in
an emission reduction.
7.2 Emission Control Technology
New and improved emission control technology will
continue to appear in petroleum refining. Significant reductions
could be achieved by widespread application of recently developed
technology, such as covers for API separators, scrubbers for flue
gas from fluid catalytic cracking, and combustion modifications
to reduce N0X< These effects will be complemented by progess on
developing technologies like the fluid cracking catalyst which
adsorbs S0x from the regenerator.
One area with great potential for improved technology
is in fugitive emission control. The manufacturers of seals,
packing, and gaskets for process equipment have designed their
products to meet the users' needs. Up until now, those needs have
been to limit product loss and maintain safe operation. No one
was aware or concerned about "low level" fugitive vapor leaks
which could not be detected visibly. Fugitive emission regula-
tions will provide the incentive to develop more effective seals,
packing, etc., and will result in lower emissions and lower costs
for monitoring and maintenance programs.
D-77
-------
CONVERSION FACTORS
To Convert From
To
Btu
kcal
bbl
£
gal
£
ton
kg
lbs
kg
cm
in
ft3
m3
psi
kg/cm2
g/gal
g/A
Btu/bbl
kcal/£
kWh/bbl
kWh/£
lb/bbl
kg/i
lb/106 Btu
g/Hcal
grain/ft3
g/m3
gal/106ft3
£/106m
gpm
m3 /hr
lb/1000 gal
mg/iI
Multiply By
0.252
159.0
3.785
907.2
0.454
0.394
0.0283
14.223
0.264
0.0016
0.0063
0.0285
18.0
2.29
133.7
0.227
119.8
D-78
-------
REFERENCES
1. Serth, R.W., et al., Source Assessment: Analysis of
Uncertainty, Vol. II: Application to Air Source Assessment
Program, EPA-600/2-77-107, EPA, Research Triangle Park, N.C.,
1977.
2. Powell, D., et al., Development of Petroleum Refinery Plot
Plans, EPA-450/3-78-025, EPA, Research Triangle Park, N.C.,
June, 1978.
3. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Compilation of Air Pollutant Emission
Factors, AP-42, Third Edition plus Supplements, Research
Triangle Park, N.C.
4. Wall, J., Gas Processing Handbook, Hydrocarbon Processing,
April, 1975.
5. Wetherold, R.G., and Provost, L.P., Emission Factors and
Frequency of Leak Occurrence for Fittings in Refinery Process
Units, EPA-600/2-79-044, EPA, Research Triangle Park, N.C.,
February, 1979.
6. Bombaugh, K. J., et al., Sampling and Analytical Strategies
for Compounds in Petroleum Refinery Streams, Vol. II.
Radian Corporation, Austin, Texas, September, 1975.
7. American Petroleum Institute, Medical Research Report #EA-7103,
Petroleum Asphalt and Health. Reinhold Publishing Company, 1966.
8. EPA, Office of Air and Waste Management, Guideline on Air
Quality Models, Research Triangle Park, N.C., 1978.
D-79
-------
REFERENCES (Continued)
9. Turner, B. Workbook of Atmospheric Dispersion Estimates,
No. AP-26 EPA, Office of Air Programs, Research Triangle
Park, N.C., 1969.
10. Larsen, R. A Mathematical Model for Relating Air Quality
Measurements to Air Quality Standards EPA, Office of Air
Programs, Research Triangle Park, N.D., 1971
11. Novak, J.H., et al. An Efficient Gaussian-Plume Multiple-
Source Air Quality Algorithm APCA Journal, 26(6), June 1976.
12. Gifford, F. and Hanna, S. Urban Air Pollution Modeling. In:
Proceedings of the Second International Clean Air Congress,
H. Englund and W. Beery, ed. Academic Press, New York, 1971,
pp 1146-1151.
13. Thayer, D.S. and Koch, R.C. Validity of the Multiple-Source,
Gaussian Plume Urban Diffusion Model Using Hourly Estimates
of Input, Preprint Volume of the Conference on Urban Environ-
ment and Second Conference of Biometeorology. American Meteor-
ological Society, Boston, Mass., 1972.
14. Environment Reporter, State Air Laws, Bureau of National
Affairs, Inc., 1231 Twenty-fifth Street, N.W. , Washington, D.C.
D-&0
-------
APPENDIX E:
CONTROL TECHNOLOGY REVIEW
AND EVALUATION
-------
APPENDIX E: CONTROL TECHNOLOGY REVIEW
AND EVALUATION
TABLE OF CONTENTS
Section Page
1.0 INTRODUCTION E-l
1.1 Objectives . . . E-l
1.2 Scope E-2
2.0 CONTROL OF FUGITIVE EMISSIONS E-4
2.1 Sources of Fugitive Emissions E-4
2.2 Control Technology for Fugitive Emissions. E-4
2.2.1 Valves E-5
2.2.2 Flanges E-18
2.2.3 Pumps E-19
2.2.4 Compressors E-43
2.2.5 Agitators E-53
2.2.6 Pressure Relief Devices E-54
2.2.7 Sampling Connections E-62
2.2.8 Wastewater Systems E-64
2.2.9 Cooling Towers E-73
2.2.10 Solid Waste System Alternatives. . E-75
2.3 References E-84
3.0 CONTROL OF STACK AND OTHER PROCESS EMISSIONS. . E-91
3.1 Sources of Emissions E-91
3.1.1 Sulfur Recovery E-91
3.1.2 Catalyst Regeneration E-94
3.1.3 Boilers and Process Heaters10. . . E-99
E-ii
\
-------
TABLE OF CONTENTS (Continued)
Section Page
3.1.4 Vacuum Distillation E-101
3.1.5 Coking E-102
3.1.6 Air Blowing E-103
3.1.7 Chemical Sweetening E-104
3.1.8 Acid Treating E-104
3.1.9 Blowdown E-105
3.1.10 Compressor Engines E-106
3.2 Control Technology . E-107
3.2.1 Existing Levels of Control
in Refineries E-107
3.2.2 Control Technology Available
in Refineries E-149
3.2.3 Control Technology from Other
Industries E-169
3.3 References E-181
4.0 EMISSION REDUCTION THROUGH PROCESS MODIFICATION E-187
4.1 Alternative Operating Practices and
Conditions E-187
4.1.1 Catalytic Cracking Catalysts
Regeneration E-187
4.1.2 S0X Removal in the FCC
Regenerators E-190
4.1.3 Combustion Modification for
Control of NOx E-190
4.1.4 General Refinery Operations. . . . E-202
4.2 Alternative Fuels. E-206
4.3 Hydroprocessing as Feedstock
Pretreatment E-210
4.4 References E-215
5.0 CONVERSION FACTORS E-218
E-iii
-------
LIST OF TABLES
Table Page
E2-1 Approximate Distribution of Refinery
Process Valves by Type and Service E-7
E2-2 Packing Materials - Process Valves E-10
E2-3 Service Limits for Selected Mechanical
Packings E-23
E2-4 Distribution of Pump Seals in Radian
Refinery Study E-39
E2-5 Centrifugal Pump Seals - Cost Contribution
to Total Pump Cost E-40
E2-6 Estimated Energy Losses - Pump Seals E-41
E2-7 Types of Compressor Seals E-44
E2-8 API Recommended Cooling Standards E-45
E2-9 Compressor Seal Leakage ' E-51
E2-10 Basic Agitator Seals E-56
E2-11 Description of SRV Types E-57
E2-12 Safety Relief Valve (SRV) and Rupture
Disk (RD) Assembly Costs E-63
E2-13 Degree of Adoption of Various Wastewater
Treatment Processes E-67
E2-14 Typical Removal Efficiencies for Oil
Refinery Treatment Processes E-69
E2-15 Wastewater System Emission Factors E-70
E2-16 Radian-Generated Cooling Tower Emission
Factors E-74
E2-17 Total Estimated U. S. Refinery Solid
Waste Generation Rate E-83
E-iv
-------
LIST OF TABLES (Continued)
Table Page
E3-1 Process Emissions by Source and Type E-92
E3-2 Typical Compositions of Feed Stream and
Tail Gas for a 94 Percent Efficient Claus
Unit E-95
E3-3 Emission Factors for Uncontrolled Regeneration
of the Catalytic Cracking Catalyst E-96
E3-4 Emission Rates from FCCU Regenerators, Before
and After CO Boiler E-97
E3-5 Emission Rates from FCCU Regenerators, Equipped
with CO Boilers E-98
E3-6 Emissions from Refinery Boilers and Heaters. . . E-100
E3-7 Emission Factors for Reciprocating and Gas
Turbine Compressor Fueled with Natural Gas . . . E-106
E3-8 Existing Methods for Removal of Sulfur from
Claus Tail Gas E-120
E3-9 Typical Claus Plant Sulfur Recovery for
Various Feed Compositions E-124
E3-10 Typical Compositions of Feed Stream and Tail
Gas Streams from a 94 Percent Efficient Claus
Unit and Incineration E-126
E3-11 Available Methods for Removal of Sulfur from
Claus Tail Gas E-152
E4-1 Refining Sources of Thermal N0x and Fuel N0x . . E-191
E4-2 Boiler Combustion Modifications for Reduction
of N0X Emissions E-195
E4-3 Reductions of N0X Emissions with Combustion
Modifications at Various Boiler Loads E-197
E4-4 Engine Modifications Which Reduce N0X Emissions
from Internal Combustion Engines E-200
E -v
-------
LIST OF TABLES (Continued)
Table Page
E4-5 Emission Factors for Natural'Gas Combustion . . E-207
E4-6 Emission Factors for Fuel Oil Combustion in
Industrial Boilers E-208
E4-7 Refinery Fuel Emissions at Equivalent Heat
Release E- 209
E4-8 Typical Hydroprocessing Reactions E- 212
E-vi
-------
LIST OF FIGURES
Figure Page
E2-1 Process Valves E-8
E2-2 Globe Valves with Isolated Stem Seals E-17
E2-3 Raised-face Flange (gasket not shown) E-19
E2-4 Packed Seals E-21
E2-5 Single Mechanical Seals E-26
E2-6 Balanced and Unbalanced Mechanical Seals. . . . E-28
E2-7 Balanced Internal Seal Showing Seal
Coolant Inlet E-30
E2-8 Back-up Systems for Seal Failure E-32
E2-9 Double Mechanical Seal: Seal Liquid at
Higher Pressure Than Pumped Liquid E-34
E2-10 Tandem Mechanical Seal: Buffer Liquid is
at Lower Pressure Than Pumped Liquid E-35
E2-11 Tandem Seal with "Tell-Tale" Pressure
Gauge/Alarm E-37
E2-12 Straight Pass Labyrinth Compressor Shaft
Seal E-46
E2-13 Restrictive-ring Compressor Shaft Seal Et47
E2-14 Liquid-film Compressor Shaft Seal with
Cylindrical Bushing E-48
E2-15 Mechanical Contact Compressor Shaft Seal. . . . E-49
E2-16 Seals for Agitator Shafts E-55
E2-17 Typical Safety-Relief Valves E-58
E2-18 Rupture Disk and Hold-Down Flanges E-60
E2-19 Relief Valve/Rupture Disk Assembly E-60
E-vii
-------
LIST OF FIGURES (Continued)
Figure Page
E3-1 The Claus Process • E-122
E3-2 The CBA Process E-127
E3-3 The SNPA/Lurgi Sulfreen Process E-129
E3-4 The IFP-1500 Process E-132
E3-5 The BSRP Process E-134
E3-6 The SCOT Process E-137
E3-7 The Wellman-Lord Process E-140
E3-8 Total Sulfur Recovery with the IFP-150 Process E-158
E-viii
-------
1.0
INTRODUCTION
1.1 Obj ectives
In this appendix, techniques for controlling
atmospheric emissions are presented. These techniques include
controls that are typically used in the refining industry,
controls used in some areas in the refining industry that may
be applicable to other areas, and controls used in other re-
lated industries that could be applied to similar emission
sources in the refining industry. The effectiveness, cost,
energy requirement, and applicability of some of the emission
control technique are presented. Techniques for controlling
fugitive emissions and process emissions are included in this
appendix. Fugitive hydrocarbon emissions may be caused by leaks
from process equipment sealing devices such as pump and com-
pressor seals, valve seals, and flange gaskets or may be caused
by atmospheric exposure of hydrocarbon-containing substances
that result in evaporative hydrocarbon emissions. Fugitive
particulate emissions are generally the result of wind action
on unpaved areas and storage piles. Process emissions are
released to the atmosphere via a stack, flue, or vent duct.
Control techniques or control technologies include
equipment controls, process controls, feedstock and fuel
controls, and work practice controls. The term equipment
controls pertains to the substitution or modification of unit
operations hardware or addition of emission control hardware.
Emission control hardware can generally be described as device
which capture, collect, and/or destroy pollutants. Substitution
or modification of unit operations hardware can be characterized
as methods that reduce the likelihood of emissions from the unit
E-l
-------
operations process equipment. Process controls include any
changes in operating parameters that result in a net decrease
in emissions. Feedstock and fuel controls consist of modifi-
cations such as treating processes to remove pollutants before
utilization and substitution of materials containing no
pollutants or lower levels of pollutants. Work practice
controls consist of maintenance and housekeeping activities
that either reduce the emission potential of a source or that
identify emitting sources and subsequently reduce the emission
rate from the source.
1.2 Scope
Discussions of emission sources and emission controls
are limited to the battery limits of process units in refiner-
ies. Emissions from transfer facilities, storage vessels, or
other auxiliary processes such as electricity generation are
not included. Coal combustion is not discussed as a fuel
substitution alternative because of its complexity and its
economically questionable application to petroleum refining.
Electricity/steam cogeneration is not considered as a fuel
substitution alternative. Petroleum based fuels and natural
gas are considered as fuel substitution alternatives.
Controls for fugitive emissions discussed in Section 2
include equipment and work practice controls. Equipment
controls for process emissions are discussed in Section 3, and
process, fuel, and feedstock controls for process emissions
are discussed in Section 4. Controls for fugitive emission
sources are generally applicable to a particular source type
(valve, pump, etc.) regardless of the type of process unit.
E-2
-------
Fugitive emission controls are, therefore, discussed by source
type. Process emission controls are discussed on the basis
of the type of process unit, because of the differences in
emissions and controls between processes.
E-3
-------
2.0 CONTROL OF FUGITIVE EMISSIONS
2.1 Sources of Fugitive Emissions
Sources of fugitive emissions include process equipment
that can "leak" hydrocarbons such as valves, flanges, pumps,
compressors, agitators, and relief valves. Fugitive emissions
may also result from atmospheric exposure of hydrocarbons from
sample purging, drains, wastewater systems, and cooling towers.
Unpaved roads and outdoor storage piles may be sources of fugitive
emissions of particulate matter. Because the primary emphasis of
the sampling effort was measurement of hydrocarbon emissions,
these particulate emission sources are not discussed. Controls
for these two types of fugitive emission sources are practiced
in many other industries and include ordinary techniques such
as watering or paving of roads and covering or coating storage
piles.
Solid wastes generated in refineries may also be a
source of fugitive emissions of hydrocarbons, particulates, or
other pollutants depending on methods for handling and disposal
of the solid wastes. A brief description of sources of solid
wastes and disposal alternatives is included in Section 2.2.10.
2.2 Control Technology for Fugitive Emissions
For all of the sources that may "leak" hydrocarbons
(valves, pumps, etc.) one control option is a leak detection
and repair strategy. This type of work practice emission con-
trol is described in total for valves. For each of the
succeeding leak sources, only the specific features that would
be different for that type of source are described. Various
types of equipment controls may be applicable to more than one
E-4
-------
source such as mechanical seals for pumps, compressors, and
agitators. In these cases, reference may be made to a preceding
description of the control technology.
The descriptions of fugitive emission control technology
are presented for each type of emission source (valve, pump, etc.).
The order of presentation is such that sources with similar types
of controls are discussed in sequence. The relative contribution
of source types for a hypothetical refinery is presented in
Section 2.7.3 of Appendix B of this report.
Three levels of control are described for each source.
Existing controls are those in general refinery use, although
the extent of application may be variable. Available control
technology may be used in some areas of the refining industry
due to regulatory or other constraints, and control technology
transfer includes any types of emission controls that have been
applied to similar types of emission sources in other industries.
2.2.1 Valves
Valves can leak hydrocarbons through the junction where
the activating stem penetrates the valve body. Excessive leakage
from this junction is generally prevented by a packing gland or
a pressurized grease seal. If a valve is operated with one side
of the valve seat open to the atmosphere, such as for draining
or sampling operations, hydrocarbons may leak through the valve
seat. The following discussions pertain to these two potential
leak sources for valves. Recent review articles are available
that describe the overall features and functions of valves in
detail.1
E-5
-------
Table E2-1 contains the approximate distribution of
refinery valves screened by Radian within the battery limits of
major process units during the thirteen refinery sampling pro-
grams. The distribution of each type of valve is shown for
manually operated and automatically controlled service cate-
gories. The Radian survey results indicate that 88 percent of
all refinery valves screened were either manual gate valves (65
percent) or control globe valves (23 percent). Figure E2-1 shows
the internal features of the various types of valves. Check
valves are not discussed because they do not have a potential
leak junction, and relief valves are discussed in Section 2.2.6.
2.2.1.1 Existing Controls for Valves
Existing controls for valves include the valve stem
seal, inspection and maintenance practices, and closure of the
atmospheric side of open-ended valves.
Valve Stem Seals—The valve stem seal is designed to
prevent leakage of the contained fluid and is therefore a fugi-
tive emission control. All gate, globe, and butterfly valves
screened by Radian had a packed gland stem seal. These packed
stem valves represent approximately 94 percent of all refinery
valves. Plug valves typically have a grease-lubricated, tapered
plug to prevent leakage. Grease may be added periodically to
prevent leakage and to assure proper operation of the plug valve.
Packed stem seals consist of a stuffing box that
surrounds the stem, rings of compliant packing material in the
annular space, and a gland or follower that is used to compress
the packing against the stem to form a seal. Figure E2-1 shows
packing nuts that force the gland against the packing, but many
valves may have bolts or studs that pass through the gland in
E-6
-------
TABLE E2-1.
APPROXIMATE
VALVES3 BY
DISTRIBUTION OF REFINERY
TYPE AND SERVICE
PROCESS
Type
Valve
Manual
Service
Control
Total
Gate
64.7
0.0
64.7
Globe
3.8
23.3
27.0
Plug
5.7
0.0
5.7
Butterfly
0.6
1.8
2.5
Diaphragm
0.0
0.1
0.1
Total
00
25.2
100.0
g
Check and sample system valves excluded. No dry-service slide valves
surveyed.
E-7
-------
mting
6l»nd*x
Stam ^
Packing m/t
Packing gland
Packing--
-fconnat*..
Sailing.^
diaphragm *
check niva
Ptug
Unw nut »v
Disk nut
. Disk
Saat nng
Inrilnad
typt OiU bennvt.
% fttm
Gland
Packing
.Handle
Ditk -
Spring washer
'-•Compfwlon
ring
- S*#t
Dilk nut
Ollk
Body Mt
ring
Pviioort (io
--Stm
Ltvet
devil
~St*n
'"Packing nut
~ Yoke
Packing gland
'Packing
Stuffing box
Packing
Union nut
Olik hold*
Disk nut
Ollk
Figure E2-1. Process Valves
Reprinted by special permission from CHEMICAL ENGINEERING1
E-8
-------
order to permit packing adjustment by tightening nuts on the
bolts or studs. Valve packing may lose its resiliency due to
age or overtightening, and replacement of the packing is then
required.
The fluid may be further prevented from diffusing
through stranded type packing by dispersion of lubricant through
the packing. The lubricant also alleviates galling, heating,
and scoring of the stem or shaft. In most cases, a lubricant
must be compatible with the packing and the working fluid. In
refining, this lubricant might be a silicone oil, a petroleum
grease, or a TFE or graphite dispersion in an oil or grease.
Lubricants are often present in the coils or rings of
packing as received. They may also be introduced into the gland
through a "grease" fitting which passes lubricant into a "lantern
ring" positioned in the stuffing box. A "lantern ring" replaces
a ring of packing. It is grooved to allow free flow of lubricant
completely around the stem or shaft. As packing wear or lubri-
cant loss occurs, the packing gland is tightened and/or additional
lubricant is injected to provide a tight seal face and relatively
impervious packing.
Table E2-2 shows the diversity of valve packing
materials used alone or in combination.1 Most of these materials
may be purchased in coils or in preformed rings. They may be
solid or stranded and may have a round, square, "U," or chevron
cross-section.
Inspection and Maintenance--All refineries have
operating practices that require repair of any leaks detected.
These practices are primarily aimed at preventing fires or other
safety hazards that could result from large amounts of hydro-
carbon leakage. The leak detection methods used are generally
E-9
-------
TABLE E2-2. PACKING MATERIALS - PROCESS VALVES
Packing Material Form Use Temperature
Flexible, all metallic
Spiral wrapping. Thin
ribbons of soft babbit
foil.
Valve stem
packing
Up
to 450°F.
Flexible metallic packing
(aluminum).
Spiral wrapping. Thin
ribbons of soft annealed
aluminum foil loosely
around a small core of
pure dry asbestos.
Hot oil valves,
diphenyl valves.
Up
to 1000°F.
Flexible metallic packing (copper).
Soft annealed copper
foil loosely around a
small core of pure dry
asbestos.
Hot oil valves,
diphenyl valves.
Up
0
O
o
o
o
4J
Long-fiber pure asbestos and fine
lubricating graphite (nonmetallic).
Graphite special long-
fiber asbestos binder.
Extreme
resilience.
Up
to 750°F.
Closely braided asbestos yarn; top
jacket reinforced with Inconel
wire; core: long fiber asbestos.
Spools, die-formed
rings.
High-temperature
valves.
Up
to 1200°F.
Pure asbestos yarn with an Inconel
wire insert around a resilient
asbestos core impregnated with
graphite.
Spool form, die formed.
Valve stem for
air, water, steam
and mineral oil.
Stuffing box
temperature up
to 1200°F.
Twisted long fiber Canadian
asbestos.
Spool form, die formed.
Valves handling,
high and low
pressure steam.
Up
to 500°F.
(Continued)
-------
TABLE E2-2. Continued
Packing Material
Form
Use
Temperature
W
i
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 550°F.
Valve shaft for 100°F to 500°F.
highly corrosive
service.
100°F to 600°F.
Up to 1200°F.
Valve stems in
mild chemical or
solvent service.
Valve stems,
steam, air,
mineral oil.
Stainless-steel Up to 1200°F.
valve stems, air,
steam, water.
Valves for steam, Up to 550°F.
air, gas and mild
chemicals.
Refinery valves. To 750°F.
(Continued)
-------
TABLE E2-2. Continued
Packing Material
Form
Use
Temperature
Braided/overbraided, 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.
Spools and coils.
Valve stems, for Up to 750°F.
valves handling
steam, air, gas,
cresylic acid.
Valve stems.
100°F to 600°F.
Valve stems for 12°F to 500°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.
Source: Reference 1.
-------
the eyes, ears, and nose of employees in the process unit.
Radian's emission sampling program revealed that many leaks from
valves and other sources may not be detected by sight, hearing,
or smell. It is also a common refinery practice to lubricate
valves and tighten packing glands periipdically.2'3 For some
valves, the packing can be replaced while the valve is in service.
These practices are generally applied in order to maintain the
operability of the valves, rather than for fugitive emission con-
trol. Malfunctioning valves are usually replaced or rebuilt
during process unit turnarounds, but malfunctioning is not
usually defined as a valve with excessive fugitive emissions.
Open-Ended Valves-Open-ended valves may be used for
draining, venting, or sampling operations. In addition to fugi-
tive emissions from the stem seal, the valve seat may be a source
of fugitive emissions. In order to prevent emissions through
the seat, the open-end can be sealed with a cap, plug, blind
flange, or a second valve. In most refineries, sampling connec-
tions have two valves in series (double block and bleed). This
provides a second valve seat to resist emissions of the process
fluid to the atmosphere. After operations requiring flow through
the valves, the upstream valve should be closed first in order to
avoid trapping hydrocarbons between the two valves, which would
be essentially the same as one open-ended valve. Some refineries
also seal drain and vent valves with caps, plugs, or blind
flanges when the valves are not in use.
Effectiveness of Existing Controls—The absolute
effectiveness of existing controls is reflected in the emission
factors shown in Section 2.6 of Appendix B of this report.
These emission factors were derived from test data collected
from a broad cross-section of thirteen refineries, and all levels
of the existing types of control were included. The relative
E-13
-------
effectiveness of existing controls - whether or not one type of
packing or maintenance schedule is better than another - cannot
be determined from available data. The costs of any type of
increased emission control, such as using caps for open-ended
valves or more frequent maintenance schedules, would be at best
partially offset by the value of the product losses saved.
2.2.1.2 Available Control Technology for Valves
Leak detection and repair programs are the available
controls for valves. For future construction, process design
evaluation could be used to eliminate any superfluous valves.
Leak detection and repair programs are a regulatory requirement
in some areas. The frequency of application of leak detection
and repair will probably increase due to additional regulatory
requirements and due to the increasing value of the products
lost as fugitive emissions.
Leak detection and leak repair programs consist of
strategies to identify sources that are leaking significant
amounts of hydrocarbons and methods to reduce or eliminate the
leakage. At a specified interval, each valve would be checked
with a portable hydrocarbon detector. The probe of the detector
would be traversed around the potential leak areas (stem, gland,
plug). If a predetermined hydrocarbon concentration limit
(action level) were exceeded, the valve would be repaired. The
repair could consist of tightening the packing, injecting
grease, replacing the packing, or replacing the valve. During
repairs such as tightening or greasing, the hydrocarbon detector
should be used to permit assessment of the effect of the repair
attempt. This type of repair is called "directed" maintenance.
E-14
-------
Effectiveness--In the limited valve repair study
conducted by Radian, the average emission reduction immediately
after "directed" maintenance was 91 percent. The emission
reduction after "undirected" maintenance was 54 percent, where
"normal" maintenance procedures were used without using the
hydrocarbon detector to optimize emission reduction at the time
of maintenance. The detailed test results of the maintenance
study are shown in Section 6.0 of Appendix B of this report.
Data on the long-term effects of maintenance are not available.
In addition to "normal" maintenance procedures such as
tightening or greasing, injection of sealing fluids may be used
to reduce fugitive emissions. The valve body could be drilled
and tapped, and sealing fluid could be injected in order to
seal the leak. The long-term effects on emission reduction and
valve operability are not known. For some control valves,
operating procedures may prohibit excessive in-service adjust-
ment in order to prevent malfunction of vital process control
valves.
The required frequency of leak detection is dependent
on the rate of recurrence of repaired leaks and the rate of
occurrence of new leaks. The selection of an appropriate action
level is dependent on the demonstrated ability to repair leaks
of a given magnitude. Radian test results show that the smaller
the initial leak rate, the more likely it is that repair efforts
will increase rather than decrease the leak rate (Appendix B,
Section 6.0). The other factors that influence the overall
effectiveness of a leak detection and repair program is the
average leak rate after repair. Limited test results show that
for some valves the leak can be completely eliminated but for
some others, the leak may be only slightly reduced or even
increased by the attempted repair. Because these factors
E-15
-------
contributed to the effectiveness of leak detection and repair
programs, no quantitative estimate of emission reduction can
be made.
Costs—The major costs for leak detection and repair
are for labor expenses. The hydrocarbon detector can cost up
to $4,250 per instrument, ** and if leak surveys were conducted
frequently (monthly) each process unit would probably need to
have one instrument. Actual labor costs are dependent on the
wage rate of the persons performing the leak survey and leak
repairs. Estimates have been made for the time needed to con-
duct leak surveys. One petroleum refining company has estimated
that one minute per valve is the average time required for leak
detection.5 The time needed to repair a leak will be dependent
on the type of repair attempted. Simple tightening of packing
by refinery employees would be much cheaper than injection of a
sealing fluid by leak repair contractors. The total cost of a
leak detection and repair program would be reduced by the value
of the product that was prevented from leaving the process as
an emission. The emission reduction would also represent an
energy saving.
2.2.1.3 Control Technology Transfer for Valves
Fugitive emissions of some process fluids may be
hazardous or toxic. In industries with these constraints,
valves with isolated stem seals may be used. The diaphragm and
bellows-sealed valve are shown in Figure E2-2. Because the
process fluid is prevented from contacting the stem/body junction
by a bellows or diaphragm, the potential for fugitive leakage is
reduced. Isolated stem seal valves are not applicable to general
refinery use because of several limitations.
E-16
-------
Stem
Packing
Metallic
Bellows
Dl*>hr*gm
Diaphragm Bonnet Seal Bonnet Bellows Seal
Figure E2-2. Globe Valves With Isolated Stem Seals
Reprinted by special permission from CHEMICAL ENGINEERING6
i
I
i
-------
The diaphragm material in the diaphragm valve limits
operation to about 50 psi pressure differential.7 This type
valve has definite limitations in refinery use. It would fail
catastrophically upon overheating of the elastomer diaphragm, so
it should not be used in hydrocarbon service where a fire could
be fed by its failure. The bellows-sealed valve, because of the
corrosion and fatigue failure potential of the bellows, is sub-
ject to combined temperature-pressure-corrosivity stress. Its
usage is best defined by the valve manufacturer. Bellows-sealed
valves should have stem seal packing as back-up protection against
bellows failure.
Because use of these special valve stem seals will
probably be limited, the impact of their use on emission control
should also be limited, as would any economic impact. No primary
energy cost would result from substitution of a very limited
number of packless valves for conventional packed-stem, bonnet-
sealed valves.
Diaphragm and bellows valves are approximately 1.5 to
3.7 times as expensive as gate valves according to the CARB
report.8 Another source estimated that bellows valves might
cost 10 to 20 times as much as packed-stem valved, but would
have a lower cost multiple if purchased in volume.9
2.2.2 Flanges
Flanges are paired junctions between sections of pipe
and pieces of equipment. They are sealed against leakage by the
tightening of bolts or studs which compress a flat gasket between
the flat faces of the mating flanges, or compress an "o" ring
set in the grooved faces of special flanges. The most common
flanges have raised faces to accommodate tightening of the bolt
E-18
-------
and centering of the gasket, as shown in Figure E2-3. Typical
gasket materials are asbestos composition or spiral, metal strip-
reinforced asbestos or TFE. "0" rings may be made of neoprene,
TFE or soft metals, depending upon temperature and pressure
limits.
Figure E2-3. Raised-face Flange (gasket not shown).
The results of the refinery sampling program showed
that flanges have a very low emission factor, and even though
there are many of them, their overall contribution is small.
The only real controls available for flanges are leak detection
and repair programs. If a leak is found, the only repair options
are tightening the flange bolts or injection of a sealing fluid,
since most flanges cannot be isolated from the process in order
to permit gasket replacement. Because of their lower emission
contribution, flanges would not warrant as frequent leak testing
as other sources such as valves.
Pumps can leak hydrocarbons at the junction of the
shaft and body of the pump. Two basic types of seals are used
to prevent massive leakage from this junction. Packed seals
2.2.3 Pumps
E-19
-------
can be used on pumps with reciprocating or rotating shaft motion,
and mechanical seals are applicable only to rotating shafts.
2.2.3.1 Existing Controls for Pumps
The two types of existing controls for pumps are the
pump seal itself, and inspection and maintenance of the pump
seal. The packed seal and mechanical seal resist leakage of the
pumped fluid by different mechanisms, and are described
separately.
Packed Seals--The packed seal, Figure E2-4, is used to
seal both rotary and reciprocating shafts against leakage of
liquid from the "working fluid" end of the shafts to the atmos-
phere. Compressed packing in the stuffing box forms a contact
seal against the moving drive shaft. 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.
As Figure E2-4 shows, a packed seal system may require
injection of heat removal fluid, which is distributed in the
stuffing box by a slotted "lantern ring." The fluid may be
injected to not only remove heat but also to (1) seal the shaft
for vacuum service, (2) flush the stuffing box lip of slurry
solids, (3) provide a high flush rate to prevent shaft damage
from abrasives, or (4) prevent hazardous materials from being
released to the atmosphere.
Packings for the compression-type seals shown in
Figure E2-4 may be solid or braided, twisted or ribbon-form
(the latter form in graphite only). They may be obtained in
continuous rolls or preformed rings. The shape is not restricted
E-20
-------
Fluid inlet connected
to pump discharge
Fluid inlet connected
to external source
fluid inlet connected
•' to external source
Atmospheric
Atmospheric
Atmospheric
pressure
pressure
Leakage
Leakage
Leakage
a. Nayativa suction Mrvtoe
To tt*ur« fluid in ftufting box
b. Starry •nrtoa
Own UQuid to lantern ring
c. Abrmbm Mrrica
Ckin liquid flush to ring
Figure E2-A. Packed Seals
Reprinted by special permission from CHEMICAL ENGINEERING13
E-21
-------
to the square cross-sections used for illustration. The cross-
section may be round, plaited or braided square or rectangular,
ot it may be "u," "v," or chevron-shaped. The last three shapes
typically maintain their seal against a shaft after minimal
initial tightening of the gland, and require little adjustment
thereafter. For this reason, these packings are referred to as
"automatic" packings.
Service limits for selected packings are found in
Table E2-3. Note the inclusion of asbestos in the table.
Asbestos is an OSHA-controlled substance, but is deemed safe
when (1) modified by the lubricants and/or bonding agents
impregnating the packings, and (2) cut for fitting purposes
rather than sawed or abraded. Under moderate conditions the
trend in braided packings is away from asbestos and toward TFE
because of the latter's low coefficient of friction and its
chemical inertness.
Lubricants for packings include the following
substances:
• Mica and talc - high friction value lubricants,
but do not discolor product.
• Graphite - the most common lubricant. May
contribute to electrolytic or galvanic
corrosion, especially in high-pressure steam
service. May color product.
• Molybdenum disulfide (M0S2) - dry lubricant
similar to graphite. Does not cause electro-
lytic corrosion. Prevents metal galling.
Oxidizes above 650°F and is then no longer a
lubricant.
E-22
-------
TABLE E2-3. SERVICE LIMITS FOR SELECTED
MECHANICAL PACKINGS3
Packing
Maximum ^
Temperature
(°F)
Pressure at
Maximum
Temperature
(psig)
Maximum
Pressure
(psig)
Temperature
at Maximum
Pressure*3
(°F)
Asbestos/TFE
500
50
200
100
TFE (lubed)
500
50
200
100
Asbestos/Graphite
400
50
250
100
Graphite-Fiber
1000 (600)C
50
350
300
Graphite-Ribbon
1000 (600)C
50
350
300
Lead
350
50
400d
100
Aluminum
800 (500)d
50
400d
200
Basic data: 2-in shaft, 3550 rpm. Controlled leakage for 720 h. 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 ill.
Source: Reference 10.
^Temperature is product temperature; pressure is stuffing-box pressure.
cLarger number is nonoxidizing environment; smaller number is oxidizing
environment.
^Assumes rings are die-formed.
E-23
-------
• Hydrocarbon type lubricants (greases, tallow,
petroleum oils) - have low chemical resistance;
carbonize at elevated temperatures; may be sol-
uble in product being pumped.
• Tungsten disulfide - not as good a lubricant
as M0S2 or graphite, but is stable to 2400*F.
• TFE - a significant factor in lubricants as
well as in woven packings. Temperature limit
between 400°F and 500°F.10 TFE is inert to all
common chemicals except molten alkali metals and
some rare, halogenated compounds.
• Silicone oils - temperature-resistant, corrosion-
resistant oils are often injected at the lantern
ring in stuffing box. Some silicones will
oxidize and hence lose lubricating properties.
Mechanical Seals—The mechanical seal in its many forms
is the predominant pump seal today. Contrary to the broader
application of packed seals to both rotating and reciprocating
shafts, however, mechanical seals are used only on rotary shafts.
Mechanical seals may be used to seal both pump and compressor
shafts, but are more universally applied to pumps, specifically
centrifugal pumps. The American Petroleum Institute (API)
recommends mechanical seals as particularly advantageous for
hydrocarbon emission control in the following cases:11
• "... more-or-less continuous pumping of products
having a Reid Vapor Pressure of 5 pounds per
square inch..."
E-24
-------
• "... when fluids are under substantial pressure
and when the pump or compressor is in continuous
service. For pumps operating on stand-by service
either packed or mechanical seals may be used."
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.12 In Radian's survey approximately
82 percent of the screened refinery pump seals were mechanical.
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 the rings is
attached to the pumpshaft, and the other to the gland plate or
its equivalent. Seal faces are perpendicular to the shaft and
are typically lapped to a precise flatness that accounts for
their typically low leak rate when carefully installed and
started up.
Single Mechanical Seals - A single mechanical seal
may be an outside seal or an inside seal depending
on whether the primary ring is outside or inside
the stuffing box housing. Figure E2-5(a) depicts
an outside seal; Figure E2-5(b) represents an
inside seal. In the majority of cases, an internal
seal is used; the primary ring is attached to the
shaft, and the mating ring to the gland plate. The
inside seal cannot be as easily "blown" as the out- .
side seal.
Secondary seals, often "o" ring-type but sometimes
bellows, prevent leakage between housing and gland
plate, primary ring and shaft, or mating ring and
E-25
-------
Stuffing-box
Gland
Gland
Stuffing-box
Pumped
liquid s
Primary ring-"
•Primary
ring
Pumped liquid
Figure E2-5. Single Mechanical Seals
Reprinted by special permission from CHEMICAL ENGINEERING13
E-26
-------
gland plate/stuffing box housing. The secondary
seal must be sufficiently flexible so that the
primary ring may move relative to the "sacrificial"
mating ring when wear occurs or when one of the
faces is not quite true. Typically, the primary
ring is spring-loaded to maintain interfacial
contact. Secondary seals are made of elastomers,
plastics, glass-filled TFE or pure TFE. If the
latter is used, a "v" or "u" shaped ring is
necessary because of characteristic TFE cold-flow.13
The properties of the secondary seal material are
usually the limiting consideration in the applica-
tion of mechanical seals. Figure E2-6 illustrates
unbalanced and balanced inside seals. In the
unbalanced seal, (a), the closing forces are the
sum of the spring force (Fg) (spring not shown)
and the force (Fgg) of the pumped liquid acting
against the area of the left-hand side of the
primary ring. Force Fgg is the product of the
stuffing box pressure and the area of the primary
ring perpendicular to the drive shaft, A^. The
closing force, F^,, is opposed by the dynamic
force, Fq, at the seal interface. Although leakage
at the interface is small if the mechanical seal is
installed and maintained correctly, an average
pressure tending to force the lapped ring faces
apart does exist, and its value lies between
stuffing box pressure, Pgg, and ambient pressure,
*AMB' *"or ca^cu^-ati°ns» t^ie resulting average
force is often taken to be linear and is therefore
the product of the seal contact area,, Agp, and the
average pressure at the face:
E-27
-------
Gland
Gland
M l TVri All
Pumpad liquid
Mating
ring
Pumped liquid
Mating
ring
Stuffing-box:
praaaura ;
Stuffing-box
praaaura —
Primary ring'
Primary ring
Figure E2-6. Balanced and Unbalanced Mechanical Seals
Reprinted by special permission from CHEMICAL ENGINEERING13
E-28
-------
F = A x
0 SF
(PSB + PAMB)
Thus the net closing force, is
fnc = Fc - Fo
= (Fs + fsb} - Fo
F + (P x A )
S SB DR
(PSB + PAMB)
x A
SF
Because of the typically high wear rate
associated with the unbalanced seal operating
at high pressures, stuffing box pressures for
this seal are limited to 25-100 psig, depending
upon seal diameter and shaft speed.13 By stepping
or sleeving the drive shaft as shown in Figure
E2-6(b), the hydraulic force Fgg acting on the
exposed face of the primary ring can be reduced
from that of an unbalanced seal. Differences in
frictional forces favor the balanced seal by
about 20 percent.11*
Figure E2-7 is a more complete representation
of a balanced internal seal. The spring-loading
arrangement for the primary ring is shown. The
seal coolant flow path shown allows liquid to
pass over the mating rings. The liquid flushes
any foreign matter toward the throat of the
stuffing box and cools the rings near the contact
surface. This type of flushing is not possible
with the outside seal shown in Figure E2-5.
E-29
-------
S£3l Coolant Secondary Gland
Collar with
set screwv
Thro8t of y
fluffing box |
'S
Pumped \
liquid t
Mating
ring
Primary ring
Figure E2-7. Balanced Internal Seal Showing
Seal Coolant Inlet.
Reprinted by special permission from CHEMICAL ENGINEERING13
E-30
-------
Thus liquid for a single inside seal is often
the pump discharge liquid delivered via an
auxiliary cooler and/or filter (if needed).
If the pumped liquid flashes while passing
across the seal faces or is otherwise a poor
lubricant, a dissimilar but compatible liquid
may be injected as coolant. Seal coolant is
sometimes throttled at the throat of the stuffing
box by positioning a lip seal or bushing there.
Throttling reduces the loss of pump efficiency
in the case of the recycle flush configuration;
in all cases, a throttling device reduces the
chance of debris entering the stuffing box from
the pump casing.
A variation of the single mechanical seal involves
placing a throttle bushing or auxiliary packing
outboard of the single mechanical seal (Figure
E2-8). The bushing or packing acts as a con-
striction to limit fluid escape in case of
mechanical seal failure. The gland plate is
drilled to provide vent and/or drain points for
vapors and/or liquids which are flammable or
toxic. Those fluids, should they escape, are
ducted to appropriate disposal systems. If
auxiliary packing [Figure E2-8(c) ] is used as
the back-up system, a water flush is introduced
into the outboard flush connection. The water
flush cools the packing as the water leaks out.
The single mechanical seal/outboard restriction
combination is less expensive than the double
mechanical seals to be described next.
E-31
-------
Flush connection
I. Fixed throttle
bushing
Flush connection —
c. Auxiliary
ftuffing box
Vent (top)
* and drain
(bottom)
connections
FUnh connection v
,Vent connection
> V-Fixed throttle
bushing
Flush (top) end
drain (bottom)
connections
b. Floating throttle
bushing
r-— Auxiliary
packing
gland
'/-Auxiliary packing
-Expansion collar
-Floating bushing
13rain connection
Figure E2-8. Back-up Systems for Seal Failure.
Reprinted by special permission from CHEMICAL ENGINEERING13
E-32
-------
Double Mechanical Seals - Double mechanical
seals provide a margin of protection against seal
failure not offered by single mechanical seals.
Figure E2-9 illustrates one of two types of
double mechanical seals, the'"back-to-back"
or "mirror image" arrangement.
Seal liquid pressure must be higher than
stuffing box pressure in this configuration to
prevent the inner primary ring from being blown
back from the mating ring. Seal liquid normally
lubricates both the inner and outer seal inter-
faces, passing into the stuffing box and also
out past the outer seal faces and to the atmos-
phere from the seal housing. The seal liquid
should, therefore, ideally be noncontaminating
so that it will contaminate neither the pumped
liquid nor the environment.
If the inner seal should fail, the outer seal
would prevent escaping fluid from reaching the
atmosphere. In case of accidental pressure loss
in the seal liquid system, however, the pumped
liquid would 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.
A second type of double mechanical seal is the
tandem mechanical seal (Figure E2-10). In this
form, the seals face the same direction, with the
inner seal located inside the stuffing box housing
rather than in the seal housing as in the "back-to-
back" form.
E-33
-------
Seal Housing
SmI liquid,
in {bottom)
Gland plate
Stuffing-box
housing*"--
Pumped liquid
Outer mating ring
Inner mating ring
Drive Shaft
Outer primary
ring
Inner primary-
ring
Figure E2-9. Double Mechanical Seal: Seal Liquid at
Higher Pressure Than Pumped Liquid.
Reprinted by special permission from CHEMICAL ENGINEERING13
E-34
-------
Stuffing-box Bypass Buffer liquid,
housing
out in
(top) (bottom)
Gland
pltte
/
Pumped
liquid i
Outer Outer
primary mating
ring ring
Inner Inner
primary mating
ring ring
Figure E2-10. Tandem Mechanical Seal: Buffer Liquid is
at Lower Pressure than Pimped Liquid
Reprinted by special permission from CHEMICAL ENGINEERING1
E-35
-------
By contrast with the "back-to-back" double seal,
which requires a seal liquid pressure higher
than pumped liquid pressure, the tandem seal
buffer liquid is at a lower pressure than the
stuffing box pressure. This' configuration pro-
tects the pumped liquid from buffer fluid contami-
nation. This operational mode is possible because
stuffing box pressure tends to close the seal faces,
and loss of buffer pressure only increases the face
seal pressure. In normal operation the buffer pres-
sure is set so that it balances the opening/closing
forces on the seal faces, minimizing both inter-
facial leakage and friction losses.
Should the inner seal of a tandem seal fail,
pumped liquid would contaminate the barrier fluid,
but would not be lost to the atmosphere except
as it was lost slowly across the "back-up" outer
seal face or by degassing through the reservoir
vent. If a closed system is provided with pres-
sure indicating/high pressure alarm devices as
shown in Figure E2-11, very little time should be
lost in taking action if inner seal failure
should occur.15
Selected seal face materials are capable of
service temperatures up to 750°F if coolant or
flush liquid below 200°F is provided. Face combi-
nations usually include a carbon/graphite mating
ring because of the self-lubricating quality and
softness of the graphite. Other ring materials
include stellite, tungsten carbide, stainless
steel, ceramics and Ni-Resist.13
E-36
-------
VENT TO SAFE AREA
Flow
FLUSH LINE FROM
PUMP DISCHARGE OR
EXTERNAL SOURCE
PRESSURE GAUGE
PRESSURE ALARM
FILLER PLUG ^Restrictions:
1 ORIFICE
NORMALLY
OPEN
RESERVOIR
BUFFER
FLUID
Figure E2-11. Tandem Seal with "Tell-Tale"
Pressure Gauge/Alarm
Reprinted by special permission from HYDROCARBON PROCESSING1
E-37
-------
Guidelines and specifications for the application
of packed and mechanical seals to centrifugal
pumps in refinery service are found in API
Standard 610.16 Generally, recommendations cover:
• Minimum packing dimensions and number
of packing rings,
• Location of coolant connections,
• Mechanical strength features of
stuffing box,
• Requirements for water-smothering glands
for pumps which handle high vapor pres-
sure liquids,
• Seal back-up systems (auxiliary
throttle bushings or packing),
• Piping for seals in clean service and
for seals in dirty or special service.
Frequency of Application, Effectiveness, and Cost
of Pump Seals—Application of the types of pump seals is rela-
tively uniform within the refining industry. This may be the
result of a greater uniformity of feedstocks and products in
the refining industry than in the chemical industry. The appl
cation of standards published by the American Petroleum
Institute (API) has also undoubtedly led to uniformity among
devices used to control fugitive emissions, not only from pump
but also from some of the other devices tested in this program
E-38
-------
The frequency of application of types of pump seals
that was observed in the Radian sampling program is shown in
Table E2-4. Emission factors for these seal categories were not
statistically different (see Appendix B). This may be due to the
use of mechanical seals in high volatility service. Sufficient
data are not available for comparison of effectiveness of seal
types for identical stream service.
TABLE E2-4. DISTRIBUTION OF PUMP SEALS IN
RADIAN REFINERY STUDY
Pump Type
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
Table E2-5 is a cost breakdown of pump system elements
for systems rated at 3-200 horsepower. May 1980 costs were
rolled back to mid-1979. Cost estimates of packed and mechani-
cal seals are shown in part (5) of the table in mid-1979 dollars,
and in part (6) as percentage add-on costs to bare, uninstalled
pump costs [Subtotal (4) ] . These add-on costs for seals range
from 1.2 to 3.0 percent for packed seals to 14.2 to 36.4 percent
for double mechanical seals for the most common shaft size of
1.875 inches diameter.
Table E2-6 is a comparison of seal friction losses and
hydrocarbon leak estimates for packed seals and three basic types
of mechanical seals. Friction losses and hydrocarbon losses are
known to vary widely with the fluid properties of the sealed
E-39
-------
TABLE E2-5. CENTRIFUGAL PUMP SEALS - COST
CONTRIBUTION TO TOTAL PUMP COST*
Pump Horsepower 3.0 100. 100. 200.
Shaft Diameter, Inches 1.875 1.875 2.375 2.375
1.
Pump, including shaft, coupling,
bore plate, seal/bush hardware as
required. (Installation costs
not included)3
2830
4670
4810
6370
2.
Switchgear - Switch, enclosure
lighted push botton.b
620
1940
1940
4110
3.
Driver - Electricc
230
2850
2850
8750
4.
Subtotal
3680
9460
9600
19230
5.
Seal Alternatives
a. Packed Seal**
110
110
130
130
0
b. Single Mechanical Seal
860
860
1000
1000
c. Double Mechanical Seal^
1340
1340
—
—
6.
Seal Costs - Percentage of Subtotal
(4)
a. Packed Seal
3.0
1.2
1.4
0.6!
b. Single Mechanical Seal
23.4
9.1
10.4
5.2
c. Double Mechanical Seal
36.4
14.2
—
—
*Mid-1979 Costs = May, 1980 Dollars x 0.
921
Bases:
aReference 17. Pump built to API Specification 610,16 and upon the following
conditions: 1) Low corrosion—steel pump casing, cast iron/steel impeller
2) Seal gland pressure—200 psig (-1/3 of discharge pressure
maximum)
3) Pumped Fluid—light gasoline
A) Pumped Fluid Temperature— 350°F
5) Shaft Speed—3500 RPM
^Reference 18. Switch gear—explosion-proof, locally-mounted push button
stop-start with red light for "ON" indication,
c *
Reference 17. Electric Driver—Three phase, 400 volt, explosion proof.
^Reference 19. Packed Seal—Cost of packing materials approximate.
CReferen;e 20. Single Mechanical Seal—Crane Packing Co. 48-B-l vith throttle
bushing as back-up.
^Reference 19. Double Mechanical Seal—Chesterton Seal Mo. 241.
E-40
-------
TABLE E2-6. ESTIMATED ENERGY LOSSES - PUMP SEALS
Hydrocarbon Leak Estimates, lb/hr
Seal Power
Open
b
Seal Type
Consumption, kW
Literature
This Study
Packed
1.16
0.264C
Single mechanical,
(
unbalanced
0.422
>0.0044e
I
J 0.16-0.37,d
Single mechanical,
I all pumps
balanced
0.194
>0.0044
(
Double mechanical,
balanced
0.287
=0.00
Reference 14.
^See Appendix B, p. 2-263, pumps, light liquids.
cBased upon 60 drops/min of hexane 0 20 drops/m£. Reference 10.
^Range based upon 95% confidence interval.
eBased upon as little as 1 drop/min. of hexane @ 20 drops/m£. Reference 13.
^Reference 13.
Bases: Pump shaft dia.—1.875 in.; stuffing box pressure—200 psig; barrier
fluid pressure—175 psig (double mechanical seal only); pump speed—
3500 rpm; pump horsepower range (typical)—3-100 h.p.
E-41
-------
liquid, the seal face materials, the condition of the seal,
bearings and shaft, and seal design, so these figures are pre-
sented only as approximations of expected performance.
Inspection and Maintenance—All refineries practice
inspection and maintenance of pump seals in order to prevent
fire hazards resulting from complete seal failure. Pump seals
are usually inspected visually once per day or per shift.
Packed seals can be adjusted while in service to reduce leakage,
but mechanical seals usually require removal for repair. The
effectiveness of these inspection and maintenance programs is
reflected in the emission factors presented in Appendix B.
2.2.3.2 Available Control Technology for Pumps
Leak detection and repair strategies are the available
controls for pumps. The procedures for finding leaks requiring
repair are the same as described for valves in Section 2.2.1.2.
The hydrocarbon detector probe would be traversed around all
potential leak areas, including barrier fluid degassing vents
and the shaft seal junction. The probe should be prevented from
contacting leaking liquids or the rotating shaft.
No data are available to quantify the effectiveness or
cost of leak detection and repair for pumps. Effectiveness
would be dependent on initial leak rates, the ability to repair
the leaks, and the length of time before the leaks re-occurred.
Costs would be dependent on labor rates, labor requirements, and
the value of the product saved. Average leak detection time
required for pumps has been estimated to be five minutes per
seal, and the average leak repair time has been estimated to be
80 hours per seal.8
E-42
-------
2.2.3.3 Technology Transfer for Pumps
Sealless pumps are used in other industries in cases
where the pumped fluid is toxic or otherwise hazardous and leak-
age cannot be tolerated. Sealless pumps include diaphragm pumps,
hermetically sealed "canned" pumps, and magnetically coupled
pumps. Since these pumps do not have a shaft/casing seal, their
emission potential is much lower. Emissions may result from
diaphragm failure or case failure.
Sealless pumps are not covered by API Standard 61016
from pumps, which may explain why no sealless pumps were found
in the 13 refinery survey. If sealless pumps are to be used in
the refining industry, they must be proven performers in terms
of leak-tightness, reliability, maintainability, useful life
and safety.
The original cost of a "canned" pump may be approxi-
mately 110 to 115 percent of the cost of a centrifugal pump with
conventional seals.21 No data are available to discern differ-
ences among the other true costs of running conventionally-sealed
versus sealless pumps. Sealless pumps also have a more limited
range of applicability due to limitations on temperature, through-
put, and horsepower.
2.2.4 Compressors
Compressors can leak hydrocarbons at the junction of
the shaft and body of the compressor or at the degassing reser-
voir of a seal oil system. Packed seals are the only type avail-
able for reciprocating shafts, and they are seldom used for
rotating shafts. Rotating shafts can also be equipped with
mechanical seals, labyrinth seals, restrictive ring seals, and
E-43
-------
liquid film seals. Thirty-seven percent of the compressors
screened were in hydrogen service and 63 percent were in hydro-
carbon service (Table B3-32, Appendix B).
2.2.4.1 Existing Controls for Compressors
The compressor seal and inspection and maintenance of
the seal are the two types of existing controls for compressors.
Types of Compressor Seals--The American Petroleum
Institute recognizes five basic types of seals which are applied
to refinery compressors.22'23 The frequency of application
observed in the Radian study is shown in Table E2-7.
TABLE E2-7. TYPES OF COMPRESSOR SEALS
Approximate
Shaft Type - Seal Type Frequency of Observation, %
Reciprocating - Packed 67
Rotating - Labyrinth 5
Rotating - Restrictive Ring 0
Rotating - Liquid Film or Bushing 0
Rotating - Mechanical Contact (dry/wet) 10
Unidentified3 18
TOTAL 100
0
Probably includes Restrictive Ring and Liquid Film seals
Source: Table B2-32, Appendix B.
Cooling of friction-type compressor seals differs
from cooling of pump seals of similar construction in that the
gaseous compressor working fluid provides negligible lubrication
E-44
-------
and has a lower heat capacity than does liquid. Therefore most,
but not all, contact-type compressor seals use some form of
liquid seal coolant which may also serve to reduce gas emissions.
Packed seals may be lubricated or non-lubricated. API
Standard 618 suggests the increasingly higher cooling standards
given in Table E2-8 for seal glands as discharge pressures
increase.
TABLE E2-8. API RECOMMENDED COOLING STANDARDS
Pressure Limits, psig
Lubricated Non-lubricated
Degree of Cooling Gland Gland*
Manufacturers Standard
Indirect Cooling - Gland Jacket
Direct Cooling through Packing
Cup + Gland Jacket Cooling
< 1500 < 250
1500-2500
> 2500 > 250
*Fluorocarbon packing assumed.
Source: Reference 23.
Also, according to API Standard 618, gases escaping the
packing should be vented and piped from the "distance piece"
separating the driver and compressor cylinder through an opening
below the reciprocating rod.23
The labyrinth shaft seal is shown in its simplest form
in Figure E2-12. A given gas at internal pressure escapes slowly
to the atmosphere (or to an adjacent stage) at a rate which
depends on shaft diameter, pressure differential, the closeness
of fit of the labyrinth to the rotating shaft, and the number of
E-45
-------
"teeth" in the labyrinth. 2" In some designs, the edges score a
sacrificial sleeve on the shaft to establish a tight fit and
reduce leakage. As seen in Figure E2-12 a port may be provided
which can be used for withdrawal of escaping gas (scavenging) or
pressure purging inward and outward from the port. Nelson21'
states that the loss rate or recycle rate from this type opera- .
tion is not generally acceptable today for energy and environ-
mental reasons. For this reason, labyrinths are now more often
seen in outboard seals in combination with other sealing devices,
as seen in the illustrations to follow.
PORT MAY BE ADDED
FOR SCAVENGING OP
inert-gas sealing-
ATMOSPHERE
INTERNAL
GAS PRESSURE
Figure E2-12. Straight Pass Labyrinth
Compressor Shaft Seal.
Reprinted by courtesy of the American Petroleum Institute
Source: Reference 22, p.9.
E-46
-------
A restrictive ring seal is illustrated in Figure E2-13.
Fixed rings, often of carbon, are closely fitted without actually
contacting the rotating shaft or shaft sleeve. The seal may be
operated dry, with sealing liquid, or with a buffer gas. This
type of seal is superior to the labyrinth seal alone, but is
limited to about 200 psi and relatively clean gas service.22'2"
Sealing and scavenging ports may be used as for labyrinth seals.
SCAVENGING
/J PORT MAY BE
DADOED FOR VAC-
UUM APPLICATION
PORT MAY BE
added for
SEALING
INTERNAL
GAS
PRESSURE
Figure E2-13. Restrictive-ring Compressor
Shaft Seal.
Reprinted by courtesy of the American Petroleum Institute
Source: Reference 22, p.9.
E-47
-------
Figure E2-14 depicts a liquid film, or bushing, seal.
In this design, bushings are fitted within four to six thousandths
of an inch of the shaft sleeve; they float with lateral shaft
movement. Float is allowed by secondary elastomer seals on the
outside of the bushings. Fresh sealant liquid entering the
space between the inner and outer bushings flows both inward and
outward. Inward flow is at a lower rate than outward flow
because of the smaller available P driving force. If the gas
being compressed is sour, the contaminated sealant oil must be
treated. External piping and controls for surge, supply, recycle
and makeup are diagrammed and described in API Standard 614.25
internal
GAS PRESSURE
CUEAN ttl IN
OUTER tUSHING
INNER VUSHIN6
SHAFT SLEEVE
atmosphere
CONTAMINATED
0»L OUT
OIL OUT
Figure E2-14. Liquid-film Compressor Shaft Seal
with Cylindrical Bushing.
Reprinted by courtesy of American Petroleum Institute
Source: Reference 22, p. 10.
E-48
-------
The liquid-film seal itself is relatively simple, and
is not subject to significant wear. It is capable of operating
at pressures of up to 5000 psi in a multiple seal configuration,
but has, in all configurations, a relatively complicated piping
and control system.21*
The mechanical contact seal shown in Figure E2-15
differs significantly from a mechanical pump seal, but utilizes
the identical concept of zero clearance at closely-lapped wear
surfaces to limit leakage. The compressor seal has a floating
carbon ring which rotates at about half compressor speed between
a stationary and a rotating seat. The mechanical seal may be
run "dry" (cooled and gas film lubricated by a seal gas), but
interfacial pressures must be carefully regulated for "dry"
operation.
INTERNAL
CAS PRESSURE
CLEAN OIL IN
- PRESSURE
BREAKDOWN
SLEEVE
iJJsi
STATIONARY seat' \
CARBON RING
ROTATING SEAT
RUNNING FACE,
CONTAMINATED
OIL OUT
Figure E2-15. Mechanical Contact Compressor Shaft Seal.
Reprinted by courtesy of American Petroleum Institute
Source: Reference 22, p.9
E-49
-------
If the mechanical contact seal is liquid-lubricated,
clean lubricant is introduced into the seal casing near the two
friction surfaces of the carbon ring to aid heat-removal. The
bulk of the oil is removed uncontaminated for recycle; a small
portion passes the running faces and between the stationary seat
and shaft, where it becomes contaminated and is withdrawn.
The mechanical contact seal is limited to pressures of
< 500 psi. Like mechanical seals for pumps, mechanical contact
seals are subject to catastrophic failure. Their oil supply
systems, where used, are simpler than oil supply systems for
liquid-film seals.2u Mechanical contact seals form a nearly
perfect seal when at rest, which contrasts with pump mechanical
seals that are believed to seal better when the faces are
rotating.211
Effectiveness of Compressor Seals—Table E2-9 shows a
comparison of seal leakage. The worst, the straight pass laby-
rinth, is given a gas leakage index of 100. It is not clear
from the table, which includes both dry and lubricated seals,
where the oil film seal fits in according to the gas leakage
index. The liquid film seal is shown, however, to lose more
lubricant than the lubricated mechanical contact seal by a
factor of 55. It is not clear if this refers to oil loss into
the compressed gas stream or if it refers to loss of oil (and
dissolved gas) to the atmosphere.
The packed seal is the only seal available for a
reciprocating compressor application. The mechanical contact
seal, wet or dry depending upon design needs, would appear to
rank the best among centrifugal compressor seals for pressures
up to about 500 psi. However, these seals are said to be
fragile and prone to failure, as well as complex and difficult
E-50
-------
to install correctly. It may be wise to limit applications
to spare equipment or to investigate the history of a particular
seal manufacturer's product before choosing this device.
TABLE E2-9. COMPRESSOR SEAL LEAKAGE
Compressor Seal
Dry Types
Straight Pass Labyrinth
Staggered Labyrinth
Honeycomb Labyrinth
Restrictive Ring
Mechanical Contact (Running Dry)
Oil Types
Mechanical Contact
6% in. Face Diameter
30 psi Differential
500 rpm
Liquid Film or Bushing
5V2 in. Bore Diameter
0.007 in. Clearance
5000 rpm
60°F Oil Rise
Source: Reference 24.
A more flexible device in terms of broad pressure range
application (to 5000 psi) and suitability for dirty gas service
is the liquid film seal. The complexity of its external circula-
tion and control system would be perhaps its most costly feature.
Gas Leakage Index
100
56
AO
20
2
Oil Loss
0.03 gal/hr
Lubricant Loss
1.75 gal/hr or
55 times the
contact type
E-51
-------
Acid gas stripping from circulating seal oil is a must with the
use of liquid film seals if the working gas is sour. The oil
reservoir degassing vent may be a source of hydrocarbon emissions.
Seal Energy Requirements and Cost—Compressor seal
design is traditionally an integral part of overall compressor
design. As a result, data are not available to allow independent
seal energy usage and cost analysis.
Inspection and Maintenance—Inspection and maintenance
procedures for compressors are similar to those described for
pumps. Leakage may be more difficult to detect because some
compressors have enclosed seal areas that transport leakage to
an elevated vent pipe. The effectiveness of these procedures is
reflected in the emission factors for compressors shown in
Appendix B.
2.2.4.2 Available Controls for Compressors
Closed vent systems and leak detection and repair
programs are the available controls for compressors. A closed
vent system consists of piping and, if necessary, flow inducing
devices that transport compressor seal leakage to a control
device. Control devices could include fired heaters or boilers,
incinerators, flares, or vapor recovery systems. For compressors
with seal oil systems, the closed vent system would be connected
to the seal oil reservoir degassing unit. For other compressor
seals, the seal area itself would be enclosed and connected to
the closed vent system.
Leak detection and repair for compressors is similar
to the program described for pumps. The hydrocarbon detector
probe would be traversed around all potential leak areas. These
E-52
-------
areas would include the seal itself (if accessible), the seal
vent pipe, and the seal oil reservoir vent depending on the
physical configuration of the compressor. No data are available
to quantify effectiveness of the leak detection and repair for
compressors. Effectiveness and cost would be dependent on the
same factors discussed for pumps. Average leak detection time
required for compressors has been estimated as 10 minutes per
seal, and repair time has been estimated as 40 hours per seal.8
One major difference between repair of pump and compressor seals
is that most refinery pumps have spares and most compressors do
not. Therefore any repair that required compressor shutdown
might also require shutdown of the process unit. Depending on
the type of process unit, the unit shutdown could cause more
emissions than allowing the compressor seal to leak until the
next turnaround for repair.
2.2.4.3 Technology Transfer for Compressors
No other controls were identified for compressor seal
leakage. Sealless compressors are not available in the range
of throughput that would be required in almost any refinery
application.
2.2.5 Agitators
Agitators may leak hydrocarbons at the junction of the
vessel and the rotating agitator shaft. The agitator seal may
be in liquid service if the agitator is located at the side of
a storage tank, or the seal may be in vapor service if the agita-
tor is located at the top of alkylation reactors. In some types
of refinery operations, in-line blending has replaced the use of
agitated mixing vessels.
E-53
-------
2.2.5.1 Existing Controls for Agitators
The four basic types of agitator seals are shown in
Figure E2-16. Some of the seals are similar to pump seals
(packed and mechanical). The limitations of the four seal types
are shown in Table E2-10. No data are available to establish
the magnitude of leakage from agitator seals. The seals are
listed in Table E2-10 in order of increasing cost.
2.2.5.2 Available Controls for Agitators
Leak detection and repair strategies for agitators
should be similar to those described for pumps and compressors.
The time required to detect leaks is probably about the same
as for pumps and compressors. The time requirements for repair
are not quantified.
2.2.6 Pressure Relief Devices
Pressure relief devices are required in refining opera-
tions in order to protect process equipment from dangerous
over-pressure conditions. The following terminology is used for
valve-type pressure-relieving devices used in industry:27
• Relief Valve - primary liquid service,
• Safety Valve - for steam, gas and vapor
services,
• Safety relief valve - for liquid or vapor
services.
E-54
-------
Inverted cup..,.
Annular cup
a. Hydraulic aeel
Outboard
seal
elastomer
| b. Lip f*al
Inboard
seal
Springs
Retainer :
Gland plate—
Split gland
Lantern ring
Grease fitting
Secondary
Packing,.
Primary
•eat ring
Mating ring—
Static teal
Alloy face and core-
**-Agitator aha ft
c. Packing gland
Figure E2-16. Seals for Agitator Shafts.
Reprinted by special permission from CHEMICAL ENGINEERING2®
E-55
-------
TABLE E2-10. BASIC AGITATOR SEALS
Seal Type
Limitations
Comments
a. Hydraulic
b. Lip
c. Packing Gland
d. Mechanical Face
Low pressure and
temperature
2-3 psi;
unlubricated
150 psi
0 psia to 5,000 psia
if housed and
pressured to working
fluid pressure
Least-used agitator seal.
Dust or vapor seal only;
temperature limited by
elastomer lip melting point.
Six packing rings and lantern
ring required for 150 psi
capability.
Externally lubricated so as to
leak in if inboard seal fails
(double seal configuration).
Single seals also used.
Source: Reference 26.
E-56
-------
In this appendix, the term "safety relief valve" (abbreviated SRV)
will be used to designate all valves discussed. Fugitive emis-
sions from SRV's may be the result of deterioration of the valve
seat or incomplete closure of the valve seat after over-pressure
release.
2.2.6.1 Existing Controls for Pressure Relief Devices
service are shown in Figure E2-17, and are described in Table
E2-11. The type A SRV is probably the most common configuration
found in refining. The Radian survey did not differentiate
among SRV types.
The three types of SRV's for use for hydrocarbon
TABLE E2-11. DESCRIPTION OF SRV TYPES
Type
Description
Application
A. Standard spring-loaded angle-valve
design, with top of relief disk at
discharge-side pressure.
Discharge to
atmosphere.
B. Same as A above, except bellows
isolate the bonnet so that top of
disk is at atmospheric pressure when
bonnet is vented to atmosphere.
Discharge to
back-pressured
header.
C. Pilot operated SRV. Pilot valve opens Valve is capable
at set pressure, relieving the pressure of operating near
on top of main valve. Main valve then the set pressure
opens wide until blowdown pressure is without "bubbling"
reached, at which time the pilot valve or "simmering"
closes, pressuring the top of the main type leaks,
valve by way of the supply tube pressure
so that the main valve closes tightly.
E-57
-------
Standard nozzle - b. Balanced-bellows c. Pilot-operated
guide vented to nozzle - bonnet SRV.
exhaust, bonnet vented to atmosphere,
plugged.
M
i
Ln
Co
—Ad|usting bolt—-
Bonnet
' — Spring--^
Spindle
_ - Guide
Spindle guide vN
Bellows
Bellows tailpiece^
Set screw plug N
Disk--,.
Disk ring—
Nozzle ring
— set screw-
- - Nozzle ring-
—Body'
Nozzle--"
Exhaust tube
Lift adjustment \
screw X
Riot vahrt
Supply tube
Cap
Shipping
spring
- Piston seal
. ¦ • Piston
- Seat
Nozzle
Dipper tutM
retainer
Figure E2-17. Typical Safety-Relief Valves.
Reprinted by special permission from CHEMICAL ENGINEERING28
-------
Inspection and maintenance is one existing control for
SRV's. The main objective of most inspection and maintenance
programs is to make sure the SRV will provide proper over-pressure
protection. Some companies remove and test SRV's after every
overpressure release.8 This procedure requires that a means be
provided to install a spare SRV while the other one is tested.
Although this testing is primarily to check the set pressure of
the SRV, it may also detect fugitive leakage. The other existing
control for SRV's is discharge header systems that transport over-
pressure releases (and fugitive leakage) to a flare.
2.2.6.2 Available Controls for Pressure Relief Devices
Leak detection and repair programs and upstream rupture
disks are the available controls for SRV's. Leak detection would
require periodic testing of SRV's that discharge to the atmos-
phere. The hydrocarbon detector probe would be placed at the
exit of the discharge "horn" or at the weep hole at the bottom
of the "horn". Repair of the SRV would probably require removal
of the SRV, and therefore a means of replacing the SRV while the
process unit was operating would be needed. Data on costs and
effectiveness of leak detection and repair for SRV's are not
available.
Although most SRV's are used alone or in pressure-
stepped combinations, some are used with rupture disks mounted
under them (i.e., between the process fluid and the SRV), as
seen in Figures E2-18 and E2-19. Rupture disks (RD's) are some-
what prone to age-induced fatigue or corrosion failure, and
therefore are not ordinarily used alone except where complete
loss of process fluid is acceptable economically and environ-
mentally. Such acceptable cases probably no longer exist in any
organic chemicals or fuels manufacturing facility.
E-59
-------
t
INLCT
Figure E2-18. Rupture Disk and Hold-Down Flanges
Flange for
Blind Flange SRV Mount
Connection for
Pressure Gauge
and Bleed Valve
Rupture Disc
Assembly
From System
Figure E2-19. Relief Valve/Rupture Disk Assembly.
E-60
-------
Rupture disks may be used upstream of SRV's as illus-
trated in Figure E2-19 to help prevent escape of refinery fluids
provided means are established to monitor for small rupture disk
leaks. Alternatively, they may be positioned downstream of SRV's
to protect working parts from weather or other corrosive atmos-
phere, as when connected to a relief header.
Rupture disk leaks may be detected by "tell-tale"
bubblers or pressure gauges, and by excess flow valves connected
to the inside piping space between the RD and the SRV. This
arrangement is covered by ASME code.29 Such a pressure connec-
tion is shown in Figure E2-19 on the upper flange of the RD
assembly. If small RD leaks are not monitored, there is a
chance that the pressure between the RD and SRV might build to
system pressure. Then, with a rapid rise in pressure, as in an
emergency, working pressure would almost double before the disk
and SRV would release, depending upon the rate of increase and
size of the RD leak.
Effectiveness--As long as the integrity of the rupture
disk is maintained, fugitive emissions are completely eliminated.
The disk would require replacement after over-pressure release,
and therefore a means for replacing it while the process unit was
in service would be needed. Although there is controversy within
the industry concerning the use of rupture disk-safety relief
valve combinations, some feel that the combination may be
operated safely.29 Others consider RD use upstream or downstream
of the SRV only as necessary for either 1) added isolation of
particularly toxic materials from the air, or 2) as a means of
isolating the SRV from a corrosive atmosphere. This atmosphere
might be, for example, a header with sulfur compounds present,
or simply salt air near the ocean.28
E-61
-------
Costs—The addition of an inlet or outlet side rupture
disk to an SRV adds between three percent and 50 percent to the
materials cost of the SRV, depending on size and service.
Materials costs for SRV and RD assemblies (excluding piping)are
shown in Table E2-12. The net cost of the system would take into
account a value for the product saved by eliminating fugitive
emissions.
2.2.6.3 Technology Transfer for Pressure Relief Devices
Fugitive leakage caused by improper reseating after
over-pressure release may be minimized by using pilot operated
SRV's with resilient (0-ring) seats. No data are available to
quantify the effectiveness of this type of control. Another
potential improvement in SRV design would be to install
parallel SRV's in all applications. This would allow an SRV to
be in service, with the other blocked off as a spare. This would
permit SRV removal for testing and rupture disk replacement after
overpressure releases.
2.2.7 Sampling Connections
Fugitive emissions from sampling connections are the
result of purging the sample line in order to obtain a repre-
sentative sample. Atmospheric exposure of the purged fluid can
result in evaporative hydrocarbon emissions.
2.2.7.1 Existing Controls for Sampling Connections
Existing practices for obtaining process samples vary
considerably. They may range from draining process fluid onto
the ground to collection of the purge in step oil systems. " All
E-62
-------
TABLE E2-12. SAFETY RELIEF VALVE (SRV) AND RUPTURE
DISK (RD) ASSEMBLY COSTS
May, 1980 Dollars
System Size, SRV RD Assembly
Inlet x Outlet
diameter, inches 150 pri flanges 300 pri flanges Inlet Outlet
1x2 637 681 ' 321 124
3x4 1,035 1,148 517 159
8 x 10 5,943 7,802 1,058 220a
Basis: Materials only; piping excluded. May, 1980 prices.
RD assembly includes cost of safety head and one disk.
3
Interpolated from 4 inch and 12 inch diameter RD costs.
Source: Reference 30
E-63
-------
existing practices result in some atmospheric exposure and
emissions, but the magnitude has not been quantified.
2.2.7.2 Available Controls for Sampling Connections
Closed loop sampling systems are the primary control
available to reduce sample purge emissions. A closed loop
sampling system, consists of a network of piping and valves that
either returns the purged material directly to the process, or
that transports the purge to a closed collection system for
recycle.
2.2.7.3 Technology Transfer for Sampling Connections
The main innovations that are likely to reduce sample
purge emissions are the increasing availability of on-line
continuous analytical instruments that do not. require discrete
samples.
2.2.8 Wastewater Systems
Refinery wastewater systems include wastewater trans-
port and wastewater treatment. Both operations can result in
atmospheric exposure of hydrocarbon-containing streams and
evaporative emissions.
Refinery wastewater treatment 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:31
E-64
-------
Primary Separation - The removal of oil by
gravity separation. Normally, an API- or a
CPI-type separator is used. These separators
effectively remove free.oil from water, but
will not separate substances in solution nor
break up emulsions.32
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 bio-
logical treatment methods. Carbon adsorp-
tion is the most common form of tertiary
treatment.
Classification of the numerous treatment processes
into these categories is shown below.33
Primary: API Separators
Tilted-Plate Separators (CPI)
Filtration for Oil Removal
pH Control
Stripping Processes
Intermediate: Dissolved Air Flotation
Coagulation-Precipitation
Equalization
E-65
-------
• Secondary-Tertiary: Carbon Adsorption
Activated Sludge
Aerated Lagoons
Trickling Filters
Waste Stabilization Ponds
Cooling Tower Oxidation
Chemical Oxidation
Filtration
In addition, there is a wastewater collection system which con-
sists of process drains, sewers, holding basins, and pumps. See
Section 4.8.2.2 of Appendix F for a complete review of wastewater
treatment systems.
Table E2-13 gives an estimate of the degree of adoption
of various wastewater treatment processes for 1950, 1963, 1967,
1972, and 1977. While this table utilizes the author's judgement
in many areas due to the "dearth of usable information," the data
on API separators are reliable and confirm that by 1977 nearly
all refineries had an oil and water separator of the API or the
CPI type.3*
Table E2-13 also shows an increasing use of intermediate
secondary and tertiary treatment methods. This trend is a result
in part, of Federal and state laws pertaining to water pollution.
The Federal Water Pollution Control Act Amendments of 1972
established the need to install "best practicable control tech-
nology" by July 1, 1977 and to install "best available control
technology economically achievable" by July 1, 1983.
A new set of Federal guidelines, developed in response
to a successful legal challenge by environmental groups, requires
that by 1984 industries must use "best available technology" to
remove those compounds that have been deemed toxic to humans or
E-66
-------
TABLE E2-13. DEGREE OF ADOPTION OF VARIOUS WASTEWATER
TREATMENT PROCESSES
Z of Refineries Using the Process
Processes and Subprocesses 1950 1963 1967 1972 1977
API Separators
40
50
60
70
80
Earthen Basin Separators a
60
50
40
30
20
Evaporation
0-1
0-1
1
1-2
2-5
Air Flotation
0-1
10
15
18
20
Neutralization (Total Wastewater)
0-1
0-1
0-1
0-1
0-1
Chemical Coagulation and Precipitation
1-5
1-5
5-10
10-15
10-15
Activated Sludge
0
5
10
40
55
Aerated Lagoons
0
5
10
25
30
Trickling Filters
1-2
7
10
10
10
Oxidation Ponds
10
25
25
25
20
Activated Carbon
0
0.5
0.5
3
5
Ozonation
0
1
1
3
5
Ballast Water Treatment - Physical
9
9
8
5
5
Ballast Water Treatment - Chemical
1
1
2
5
5
Slop Oil - Vacuum Filtration
0
5
7
12
15
Slop Oil - Centrifugation
0
2
3
10
15
Slop Oil - Separation
100
93
90
80
70
Sour Water - Steam Stripping
- Flue Gas Strippers
- Natural Gas
60
70
85
90
90
Sour Water - Air Oxidation
0
3
3-5
7
10
Sour Water - Vaporization
1
1-2
1
0
0
Sour Water - Incineration8
35-40
40
50
30
20
Neutralization of Spent Caustics
Flue Gas
20
30
35
20
20
Spent Acid (including
springing and stripping)
15
25
30
25
20
Oxidation
0
3
5
5
5
Incineration
25
40
50
20
15
'incineration Includes flaring, boiler furnaces, and separate incinerators used only in
conjunction with stripping and vaporization.
Source: Reference 34.
E-67
-------
animals, or that are environmentally unacceptable. There are
now 129 compounds that may eventually be subject to these specific
removal guidelines.3 5
The relationship between wastewater flowrate and crude
oil throughput has been shown to vary widely among refineries.
Newer or updated refineries do a better job of segregating pro-
cess water from storm water. Wastewater flow rates may range
from six barrels per barrel of crude in old refineries to as
little as 0.8 barrels per barrel of crude in new refineries.37
Table E2-14 shows typical removal efficiencies for the various
refinery wastewater treatment processes.
Higher water/oil ratios require more oil and water
separator surface area and thus increase the fugitive emissions
from the separators. The required surface area is a function of
the wastewater flow rate, the oil density, and the oil content .•
of the effluent water. In designing an API separator, the total
surface area is obtained from these three factors, then the con-
figuration of the separator is determined by using the following
guidelines:3 8
• Velocity should not exceed two feet per minute.
• Length-to-width ratio should be at least five to
avoid dead spots in the separator.
• Minimum depth should be four feet.
2.2.8.1 Existing Levels of Control for Wastewater Systems
Current Emissions—The problem of determining and
controlling fugitive emissions to the atmosphere from refinery
wastewater systems is clouded by several factors. First,
E-68
-------
t
TABLE E2-14. TYPICAL REMOVAL EFFICIENCIES FOR OIL
REFINERY TREATMENT PROCESSES
Process Number
Removal Efficiency, X
W
I
o>
vo
Number Process Process Influent
BOD 5
COD
TOC
SS
Oil
Phenol
Amnion la
Sulfide
1 API Separator
Raw Haste
5-40
5-30
NA
10-50
60-99
0-50
NA
NA
2 Clarlfler
1
30-60
20-50
NA
50-80
60-95
0-50
NA
NA
3 Dissolved Air
Flotation
1
20-70
10-60
NA
50-85
70-85
10-75
NA
NA
4 Filter
1
40-70
20-55
NA
75-95
65-90
5-20
NA
NA
S Oxidation Pond
1
40-95
30-65
60
20-70
50-90
60-99
0-15
70-100
6 Aerated Lagoon
2,3,4
75-95
60-85
NA
40-65
70-90
90-99
10-45
95-100
7 Activated Sludge
2,3,4
80-99
50-95
40-90
60-85
80-99
95-99+
33-99
97-100
8 Trickling Filter
1
60-85
30-70
NA
60-85
50-80
70-98
15-90
70-100
9 Cooling Tower
2,3,4
50-90
40-90
10-70
50-85
60-75
75-99+
60-95 •
NA
10 Activated Carbon
2,3,4
70-95
70-90
50-80
60-90
75-95
90-100
7-33
NA
11 Filter Granular
Media
5-9
NA
NA
50-65
75-95
65-95
5-20
NA
NA
12 Activated Carbon
5-9 plus 11
91-98
86-94
50-80
60-90
70-95
90-99
33-87
NA
HA " Data not available.
Source: Reference 36.
-------
wastewater systems are designed to treat wastewater and it is
that function which receives the most attention. Second,
refinery wastewater systems vary tremendously as to volumes of
process water, storm water, particulates, oil and grease, and
other contaminants. Third, refinery wastewater systems vary
from one refinery to the next. The only common denominator is
an oil and water separator of the API or CPI type, although the
actual configuration of these separators may vary considerably.
Fourth, reliable test data on hydrocarbon emissions from refinery
wastewater systems are not available.
Most sources give emission figures which stem from
the estimates made for the Los Angeles area in 1958. In that
early study, hydrocarbon emissions from wastewater separators
from existing refineries in the Los Angeles County area were
estimated to range from 10 lb/1000 bbls refinery capacity to
200 lb/1000 bbls refinery capacity.39 Supplement 8 of AP-42
(May, 1978) lists the hydrocarbon emission factors shown in
Table E2-15. These emission factors are for the combined
emissions of drains and wastewater separators.
TABLE E2-15. WASTEWATER SYSTEM EMISSION FACTORS
Control Emission Factor (lb/103 bbl refinery feed)
Uncontrolled 200
Vapor recovery; separator covers 10
The results of Radian's oil/water separator emission
sampling are shown in Appendix B, Tables B3-12 through B3-20.
The results are not shown here because it appears that some of
the data are invalid due to difficulties in obtaining representa-
tive samples and in applying the test methods used. The emission
E-70
-------
factor for process drains is shown in Table B2-23. The mean
value is 0.07 pounds of hydrocarbon per hour for each drain.
The 95 percent confidence intervals for the mean estimate
are (0.023, 0.20).
Current Controls — Covered oil/water separators and
trapped drain systems are two types of emission controls used in
some refineries. Some state regulations require covers for
separators, and as of January 1977, 80 percent of the U.S.
refining capacity was located in states where covers are
required.1,1 The extent of application of trapped drain systems
is not known. Because of the lack of emission data, effective-
ness of those controls cannot be assessed. Costs would vary,
depending on site specific conditions.
The current AP-42 emission factors for drains and
oil/water separators, uncovered versus covered, imply a 96
percent fugitive hydrocarbon emission reduction. The original
reference upon which the AP - 42 emissions are based is, accord-
ing to the California Air Resources Board (CARB) no longer
available.42 In a laboratory study using a simulated API
separator, the covered separator provided 89 percent emission
reduction. **3
2.2.8.2 Available Controls for Wastewater Systems
In general, available controls for reducing fugitive
emissions from existing process and storm sewers and collection
systems consist of relatively minor modifications such as seal-
ing open sewer systems, altering pump bases, recurbing some
process areas, and improving housekeeping practices.
E-71
-------
Changes which involve substantial capital outlays (or
which may be infeasible from a construction standpoint) such as
major revisions to existing underground sewer systems or installa-
tion of vapor recovery systems probably do not represent best
available technology economically achievable. Techniques which
are considered reasonable controls for emissions from the collec-
tion system are listed below.
• Open drains, sewers, or holding basins upstream
of the oil and water separator: these sources
of emissions in the U.S. refining industry are
now fairly rare. The evaporation of significant
volumes of oil at current world scale prices is
a readily apparent financial burden. Obviously,
holding basins should be eliminated and process
drains and sewers sealed or vented through
liquid seals.
• Pump bases which do not drain completely by
gravity: many pump bases are designed so that
a slight level of oil (from a leaking seal)
must build up before the base drains to the
sewer. When new pumps are to be installed,
bases should be selected which allow proper
drainage. Existing pump bases can be modi-
fied at minor cost.
• Segregation of process water from storm water:
oily water volumes should be minimized. Curbing
should be installed so that only those areas
which are subject to oil spills drain into the
oily water sewer system. Storm sewers should
be sized' so that overflow into process sewers
E-72
-------
during peak runoff is avoided. In many cases,
however, substantial revisions to the sewer
systems of older plants will be prohibitively
expensive.
• General housekeeping: an undefined but in some
cases significant source of emissions is the lack
of good housekeeping practices concerning oil
spills and leaks. A quantitative control techni-
que in the area of oil spills and leaks could
probably not be formulated, but an awareness of
the problem would be beneficial.
2.2.9 Cooling Towers
Cooling towers can emit hydrocarbons that contaminate
the circulating cooling water. The hydrocarbons may enter the
cooling water as a result of leaks in heat exchangers or as a
result of using contaminated process water as cooling tower
make-up water. When the contaminated water is cascaded through
the cooling tower, hydrocarbons can be evaporated and emitted
to the atmosphere.
2.2.9.1 Existing Control Levels for Cooling Towers
At the time of the 1958 Los Angeles County California
Emissions Study,* * "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 bio-
logical growth, respectively. The emission factor for cooling
towers was estimated to be 6.2 lb hydrocarbons/10s gal cooling
water circulation.
E-73
-------
By 1978, wetted "atmospheric" sections had generally
been phased out of refineries. Organo-phosphates had replaced
chromates for corrosion control; biological growth was being
controlled by combinations of chlorine and, often, nonoxidizing
biocides.
Today, as in 1958, make-up 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. During recycling in a cooling tower, the aeration
encourages oxidation of phenols in the stripper bottoms water.
Emission factors determined during this study were
based on two analytical methods: Total Organic Carbon (TOC)
Analysis and a purge technique (see Appendix A). These results
bracket the 1958 emission factor of 6.2 lb/106 gal cooling water,
as shown in Table E2-16.
TABLE E2-16. RADIAN-GENERATED COOLING TOWER
EMISSION FACTORS
Emission Factor
Analytical Technique lb HC/106 gal C.W.
(Appendix B)
TOC 12.4
Purge 0.108
Because of variations in sampling and analytical techniques, these
results are inconclusive. There is also a wide variation in the
quality of water that is used as cooling tower make-up, and there-
fore large differences may be expected at different facilities.
E-74
-------
2.2.9.2 Available Controls for Cooling Towers
The best control for cooling towers is to minimize the
amount of hydrocarbons entering the tower. One method to
achieve this goal is to eliminate the use of contaminated process
water as cooling tower make-up. This may be difficult, since
efforts to reduce water discharges may require the use of pro-
cess water for cooling towers. Another control option is to
monitor the hydrocarbon content of the cooling tower input. If
elevated concentrations were detected, a leak in the process
equipment would be indicated. The problem then is to identify
the specific leak and to repair it. Ambient air monitors could
also be used to measure the hydrocarbon concentration in the
cooling tower plume. Because of variable meteorological condi-
tions, this method would be difficult to apply.
2.2.10 Solid Waste System Alternatives
Petroleum refineries generate numerous solid waste
streams, which may contain many substances. These streams may
be divided into two main groups: intermittent and continuous.
Typical intermittent waste streams are as follows:36
• process vessel sludges, vessel scale and
piping deposits,
• storage tank sediments,
• produce treatment wastes, such as filter
clay and spent catalysts from certain
processing units.
E-75
-------
Typical continuous waste streams are as follows:36
• coke fines from cokers and spilled coke
from unloading facilities,
• fluid catalytic cracker catalyst,
• spent and spilled grease and wax from
lube oil processing plants,
• waste biological sludges from activated
sludge units,
• floating solids from dissolved air
flotation unit.
Most solid wastes are residuals from wastewater treat-
ment. The exceptions to this are some spent catalysts which are
recovered in segregated containers, spent acids and caustic, and
other spills and sediments which can be segregated. Normally
these exceptions are handled separately from other solid wastes.
Fugitive emissions from solid waste disposal could
be particulate matter or evaporative emissions of residual hydro-
carbons. The only control option available is to minimize the
amount of atmospheric contact of the wastes during handling and
disposal. Therefore, the following sections only describe the
types and amounts of solid wastes produced in refineries, and
the waste disposal practices used.
E-76
-------
2.2.10.1 Existing Solid Waste Disposal Practices
The five general categories of solid waste disposal
alternatives are listed below."5 The first three of these
disposal alternatives can create emissions to the atmosphere.
• Landfarming
• Incineration (with landfilling of the ash)
• Landfilling
• Deep-well injection
• Solidification (producing relatively inert
substances which chemically or physically
isolate the pollutant, or surrounding the
pollutant by encapsulation)
Landfarming, incineration, and landfilling are the most
common methods of disposal for refinery solid wastes. A 1976
summary of solid waste disposal practices by Jacobs Engineering
Company describes disposal practices of the U. S. refining
industry.1*6 Unless otherwise noted, percentages of adoption
figures are for the year 1973.
Crude Tank Bottoms--Crude tank bottoms are oil-water
emulsions containing large quantities of solids and colloidal
material which have settled to the bottom of the tank. In
general crude tank bottoms are not flammable at ambient conditions,
but can be hazardous due to their heavy metals and oil content.
About 14 percent of the U. S. refineries are known to be land-
filling these wastes, 23 percent are landfarming, 6 percent are
landfilling in a "secure" site (California Classification Class 1),
E-77
-------
and less than 1 percent are landfarming on a "secure" site.
Practices at the remaining refineries are not known.
Straight landfilling is an inadequate method of
disposal due to the danger of migration of oil and/or heavy
metals from the site. Landfarming appears promising, but more
information is needed about best operating methods and degrada-
tion products.
Leaded Gasoline Sludges — Sludge from leaded gasoline
storage tanks contains precipitated organic lead compounds in a
mixture of gasoline, other additives and tank scale. The typical
organic lead concentration in fresh sludge is 100 to 200 ppm.
This sludge produces organic lead vapors which are toxic at very
low concentration (OSHA limits are 0.075 to 0.15 mg/m3). Other
hazardous constituents are phenols, arsenic, selenium, mercury,
cobalt, nickel, zinc, cadmium, and molybdenum.
About 61 percent of U.S. refineries pump leaded gaso-
line tank bottoms into an excavated pit, allow it to evaporate
and weather for one to four months, and then cover the pit. The
disposal method is often inadequate since dangerous organic lead
concentrations remain years after covering. An alternative pro-
cedure used by 39 percent of U. S. refineries is to spread the
sludge two to four inches thick on a curbed concrete pad and
allow it to weather until the organic lead concentration is
below 20 ppm. The residue is then placed in a sanitary landfill.
Non-leaded Tank Bottoms--Non-leaded tank bottoms sludge
waste is generated when tank sediment levels become too high
(once every one to five years) or when tank service is changed
(which occurs annually for many tanks such as diesel/gasoline
switch tanks). This sludge is put in an approved sanitary
E-78
-------
landfill in 57 percent of U. S. refineries. This disposal tech-
nique merely allows the petroleum to be adsorbed into the soil
and stored until it leaches out or is degraded. About 23 percent
of the refineries landfarm this sludge. This method is promising.
API Separator Sludge--API separator sludge is a heavy
black oily mud which contains relatively high concentrations of
phenols, Cr, Se, Hg, Co, Cu, Zn, Cd, Pb and Mo. In 1973 about
53 percent of U. S. refineries landfilled API separator sludge
in unsecured sites, while another seven percent landfilled in
secured sites. Landfilling in an unsecured site will not be
acceptable by future standards. In 1976, seven percent of these
sludges were being landfarmed.117 By 1979 about half of these
wastes were being landf armed.118
Hydrofluoric Acid Alkylation Sludge--A cream-brown
slurry of insoluble CaF2 sludge is produced when HF acid is
neutralized with spent lime. While HF is extremely hazardous,
when it has been properly neutralized with lime the resultant
precipitate can be landfilled without much risk to ground waters.
Landfilling is the normal disposal technique.
Kerosene Filter Clays--Kerosene filter clays contain
oil and heavy metals. These sludges are landfilled in secure or
unsecure sites. Danger of leakage from unsecured landfill is
high. Landfarming is an option.
Once-Through Cooling Water Sludges--Silt from the raw
water supply, process leakage, and corrosion products are settled
in clarification basins. The silt or mud has occluded oil and
heavy metals such as As, Hg, Mo, Pb and Se. These basins are
periodically drained and dredged and the silt landfilled or land-
farmed in secured or unsecured sites. Unsecured sites present
groundwater contamination hazards in landfill disposal.
E-79
-------
Dissolved Air Flotation (DAF) Float--DAF float contains
oil and heavy metals. It is normally trucked to a landfill or a
landfarm site. Once again, unsecured landfill sites may be
hazardous to groundwaters.
Slop Oil Emulsion Solids—Crude and other petroleum
fractions recovered from spills and interception points within
the refinery are often too contaminated to be rerun in the plant.
These wastes can be landfarmed or landfilled in secure or
unsecure sites. Potential exists for quantity reduction through
centrifugation or filtration.
Spent Lime from Boiler Feedwater Treatment--The spent
lime slurry from boiler feedwater softeners contains mixtures of
calcium, magnesium and other carbonates; hydroxides; silt;
organic matter and water. The metals are effectively tied up
in the sludge and minimal hazard is present.
Cooling Tower Sludge—Cooling tower sludge contains
high concentrations of several heavy metals, particularly hex-
avalent chromium and zinc. In addition, copper, lead, arsenic
and selenium are present in relatively high concentrations.
In 1973, 81 percent of U. S. refineries trucked this
sludge to unsecured landfill sites while only eight percent were
known to landfill in secured sites. Unsecured site landfilling
presents possible danger to groundwaters.
Exchanger Bundle Cleaning Sludge--Exchanger bundle
cleaning creates a residue of metal scale, coke, particulate
matter, and oil. These wastes are landfilled in secure or
unsecure sites. The oil and heavy metals content of the material
makes unsecure landfilling potentially hazardous to groundwaters.
E-80
-------
Waste Bio Sludge--The organic cell material synthesized
by microorganisms during secondary wastewater treatment is
periodically wasted to maintain steady state conditions. This
material is landfilled in secured sites (19 percent of U. S.
refineries in 1973), landfarmed (25 percent) or landfilled in
unsecure sites (15 percent). Landfarming this waste is becoming
more frequent as techniques are improved.
Storm Water Silt--Storm water runoff picks up soil,
oil, spilled solutions, metals, coke fines and other miscellane-
ous substances. Twenty-four percent of U. S. plants were known
to landfill this material in unsecured sites in 1973, four
percent landfilled in secure sites. Secured sites are needed
for adequate disposal.
FCC Catalyst Fines--FCC catalyst fines consist of a
fine grey powder of aluminum silicate with absorbed heavy metals,
primarily vanadium and nickel. During transfer operations, the
dust may become airborne, so workers should wear goggles and
filtering masks. In 1973, 30 percent of the refineries were
known to be transporting these fines to unsecured landfill sites
and eight percent were known to use secured sites. Unsecured
sites are subject to leaching of heavy metals into groundwaters.
Many refineries contract for removal of this waste by companies
which specialize in handling this material. Several companies
are investigating the feasibility of using FCC fines as alumina
cement additive and of recovering aluminum and vanadium.
Coke Fines--Grey-black chunks and particle fines from
coking units are generated by spillage during transport and load-
ing of coke, by contamination of coke with earth during stock-
piling, and by mixing with water during removal from the drum.
E-81
-------
Most refineries landfill this waste. High concentrations of
vanadium and nickel make a secured landfill the preferred
disposal technique.
2.2.10.2 Amount of Wastes Produced
Table E2-17 gives the total estimated solid waste
generation rate for U. S. refineries. Boiler feedwater lime
sludge accounts for over 60 percent of the tonnage of waste in
this 1975 survey. This source is not a major problem since the
sludge is not hazardous, and since lime treating of boiler feed-
water is gradually being replaced with demineralization processes.
E-32
-------
TABLE E2-17. TOTAL ESTIMATED U. S. REFINERY
SOLID WASTE GENERATION RATE
Total
Solid Waste Stream Tons/yr lb/103 BBL Crude
Crude Tank Bottoms
830
0.3
Leaded Gasoline Sludge
1,840
0.7
Non-Leaded Tank Bottoms
91,500
35.8
API Separator Sludge
76,500
29.9
HF Acid Alkylation Sludge
18,430
7.2
Kerosene Filter Clays
4,800
1.9
Once-Through Tooling Water Sludge
41,300
16.2
Dissolved Air Flotation Float
66,200
25.9
Slop Oil Emulsion Solids
37,200
14.6
Boiler Feedwater Lime Sludge
858,700
336.1
Cooling Tower Sludge
500
0.2
Exchanger Bundle Cleaning Sludge
1,450
0.6
Waste Bio Sludge
89,500
35.0
Storm Water Silt
33,200
13.0
FCC Catalyst Fines
34,100
13.3
Coke Fines
3,820
1.5
TOTAL
1,359,870
532.2
Note: Unit loads are based on a total U.S. refinery capacity of 14xl06 BPD.
Source: Reference 45.
E-83
-------
2.3 References
1. Pikulik, A. Selecting and Specifying Valves for New
Plants. Chem. Eng., 83(19):168, 1976.
2. Scott, Carleton B. Union Oil Company of California.
Letter to Don R. Goodwin, EPA:ESED, December 3, 1976,
response to EPA request for information on miscellaneous
hydrocarbon emission sources from refineries.
3. Van Ingen, R. E. Shell Oil Company. Letter to Don R.
Goodwin, EPA:ESED, January 10, 1977, response to EPA
request for information on miscellaneous hydrocarbon
emission sources from refineries.
4. Amey, Guy C., Century Systems Corporation. Letter to
Jim Serne, Pacific Environmental Services, Inc., October
17, 1979.
5. Johnson, J. M. , Exxon Company, Letter to Robert T. Walsh,
EPA:CPB, July 28, 1977, comments on 1st draft report,
"Control of Hydrocarbons from Miscellaneous Refinery Sources."
6. Pikulik, A. Manually Operated Valves. Chem. Eng., 85(8):
119, 1978.
7. Perry, R. H. and C. H. Chilton. Chemical Engineers Hand-
book, 5th edition. McGraw-Hill, New York, New York, 1973.
E-84
-------
8. California, State of, Air Resources 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, Sacramento, California. April 1978. Page 1-2.
9. Frazier, William, Crane Co., Houston, Texas. Telephone
conversation with W. R. Phillips, Radian Corporation,
Austin, Texas, June 18, 1979.
10. Hoyle, R. How to Select and Use Mechanical Packing.
Chem. Eng., 85(22) : 103, 1978.
11. American Petroleum Institute, Division of Refining.
Manual on Disposal of Refining Wastes, Volume on
Atmospheric Emissions. API Pub. No. 931, Washington,
D.C., 1976.
12. Steigerwald, B. J. Emission of Hydrocarbons to the
Atmosphere from Seals on PUmps and Compressors. Report
No. 6. Joint District, Federal and State Project for
the Evaluation of Refinery Emissions and Los Angeles
County Air Pollution Control District, Los Angeles,
California, April 1958.
13. Ramsden, J.. H. How to Choose and Install Mechanical
Seals. Chem. Eng., 85(22):102, 1978.
14. Center for Professional Advancement. Mechanical Seal
Technology for the Process Industries. East Brunswick,
New Jersey, March 1978.
15. Block, H. P. Improve Safety and Reliability of Pumps and
Drivers, Part 4. Hydrocarbon Processing, 56(4):181, 1977.
E-85
-------
16. American Petroleum Institute, Refining Dept. Centrifugal
Pumps for General Refining Services, 5th edition. API
Standard 610. Washington, D.C., March 1971.
17. Richards, C. J. Pacific Pump Division of Dresser Indus-
tries, Houston, Texas. Private communication with W. R.
Phillips, Radian Corporation, Austin, Texas, May 15, 1980,
regarding Pump and Driver costs for 3 - 100 h.p. pumps with
1.875 - 2.375 in. dia. shafts.
18. Voden, James, Allen-Bradley Co., Houston, Texas. Private
communication with W. R. Phillips, Radian Corporation,
Austin, Texas, May 15, 1980, regarding electric switch-
gear for pump motors.
19. Adams, C. S., Gulf Coast Packing and Seal Co., Inc.,
Houston, Texas. Private communication with W. R. Phillips,
Radian Corporation, Austin, Texas, May 1980, regarding
A. W. Chesterton pump seals and packings.
20. Fadner, D. D., Crane Packing Co., Houston, Texas. Private
communication with W. R. Phillips, Radian Corporation,
Austin, Texas, May 1980, regarding costs and applications
of John Crane Mechanical Seals.
21. Potter, Charles, Crane-Deming Pump Co., Houston, Texas.
Telephone conversation with W. R. Phillips, Radian Corpora-
tion, Austin, Texas, September 27, 1979.
22. American Petroleum Institute, Refining Dept. Centrifugal
Compressors for General Refinery Sources, 3rd edition.
API Standard 618. Washington, D.C., October 1973.
E-86
-------
23. American Petroleum Institute, Refining Dept. Reciprocating
Compressors for General Refinery Service, 2nd edition. API
Standard 618. Washington, D.C., July 1974.
24. Nelson, W. E. Compressor Seal Fundamentals. Hydrocarbon
Processing, 56(12):91, 1977.
25. American Petroleum Institute, Refining Dept.. Lubrication,
Shaft-Sealing, and Control Oil Systems for Special Purpose
Applications. API Standard 614. Washington, D.C.,
September 1973.
26. Ramsey, W. D. and G. C. Zoller. How the Design of Shafts,
Seals and Impellors Affects Agitator Performance. Chem.
Eng., 83(18):101, 1976.
27. Kem, R. Pressure Relief Valves for Process Plants. Chem.
Eng., 84(5):187, 1977.
28. Isaacs, M. Pressure-Relief Systems. Chem. Eng., 78(5):113,
1971.
29. Kayser, D. S. Rupture Disc Selection. CEP, 68(5):61, 1972.
30. Britton, Stephen, Groth Equipment Corporation, Houston,
Texas. Private communication with W. R. Phillips, Radian
Corporation, Austin, Texas, May 19, 1980.
31. Beychok, Milton R. Wastewater Treatment. Hydrocarbon
Processing, 50(12):110, 1971.
E-87
-------
32. U.S. 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, Washington, D.C., April 1974.
33. Azad, H. S. ed. Industrial Wastewater Management Hand-
book. McGraw-Hill, New York, New York, 1976.
34. Jones, H. R. Pollution Control in the Petroleum Industry.
Pollution Technology Review No. 4. Noyes Data Corporation,
Park Ridge, New Jersey, 1973.
35. Long, J. and C. Murray. Industry Braces for New Water
Cleanup Rules. C&EN, 57(20):31, 1979.
36. Sittig, M. Petroleum Refining Industry-Energy Saving and
Environmental Control. Noyes Data Corporation, Park
Ridge, New Jersey, 1978.
37. Little (Arthur D.), Inc. Environmental Consideration of
Selected Energy Conserving Manufacturing Process Options,
final report. Volume 4, Petroleum Refining Industry
Report. EPA-600/7-076-034d, Washington, D.C., December
1976.
38. Thompson, S. J. Data Improves Separator Design. Hydro-
carbon Processing, 52(10):81-83, 1973.
E-88
-------
39. 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.
AO. U.S. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Monitoring and Data Analysis
Division. Compilation of Air Pollution Emission Factors,
3rd edition, Supplement 8. AP-42, PB-275-525. Research
Triangle Park, North Carolina, May 1978.
41. 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. U.S. Environmental Protection Agency, Office
of Air Quality Planning and Standards, Office of Air and
Waste Management, Research Triangle Park, North Carolina,
October 1977.
42. Shaffer, N. R. and C. J. Seymour. Emissions of Hydro-
carbons from Refineries in Los Angeles County. Los Angeles
County, Air Pollution Control District, Los Angeles,
California, April 1957.
43. Litchfield, D. K. Controlling Odors and Vapors from API
Separators. Oil and Gas Journal, 69(44):60-62, November
1971.
44. Los Angeles, County of, Air Pollution Control District,
et al. Emissions to the Atmosphere from Petroleum Refin-
eries in Los Angeles County. Los Angeles, California, 1958.
E-89
-------
45. Pojacek, R. B. Solid Waste Disposal-Solidification.
Chem. Eng.,86(17):141-145, 1979.
46. Rosenberg, D. G., et al. Assessment of Hazardous Waste
Practices in the Petroleum REfining Industry, final report.
EPA/SW-129c. U.S. Environmental Protection Agency, June
1976.
47. Grove, G. W. Use Landfarming for Oily Waste Disposal
Hydrocarbon Processing, 57(5):138-140, 1978.
48. Huddleston, R. L. Solid-Waste Disposal: Landfarming.
Chem. Eng., 86(5):119-124, 1979.
E-90
-------
3.0
CONTROL OF STACK AND OTHER PROCESS EMISSIONS
3.1 Sources of Emissions
Since no two refineries are exactly alike, refining
process emissions vary considerably from site to site. In
general, however, the major sources of atmospheric process
emissions are sulfur recovery, fluid catalytic cracker catalyst
regeneration, and process heaters and boilers.
The major types of atmospheric process emissions from
refineries are hydrocarbons, sulfur oxides (S0x), particulates,
and carbon monoxide (CO). Process emission sources are cate-
gorized in Table E3-1 according to type of emissions.
Process heaters and boilers are used in a number of
refinery processes. Instead of being discussed with each pro-
cess, they are discussed collectively as a separate emission
source.
3.1.1 Sulfur Recovery
The amount of sulfur in various product streams
depends directly on the sulfur content of the crude oil.
Sulfur tends to be more concentrated in the heavier cuts
because of the low volatility of its 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 sour 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 petrochemical feedstocks.
E-91
-------
TABLE E3-1. PROCESS EMISSIONS BY SOURCE AND TYPE
Emissions
Source
HC Particulate SO CO Aldehydes NH3 NO
X X
Sulfur
Recovery1
/ /
Catalyst
Regeneration
(CO Boiler
Vent)3
/
/
/ /
/
/ /
Boilers and
Process
Heaters3
/
/
/ /
/
/
Vacuum ,
Distillation
Coking*5
Air
Blowing
Chemical ^
Sweetening
Acid k
Treating
/
/
/
/
/
/
/
Blowdown
Compressor
Engines
/
/
/
/ /
Detailed results are in Appendix B.
Not measured in this study.
E-92
-------
Sulfur removal from whole crude is not generally
economical.2 Various intermediate stock streams routinely
subjected to sulfur removal include the outlet streams from
crude distillation and the cracking units.3 The sulfur com-
ponents in these streams are converted to hydrogen sulfide by
contact with hydrogen over a nickel-molybdenum catalyst at an
elevated temperature. The resulting H2S boils between ethane
and propane, so may be removed from the stream and concentrated
by one of several means, the most common of which is absorption
by monoethanolamine (MEA) or diethanolamine (DEA) followed by
steam stripping.
At one time this H2S was simply burned with other
light gases as refinery fuel. In recent years, to minimize S0y
emissions and to produce elemental sulfur for sale to other
industries, the Claus process has been used. The tail gas from
a Claus unit is the main source of S0x emissions in a refinery
today. In the Claus process, some H2S is oxidized to form S02
and water. Additional H2S reacts with this S02 to form elemen-
tal sulfur and water.
The tail gas from a Claus unit contains H2S, S02, CS2,
COS and sulfur. The emission rates of these sulfur compounds
depend on the concentration of the H2S stream to the Claus unit
and the efficiency of that unit. Tail gas from a typical three-
stage Claus unit, 95 to 96 percent efficient, can be expected
to contain about 7,000 - 12,000 ppmv sulfur compounds. 5 The
tail gas also contains carbon monoxide formed from small
amounts of hydrocarbons and carbon dioxide in the feed stream.
Since only partial oxidation of H2S is desired, not enough
oxygen is supplied to convert all the CO formed to C02.
Typical compositions of the feed to a 94 percent efficient
Claus unit and the tail gas stream from this unit are shown
E-93
-------
in Table E3-2. Claus tail gas is typically incinerated so that
S0x constitute the only sulfur compounds measurable in the
final stack gas.
3.1.2 Catalyst Regeneration
Catalysts are used in several petroleum refining
operations, namely, fluid catalytic cracking (FCC), moving bed
catalytic cracking (TCC), catalytic hydrocracking, reforming,
and various oil desulfurizations. These catalysts become
coated with carbon and metals and must be regenerated to
restore their activity. During regeneration, the carbon is
oxidized to mixed carbon oxides and the hydrocarbons are
burned incompletely.
In some applications, a catalyst must be regenerated
only a few times a year. Emissions during regeneration may
include catalyst fines, oil mist, hydrocarbons, ammonia, S0x,
chlorides, cyanides, N0X, 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 a year
probably are not significant.
Catalytic cracking catalyst regeneration is a con-
tinuous process. Uncontrolled cracking catalyst regeneration
emissions are one of the major sources of air pollution in a
petroleum refinery. Flue gases from catalytic cracker regen-
erators contain particulates, S0X, carbon monoxide, hydro-
carbons, N0x, aldehydes and ammonia.
Emission factors for the uncontrolled regeneration
of FCC and TCC catalysts are reported in AP-42 and are listed
here in Table E3-3. These factors are from a 1956 stack
E-94
-------
TABLE E3-2
TYPICAL COMPOSITIONS OF FEED STREAM AND
TAIL GAS FOR A 94 PERCENT EFFICIENT
CLAUS UNIT
Component
Sour Gas Feed
Volume %
Claus Tail Gas
Volume %
H2S
S02
S8 vapor
S8 aerosol
COS
CS2
CO
CO 2
02
N2
h2
h2o
H.C.
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
0.85
0.42b
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
Temperature, °F
Pressure, psig
Total Gas Volume'
104
6.6
284
1.5
3.0 * feed
gas volume
3
Gas volumes compared at standard conditions.
^NSPS requires an emission of less than 250 ppmv (0.025%) S02, zero percent
02, dry basis if Claus Unit Tail Gas is oxidized as the last control step,
or, 300 ppmv S02 equivalent reduced compounds (H2S, COS, CS2) and only
10 ppm H2S as S02, zero percent 02, dry basis, if the Tail Gas is reduced
as the last control step.
Source: Reference 6.
E-95
-------
sampling survey of FCC and TCC units in Los Angeles County.8
The survey involved six FCC units (with a total fresh feed
rate of 157,000 BPD) and nine TCC units (with a total fresh
feed rate of 69,300 BPD).
TABLE E3-3.
EMISSION FACTORS FOR UNCONTROLLED REGENERATION
OF THE CATALYTIC CRACKING CATALYST
Emission Factor (lb/1,000 bbl Fresh Feed)
OJ
Source
4->
N
M
to
o
o
CO
t-H
cn
a
z
a)
3
EE
•a
O
Cfi
CO
•H
to
rH
TO
.c
4-1
TO
a)
X
4J
X
•a
to
o
o
o
o
t-H
PL,
W
o
H
z
<
-------
TABLE E3-4. EMISSION RATES FROM FCCU REGENERATORS,
BEFORE AND AFTER CO BOILER
Composition of Emissions from FCCU Equipped
Flue Gas from FCCU With CO Boilers _
Without CO Regenerator Flue Gas
Boiler Composition
Chemical Species
(Reference 9)
Reference 9
This
Study
SO2, ppmv
130-3300
£ 2700
14-871
SO 3, ppmv
NAC
NAC
0.7- 13.5
N0X (as NO2), ppmv
8-394
£ 500
94-453
CO, Vol.%
7.2-12.0
0-14
0.0
C02, Vol.%
10.5-11.3
11.2-14.0
13.5-16.1
02, Vol.%
0.2-2.4
2.0-6.4
3.2-7.0
N2, Vol.%
78.5-80.3
82.0-84.2
77.0-82.7
H20, Vol.%
13.9-26.3
13.4-23.9
9.2-22.7
Hydrocarbons, ppmv
98-1213
NAC
0-46
Ammonia, ppmv
0-675
NAC
0-15
Aldehydes, ppmv
3-130
NAC
0-20
Cyanides, ppmv
0.19-0.94
NAC
0-19
Particulates, gr/SCF
0.08-1.39
0.017-1.03
0.012-0.
Temp., °F
1000-1200
458-820
386-727
a
^All considerations on dry basis
Based on sampling of 6 stacks
Not available
E-97
-------
TABLE E3-5. EMISSION RATES FROM FCCU REGENERATORS
EQUIPPED WITH CO BOILERS
Gas Components
Total Emission Rate
Pounds Emitted per
Based on Data From This Study,3
1000 bbl Fresh Feed to FCCU
Range of
Emission Rates
Mean Emission
Rate
Median Emission
Rate
SO 2
8-382
196
162
SO 3
0.5-9.0
2.9
1.0
N0X (as N02)
41-193
89
71
CO
0
0
0
Hydrocarbons
1.1-12.0
5.1
2.5
Ammonia
0.06-1.65
0.56
0.21
Aldehydes
0.0-4.6
2.0
1.8
Cyanides
0.001-4.54
0.84
0.11
Particulates
7.9-45.2
23.2
17.0
Based on Sampling of 6 Stacks
E-98
-------
The amount of CO is reduced, of course. However, another
noticeable result is the substantial reduction in emissions
of hydrocarbons, ammonia, and aldehydes. The exiting flue
gas temperature is much lower after passing through the CO
boiler.
3.1.3 Boilers and Process Heaters10
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 end 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. 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. Most steam production
facilities associated with processes produce low pressure steam.
Process heaters are used extensively in refining
operations. They are the largest combustion source of hydro-
carbons in a refinery. The total process heater demand in a
modern refinery is approximately 270,000 Btu per barrel of
refining feed. Older, less efficient refineries may require
up to 600,000 Btu per barrel of refining feed.
Refinery 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 about half the fuel needs. Natural gas is
E-99
-------
used in the summer months and residual oil in the cooler months
when natural gas supplies go to preferred residential customers.
Emissions from boilers and process heaters depend on
the operating parameters of the unit and the fuel burned. Emis-
sion factors for burning natural gas and residual fuel oil are
given in Table E3-6.
TABLE E3-6. EMISSIONS FROM REFINERY BOILERS AND HEATERS
Pollutant
Fuel
Natural Gas
lb/106 SCF
Fuel Oil
lb/103 gal
Hydrocarbons (as CHi*)
Particulates
SO^ as SO2
CO
N0X as NO2
5-15
0.6b
17
120-130C
157S
5
60e
A function of fuel oil grade and sulfur content - For Grade 6: lb/10
gal - 10S + 3; For Grade 4: 7 lb/103 gal; For Grade 5: 10 lb/103 gal.
Based on average sulfur content of natural gas of 2,000 gr/10 Std Ft .
~S equals percent by weight of sulfur in fuel.
^Use first number for tangentially fired units, second for horizontally
fired units.
'Strongly dependent on the fuel nitrogen content.
Source: Reference 5
E-100
-------
In addition to the combustion emissions, there are
also emissions associated with the decoking of the heaters.
At intervals of about six months to three years, each heater
must be flushed with a steam-air mixture to remove the interior
coke deposits. Emissions are similar to those from decoking
the delayed coking unit, but they are smaller and more in-
frequent .
3.1.4 Vacuum Distillation
After the crude oil has been separated in an
atmospheric tower, the bottoms are transferred to a vacuum
distillation tower for further separation. Vacuum distillation
separates the heavy residue from the atmospheric distillation
into a heavy residual oil and one or more heavy gas oil streams.
This is accomplished at lower temperatures than would be
required at atmospheric pressure and avoids excess thermal
cracking of heavier materials.11
Vacuum fractionators are maintained at approximately
0.2 to 0.8 psia by either steam ejectors or mechanical vacuum
pumps.1 A major portion of the vapors withdrawn from the
column by the ejectors or pumps is recovered in condensers.
Approximately 50 pounds of noncondensable hydrocarbons are
emitted per 1,000 barrels of vacuum unit charge when barometric
condensers are used. This corresponds to approximately 18
pounds of hydrocarbon emissions per 1,000 barrels of refinery
feed.5 Quantities of noncondensable vapors as high as 130
pounds per 1,000 barrels of charge have been recorded.12
E-101
-------
3.1.5 Coking
Coking processes convert low value residual oil into
higher value gas oil and coke by cracking the residual oil
at a high temperature and atmospheric pressure. There are
two accepted methods for coking: fluid coking and delayed
coking. Delayed coking is the long standing, more widely
used method; fluid coking is expected to increase in importance
in the next few years. Each process has advantages and dis-
advantages relative to the other.
In the delayed coking process, the feed stream is
heated and transferred to a coke drum which provides the proper
residence time, pressure, and temperature for coking. When
the coke drum has been filled to capacity with coke, the
coke is cut from the walls with high-pressure water. Hydro-
carbons and particulates are emitted when coke is removed
from the drums.
Fluid coking is a continuous process in which the
feed is injected into a fluidized bed of hot coke particles.
The hot oil is cracked and a thin layer of new coke is
deposited on the particles. The coke particles travel to
a burning chamber where approximately one fourth of the
coke is burned for process heat. Since more coke is produced
than burned, a net coke product stream is produced. Approxi-
mately 30 pounds of carbon monoxide and about 520 pounds
of particulates per 1,000 barrels of feed are emitted from an
uncontrolled fluid coking unit.5'13 There are often additional
pollutants from coke combustion such as S0X, N0x, and organics.
E-102
-------
3.1.6 Air Blowing
Blowing air through a material may serve one of
several purposes: to oxidize, to remove moisture, to strip
spent chemicals, or to mix. The amount of emissions produced
by air blowing depends on the amount of air used per ton of
charge, the volatility of the charge, and the temperature of
the operation. In all of its uses,.uncontrolled air .blowing
produces noxious odors.
3.1.6.1 Asphalt Blowing
Air is sometimes blown through asphalt to oxidize it
and, therefore, increase its melting temperature and its
hardness. High temperatures are used to promote the oxidation.
The vent gases from asphalt blowing were the most
objectionable form of air pollution in a petroleum refinery
before control procedures became commonplace.3 The operating
conditions are favorable for the production of extremely unde-
sirable polynuclear aromatics.
Emissions from asphalt blowing are lessened by the
fact that asphalt material is distilled at high temperatures
before it is subjected to asphalt blowing. Available data
indicate uncontrolled emissions from asphalt blowing to be 40
to 80 pounds of hydrocarbons per ton of asphalt treated.12
3.1.6.2 Air Blowing of Gas Oils
Air blowing of gas oil products to remove moisture
takes place in a packed tower or vessel. Operating temperatures
E-103
-------
are kept low to minimize loss of the gas oil and to prevent
oxidation or degradation of the product. The exhausted gas
does, however, contain the lighter hydrocarbon components of
the gas oil.
3.1.7 Chemical Sweetening
Chemical sweetening rids hydrocarbons of odorous
mercaptans. Only low-sulfur (sweet) materials are subjected
to this treatment; more drastic sulfur removal methods such
as hydrodesulfurization are used for high sulfur (sour)
materials. The mercaptans may be removed from the material
(extractive sweetening) or converted to disulfides (oxidative
sweetening).
In extractive sweetening, an aqueous NaOH or KOH
solution extracts the sulfur. The spent caustic solution
may be regenerated by steam blowing or steam-air blowing, or
it may be disposed of. Before disposal, hydrocarbons are
removed from the solution by inert-gas stripping. This inert-
gas stripping may be a source of hydrocarbon emissions.
Catalysts are used to promote oxidative sweetening;
air is the oxidizing agent. Air is also used to regenerate
the catalyst. Hydrocarbon emissions may results from both
the oxidation and the regeneration steps.
3.1.8 Acid Treating
Hydrocarbon streams may be treated with acid to
remove or dissolve undesirable materials, or convert them to a
more desirable form. The stream is contacted with the acid
E-104
-------
and mixed thoroughly to form an emulsion. Air blowing is
sometimes used to agitate the mixture.
The use of sulfuric acid results in a hydrocarbon/
acid sludge which is removed by clay filtration. To recover
the acid, the sludge may be incinerated and the resultant S02
used to produce more sulfuric acid. Alternatively the
hydrolysis-concentration process may be used. In this process,
hot gases from the combustion of oil or gas are bubbled through
the sludge to volatilize the hydrocarbon dilutent and to
concentrate the acid. Off-gases pass through a mist eliminator
to the atmosphere. These gases may contain hydrocarbons and
S02 .
3.1.9 Blowdown
Periodic maintenance and repair of equipment is
essential to the refining operation. All units and equipment
subject to shutdowns, upsets, emergency venting, or purging
are manifolded into a multi-pressure collection system.
Discharges into the system are segregated according to
their operating pressures, then separated into vapor and
liquid cuts.
Because the blowdown system receives materials from
all processing units within the plant, any volatile material
found in any process stream may be emitted from an uncontrolled
blowdown system. The emission rate is a function of the
number of equipment items manifolded into the system and the
frequency of equipment discharges. It is estimated that 580
pounds of hydrocarbons are emitted from an uncontrolled blow-
down system per 1,000 barrels of refining feed.5
E-105
-------
3.1.10 Compressor Engines
Reciprocating and gas turbine engines fired with
natural gas or refinery fuel gas are often used in older
refineries to run high-pressure compressors. These units are
less reliable and harder to monitor than steam turbines or
electric motors. These factors, along with increased prices
of natural gas, will probably cause a decline in the use of
these engines.10
The exhaust emissions from these engines include
carbon monoxide, hydrocarbons, nitrogen oxides, aldehydes, and
depending on the sulfur content of the fuel, sulfur compounds.
Emission factors for reciprocating and gas turbine compressor
engines fired with natural gas are given in Table E3-7.
Particulate values were not available.
TABLE E3-7. EMISSION FACTORS FOR RECIPROCATING AND GAS
TURBINE COMPRESSOR FUELED WITH NATURAL GAS
Pollutant,
lb/103 ft3 gas burned
Engine Type
N0x as N02a ' CO
HC as Cb S0x
as S02C
Reciprocating
3.4 0.A3
1.4
2S
Gas Turbine
0.3 0.12
0.02
2S
aAt rated load. In general, N0X emissions increase with increasing load
and intake air temperature. They generally decrease with increasing
air-fuel ratios and absolute humidity.
^Overall less than one percent by weight is methane.
CS = Refinery gas sulfur content (lb/1,000 SCF): factors based on 100%
combustion of S + SO2.
Source: Reference 5
E-106
-------
3.2 Control Technology
Most of the process emissions described in Section
3.1 can be controlled. This section describes control methods
that are used now or that could be used by the industry.
A
"Existing" controls included in Section 3.2.1 are
those considered to be in relatively common usage. "Available"
controls included in Section 3.2.2 are those which have had
only limited application and those which have not yet been
applied in the refining industry. Controls which have been
used in other industries and which might be applicable for the
refining industry are included in Section 3.2.3.
3.2.1 Existing Levels of Control in Refineries
Existing controls for each process emission source
are discussed in this section. Because incinerators and flares
are applicable to several emission sources, they are discussed
in detail in Section 3.2.1.1 and 3.2.1.2.
3.2.1.1 Incineration
Incineration is an accepted method for disposal of
combustible wastes. It is most useful when the heat generated
during combustion can be used to offset other plant energy
needs.
Three types of incinerators used in refineries are
existing process heaters and boilers, thermal incinerators, and
catalytic incinerators. Particle collection devices are often
used on the flue gases from incinerators.
E-107
-------
Boilers Used as Incinerators—Fireboxes of boilers
and fired heaters can be potential incinerators (afterburners)
if the temperature, turbulence, and flame contact are adequate
to volatilize and burn the combustible contaminants. If the
waste volatile organics have appreciable heating value, the
firebox must be specially designed to take advantage of this
heat potential; such units are known as waste heat boilers.
If the heat content of the waste gas is low, common steam and
hot water heaters and boilers are used.
Successful adaptation of boilers for use as after-
burners is not common. The primary function of a boiler is
to supply steam or hot water. Its use as an air pollution
control unit may interfere with that function.
Satisfactory use of boilers as afterburners is
possible only if the following conditions exist:
• The boiler operates when the pollution
source is operated.
• Temperature, turbulence, and residence
time within the firebox are sufficient
for complete combustion.
• The air contaminants are wholly combustible
(otherwise, boiler efficiency and steam
generation may be reduced from deposits
on the process heater internals).
• The products of combustion are compatible
with the boiler construction materials.
E-108
-------
Thermal Incineration--Thermal incinerators are used
for most refinery waste gases. They usually eliminate more
than 95 percent of the organic vapor present.10 For combustion
of organic vapors and liquids, the concentrations of vapor
and air must be within the limits of flammability, termed the
upper and lower explosive limits (UEL and LEL, respectively).
These limits differ for various hydrocarbon compounds. When
concentrations of combustibles are less than LEL, supplemental
heat is required to initiate combustion.
An efficient thermal afterburner design must provide
for:
• An adequate residence time for completion
of the combustion process.
• Sufficiently high temperature in the after-
burner for the complete oxidation of the
combustibles.
• Adequate velocities to insure good mixing
without quenching combustion.
If combustion is inhibited by insufficient temperature, insuf-
ficient residence time, or poor mixing, then carbon monoxide,
aldehydes, and other products of incomplete combustion may
result'.
Burner type and arrangement affect combustion rates
and residence time. The more thorough the contact of flame
with the waste stream, the shorter the time required for com-
plete combustion. Burner placement depends not only on the
burner type, but also on the design requirement for intimate
E-109
-------
contact of the combustible gases with the burner flame.
Maximum efficiency occurs when all of the combustible matter
passes through the burner. Multijet and mixing-plate
burners provide the most effective flame contact. Maintaining
high turbulence or injecting steam also promotes the intimate
contact necessary for complete combustion.
Most refinery waste gases can be satisfactorily in-
cinerated at 1,150° to 1,400°F with a residence time of 0.2
to 0.7 seconds.14
In thermal afterburners, the organic vapor stream is
delivered to the refractory-lined burner area by either the
process exhaust system or by a blower. The combustible gases
are mixed thoroughly with the burner flames in the upstream
part of the chamber and then pass through the remaining part
of the chamber where the combustion process is completed.
Natural gas, LPG, distillate fuel oil, and residual
fuel oils are used to fuel afterburners. Oil flames are
longer than gas flames and thus require longer fireboxes. The
combustion of fuel oils produces sulfur oxides and particulates
which may cause corrosion and soot accumulation on afterburner
internals and heat transfer surfaces.
Heat recovery from the hot cleaned gases offers a
way to reduce the afterburner energy requirements at the
expense of increased equipment costs. The simplest application
is to use the hot cleaned gases exiting the afterburner to
preheat cooler process gases entering the afterburner. This
arrangement is termed primary heat recovery. In designing
E-110
-------
heat recovery systems, consideration should be given to
potential safety problems. Explosions or fires may occur
in the heat exchanger from process upsets or from accumulation
of flammable liquids or dust in the exchanger.
With secondary heat recovery, exhaust from the
primary heat exchanger supplies energy which can be used
elsewhere in the plant. Secondary heat recovery can have
a major effect on energy economics when used to supply steam,
hot water, or process heat to facilities located near the
incinerator. Optimum application occurs when heat utilizing
equipment is operated on the same schedule as the afterburner.
Catalytic Incineration--The use of catalytic incinera-
tion is not widespread in refineries.11* Cost of catalyst
versus cost of fuel must be considered along with possible
fouling of the catalyst by components of the waste gas. Very
dilute streams with foul odors are the most likely candidates
for catalytic incineration.
When a preheated gas stream is passed through a
catalytic afterburner, the catalyst bed initiates and promotes
oxidation. The combustion reaction occurs at a significantly
lower temperature than that of direct flame combustion; however,
care must be taken to assure that combustion is complete.
Catalytic afterburners offer the advantage of lower fuel costs
in some applications; however, the relative fuel savings
diminish as primary and secondary heat recovery are added to
thermal afterburners. Construction materials for catalytic
incinerators costs may also be lower because of the reduced
temperatures.
E-lll
-------
Combustion catalysts include platinum, platinum
alloys, copper chromite, copper oxide, chromium, manganese,
nickel, and cobalt. These are deposited in thin layers on an
inert substrate. Available substrate shapes include rods,
honeycombs, and ribbons designed to provide catalyst surface
A
area.
For.a catalyst to be effective, "active sites" upon
which the organic gas molecules can react must be accessible.
The buildup of condensed polymerized material or particles
prevents contact between active sites and the gases ("deactiva-
tion") . A catalyst can be "reactivated" by removing the
coating; cleaning methods vary with the catalyst. Deactivation
also occurs through reaction of the catalyst metal with
phosphorous, bismuth, arsenic, antimony, mercury, lead, zinc",
or tin. Gas streams containing these elements are best treated
by thermal incineration; Sulfur and halogens are also
"poisonous" to the catalyst, but their effects are reversible.
Catalyst material can be lost from the support by
erosion and attrition and by vaporization at high temperatures.
To protect the catalyst from overheating, volatile organic
concentrations are usually limited to 25 percent of the
lower flammability limit. Most combustion catalysts cannot
be operated at temperatures greater than 1,000 - 1,200°F.
Particulate Control Devices—Many kinds of wastes
other than refining gases may be incinerated, such as garbage,
contaminated and combustible liquids and semifluid sludges.
Flue gases from the combustion of these materials may contain
excessive amounts of particles which must be removed. The
methods commonly used to remove particles from flue gases are
cyclones and electrostatic precipitators.
E-112
-------
Cyclones - A cyclone consists of a cylinder with a
tangential opening through which the particle-
bearing gas enters at a high velocity, creating a
centrifugal force. As the particles strike the
walls of the cylinder, they slide down it. At the
bottom of the cyclone is the dipleg, essentially a
hollow pipe through which the particles exit the
cyclone.
Cyclones have no moving parts and use no fuel or
electricity. However, they cannot handle large
volumes of air effectively. When large volumes
of air must be processed, many small cyclones may
be used or air may be sent through several cyclones
in series.
Cyclones are most effective for particles larger than
40 microns. Efficiency drops drastically for smaller
particles: cyclones can collect 90 percent by
weight of particles larger than 10 microns, but only
60 percent by weight of particles smaller than 5
microns.15
Electrostatic precipitators - An electrostatic
precipitator is a device that removes particles from
a gas stream by giving them electrical charges
(from a discharge electrode) and then applying an
electrostatic force that causes them to move toward
electrically grounded collecting electrodes. There,
the fine particles are agglomerated and dislodged to
fall into a hopper below.
E-113
-------
As opposed to a cyclone, an electrostatic precipitator
can process very large volumes of gas (50,000 to
2,000,000 acfm) and very efficiently remove particles
as small as 0.05 micron.1"
A
However, electrostatic precipitators are very large
and their initial and total annual operating costs
are quite high. They are used only when essential
for meeting air pollution control regulations and
usually are installed in series after several high
efficiency cyclones.11*
3.2.1.2 Flaring
Flares are used most often as safety devices to dis-
pose of waste streams which cannot be economically recovered or
used as fuel. They are usually unsuitable for the treatment
of dilute gas streams because of the cost of supplemental fuel
needed to attain the minimum combustion temperature. They are
also generally less effective than other devices for controlling
organic vapors because of the considerable quantities of carbon
monoxide that may be produced.
Almost complete combustion of organic gases and vapors
can be achieved with a flare if the gas has sufficient heat
value to attain the minimum temperature necessary for combustion
and if it is adequately mixed with an adequate supply of air.
However, not all of the hydrocarbons are completely oxidized to
carbon dioxide and water. As much as ten percent of the
combustion mixture may be carbon monoxide.16
E-114
-------
N0X emissions from flares are also common due to
direct contact of nitrogen with oxygen at the flame temperature.
A typical emission rate for a flare system in a petroleum
refinery is 19 lb N0x/103 bbl refinery feed.5
Other air contaminants emitted from flares vary with
the composition of gases burned. Sulfur dioxide is produced
from the combustion of sulfur compounds such as hydrogen
sulfide in the flare gas. However, high toxicity and low
odor threshold make venting of hydrogen sulfide to a flare an
undesirable and sometimes dangerous method of disposal.
An insufficient air supply produces a smoky flame.
Within the reducing atmosphere of the smoke, hydrocarbons can
crack to elemental hydrogen and carbon or can react to form
polymers. Side reactions become more pronounced as the
molecular weight and unsaturation of the inlet gas increase.
Olefins, diolefins, and aromatics characteristically burn
with smoky, sooty flames as compared to paraffins and naphthenes.
There are three types of flares for the disposal of
waste gases: elevated flares, ground-level flares, and
burning pits.
Elevated Flares--An inert gas is injected into the
combustion zone of elevated flares to provide smokeless
operation. Steam is most commonly used for this purpose for
elevated flares. Steam injection provides the following
benefits:
• Energy available at relatively low cost
can be used to inject air and provide
turbulence within flame.
E-115
-------
• Steam and/or water react with the gas to
form oxygenated compounds that burn readily
at relatively low temperatures.
• Steam retards polymerization by reducing
the partial pressure of the fuel.
There are alternate methods for obtaining smokeless
operation in an elevated flare when the use of steam is uneco-
nomical or presents problems such as freezing in cold climates.
The alternative methods are compressed air injection, water
spray, high-pressure gas injection, air blower, and high
turbulence tips. In general, these methods are not used if
the use of steam is practical.
There are three types of elevated steam-injected
flares which vary with the manner in which the steam is in-
jected into the combustion zone.
In the first type, steam is injected by several small
jets placed concentrically around the flare tip. The jets are
installed at an angle and cause the steam to discharge in a
converging pattern immediately above the flare tip.
A second type has a flare tip with no obstruction
to flow. The flare tip is the same diameter as the stack. The
steam is injected by a single nozzle located concentrically
within the burner tip. In this type of flare, the steam is
premixed with the gas before ignition and discharge.
A third type is equipped with a flare tip that
promotes turbulence by causing the gases to flow through
E-116
-------
several tangential openings. A steam ring at the top of the
stack has numerous equally spaced holes for injecting steam
into the gas stream.
The injection of steam into a flare can be controlled
either manually or automatically. In some installations, the
steam is supplied at maximum rates, and manual throttling of the
steam is required for a particular gas flow rate. For the best
combustion with minimum steam consumption, instrumentation should
be provided which automatically controls the steam rate based
on the gas flow rate.
Ground-level Flares--A ground-level flare is usually
designed for daily process needs with the high flows during
major emergencies routed to an accompanying elevated flare.
Ground-level flares are of four principal types: horizontal
venturi, water injection, multijet, and vertical venturi.
A horizontal venturi flare system utilizes groups of
standard venturi burners. In this type of burner, the gas
pressure inspires combustion air for smokeless operation.
A water-injection flare consists of a single burner
with a water spray ring around the burner nozzle. Air is drawn
in as a result of the spray action and the water vapor provides
for the smokeless combustion of gases. Water is not as effect-
ive as steam for controlling smoke with high gas-flow rates,
unsaturated material, or wet gases.
A multijet ground flare uses two sets of burners, one
for normal gas release rates and both for higher flaring rates.
fe-117
-------
A vertical, venturi ground flare also uses commerical-
type venturi burners. This type of flare is suitable for
relatively small flows of gas at a constant rate.
Burning Pits--Burning pits are reserved for extremely
large gas flows caused by catastrophic emergencies ifi which the
capacity of the primary smokeless flares is exceeded. Ordinarily,
the main gas header to the flare system has a water seal bypass
to a burning pit. Excessive pressure in the header blows the
water seal and vents the vapors and gases to the burning pit
for combustion.
3.2.1.3 Sulfur Recovery
The Claus unit is the accepted method for sulfur
recovery 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 Claus .¦.units'. This '.dis-
cussion will consider first the Claus unit itself, then methods
for cleaning up the Claus unit tail gas.
More than 70 methods have been proposed for treatment
of the Claus unit tail gas.17 These methods may be continua-
tions of the Claus reaction or add-on processes with chemistry
quite different from that of the Claus reaction. Incineration
is sometimes used alone to clean Claus unit tail gas, sometimes
to prepare the tail gas for further treatment, and sometimes
after that treatment.
E-118
-------
The six tail gas clean-up methods discussed in this
section are those considered the most viable at present in
light of energy demands, economics, and effectiveness. The
first three processes--Amoco1s CBA Process, the Sulfreen Process,
and the IFP Process—are continuations of the Claus 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. These six methods are outlined briefly in
Table E3-8.
Because incineration is used with several of these
methods, it is discussed in the section "Incineration of Claus
Unit Tail Gas," immediately after the discussion of the Claus
unit itself.
The Claus Process—Because of its economic advantages,
a Claus unit for the conversion of H2S to elemental sulfur is
often considered as simply part of normal refining operations.
It is the source of the purest elemental sulfur. It is cited as
a major source of air pollution; but, it must be recognized as
a very effective control device.
The Claus process works best for gas streams contain-
ing greater than 20 volume percent H2S and less than 5 volume
percent hydrocarbons.1 There are several flow schemes avail-
able according to the H2S content of the feed stream of the
unit. In any case, the overall Claus reaction is as follows:
H2S + U- 5S +H20 (1)
l z n
E-119
-------
TABLE E3-8. EXISTING METHODS FOR REMOVAL OF SULFUR FROM CLAUS TAIL GAS
Name
Developer
Description
Final Tail Gas
S Concentration
Coat
Product (Z Cost of Claus)
CBA
Amoco
Sulfreen
IFP-1500
W
l
N3
BSRP
SCOT
Wellman-
Lord
SNPA/Lurgl
Instltut
Francals
du Petrole
Ralph M. Parsons
& Union Oil Co.
of California
Shell
Wellnan Power
Gas
Claus reaction continued
at low temperature; re-
moval of condensed sul-
fur drives reaction.
Bed regenerated with hot
gas from Claus unit.
Claus reaction continued
at low temperature as in
CBA. Bed regenerated with
hot nitrogen.
Claus reaction occurs In
a solvent.
All sulfur compounds re-
duced to HjS which Is
processed In a Stretford
unit.
All sulfur compounds re-
duced to H2S which Is
recycled to Claus
SO2 in incinerator gas
contacted with Na2SOj to
form NaHSOj. Na2S03 regen-
erated In evaporator/
crystalllzer.
S.
50-150*
1500 ppmv S
1500-2000 ppm S
1000-2000 ppm S
250 ppm S
or less
200-500 ppmv P.2S
<200 ppmv S0j
S.
S.
50-150Z
variable
Feed
to
Claus
Na2SO„/Na2S0j
crystals
100Z
75-100Z
130-150Z
-------
where n represents the various molecular forms of sulfur vapor.
The two most popular designs are illustrated in Figure E3-1.
In the "once-through" design, the incoming HaS-rich stream is
burned in a limited amount of air to convert one-third of the
H2S to S02 according to the following reaction:
2 h2 s + 2 o2 ->• SO2 + s + 2 h2 o (2)
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:
2 h2 s + so2 -»• 3 s + 2 h2 0 (3)
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:
h2 S + | o2 -»• so2 + h2 0 (4)
The remaining H2S is reacted over a catalyst with the hot gas
from the furnace to form elemental sulfur according to reaction
(3).
The "direct oxidation" design is for streams with
lower concentrations of H2S. In this design, the incoming
stream is preheated, mixed with air, and then passed over the
bauxite or alumina catalyst.
E-121
-------
tail gas
STEAM
MASH HEAT
BOILER
STEAM
FEED HATER
CONDENSERS
n L.P. STEAM
SULFUR PIT
LIQUID SULRJR
STRAIGHT THROUGH CLAUS PROCESS
CATALYTIC,"
REACTORS
L.P. STEAM
CONOENSERS
AIR*
SULFUR PIT
•LIQUID SULFUR
SPLIT FICU ClAUS PROCESS
70-1495-2
Figure E3-1. The Claus Process,
E-122
-------
Two side reactions produce relatively inert forms of
sulfur when any of the designs are used:
CH4 + 4S -+ cs2 + 2 h2 S (5)
a
CO2 + h2 s -»• cos + h2 0 (6)
The CS2 and COS are usually passed unchanged to the tail gas.
They can account for 0.25 to 2.5 percent of the sulfur content
of the tail gas.18 However, with proper design, including
the use of the cobalt molybdenum catalyst and a higher inlet
temperature for the first reactor, the CS2 and COS concentra-
tions in the tail gas can be minimized.'1
The choice of Claus unit design depends on the con-
centrations of H2S and other components such as hydrocarbons,
water vapor, and CO2 in the feed streams. For instance, the
"split-stream" design efficiently suppresses the undesirable
formation of COS when the incoming stream contains 30 percent
or more C02.19 Hydrocarbons can seriously affect the operation
of the unit.2 0
The Claus designs described above with one pass
through the catalyst reactor convert 80 to 86 percent of the
H2S to elemental sulfur.1'21 This efficiency can be greatly
enhanced by repeating the catalytic stage one or more times.
Two-stage Claus units can achieve 92 to 95 percent efficiency;
three stages, 95 to 96 percent; and four stages, 96 to 97
percent.5 Conversion is ultimately limited by the reverse
reaction. Recovery percentages for various feed compositions
are given in Table E3-9.
E-123
-------
TABLE E3-9. TYPICAL CLAUS PLANT SULFUR RECOVERY
FOR VARIOUS FEED COMPOSITIONS
Hydrogen Sulfide in Sulfur Calculated Percentage Recovery3
Plant Feed (dry basis) (%) Two Reactors Three Reactors Four Reactors
20
92.7
93.8
95.0
30
93.1
94.4
95.7
40
93.5
94.8
95.1
50
93.9
95.3
96.5
60
94.4
95.7
96.7
70
94.7
96.1
96.3
80
95.0
96.4
97.0
90
95.3
96.6
97.1
Assumes 1 mole percent hydrocarbon contamination, conventional temperature
and reheat techniques, average organic by-products and entrainment allow-
ance.
Source: Reference 20.
These efficiencies, once considered sufficient, do
not meet new regulations. Further treatment of the Claus unit
tail gas is discussed in succeeding sections.
Claus plant costs are sensitive to the flow rate and
composition of the input stream as well as the sulfur removal
efficiency. It is difficult to generalize the costs. As an
example, however, the capital investment costs for a Claus plant
having a capacity of 250 * 106 ft3/day of gas are $14 x 106
(construction period is 4th quarter 1979 through 4th quarter
1980). This plant has a sulfur removal efficiency of about
95 percent.
E-124
-------
Incineration of Claus Unit Tail Gas--The tail gas
from the Claus unit is often incinerated either before it
passes to the atmosphere or is subjected to further treatment.
This incineration takes place at temperatures of about 950 to
1,200°F or above in refractory lined vessels with one or
more burners.
An auxiliary fuel supply 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 100 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 not complete. Typical compositions of a sour
gas feed stream and the corresponding Claus tail gas before
and after incineration are given in Table E3-10.
The Amoco Cold Bed Absorption (CBA) Process17'2 2'2 3--
As stated in "The Claus Process" section, the completion of
the Claus reaction is ultimately limited by the reverse re-
action. The Amoco CBA process carries the Claus reaction
nearer to completion by condensing out the elemental sulfur
as it is formed. The Claus reaction is continued in a
separate reactor over the conventional Claus catalyst at
temperatures (260 to 300°F) below the dew point of sulfur.
Gas from the reactor is then incinerated.
The condensation of sulfur gradually reduces the
activity of the catalyst. There are two reactors so that one
can be regenerated while the other is operating. To regenerate
the catalyst, hot gas from the first Claus reactor is passed
through the bed. The sulfur is vaporized and condensed in
E-125
-------
TABLE E3-10.
TYPICAL COMPOSITIONS OF FEED STREAM AND
TAIL GAS STREAMS FROM A 94 PERCENT
EFFICIENT CLAUS UNIT AND INCINERATION
Component
Sour Cas Feed
Volume Z
Claus Tail Gas
Volume Z
Thermally Incinerated
Tail Gas
Volume Z
H2S
SO 2
Ss vapor
Ss aerosol
COS
CS2
CO
C02
02
N2
h2
H20
H.C.
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
0.85
0.42
0.10 as Si
0.30 as Sj
0.05
0.05
0.22
2.37
0.00
61. OA
1.60
33.00
0.00
100.00
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
Temperature, °F
104
284
752
Pressure, Fsig
6.6
1.5
Total gas volume
3.0 z feed
gas volume
5,8 x feed
gas volume
Gas volumes compared at standard conditions
Source: Reference 6
E-126
-------
REGENERATION CLEANUP
COOLING
CLEANUP
CLAUS TAIL GAS
CLAUS TAIL GAS
CLAUS FEED GAS
CBA
CBA
CBA
CBA
CONDENSER
CONDENSER
TO CLAUS
UNIT
TREATED GAS
TREATED GAS
START
START
SULFUR
SULFUR
Figure E3-2. The CBA Process
E-127
-------
a sulfur condenser. The hot gas is returned to the Claus
cycle just downstream of the first Claus sulfur condenser. The
flow diagram for a typical CBA unit is shown in Figure E3-2.
Construction and operation of a CBA unit are similar
to those for a conventional Claus unit; energy consumption is
low because no additional fuel is required. The process is
suitable for retrofit installations because only a small plot
area and minor modifications to the Claus unit are required.
Capital costs for adding a CBA unit to an existing
Claus unit are estimated to be equal to the cost of the Claus
unit. The CBA process is estimated to cost about half as
much as a standard three-stage Claus unit, when installed
at the same time as the Claus unit.
As of April 1979, there were three operational CBA
units. Amoco claims an overall efficiency of 98 to 99 percent
(Claus plus CBA). A 99 percent efficiency corresponds to
about 1,500 ppmv S02 in the incinerated tail gas. Levels of
COS and CS2 are not reduced by the CBA process; therefore,
formation of these compounds must be carefully minimized in
the Claus unit.
The SNPA/Lurgi Sulfreen Process1 7 '2 2 ' 2 3'211 - -When
the Sulfreen process was first introduced in 1968, six carbon
beds in parallel were used to absorb the elemental sulfur
formed in the Claus reaction and thus drive the reaction to
completion. Later developments have led to the use of fewer
standard Claus catalyst beds (alumina) with a lowered temp-
erature as in the Amoco CBA process. Gas from the reactor is
incinerated. The flow diagram of a typical Sulfreen unit is
shown in Figure E3-3.
E-128
-------
REGENERATION GAS IN
4—&
COOLING
GAS IN
REACTORS
D
TAIL
GAS IN
COOLING & REGENERATION
GAS OUT
ri/
FEED GAS
TAIL GAS
TO INCINERATOR
/ \
HEATER
CONDENSER
BLOWER
SULFUR WASHING TOWER
LIQUID SULFUR
70-1496-1
Figure E3-3. The SNPA/Lurgi Sulfreen Process
E-129
-------
In the Sulfreen process, hot gas, mostly nitrogen,
is blown through the beds to vaporize the sulfur and thus
regenerate the bed. The sulfur is then collected in a sulfur
condenser. The condensing sulfur generates 50 psig steam to
be used elsewhere in the plant. Reactors are cooled with un-
heated gas before being returned to service. To provide more
continuity, an improved flow scheme with three reactors is now
being offered. Exchange of the reactor loop from one reactor
to another is done automatically in 48-hour cycles.
Oxygen can cause a buildup of sulfates on the cata-
lyst. These sulfates decrease the effectiveness of the cata-
lyst. Because the regeneration gas in the Sulfreen process
is in a closed loop, oxygen in the regeneration gas can be
avoided and H2S can be added to reduce a major portion of the'
sulfates formed during absorption. The ability to attack these
sulfates and their formation is a principal characteristic
of the Sulfreen process.
As of April 1979, there were 19 Sulfreen plants in
operation and three under construction. They are located in
France, Canada, and West Germany. The original Sulfreen plant,
started up in 1970 and still using its original charge of
carbon catalyst, removes 75 percent of the SO2 from the Claus
unit tail gas. The Claus unit itself has a 91 percent ef-
ficiency for an overall efficiency of 97.8 percent.
The effectiveness of a Sulfreen unit is dependent on
the H2S and S02 concentrations in the feed gas to the unit.
At 15,000 ppmv H2S and S02 in the feed gas, 85 percent con-
version in the Sulfreen unit can be expected. This value
drops to 70 percent for 8,000 ppmv H2S and S02. For typical
Claus-Sulfreen installations, an overall theoretical conversion
of 99 percent is attainable with 2,500 ppmv H2S and S02 in the
E-130
-------
treated tail gas and 1,500 ppmv or less SO2 after incineration.
COS and CS2 levels are not affected by the Sulfreen process.
As in the case of the CBA process, construction and
operation are similar to those for a conventional Claus unit.
The facility is compact and can easily be retrofitted to an
existing sulfur plant. Though the Sulfreen process uses fuel
where the CBA process does not, it also produces more usable
steam. Capital and operating costs for the Sulfreen process
are comparable to those of the CBA process.
The Sulfreen process is attractive in large refineries
when it can eliminate the need for a third or fourth Claus con-
verter and reduce the required height and size of the tail gas
disposal site. It is also attractive in smaller refineries
where emissions of 1,500 - 2,000 ppmv S02 are allowable.
The Institut Francais du Petrole (IFP-1500)
Process17,2 3--The IFP-1500 is a liquid phase continuation of
the Claus reaction. It may be followed by an incineration unit.
With this process, S02 emissions from a two-stage Claus plant
can be reduced to 1,500 - 2,000 ppm. The process has no
effect on COS or CS2 emissions.
A schematic diagram of the basic IFP-1500 process
is given in Figure E3-4. In the packed tower, a low vapor
pressure polyethylene solvent which contains a proprietary
carboxylic acid salt catalyst in solution circulates counter-
current to the incoming tail gas. The catalyst complexes
with H2S and SO2 in an exothemic reaction; the complex then
reacts with more H2S and S02 to form elemental sulfur and
to regenerate the catalyst. Heat is removed by injecting and
vaporizing steam condensate to maintain a temperature of about
250 - 270°F. At this temperature, the sulfur is molten and
E-131
-------
TREATED GAS
TO INCINERATOR
-—2. STEAM CONDENSATE
¦2SOLVENT MAKEUP
STEAM FOR
START UP
PACKED TOWER
TAIL GAS FROM
CLAUS PLANT
LIQUID SULFUR
70-1497-1
PRODUCT
Figure E3-4. The IFP-1500 Process.
E-132
-------
there is little loss of sulfur or glycol. Treated gas is
incinerated, treated further or exhausted. Solvent is
recirculated.
Corrosion is not a problem in IFP-1500 units because
there is no water buildup. The entire unit can be made of
carbon steel. The unit can also go as long as two years be-
tween shutdowns. After about two years, the unit is washed
with water to flush out catalyst that has been converted to
sulfates and has deposited on the packing in the tower. No
extraordinary maintenance is required.
Because little plot area is required and there is
no recycle of gas to the Claus unit feed, an IFP-1500 unit
can easily be retrofitted to an existing Claus unit.
Of all Claus tail gas clean-up methods which can
reduce S02 levels to 1,000 - 2,000 ppmv, the IFP-1500
process is probably the simplest in operation and design
and requires the lowest capital investment. It can be
economically adapted to even the smallest of Claus units. As
of April 1979, there were 25 IFP-1500 units in operation,
under construction, or being designed.
The Eeavon Sulfur Removal Process23'2 5'26--The
Beavon Sulfur Removal Process (BSRP) is one of the most com-
monly used designs for controlling the emissions from a Claus
unit. Thirty-six BSRP units were in operation or under
construction as of April 1979. In this process, the sulfur
compounds in the tail gas are converted to H2S and recovered
as elemental sulfur by the Stretford process. The BSRP pro-
cess is illustrated in Figure E3-5.
E-133
-------
CLEAN GAS TO
•ATMOSPHERE
HYDROGENATED
COOLED TAIL GAS
REACTOR
AIR
FUEL GAS
TO H-S RECOVERY
SCRUBBER
STRETFORD SOLUTION
STRETFORD ABSORBER
SULFUR PLANT
TAIL GAS
COOLER
OXIDIZER
SULFUR
LIQUOR RETURN
FILTER
Q
SULFUR
70-1500-1
Figure E3-5. The BSRP Process.
-------
The tail gas is reduced over a cobalt-molybdenum
catalyst in a reducing gas atmosphere. Temperature and pres-
sure are moderate. Sulfur plant tail gas usually contains
more than enough hydrogen (about 2.5 percent) for hydrogenation.
Sulfur dioxide and elemental sulfur are reduced to H2S by the
following reactions:
S02 + 3H2 -* 2 h2 0 + h2 s (1)
S2 + 2 H2 -* 2 H2 S (2)
The catalyst also reacts water vapor with CO, COS, and CS2 in
the following hydrolysis reactions:
cos + h2 0 -* h2 s + CO2 (3)
cs2 + 2 h2 0 -+ 2 h2 s + co2 (4)
co + h2o -* h2 + co2 (5)
After hydrogenation and hydrolysis, the stream is cooled by
water scrubbing. The vapor leaving the cooling unit contains
about 2-4 percent H2S, 12 percent CO2, and the rest mainly
nitrogen.
The H2S-rich stream from the Beavon unit is then sent
to the H2S absorber column of the Stretford unit where it is
contacted with the "Stretford Solution" of sodium carbonate and
sodium vanadate. The H2S is absorbed by the sodium carbonate
and precipitated by the sodium vanadate:
H2S + NA2C03 -* NaHS + NaHC03 (6)
NaHS + NaHC03 + 2 NaV03 S + Na2V2Os + Na2C03 + H20 (7)
E-135
-------
Complexing agents such as sodium potassium tartrate or citric
acid are sometimes used in the absorber to prevent deposition
of vanadium in systems operating beyond their capacity; solu-
bilized iron with Bellasol S.C.S. or EDTA may be used to speed
up the reoxidation of some unwanted colored by-products.
The treated gas from the absorber is odorless and
essentially free of COS and H2S. Sulfur compounds, mainly COS,
are generally equivalent to less than 100 ppm SO2. No inciner-
ation of this stream is required during normal operation, but
provisions must be made for upsets.17
The liquor from the absorber is sent to an oxidizer
where air is blown into the stream and the sodium vanadate is
regenerated by sodium anthraquinone disulfonate (ADA):
Na2V2Os + | 02 2 NaV03 (8)
The finely divided sulfur forms a heavy froth which is skimmed
off, filtered, and washed. Reoxidized solution is pumped to
the absorber.
The total investment for a BSRP unit is approximately
equal to that of the parent Claus unit. Normally, the BSRP
process should be considered for larger (>100 lt/day) Claus
plants. If adequate space is available, the process is
suitable for retrofitting since no streams are recycled to
the Claus unit.
The Shell Claus Off-Gas Treating (SCOT) Process6'17'
23,27--The SCOT process, like the BSRP, entails the reduction
of all sulfur compounds in the Claus unit tail gas to H2S. It
is shown in Figure E3-6. In a manner similar to that of the
E-136
-------
CLAUS UNIT
OFF G/\S
LEAN AMINE £
M
i
i-»
co
-vl
REDUCING GAS
FUEL GAS?
HEATER
COOLING
TOWER
REACTOR
STEAM
TO CLAUS UNIT
INCINERATOR
-------
BSRP, the tail gas, along with a reducing gas, is passed over
a cobalt-molybdenum catalyst. Sulfur dioxide, elemental sulfur,
COS and CS2 are converted to H2S according to reactions (1),
(2), (3), and (4) of the section "The Beavon Sulfur Removal
Process." The gas which leaves the reactor is cooled, and the
water is condensed.
The cooled gas, which normally contains up to 3 vol-
ume percent H2S and up to 20 volume percent CO2, is then
scrubbed with an alkanolamine solution in an absorption column.
The absorber is designed so that H2S is almost completely
absorbed while relatively little (30 percent) of the C02
present in the absorber feed gas is absorbed. The overhead
stream from the absorber, which contains 200 - 500 ppmv H2S,
is routed to the Claus incinerator.
The rich amine solution from the absorber is then
steam-stripped to release H2S and C02; steam is removed by
condensation. The top gas containing the H2S and C02 is re-
cycled to the front end of the Claus unit.
Because of the recycle of CO2, the SCOT process is
not suitable for streams rich in C02 and low in H2S. It is
also sensitive to the H2S/SO2 ratio in the Claus unit tail gas.
It is, however, considered one of the most flexible processes
available for Claus tail gas cleanup with respect to turndown
capability. It can be operated at from 20 - 100 percent of
design feed ratio.
Nearly all mechanical equipment for a SCOT unit can
be fabricated of carbon steel. Minimum pressure drop can be
designed into the system to facilitate ultra-filtering. If a
E-138
-------
SCOT unit is retrofitted, the cost is about equal to that of
the parent Claus unit. A SCOT unit can be incorporated into
a new facility for about 75 - 80 percent of the cost of the
Claus unit.
As of April 1979, there were more than 35 SCOT units
in operation in the United States and Canada. More than 25
others were in design, construction, or commissioning around
the world. Claus units using SCOT units were as large as
2,100 lt/day and as small as 3 lt/day.
The Wellman-Lord (W-L) Process17'2 3'2 7'28--The
Wellman-Lord process is applied to the cooled fuel gas from
the Claus plant incinerator. It can also be used to desulfur-
ize a number of other types of flue gases. This process is
illustrated in Figure E3-7.
In the Wellman-Lord process, the S0 2~rich gas is
contacted with a sodium sulfite solution in an absorber.
Sodium bisulfite is formed according to the following reaction:
S02 + Na2S03 + H20 -~ 2 NaHS03 (1)
Some sodium bisulfite is oxidized to sodium sulfate:
Na2 S03 + | 02 -~ Na2 S04 (2)
Desulfurized flue gas (about 200 ppmv S02 or less) is released
to the atmosphere. The absorber solution is fed to an evapor-
ator where SO2 and steam are evaporated. The SO2 can be fed
to the Claus plant or to a sulfuric acid plant.
E-139
-------
f
Net S02 to
Claus or
sulfuric
acid plant
CONDENSER
AIR
Dry SOj
I I DRYER
J (optional)
CLEAN GAS
TO STACK
EVAPORATOR/
CRYSTALLIZER
CENTRIFUGAL
SEPARATOR
NaOH Makeup
Makeup HgO
Steam,
HOLDING
TANK
Mother
liquor
return
DISSOLVING
TANK
A8S0RBER
Purge
Flue gas
Figure E3-7. The Wellman-Lord Process.
-------
About 15 percent of the effluent from the evaporator
is routed to a purge treatment area where a mixture of sodium
sulfite and sodium sulfate is crystallized out of solution.
The sodium sulfate/sodium sulfite crystals are dried for sale
or disposal. Sodium lost during the purge operation is
replenished by the addition of sodium hydroxide to the regen-
erated liquor.
Sodium loss through sulfate formation is a major
concern. This loss can be minimized by maintaining a high
sulfite concentration in the absorber. In one installation,
a proprietary chemical has been added to the absorber to halve
sodium loss. Fractional crystallization of the purge steam
can increase the sulfate content of the purged solids to 70
percent.
Because the Wellman-Lord process treats the tail gas
after incineration, the H2S/S02 ratio from the Claus unit is
not critical. Streams high in C02 are well suited to the
process because C02 is not recycled.
Cost of a Wellman-Lord unit is high, however, partly
because of the corrosion-resistant metals required. For a 100
lt/day Claus plant, a Wellman-Lord unit would cost about 130 -
150 percent of the cost of the parent Claus unit. Cost penal-
ties are particularly severe for smaller installations; the
Wellman-Lord process probably should not be considered for
Claus plants 50 lt/day or smaller.
As of April 1979, there were seven Wellman-Lord units
being used for Claus tail gas cleanup in the United States and
Japan. Nineteen others were treating other sulfur-bearing flue
gases.
E-141
-------
Costs--The actual costs of sulfur recovery processes
are very sensitive to the flow rate and composition of the
input gas stream. Some processes are available in capacities
ranging from skid-mounted units to full scale process units.
Most of the cost information available in the literature is
three or four years old. Changes in process technology have
probably changed these costs to a greater extent than the
effects of overall process cost escalations and some costs
may have decreased. Factors other than costs or energy re-
quirements may be responsible for the selection of the
appropriate process for a particular installation.
3.2.1.4 Catalyst Regeneration
FCCU Catalyst Regeneration—Regeneration of the cata-
lyst in fluid catalytic cracking units (FCCU's) produces three
principal types of atmospheric emissions: S0x particulates,
and CO. Lesser emissions include hydrocarbons, N0x, aldehydes,
and ammonia. S0x is typically controlled by feedstock de-
dulfurization; particulates by cyclones and electrostatic
precipitators; and CO by a CO boiler. No single process can
control all three.
SQx Emissions - Hydrodesulfurization (HDS) of feed
stock 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 feedstocks increases with the
degree of desulfurization. The result is that very
high levels of hydrodesulfurization are needed
to achieve 90 percent of higher reduction of S0x
emissions. For a feedstock with 2.3 weight percent
E-142
-------
sulfur, 92 - 95 percent desulfurization of the
feed is necessary for a 250 ppm S0X concentration
in the flue gas.29 A more complete description
of HDS processes is given in Section 4.4 of
Appendix F.
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 precipitation to remove these
catalyst fines from the flue gas. Cyclones and
electrostatic precipitators are discussed in Section
3.2.1.1 under the headings "Cyclones" and "Electro-
static Precipitators."
The collecting efficienty of ESP's for catalyst fines
is commonly 90 - 99 percent of the particles that
escape the cyclones. 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
"conventional" regeneration burns off the coke from
spent catalyst to, roughly, a 50 - 50 mixture of
CO and CO2.
E-142
-------
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.30 The
flue gas goes to the furnace of a CO boiler and
external heat is applied to raise the temperature
high enough (~1,300°F) to achieve 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
TCCU 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 regenerators may also
be incinerated in a process heater or flared, but use of these
control methods is not widespread because these emission
sources are small.
3.2.1.5 Boilers and Process Heaters
No pollution control devices for refining boilers
and process heaters are in current use.31 Combustion modifi-
cation is discussed in Section 4. When low emissions are
E-144
-------
required, only clean fuels are burned and high-efficiency
burners are used. It should be recognized that these boilers
and process heaters themselves can act as control devices
when they are used to incinerate waste process streams.
3.2.1.6 Vacuum Distillation10
Emissions of noncondensable vapors from vacuum
distillation are controlled by venting into a blowdown
system or by incineration. The vapors may be used as sup-
plemental fuel in process heaters and boilers. Oily conden-
sate emissions can be eliminated by the use of mechanical
vacuum pumps or surface condensers which discharge to a
closed drainage system. Both noncondensable and condensable
emissions can be minimized by the installation of a lean-oil
absorption unit between the vacuum tower and the first stage
vacuum j et.
These control methods virtually eliminate process
emissions from vacuum distillation.
3.2.1.7 Coking
Particulate emissions from the delayed coking
process can be minimized by wetting down the coke during the
coke removal procedure. Hydrocarbon emissions can be mini-
mized by venting the quenching steam to a vapor recovery or
blowdown system. Once the drum has cooled to 212°F, the steam
purge can be replaced by a water flood. Further cooling to
approximately ambient temperatures will minimize steam and
hydrocarbon vaporization when the drum is opened.
E-145
-------
Cooling to ambient temperatures and using large
amounts of water to reduce emissions from delayed coking units
are not practical in many instances. Cooling the coke to
ambient temperatures before removing it from the drum increases
the chance that the coke can "set up." If it sets up, removal
may be possible only by tedious, time consuming methods.
Excess water on the coke once it has been removed produces
a water pollution problem, could produce an off-specification
green coke, could plug coke handling systems in the calciner
feed section, and would upset calciner operations at those
installations producing premium calcined coke.
Significant particulate emissions often occur during
the loading of calcined coke into rail cars or trucks. An
induced draft particulate control system using bag filters
could reduce these emissions, but would be expensive to
design, install, and maintain. A more reasonable approach is
to spray the combined coke with a small amount of oil (heavy
cycle oil or coker gas oil) as it leaves the cooler.
Fluid coking units can emit particulates and carbon
monoxide. The off-gas from the coker can be controlled by
incineration in process heaters or CO boilers. In some cases,
the CO boiler used is the same one that controls CO emissions
from the fluid catalytic cracking unit. Particulate emissions
from the CO boiler are controlled by the methods described in
Section 3.2.1.4.
E-146
-------
3.2.1.8 Air Blowing
In many refineries, air-blown brightening units have
been replaced with packed vessels containing solid adsorb-
ents.10 These units have slight potential for process hydro-
carbon emissions.
Emissions from asphalt blowing can be reduced by
vapor scrubbing, incineration, or a combination of both.
Vapor scrubbers condense steam, aerosols, and essentially all
of the hydrocarbon vapors. A disadvantage in water scrubbing
is the high-volume ratio of water-to-exhaust gas required to
remove the hydrocarbons. When an adequate water supply is not
available or when condensate handling may result in hydrocarbon
emissions, incineration of the vapors by direct flame contact
may be used. Incineration may be accomplished in process
heaters, boilers, or flares.
Hydrocarbon emissions from a controlled asphalt-
blowing unit are negligible.32
3.2.1.9 Chemical Sweetening
Emissions from the inert gas stripping of spent
caustic can be prevented by venting the gases to a flare
or a furnace firebox.
Emissions from air blowing regeneration of spent
oxidative sweetening solutions can be reduced by steam-
stripping the spent solutions to recover hydrocarbons before
air-blowing. The gaseous effluent from air blowing can be
incinerated to dispose of residual hydrocarbons.
E-147
-------
3.2.1.10 Acid Treating
If the acid concentration process is used, the off-
gases from the mist eliminator can be vented to caustic
scrubbers for S02 and odorant removal, and then to an incin-
erator or a flare.
Hydrocarbons escaping from acid recovery operations
can be eliminated by using acid regeneration. Regeneration
involves sludge incineration to produce SO2, which can be
converted to H2SCU. Some of the SO2 control methods discussed
in Section 3.2.2 are available for control of SO2 emissions
from acid sludge incineration.
3.2.1.11 Blowdown
Blowdown emissions can be effectively controlled
by venting into an integrated vapor-liquid recovery system.
All units and equipment subject to shutdown, upsets, emergency
venting and purging are manifolded into a multi-pressure
collection system. Discharges into the collection system are
segregated according to their operating pressures. A series
of flash drums and condensers arranged in descending operating
pressures separate the blowdown into vapor pressure cuts. The
liquid cuts are recycled to the refinery; the gaseous cuts
are recycled or flared.
Emissions from a controlled blowdown system have been
estimated to be 0.8 lb per thousand barrels of refinery feed;
compared to 580 lb per thousand barrels for uncontrolled. The
control is estimated to be 99.9 percent effective.5
E-148
-------
3.2.1.12 Compressor Engines
No pollution control devices for refinery compressor
engines are in current use.5 Combustion modification is dis-
cussed in Section 4.
3.2.2 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 technologies is often limited.
The methods chosen for inclusion in this section are
those which have been proposed for consideration by the
industry. Inclusion of a method here indicates only that it is
worthy of consideration, not that it is necessarily a good
choice for any particular application.
3.2.2.1 Particulate Control for Incinerators
Filters—There are many kinds and several types of
filters that may be used for particle removal. The types of
filters include flexible sheets, tubes, or bags of porous
material; supported semirigid fabrics or nonwoven fibrous
mats; and fixed or fluidized beds of dry granules or particles.
The filters used to make fabric or mat filters may
be wool, cotton, dacron or other synthetics, asbestos, fiber-
glass, or metal. These materials may be used in their natural
state or they may be treated to give them an electrostatic
charge or to make them adhesive. Fixed or moving beds may be
composed of ceramic, sand, coke, crushed rock, or other
materials.
E-149
-------
Filters can often remove more than 90 percent of the
particles on micron or larger in size. But fabric filters are
not generally used for gases containing more than 10 to 12
grans per cubic foot. They also cannot be used if there is
any kind of condensate in the gas stream.
It is not usually necessary to have a filter with a
high initial efficiency because the efficiency increases as
particles collect on the filter. There is an accompanying
increase in the pressure drop, however, and the filter is
cleaned or replaced when this pressure drop becomes unaccept-
able.
Wet Collectors--Wet collectors are divided into two
categories according to the particle size they collect most
efficiently. Those which collect particles down to one micron
in size include spray towers, centrifugal spray scrubbers, and
self-induced spray scrubbers. Disintegrator scrubbers, venturi
scrubbers, and foam scrubbers collect particles less than one
micron in size.
In spray towers, droplets are formed by atomizing
sprays and are collected by gravity. These droplets in a
centrifugal spray scrubber are also formed by an atomizing
spray, but are collected by centrifugal force. In a self-
induced spray scrubber, gas breaks up the liquid into droplets.
To form droplets in a disintegrator scrubber, liquid
is sheared as it passes between a rotor and a stator. The
liquid is dispersed by high velocity gas in the venturi
scrubber. Sieve trays produce foam in foam scrubbers.
E-150
-------
Wet collectors may be used on gas streams where
addition of liquid to the gas streams is not objectionable.
The gas is cooled and vapors or gaseous contaminants removed.
Particles with diameters ranging down to 0.5 micron are col-
lected with a moderately high efficiency. Particles less than
0.5 micron are collected at a lower efficiency.
When wet collectors are used, there is a reduced
danger of explosion, but there is an added potential for
corrosion and freezing. Disposal of the sludge is also a
problem.
Wet collectors are not usually advised when a less
expensive device, such as a cyclone or a centrifugal separator
will be satisfactory. Wet collectors are about as costly as
filters such as cloth bags, but are preferable to filters be-
cause they are less affected by high temperatures and high
loads. Their efficiency is not generally as good as that of
a filter, however.
3.2.2.2 Sulfur Recovery
The control of emissions from the Claus unit is a
subject of great concern in the refining industry. A number
of control methods have been proposed in recent years. Some
are applicable only to Claus tail gas while others may be
applied to other sulfur-bearing streams as well. Also being
tested are an alternate to the Claus unit and a modification .
of the unit. The alternate would produce no objectionable
tail gas stream.
Most of the tail gas treatment methods included in
this section are described briefly in Table E3-11.
E-151
-------
TABLE E3-11. AVAILABLE METHODS FOR REMOVAL OF SULFUR FROM CLAUS TAIL GAS
Efficiency or
Outlet Concentration
(Claus + Addition)
Nil
Developer
Description
Product
Cost
(Z Cost of Claus)
IFP-150
Instltut
FrancaIs
du Petrole
Gas from IFP-1500 scrubbed
with ammonia; SO2- laden
ammonia mixed with HjS In a
glycol to form elemental
sulfur and water.
<200 ppm SOj
S.
variable
Clean air Pritchard
Trencor-M Trenthaa
Claus reactors operated at
high temperature to reduce
COS & CS2 levels; S02 + el-
emental sulfur removed by
aqueous scrubbing; II2S re-
moved by Stretford process.
All sulfur compounds reduced
to HaS. H,S absorbed by
amine solution & returned to
Claus.
50 ppmv SO,
100-200 ppmv SO2
S.
100Z
feed
to
Claus
150Z
Aqua Claua Stauffer
Sulfoxide
Topaoe
SFGD
Alberta
Sulfur
Research, Ltd.
SNPA/Topaoe
Shell
SO2 from Incineration mixed
with H2S-rlch Claus feed; Claua
reaction occurs In aqueous
phase.
Claus reaction occurs in an
organic sulfoxide medium
SOi from incineration ox-
idized to SOj which is con-
verted to HjSOi,.
SO; in gas from incineration
absorbed by CUO bed which is
regenerated with hydrogen.
<100 ppmv SO2
<1000 ppmv S
typically
<500 ppmv S
97Z
97Z
S.
NajSOii
S.
H2S0»
Feed to
Claus
125-135Z
Not
available
Not
available
250Z
Continued
-------
TABLE E3-11. (Continued)
Nmm
Developer
Description
Efficiency or
Outlet Concentration
(Claus + Addition)
Product
Cost
(X Cost of Claus)
Westvaco
Westvaco
SOj In gas from lncln-
eratIon removed and
catalyzed to HjSOi, In
activated carbon bed which
Is: regenerated with H2S.
<200 ppmv SO2
S0 or feed
to Claus
Not available
Awonlua
Blsulfate/
Aaron lua
Thlosulfate
Prltchard
SOj In gas from Incin-
eration absorbed in
aqueous ammonia & con-
verted to ammonium
thlosulfate.
<900 ppmv SO2
Ammonium
thlosulfate
75Z
BSR/
Selectox I
Union Oil
All sulfur compounds re-
duced to H2S which 1b then
oxidized to S,.
>98Z
s.
£200Z
Limestone
Slurry
Mineral 6
Chemical
Resource Co.
SO2 In Incinerator gas
absorbed by limestone
slurry.
>99.9Z
CaSOj/CaSO*
limestone
bo lid
Not available
Catalytic
Incineration
Instltut
Francals
du Petrol*
Catalyst promotes
oxidation of sulfur
compounds to SOj In
incinerator.
<200 ppmv S
SO 2
Not available
-------
The UOP Sulfox Process33--The UOP Sulfox process is
an alternative to the Claus process. In this process, aqueous
ammonia, instead of an amine solution, is used to scrub H2S
from refinery streams. 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 + | o2 -»• So + h2 0 + nh3 (1)
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 bed 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.
In each reactor, some sulfide is oxidized to thio-
sulfate. A minor portion of the thiosulfate is then oxidized
to sulfate. The operating conditions and the catalyst are
chosen to minimize sulfate formation; any sulfate formed (0.1
to 0.5 percent of the sulfur) is discarded as ammonium
sulfate.
Ammonia from the reactors can be recovered and reused
as absorbent, or it can be incinerated to nitrogen and water
vapor. Ammonia in the original feed can be recovered as a
product by distillation from a slip-stream of the sulfide-
free effluent from the oxidizer.
E-154
-------
Most of the effluent from the final oxidizer is used
directly as absorbent solution. The remainder is carried to a
reducer where the thiosulfate is reduced to sulfur by reaction
with a portion of sulfide-rich solution which bypasses the
oxidizer. This reaction is noncatalytic.
The tail gas from the oxidizers is scrubbed free of
ammonia with water. Hydrogen sulfide content in the treated
gas is 10 to 100 ppm. It is possible, at increased cost, to
design a Sulfox unit which can achieve 1 ppm H2S.in the tail
gas. However, NSPS requires less than 250 ppmv S02 from a
final oxidizing step, which in this case would probably be
interpreted as, "S02 or its equivalent as reduced sulfur
compounds."
It may be possible to convert an existing Claus
system to a Sulfox System with a minimum of expense. It is
probable that the existing amine absorber could be used as
the ammonia absorber and that the existing amine stripper
could be used in the Sulfox unit proper.
Cost of a Sulfox system is considered about equal to
that of a Claus unit, not including the cost of tail gas
cleaning. Utility costs are estimated to be about 60 percent
of those of a Claus unit.
The Mineral and Chemical Resource Company (MCRC)33--
The MCRC Sulfur Recovery process is actually a modified improve-
ment of the Claus process. A proprietary scrubber is used to
improve sulfur recovery and also to remove any ammonium sul-
fate which forms in a Claus unit if the feed contains ammonia.
•E-155
-------
The feed stream flows first into a special reaction
furnace where a carefully controlled quantity of air oxidizes
a portion of the H2S to SO2. The standard Claus reactions then
occur. The hot gas from the furnace flows through a waste heat
boiler to a condenser where elemental sulfur is condensed and
withdrawn. The remaining gas flows to a converter where ad-
ditional reaction takes place and then to another condenser
for further sulfur removal. It may pass through several more
converters and condensers until the desired sulfur removal is
accomplished.
A 98 percent sulfur recovery efficiency can be
obtained with a three converter design; greater than 99 percent
efficiency can be obtained with four converters. Two MCRC
sulfur recovery plants have been operating since 1976.
The Institut Francais Du Petrole (IFP-150) Process3 **--
To increase the effectiveness of the IFP-1500 process, it can
be coupled to an ammonia scrubbing section and then an IFP-150
unit. Ammonia scrubbing can reduce the sulfur dioxide content
in the tail gas to below 200 ppm. The IFP section converts the
ammonia scrubber liquor, mostly ammonium sulfites, to elemental
sulfur and ammonia. The ammonia is recycled to the scrubber.
The liquor from the ammonia scrubber is first heated
in a forced-circulation evaporator where sulfites are decomposed
to ammonia and sulfur dioxide. Sulfates and thiosulfates in the
liquor are reduced to ammonia, sulfur dioxide and water with a
proprietary IFP process.
The ammonia-sulfur dioxide stream and hydrogen
sulfide-rich gas are then fed to the IFP reactor at a 2:1
ratio of hydrogen sulfide to sulfur dioxide. In the reactor
E-156
-------
the hydrogen sulfide and sulfur dioxide dissolve in a high-
boiling-point glycol to form elemental sulfur and water. The
reaction temperature is maintained slightly above the melting
point of sulfur and liquid sulfur is drawn off.
The IFP-150 process can be used for various other
flue gases and acid or sour streams in a refinery to maintain
S02 emission rate of less than 200 ppm at the stack mouth.
Scrubbers can be set up at many points and their effluents
piped to a central IFP-150 reaction section. Total sulfur
recovery using the IFP-150 process in a refinery is shown
in Figure E3-8.
The Pritchard Cleanair Process26'33--The Cleanair
process is similar to the BSRP in that final cleanup is in a
Stretford unit. A Cleanair unit can recover 99 percent of the
sulfur in the Claus tail gas, leaving no more than 50 ppmv
S02 equivalent in the effluent.
The first step in the Cleanair process is the
reduction of COS and CS2 levels by operating the Claus
reactors at elevated temperatures. Sulfur dioxide and
elemental sulfur are then removed from the tail gas by aqueous
scrubbing. A Stretford unit then removes the hydrogen sulfide.
The plant cost for a Cleanair unit is roughly equiva-
lent to that of a Claus unit and requires about the same amount
of space. Per daily ton of sulfur produced in the Claus tail .
gas, 4 kwh per hour, 8 lb steam per hour, 9 gpm water for
cooling, and $2.40 to $3.00 per day for chemicals are required.
E.-157
-------
TO STACK
FUEL GAS
FUEL GAS
INCINERATOR
FURNACE GAS,
•3CLAUS TAIL GAS,
BOILER OFF-GAS
INCINERATOR
SCRUBBER
nh4oh
STORAGE
RESIDUES & TARS
NH, MAKEUP "£¦
Z FUEL GAS
INCINERATOR
SCRUBBER
CATALYTIC
REACTOR
SPENT SODA
MIXER-
SEPARATOR
Na /NH4 ION
EXCHANGE
n«/nh4
ION
EXCHANGE
tn
CVJ
, ACID SLUDGE A
Sodium sulfates
PRODUCT
SULFUR
CAUSTIC
ft SODIUM
CARBONATES
CAUSTIC & SODIUM
CARBONATES
BRINE
STORAGE
STRIPPER
REDUCER
SOUR WATER
EVAPORATOR
70-1499-1
SWEET
WATER
AMMONIUM SULFATESt
Figure E3-8. Total Sulfur Recovery with the IFP-150 Process.
-------
As of April 1979, there was one Cleanair unit
providing a guaranteed 300 ppmv sulfur dioxide equivalent.
Two other plants were under construction.
The Trentham Trencor-M Processor27--The Trencor-M
process is similar to the SCOT process. All sulfur species
in the tail gas are converted to H2S over a cobalt-molybdenum
catalyst. A proprietary aqueous solution of methyldiethan-
olamine (MDEA) with small amounts of inorganic salts absorbs
the H S from the stream. The vent gas is incinerated. When
the amine solution is regenerated, the H2S-rich stream is
recycled to the front of the Claus plant.
The Trencor-M process can reduce S02 levels to 100 -
200 ppmv. The investment cost is estimated to be 159 percent
that of the parent Claus plant. As of mid-1978, there was
one known commercial Trencor-M plant.
The Stauffer Aquaclaus Process22'35--The Aquaclaus
process treats the Claus tail gas after it has been incinerated.
The stream from the incinerator is first cooled in a waste-
heat boiler and/or a direct-contact cooler. The S02 is then
absorbed from the solution by the Aquaclaus solution (aqueous
sodium phosphate) in an absorption tower. Off-gas from the
absorber can contain less than 50 ppmv S02.
The S02-rich absorbent from the tower is then mixed
with H2S feed from the front of the Claus unit. Elemental
sulfur is formed by the Claus reaction in aqueous phase.
Liquid sulfur is withdrawn; the Aquaclaus solution is returned
to the absorber.
E-159
-------
The S02-rich absorbent from the tower is then mixed
with H2S feed from the front of the Claus unit. Elemental sul-
fur is formed by the Claus reaction in aqueous phase. Liquid
sulfur is withdrawn; the Aquaclaus solution is returned to the
absorber.
A purge stream is required for control of the levels
of undesirable polythionates, thiosulfates and sulfates formed
in the absorber and the Claus reactor.
An Aquaclaus plant which would recover 20-25 lb/d sul-
fur would cost about 125-135 percent of the cost of the parent
Claus plant. As of mid-1978, there were no commercial instal-
lations of the process although it had been thoroughly tested,
in pilot plant operations.
The Aquaclaus process, like the IFP-150, has also
been designed for total S0X control in a refinery. Besides
its use for cleanup of Claus unit tail gas, it can also be used
for the tail gases from sludge acid regeneration, the CO boiler,
and boilers and process heaters, as well as any other sulfur-
bearing gas streams in the refinery. The process is not sensi-
tive to contaminants or to operational upsets.
It is supposedly possible to use the Aquaclaus process
in lieu of the Claus in a refinery. For this application,
vacuum pitch is burned in the heaters and the S02 removed from
the flue gases in phosphate absorbers. In a central facility,
the phosphate solution is regenerated by the H2S-rich streams
from the hydrodesulfurization units. The vent gas from the re-
generation is sent to the amine unit. The H2S in the amine unit
E-160
-------
is then incinerated, and S02 in the incinerator off-gas removed
in another phosphate absorber. The S02-rich gas is then sent to
the central regeneration facility.
The Alberta Sulfur Research Ltd. (ASR) Sulfoxide
Process2 7--In the Sulfoxide process, the Claus reaction occurs
in an organic sulfoxide medium. The sulfoxide acts as a cata-
lyst: the H2S and the sulfoxide form an adduct which in turn
complexes with other sulfur compounds.
Low concentrations of H2S and S02 in the Claus tail
gas can be almost completely converted to sulfur. More than 70
percent of the COS and CS2 present can be converted to sulfur.
The treated gas stream contains appreciably less than 1,000 ppmv
sulfur, typically less than 500 ppmv sulfur, when the feed
stream contains more than 30,000 ppmv sulfur compounds.
The Sulfoxide process has been tested only on the
bench-scale level. There are no pilot plants or commercial
plants.
The SNPA/Topsoe Process22--The Topsoe process uses the
SO2-laden stream from the Claus incinerator. This stream is
first cooled in a wasteheat boiler, then passed over a vanadium
oxide-base catalyst to oxidize S02 to SO3 with a 95 percent
yield. The gas is cooled further and concentrated. In an ab-
sorber, the SO3 is converted to 80 weight percent H2S0i,, which
is evaporated by heat from incoming gases to 94 weight percent
H2S0i,. The H2S0i, is formed without the sulfuric acid "mist"
usually formed in the wet-contact process for making H2S0if.
E-161
-------
One Topsoe plant was built in Lacq, France, in 1965
for a 100 lt/day Claus plant. There are no other commercial
installations. The Lacq plant is reportedly 90 percent
efficient.
The Shell Flue Gas Desulfurization (SFGD) Process36,37
The Shell flue gas desulfurization process is especially suited
for application to a refinery because of the availability of
hydrogen. The process is also being considered by the utility
industry.
In the Shell process, the cooled flue gas from a
Claus incinerator is passed over a copper oxide-on-alumina
catalyst to absorb S02 according to the following reaction:
S02 + ^ 02 + CuO -*¦ CuS04 (1)
About 90 percent of the S02 in the flue gas is absorbed.
Two reactors are alternately in the absorption or
regeneration stage. A spent bed is regenerated with hydrogen
which reduces copper sulfate to elemental copper:
CuS04 + 2 H2 Cu + S02 + 2 H20 (2)
The elemental copper is oxidized to copper oxide by oxygen in
the incoming flue gas. Because of hydrogen's extreme flam-
mability, the reactor is purged with inert gas before and after
regeneration. The S02 produced during regeneration is routed
to the front of the Claus unit.
•E-162
-------
The cost of an SFGD unit is about 250 percent that
of the parent Claus unit. The most promising application of
the process is in a larger refinery where it might be used for
the cleanup of all stack gases from process heaters (if they
can be collected and sent to one or two stacks) and fluid cata-
lytic cracking regeneration, as well as from the Claus unit
itself.
Particulates do not affect the operation of a SFGD
unit, and for this broad application the process can be de-
signed to remove S0x and N0X simultaneously. This further
application of the SFGD will be discussed more thoroughly in
Section 3.2.3.7.
The Westvaco Process22'38--The Westvaco process
utilizes a fluidized bed of activated carbon to remove S02 from
the stack gas of a Claus unit incinerator. The activated
carbon catalyzes the conversion of sulfur dioxide, oxygen and
water vapor in the incinerated gas to sulfuric acid. The
treated gas contains less than 200 ppmv S02.
The bed is regenerated with hydrogen sulfide in a
regenerator/stripper. The hydrogen sulfide is obtained from
the feed stream to the Claus unit. The regenerator/stripper
may be designed to produce elemental sulfur or a concentrated
SO2 stream for recycle to the Claus unit.
To regenerate SO2, the hydrogen sulfide first con-
verts a portion of the absorbed SO2 to elemental sulfur. The
sulfuric acid/sulfur mix is then heated to produce S02.
E-163
-------
Production of an elemental sulfur product is similar to that
in the Sulfreen process. An excess of H2S reacts with the H2S
to form elemental sulfur which is then vaporized from the bed.
No reliable cost figures are available for the
Westvaco process. The process has not yet been demonstrated
commercially.
The Pritchard Ammonium Bisulfate/Ammonium Thiosulfate
Process3 3--In this process, the sulfur compounds in the Claus
tail gas are oxidized to S02 and absorbed by contact with a
weak aqueous solution of ammonia. The resultant solution of
ammonium bisulfite, ammonium sulfite, and ammonium sulfate
salts is converted to ammonium thiosulfate in an ATS reactor.
Capital expenditure for a 20,000 t/yr factility is
about 75 percent that of the parent Claus unit. Clean gas
from the unit contains less than 900 ppmv S02. The first
commercial installation of the process is expected to start
up in December 1979.
The BSR/Selectox I Process33--Initial treatment of
the Claus tail gas in the BSR/Selectox I process is the same
as that in the BSRP: the tail gas is reacted with a hot
reducing gas in a catalyst bed to convert all sulfur species
to H2S by hydrogenation or hydrolysis. In the BSR/Selectox
I process, the H2S-rich stream is then cooled in a contact
condenser to minimize water partial pressure.
The cooled stream is then mixed with air and passed
over the BSR/Selectox catalyst to selectively oxidize the H2S
to elemental sulfur, which is removed by condensation. The
overhead gas stream is incinerated.
E-164
-------
Sulfur recovery from this process is normally greater
than 98 percent. A recovery rate greater than 99 percent can
be achieved by passing the oxidized streams through an ad-
ditional Claus stage before incineration.
Investment for a plant to treat the tail gas from a
100 lt/day sulfur plant is $1.7 to $2 million.
The first commercial BSR/Selectox I plant began
operation in 1978 in Germany.
The Mineral and Chemical Resource Company (MCRC)
Limestone Slurry Sulfur Recovery Process33--The MCRC Limestone
Slurry process treats the tail gas from the Claus plant incin-
erator by contacting it with a limestone slurry. The gas is
contacted in each of the three phases (solid, liquid and gas)
in a high energy device. Each contacting stage typically
reduces the sulfur dioxide level more than 90 percent for a
final efficiency of 99.9 percent.
The solid product consists of about 85 percent calcium
sulfite, 10 percent calcium sulfate (gypsum) and 5 percent un-
reacted limestone. It does not have to be ponded; in fact,
studies have shown it to be non-hazardous and beneficial to
crops when applied at a rate of 3-10 tons per acre.
A typical MCRC plant consists of the contactors, a
mist eliminator, a ball mill for grinding limestone, a central
pump, and a filter for dewatering the solids, plus miscellaneous
pumps, storage bins, etc. One U.S. plant has operated success-
fully since 1976.
E-165
-------
IFP Catalytic Incineration33--A catalyst can be used
to reduce the amount of energy necessary for incineration of
Claus tail gas. The RS-103 and RS-105 catalysts promote the
selective oxidation of sulfur compounds to sulfur dioxide.
Typical outlet values of H2S are less than 5 ppmv, those for
COS and CS2 combined are less than 150 ppmv.
Catalyst life is 2 - 3 years. Energy savings are about
1 MM Btu/hr. This savings corresponds to more than ten times
the cost of catalyst replacement.
Four IFP catalytic incinerators are operating on
Claus plants. The first began operation in 1975.
3.2.2.3 Catalyst Regeneration
A method has been developed which controls S02 emis-
sions from FCCU catalyst regenerators by stack gas scrubbing.
The method can also be used to control S02 and particulates
simultaneously. Exxon has four such scrubbers installed in
its coastal refineries (Texas, Louisiana, and New Jersey).
These scrubbers operate by mixing the flue gas with
a buffered aqueous scrubbing liquor. The liquor absorbs the
S02; other processes separate the liquor from the gas stream.
The clean gas is then routed to the stack. Most of the
scrubbing liquid may or may not be regenerated and recycled,
depending on the particular scrubbing process being used.
A nonregenerable process has been used for FCCU flue
gas, and the spent scrubbing liquid handled by conventional
wastewater treatment. It contains a high concentration of
E-166
-------
dissolved solids and salts and has a high chemical oxygen
demand (COD). To date, scrubbers for controlling S0x from
FCC regneration 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.29
Exxon's operations have shown that 95 percent of the
S0X and 90 percent of the particles can be removed by a
scrubber.39 They believe that the cost of controlling both
particulates and S0x 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. Actual comparisons of costs, energy requirements, and
secondary effluent disposal are not available.
3.2.2.4 Boilers and Process Heaters
SOx Removal--Emissions of S0X from boilers and
process heaters can be minimized by routing the flue gas to
an integrated sulfur removal facility. Two such units are
the IFP-150 and the Aquaclaus (dicussed in the preceding
sections).
N0X Removal--NO., emissions can be reduced by
several tail gas cleaning methods. Removal of N0X from tail
gas is, however, inherently more difficult than controlling
N0X by the combustion modifications discussed in Section 4.
The principal difficulties are the large amount of hot gas'
to be handled, the dilute concentration of N0X, interferences
by other pollutants and the high power consumption. Three
E-167
-------
methods for removal of N0X from stack gases are gas scrubbing,
catalytic reduction and thermal reduction with added ammonia.
Because .of economic considerations, only thermal reduction
with added ammonia appears promising.11*
Gas Scrubbing1''--Wet and dry scrubbing techniques
for removal of S0x also remove most of the N02 and an equiv-
alent amount of NO which react to form N2O3. Because signif-
icant amounts of N02 form only after the NO in the stack gas
meets the oxygen in the atmosphere, the low proportion of N02
in the gas limits the effectiveness of these processes. The
N0X removal efficiency can be improved by oxidizing part of
the NO to N02. Air oxidation is cheaper than the use of
oxidizing agents such as permanganate or chlorine oxides.
Reactive salts of metals such as iron can form
chelates with NO, but the chelating agents have high molecular
weights and low solubility. Sodium sulfite can reduce both
NO and NO2 to molecular nitrogen in a special rotary atomizer.
However, oxygen must be removed from the flue gas at great
expense.
Catalytic Reduction1 "^--Catalytic reduction may be
more practical than scrubbing for removing small amounts of
nitrogen. In the process, N0x is combined with a reducing
gas such as ammonia or hydrogen sulfide to form molecular
nitrogen in the presence of a catalyst. Ammonia is less
poisonous to most catalysts than is hydrogen sulfide.
Hydrocarbons, hydrogen, or carbon monoxide can also be used
as reducing agents if oxygen is removed.
E-168
-------
A new Japanese process has two beds of a base metal
catalyst on alumina and a slight excess of ammonia. The first
bed, at 572°F, reduces the N0X. The remaining 50 ppm
ammonia is removed on the second bed.. The catalyst is less
sensitive than platinum to sulfur, but particulates must be
removed.
Thermal Reduction with Added Ammonia1 ''--Controlled
addition of ammonia to an oxygen-containing flue gas under
strictly controlled conditions at 1,300° to 1,900°F can
selectively reduce 50 to 70 percent of the N0X remaining after
combustion controls. 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 care-
fully controlled, a major portion of the N0X can be reduced
with little ammonia left over. This process is more expensive
than combustion modifications but can supplement these modifi-
cations when stricter control of N0x is required.
3.2.3 Control Technology from Other Industries
Some control methods which have been developed
primarily for other industries can also be used in the petro-
leum refinery 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 accomp-
anying N0X control can be adapted for the flue gases from
process heaters; still others might be used to control sulfur
emissions from FCC regenerators.
E-169
-------
3.2.3.1 Sulfur Recovery
The Chiyoda Thoroughbred 101 Process27--The
Chiyoda Thoroughbred 101 process was developed for desulfuri-
zation of flue gases. For application to a Claus unit, the
tail gas is first incinerated, cooled, and scrubbed with water.
About 97 percent of the S02 in the tail gas is then absorbed
by dilute sulfuric acid in an absorber. The overhead stream
is passed without further treatment to the atmosphere. If
the Claus tail gas contains less than 15,000 ppmv SO2, an
emission rate of less than 500 ppmv S02 can be achieved.
A level of 250 ppmv is required to meet Federal New Source
Performance Standards.
The bottoms from the absorber pass to an oxidizer
where they are mixed with air over an inexpensive nonpoisonous
catalyst. The majority of the acid returns to the absorber;
a bleed stream is routed to a crystallizer. Limestone is
mixed into the acid in the crystallizer to form gypsum
crystals which are collected, dried, and stored.
Because there is not a good market for gypsum in
the United States, this method will probably not be a
significant one for the refining industry.
The USBM Citrate Process27--Citrate flue gas desul-
furization is a regenerable process in which an aqueous sodium
citrate solution absorbs sulfur dioxide from flue gas. The
method is being developed primarily for flue gases from utility
boilers, but appears to be applicable to the flue gas from a
Claus plant incinerator.
E-17C
-------
As the flue gas flows through the absorber, the S02
in the gas is dissolved in the citrate liquor. Once dissolved,
it reacts with water to form sulfurous acid (H2S03). The
sulfurous acid ionizes to form hydrogen ion (H+) and bisulfite
(HSO3). The citrate enhances these reactions by removing
hydrogen ions from solution. Some of the sulfite is oxidized to
sulfate. Desulfurized flue gas is reheated and vented to the
atmosphere.
The absorber liquor is regenerated in a reactor by
contacting it with hydrogen sulfide (H2S) to form elemental
sulfur. The H2S can be obtained from the feed streams to the
Claus unit. The sulfur slurry formed in the reactor then flows
to a digester where it is mixed with a small by-pass stream of
S02-rich citrate liquor from the absorber. This liquor consumes
any unreacted hydrogen sulfide. Gases vented from the reactor
and the digester contain small quantities of H2S, CS2, and COS.
These gases are incinerated.
The sulfur slurry from the digester is separated from
the citrate solution by flotation with air in a separate vessel.
A flotation agent such as kerosene can be used to increase the
sulfur solids concentration of the froth to 50 percent. "Re-
generated citrate solution is returned to the absorber.
To remove accumulated sulfates from the system, a
small bleed stream is diverted from the regenerated citrate'
stream and treated for sulfate removal. Sodium sulfate de-
cahydrate (Glauber's Salt) is continuously removed from solution
and recovered for sale or disposal. Lean citrate solution is
returned to the system.
E-171
-------
Entrained air is released from the froth by agitation
and then incinerated. The sulfur/liquor mixture is heated in a
shell and tube heat exchanger to melt the sulfur. The molten
sulfur is separated by gravity from the citrate solution.
An important environmental consideration is disposal
of the Glauber's Salt. A long-term solution for disposal of
this sulfate by-product has not been identified.
It has been estimated that a citrate plant for an 85
lt/day Claus plant would cost 250 percent of the cost of the
parent Claus unit. No commercial citrate plants have yet been
built; a commercial-scale demonstration plant for a utility was
scheduled to begin operation in mid-1979.
One of the major drawbacks of the use of the citrate
process by the utility industry is the expensive manufacture of
H2S. Because the required H2S is readily available at a Claus
plant, the process may be more suited for use by the refining
industry.
The Townsend Process22--The Townsend process was pro-
posed as a method for the desulfurization of high-pressure
natural gas. It is quite similar to the IFP-1,500 process.
The Claus reaction takes place in an organic solvent, such as
triethylene glycol, at an elevated temperature. Liquid sulfur
is drawn from the bottom of the vessel.
The process may be applied directly to the Claus tail
gas. It does not, however, remove COS or CS2 and, therefore,
can not achieve a very low sulfur emission rate.
E-172
-------
The Townsend process has been tested in pilot plants,
but as of mid-1978 no commercial plants had been built.
The Lurgi-Claus-Abgas-Schwefelgewinnung (LUCAS)
Process**0'111 --The LUCAS process will probably be used primarily
to treat lean H2S streams for which the Claus process is unsuit-
able. It also appears to be well-suited for Claus tail gas
cleanup.
The tail gas is first incinerated, then cooled and
treated with hot granulated coke to covert impurities to S02,
Co 2, or N2:
2 S03 + C -»• 2 S02 + C02 (1)
H2 S + 1 | 02 "~ S02 + H2 0 (2)
2 NO + C "~ N2 + C02 (3)
o2 + c -»• co2 (4)
The S02 is removed from the off-gases with an aqueous alkali
phosphate solution which can be regenerated by heating:
Na2HP04 + H2 0 + S02 ^ NaH2P04 + NaHS03 (5)
Because oxygen and S02 have been removed from the tail gas, no
sulfate is formed.
The off-gas from a LUCAS absorber reportedly contains
less than 200 ppm S02, less than 150 ppm COS and CS2, and prac-
tically no H2S. Recycling the S02 to the Claus unit can
increase the sulfur yield up to 99.9 percent.
E-173
-------
Investment cost for a LUCAS unit is expected to be
75 - 80 percent that of the parent Claus Unit. It is estimated
that 9.1 kwh's of electricity, 1,900 lb steam and 2,170 Btu
fuel gas are required for the production of 200 lb sulfur
from a 50 t/d Claus unit with a 75 percent H2S feed. There is
one semi-commerical LUCAS plant in West Germany.
The Takahak Process33--The Takahak process is quite
similar to the Stretford and could be used in its place in
many applications. It is now being used to treat fermented
sewage, coke oven gas and chemical plant waste gas.
The tail gas from the Claus unit may be fed to an
absorber tower which contains sodium carbonate and sodium
1,5-naphthoquinone, 2-sulfonate as a redox catalyst. Hydrogen
sulfide in the tail gas reacts with the sodium carbonate to
form sodium bisulfide and sodium bicarbonate. The bisulfide is
oxidized by the catalyst to form finely divided sulfur which
is filtered from a slip stream.
The naphthoquinone sulfonate which was reduced to
naphtho-hydroquinone sulfonate is regenerated with bubbling
air in a regenerator. The vent stream from the regenerator
contains no H2S and can be released to the atmosphere.
Up to 99.9 percent of the H2S in a gas stream can be
removed by this method. Capital costs for a Takahak plant are
relatively low and the only utility needed is electricity for
pumping. The catalyst is cheap and easily available.
E-174
-------
3.2.3.2 Catalyst Regeneration
Several FGD methods used by the utility industry have
been proposed for use on FCC regenerators.29 They are described
below. In addition, some of the regenerable processes dis-
cussed in Section 3.2.3.1 for treatment of the Claus unit tail
gas may also be applicable. One of the processes described
below simultaneously removes S0x and particulates from the
flue gas.
The Lime/Limestone Flue Gas Desulfurization Process**2
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.
In the feed preparation area, pebble lime (CaO) is
slaked with water to yield a dilute solution of calcium hydrox-
ide (Ca(0H)2). Limestone (CaC03) is ground in wet ball mills
and is then slurried with water. Both slaked lime and lime-
stone produce calcium ions.
Absorption takes place in a wet scrubber. The S02
is absorbed in water and reacts to form sulfite ions (S07).
The desulfurized flue gas is released. The sulfite ions
begin a series of reactions and eventually coprecipitate with
calcium as a solid solution.
E-175
-------
The precipitation reactions go to completion in an
effluent hold tank. The liquid from the hold tank is recircu-
lated to the absorber. The bleed stream is continuously drawn
from the hold tank for dewatering, solids concentration, and
disposal. A thickener is usually added to the bleed stream for
primary dewatering; the stream may then be vacuum filtered or
centrifuged. The sludge may be ponded on-site or trucked off-
site to a landfill.
A major design option is the choice of lime or lime-
stone. Limestone is less expensive than lime, but it is not
used as efficiently by the process; therefore, there is more
feed material consumption and more waste sludge production if
limestone is used. Procedures which improve limestone utiliza-
tion also increase capital and operating costs.
Lime systems are usually more expensive to operate,
however, because of the high cost of lime. Lime systems may be
preferred where space is limited for feed material processing
and/or waste sludge disposal. Capital and utility costs are
also lower for a lime system. The choice between lime and
limestone is also influenced by the availability of raw
materials.
Several vendor companies market processes for the
stabilization of waste sludge. They all involve the addition
of various substances to the sludge to make it structurally
stable and leach-resistant. Stabilized sludge is easier to
transport and makes a landfill which can later be used produc-
tively. An alternative to simple disposal is the oxidation of
the bleed stream to gypsum. The purified gypsum can then be
sold if a market can be found.
E-176
-------
Costs of raw materials and utilities for lime/lime-
stone systems are generally lower than for regenerable pro-
cesses, although more raw materials are required. Annual
operating costs of a lime system are about seven percent higher
than those of a limestone system.
The Dual Alkali Flue Gas Desulfurization Process**3--
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 S02 removal efficiencies
of greater than 90 percent. The capability 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.
Absorption of S02 from the flue gas takes place in a
tray tower, or a venturi scrubber if simultaneous particle
removal is desired. The S02 in the flue gas reacts primarily
with sodium sulfite (Na2S03) to form sodium bisulfite (NaHS03).
Some sulfite and bisulfite oxidize to sulfate. The desulfurized
flue gas is reheated if necessary and released to a stack. A
bleed stream of the scrubbing liquor is withdrawn continuously
from the absorber and regenerated.
The sodium alkali in the bleed stream is regenerated
in chemical mix tanks. Lime (Ca(0H)2) or limestone (CaC03) is
added to the liquor and the resulting calcium sulfite and cal-
cium sulfate are precipitated and removed as a slurry.
E-177
-------
The slurry from the regeneration step is very dilute.
It is dewatered first in a clarifier thickener and then on a
vacuum filter. The solids cake on the vacuum filter is washed
with several water sprays to remove up to 90 percent of the
sodium salts from the cake. The cake is then ponded or used as
landfill. Make-up sodium carbonate is added to the clarifier
thickener liquor which is routed to the absorber.
Slaked lime (Ca(0H)2) is generally used to regenerate
the scrubber effluent. Limestone (CaC03) has been used as the
regenerant, but impurities and the lower calcium utilization
rate are major problems.
Loss of soluble sodium and nonsulfur calcium salts
can create water pollution problems and also a loss of raw
materials. Therefore, water can be added to the system only
to replace that lost through evaporation or in the solid waste
product. Also, retention of soluble salts by the solid waste
must be minimized.
Sludge from the dual alkali process must be fixed
chemically to decrease its permeability and leachability, or
it must be disposed of in well-designed lined ponds. The
thixotropic nature of the calcium sulfite may make land
reclamation difficult.
Dual alkali systems are economically competitive with
lime/limestone systems. The equipment required for a dual
alkali system is similar to and perhaps less than that required
for a lime/limestone system. Because of the higher moisture
content of dual alkali sludge, a larger disposal area will be
required than for a lime/limestone system.
E-178
-------
3.2.3.3 Boilers and Process Heaters
In addition to the process described below, the
processes proposed for control of S02 emissions from FCCU
regenerators may also be applicable to the flue gases from
boilers and process heaters.
The Shell Flue Gas Treatment (SFGT) process can be
used to remove S0x and N0x simultaneously from all stack gas
from process heaters (if it can be collected and sent to one
or two stacks) and fluid catalytic cracking regeneration, as
well as from a Claus unit.* *'*5
The SFGT process has demonstrated S02 and N0x removal
efficiencies of greater than 90 percent. The efficiency of the
system is not affected by variations in the S02 or N0x
concentration.
Flue gas is drawn into the process upstream of any
particle removal equipment. It is then mixed with ammonia and
fed to a reactor which contains regenerated elemental copper on
a.series of parallel plates. The flue gas flows over the
plates, not through them.
Upon initial exposure to the oxygen in the flue gas,
the elemental copper is converted to copper oxide. Since
copper catalyzes the oxidation of ammonia to N0X, N0x is
actually produced in the reactor for the short time before all
the copper has been converted. To minimize these N0y emissions,
the bed can be preoxidized with air or incompletely regenerated
before it is returned to service, or injection of ammonia can
be delayed.
E-179
-------
The copper oxide then reacts with S02 to form copper
sulfate. Maximum S02 removal occurs when the bed is fully con-
verted to copper oxide. The removal efficiency steadily
decreases to a breakthrough point when all the copper oxide
has been converted to copper sulfate.-
The reduction of N0y to nitrogen begins as copper
oxide is formed and increases rapidly as copper sulfate is
formed because copper sulfate is the more effective catalyst.
Maximum N0y removal occurs at the point of S02 breakthrough.
The spent bed is regenerated by a reducing gas made
up primarily of hydrogen. The hydrogen reduces copper sulfate
to elemental copper and sulfur dioxide. Any copper oxide
present is also reduced to elemental copper. Because of
hydrogen's extreme flammability, the reactor is purged with
an inert gas such as low pressure steam before and after
regeneration. The S02 may be sold, processed into elemental
sulfur or sulfuric acid, or routed to the front of the Claus
unit. The primary waste from the process is water generated
during the recycle step. This water contains 30 to 50 ppm
dissolved S02.
The process requires approximately two moles of hydro-
gen per mole of S02 removed and one mole of ammonia per mole of
N0X removed. A heat credit may be realized by the process.
Actual costs for an integrated SFGT system are not available
because such a system has not yet been built at a U.S.
installation. Due to the complexity of the process, space
requirements are expected to be high. Retrofit application
of the SFGT process might be difficult because of the duct
work required.
E-18C
-------
3.3
References
1. Gary, J. H., and G. E. Handwerk. Petroleum Refining,
Technology & Economics. Marcel Dekker, New York, New
York, 1975.
2. Elkin, H. F., and R. A. Constable. Source/Control of Air
Emissions, Hydrocarbon Proc., 51 (10):113, 1972.
3. Dickerman, J.C., et al. Industrial Process Profiles
for Environmental Use, Chapter 3. Petroleum Refinery
Industry. EPA 600/2-77-023C, Radian Corporation, Austin,
Texas, January, 1977.
4. Bryant, H. S. Environmental Needs Guide to Refinery Sulfur
Recovery. Oil Gas J., 71(13):70-76, 1973.
5. U. S. Environmental Protection Agency. Compilation of
Air Pollutant Emission Factors. 3rd Edition with Supple-
ments 1-8. AP-42, PB 275, Research Triangle Park, North
Carolina, 1977.
6. Groenendaal, W., and H.C.A. Van Neurs. Shell Launches Its
Claus Off-Gas Desulfurization Process. Petrol. Petrochem.
Int., 12(9):54, 1972.
7. Radian Corporation. A Program to Investigate Various
Factors in Refining Siting. Federal Energy Administration,
Washington, D. C., 1974.
E-181
-------
8. Sussman, V. H. Atmospheric Emissions from Catalytic
Cracking Unit Regenerator Stacks. Report No. 4. Joint
District, Federal and State Project for the Evaluation of
Refinery Emissions and Western Oil and Gas Association, Los
Angeles, California, 1957.
9. Little, Arthur D., Inc. Screening Study to Determine Need
for S0y and Hydrocarbon NSPS for FCC Regenerators. EPA-458/3-
77-046, U. S. Environmental Protection Agency, Research
Triangle Park, North Carolina, August, 1976.
10. Radian Corporation. Control Techniques for Volatile Organic
Emissions from Stationary Sources. EPA-450/2-78-022, U. S.
Environmental Protection Agency, Research Triangle Park,
North Carolina, 1978.
11. Nelson, W. L. Petroleum Refinery Engineering. 4th Edition.
McGraw-Hill, New York, New York, 1958.
12. American Petroleum Institute, Committee on Refinery Environ-
mental Control. Hydrocarbon Emissions from Refinery. API
Pub. No. 928, Washington, D. C., 1973.
13. Laster, L. L. Atmospheric Emissions from the Petroleum
Refining Industry. EPA-650/2-73-017, PB 225 040, Control
Systems Lab, Research Triangle Park, North Carolina, 1973.
14. American Petroleum Institute, Division of Refining. Manual
on Disposal of Refinery Wastes, Volume on Atmospheric
Emissions. API Pub. No. 931, Washington, D.C., 1976.
E-182
-------
15.
Kerlin, G. and D. Rabovsky. Cracking Down. Oil Refining
and Pollution Control. Council on Economic Priorities,
New York, New York, 1975.
16. Danielson, J. A. Comp. & Ed. Air Pollution Engineering
Manual. 2nd Edition. AP-40, PB 225-132. U. S. Environ-
mental Protection Agency, Office of Air & Water Programs,
Research Triangle Park, North Carolina, 1973.
17. Laengrich, A. R., and W. L. Cameron. Tail-Gas Cleaning
Addition May Solve Sulfur-Plant Compliance Problem. Oil
Gas J., 76(13):159-162, 1978.
18. Beavon, D. K., and R. P. Vaell. The Beavon Sulfur Removal
Process for Purifying Claus Plant Tail Gas. API, Division
of Refining, 267, 1972.
19. Meisen, A. and H. A. Bennett. Consider All Claus Reactions.
Hydrocarbon Proc., 53 (11):171-174, 1974.
20. Barry, C. B. Reduce Claus Sulfur Emission. Hydrocarbon
Proc., 51(4):102-106, 1972.
21. Valdez, A. R. New Look at Sulfur Plants. Hydrocarbon Proc.,
Petrol. Refining, 43(3):104-108 , 1964.
22. More Claus Cleanup Processes Explained. Oil Gas J., 76(73):
88-91, 1978.
23. Hydroprocessing is Lively Topic. Oil Gas J., 75(28):153,
1977.
E-183
-------
24.
Grancher, P. Advances in Claus Technology, 2 Parts. Hydro-
carbon Proc., 57(7):155, 1978, Part 1; Hydrocarbon Proc.,
57 (9):257, 1978, Part 2.
25. Beavon, D. K. Add-On Process Slashes Claus Tailgas Pollu-
tion. Chem. Eng., 78 (28) : 71-73, 1971.
26. Catalytic, Inc. The Stretford Process. Philadelphia,
Pennsylvania, 1976.
27. Processes Clean Up Tail Gas. Oil Gas J., 76(35):160-166, 1978.
28. Bailey, E. E. Power Plant Flue Gas Desulfurization by the
Wellman-Lord SO2 Process. Part II - Continuing Progress
for Wellman-Lord S02 Process. In: Flue Gas Desulfurization
Symposium, Atlanta, Georgia, November 1974. EPA-650/2-74-
126a, b. U. S. Environmental Protection Agency, Office of
Reserach and Development, Research Triangle Park, North
Carolina, December 1974, pp. 745-759.
29. Vasolos, I. A., et al. New Cracking Process Controls FCCU
S0x. Oil Gas J., 75(26):141-148, 1977.
30. Rheaume, L., et al. New FCC Catalysts Cut Energy and
Increase Activity. Oil Gas J., 74(20):103-107, 110, 1976.
31. Fogiel, M. , Ed. Modem Pollution Control Technology, Vol. 1:
Air Pollution Control. Research & Educators Education
Associates, New York, New York, 1978.
32. Atmospheric Emissions from Petroleum Refineries. A Guide
for Measurement and Control. PHS No. 763, Public Health
Service, Washington, D. C., 1960.
E-184
-------
33. Conser, R. E. Here's a New Way to Clean Process Gases.
Oil Gas J., 72(13):67-68, 70, 1974.
34. Barthel, U., et al. Sulfur Recovery in Oil Refineries
Using IFP Processes. Adv. Chem. Serv., 139:100-110, 1975.
35. 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.
36. Conser, R. E. and R. F. Anderson. New Tool Combats S02
Emissions. Oil Gas J., 71 (44):81-86, 1973.
37. Dry Process for S02 Removal Due Test. Oil Gas J., 70(34):
67-70, 1972.
38. Ball, F. G., et al. Claus Process/Gaseous Wastes. Hydro-
carbon Proc., 51(10):125-127, 1972.
39. Controlling S0x Emissions from Fluid Catalytic Cracking
(FCC) Units. Wet Scrubber Newsletter, (48):6, 1978.
40. Coke Key to Cleaning Claus Unit Tail Gas. Oil Gas J.,
74(47) .-142-144, 1976.
41. Doerges, A., K. Bratzler, and J. Schlauer. LUCAS Process
Cleans Lean H2S Streams. Hydrocarbon Proc., 55(10):110-111,
1976.
E-185
-------
A2. Hulman, P. B. and J. M. Burke. The Lime/Limestone Flue Gas
Desulfurization Processes, Draft Final Report. EPA Contract
No. 68-02-26-08. Radian Corporation, Austin, Texas; U. S.
Environmental Protection Agency, Research Triangle Park,
North Carolina, 1978.
A3. Gibson, E. D., T. G. Sipes, and J. C. Lacy. The Dual
Alkali Flue Gas Desulfurization Process, Draft Final Report.
EPA Contract No. 68-02-26-08. Radian Corporation, Austin,
Texas; U. S. Environmental Protection Agency, Research
Triangle Park, North Carolina, 1978.
AA. Arneson, A. D., F. M. Nooy, and J. B. Pohlenz. The Shell
FGD Process Pilot Plant Experiences at Tampa Electric. In:
Flue Gas Desulfurization Symposium, Fourth, Hollywood,
Florida, November 1977, EPA-600/7-78-058a, b. Franklin A.
Ayer, comp. Research Triangle Institute, Research Triangle
Park, North Carolina, March 1978. pp. 752-793.
A5. Stern, R. D., et al. Interagency Flue Gas Desulfurization
Evaluation, Draft Final Report. EPA Contract No. 68-02-26-
08, Task 16. Radian Corporation, Austin, Texas; U. S.
Environmental Protection Agency, Research Triangle Park,
North Carolina, 1977.
E-186
-------
4.0
EMISSION REDUCTION THROUGH PROCESS MODIFICATION
4.1 Alternative Operating Practices and Conditions
Refinery operations are routinely modified to meet
product specification requirements, product marketing trends,
feedstock availablity constraints, and operating cost goals.
The operating choices made include both deliberate actions con-
cerning processing alternatives (such as which catalyst to use
or what cutpoint 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.
A.1.1 Catalytic Cracking Catalysts Regeneration 1 '2'3'14'5
In older FCCU regenerators, the highly exothermic
oxidation of CO to CO2 is avoided because the resulting high
temperatures can damage regenerator equipment, permanently
deactivate the catalyst, and damage downstream equipment as
well. To avoid this oxidation, flue gas from the regenerator
generally contains little oxygen and large, nearly equal,
amounts of CO and CO 2.
For several years, emphasis has been placed on promot-
ing the oxidation of CO to CO2 in the regenerator while control-
ling peak temperatures. This oxidation can be promoted by simply
raising the temperature of the regenerator and increasing the
oxygen partial pressure. A cooling system consisting of steam
and water sprays helps prevent equipment damage if combustion
control is lost.
E-187
-------
Older FCC regenerators were designed for operation at
temperatures up to 1150°F; the introduction of newer, more coke-
sensitive catalysts necessitated higher temperatures. By 1976,
30 percent of all FCC regenerators were operating at 1300°F.
High-temperature conversion of CO to C02 generally occurs at
1400-1500°F.
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; there-
fore a CO boiler is no longer necessary for CO emission control.
Complete combustion of carbon monoxide within the
regenerator offers other emissions benefits in addition to reduc-
ing carbon monoxide to less than 500 ppm: elimination of the
need for a carbon monoxide boiler reduces other emissions since
auxiliary fuel burning is not required and N0x- producing carbon
monoxide boiler temperatures are avoided. Recovery of additional
heat in the regenerator reduces (and in some cases eliminates)
the need for a FCC preheater and its associated emissions;6
Several new catalysts, or promoters, have been intro-
duced 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.
There are three types of promoters now in use. First
to be offered were catalysts modified with a small concentration
of noble-metal promoting agent. The metal content is so small
that recovery from discarded catalyst is uneconomical. The
degree of promotion can be varied. The second type of promoter
E-188
-------
is a liquid additive which is injected into the FCC regenerator
combustion zone. The third type is a solid promoter which can
be mixed with make-up catalyst in varying proportions.
In situations where partial combustion is needed, it
is possible to combine high temperature regeneration with the CO
boilers. For this combined operation, the degree of high-
temperature regeneration, and the final temperature, can be con-
trolled by the amount of promoter used. The partial loss of CO
to the CO boiler can be somewhat compensated by the high tempera-
ture of the gas that does enter the boiler. Supplemental fuel
to the boiler is still required, but its cost is offset by the
increased product yield in the FCC unit.
In 1975, the cost of converting a relatively modern
FCCU with stainless steel cyclones to high-temperature regenera-
tion was $50,000 to $300,000. Cost of a CO boiler for the unit
was perhaps $2 million to $3 million.
Other FCC operations which affect regenerator emissions
are the amount of recycle and the stripping steam rate. Higher
recycle rates produce more flue gas, but the net effect is neg-
ligible when compared to the impact of recycle on yields and
operating costs. Similarly, insufficient stripping steam allows
hydrocarbons to enter the regenerator to produce more flue gas
and possible unstable operation. In this case, the impact of
stripping steam rate on the entire unit's performance provides
incentives for the reduction of emissions.
Operators of thermofor catalytic crackers (TCC) can
achieve the lowest emissions by preventing upsets, which can
cause releases of hydrocarbons and other gases through the var-
ious vent stacks. Stable operation is important to refiners for
efficiency and product specification reasons, so there is no
conflict with environmental goals.
E-189
-------
One aspect of TCC operations which does conflict with
environmental aims is the proper operation of elutriator systems.
Elutriators remove catalyst fines from a slip stream of catalyst.
Maintaining the elutriator in proper condition does not improve
TCC performance. Fines which are not removed by the elutriator
are vented to the atmosphere through the surge separator stack,
A.1.2 S0x Removal in the FCC Regenerators7'8
Amoco has developed a catalyst which reduces the amount
of sulfur leaving the regenerator as S02. 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
Claus plant; the regenerated catalyst returns to the regenerator.
Cost for the 60 to 75 percent reduction in S0x emis-
sions with this method in a new facility is estimated at 3<£/bbl,
compared to 22-24
-------
TABLE E4-1. REFINING SOURCES OF THERMAL NOv AND FUEL NO
X X
Classification
Source
Thermal
N°x
Fuel
N0x
High Temperature
Power boilers
firing - gas
Present
Possible
Power boilers
firing - oil
Present
Possible
Power boilers
firing - coal
Present
Strong
Internal Combustion
Engines
Present
Unlikely
Turbines
Strong
Possible
Moderate Temperature
CO boilers
Present
Present
Coke and residual
fuels
Present
Present
Catalyst
regeneration
Unlikely
Present
Incineration
Present
Present
Process Heating
Gas cracking
Present
Possible
Oil cracking
Unlikely
Possible
Oil heating
Unlikely
Possible
Source: Reference 9.
E-19]
-------
Fuel N0V forms more readily than does thermal NO
X X
because of the relative weakness of the carbon-nitrogen bond as
compared to the extreme stability of the nitrogen-nitrogen triple
bond. About 80 percent of the nitrogen in fuels containing less
than 0.1 percent nitrogen goes to N0x. For fuels containing more
than 1.5 percent nitrogen, about 20 percent of the fuel nitrogen
is converted to N0X. The mechanism of formation of fuel N0X is
essentially the same for all types of bound N.
Control of fuel N0X is not a major concern in refin-
eries for several reasons. First, refinery fuels are typically
low in nitrogen content since coal is rarely used. Second, since
the percent conversion of fuel nitrogen to NOx increases as the
nitrogen content decreases, the use of low-nitrogen fuels is not
by itself an effective method for reducing N0X emissions. Fuel
N0x emissions are also relatively unaffected by combustion modi-
fications. Therefore, combustion modifications for the control
of N0x emissions from refineries are mainly concerned with the
reduction of thermal N0X. Fuel N0x may become more significant,
however, as thermal N0x values are minimized in boilers and as
air quality regulations become stricter.
E-192
-------
4.1.3.1 Boilers, Furnaces, and Process Heaters
Combustion modifications for N0X control on boilers,
furnaces, and process heaters are" of three general types: low-
ering the flame temperature, limiting the amount of excess air,
and limiting the residence time within the flame. Control of N0X
is often counter to a high thermal efficiency and contributes to
the emission of other undesirable substances; therefore, compro-
mises must often be made.
Flame temperature is always greater than furnace tem-
perature because of the immediate dissipation of heat. Flame
temperature is determined by combustion chemistry and is depen-
dent on the type of fuel used. If all other factors are equal,
flame temperature is lowest for gas, higher for oil and highest
for coal. Flame temperature also depends on the amount of aera-
tion of the fuel: a well aerated blue flame has a higher tem-
perature and thus produces more N0X than does a less-well-aerated
yellow flame. Theoretically, a reduction of only 70°F in the
flame temperature can cut N0x emissions by one-half.
Excess air contributes to N0x emissions because of. the
availability of oxygen for NOx-forming reactions. It also con-
tributes to a hotter flame. However, a lack of sufficient excess
air will lead to the emission of unburned fuel, smoke, and carbon
monoxide. A reducing zone in the fire box resulting from inade-
quate combustion air can lead to premature failure of tubes and
fittings. On the other hand, a large excess of air can dilute
the flame, decrease flame temperature, and thus decrease N0X
production.
Residence time (the time the gas is actually in the
flame) is determined by the gas flow rate and the length of the
E-193
-------
flame. The flame length is directly proportional to heat
release. Therefore, smaller burners with less heat release per
burner reduce the residence time.
A number of specific combustion modifications for N0x
control have been devised. Those for refinery boilers are sum-
marized in Table E4-2. Combinations 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 E4-3.
Low excess air firing is the simplest combustion mod-
ification for N0X control. It does require more complex ductwork
and instrumentation for monitoring emissions so that the proper
level of excess air can be maintained. Emissions of unburned
fuel, smoke, and carbon monoxide can be a problem. The applica-
tion of low excess air firing to refinery furnaces is complicated
by the extensive use of natural draft and dampers to control the
air supply. If oil burners are to be fired with low excess air,
the oil must be vaporized before ignition and combustion essen-
tially carried out in the vapor phase. Heat for vaporizing the
fuel oil can come from recirculating combustion products.
Flue gas recirculation can be a control method in its
own right, as it reduces the temperature and dilutes the oxygen
with no loss in thermal efficiency. The flue gas is injected
directly into the flame zone. In small furnace tests, recircu-
lation of 10 weight percent of the flue gas reduced N0x emissions
by 60 percent. Recirculation of 20 weight percent resulted in
a 75 percent reduction in N0x emissions.
E-194
-------
TABLE E4-2. BOILER COMBUSTION MODIFICATIONS FOR REDUCTION OF
NOx EMISSIONS
Method
Gas.Fired
Units
Oil Fired
Units
Low Excess Air
Rating
Advantages
Good
Improved effi-
ciency; less
power required
Good
Improved efficiency;
less power required;
less chance of cold
end deposits
Disadvantages More complex More complex ducts
ducts and controls and controls
Flue Gas
Recirculation
Rating
Advantages
Excellent
Very effective;
does not upset
combustion
Disadvantages High initial cost;
high operating
cost; additional
controls
Good
In moderation does
not upset combustion
High initial cost;
high operating
cost; additional
controls; works
mainly on thermal
NCI
Staged Combustion
Two-stage
combustion with
over-fire air ports
Rating
Advantages
Excellent
Inexpensive; very
effective
Very good
In exp ensive; mod-
erately effective
Disadvantages Longer flames;
slight increase
in excess air
Longer flames;
slight increase
in excess air
Continued
E-195
-------
TABLE E4-2. Continued
Method
Gas Fired
Oil Fired
Units
"Units
Staged combustion
Rating
Very good
Good
(Cont'd)
Off-stoichiometric
Advantages
No power;
No power;
or biased firing
effective
moderately
effective
Disadvantages
Slightly longer
Slightly longer
flames; small
flames; small
increase in
increase in
excess air
excess air
Direct cooling
Rating
Fair
Fair
Lower preheat or
,
water injection
Advantages
Simple; no
Simple; no
power
power
Disadvantages
Reduced unit'
Reduced unit
efficiency;
efficiency;
require equipment
require equipmer
Reduced load or
Rating
Very good
Very good
oversized fire box
Advantages
Simple; no power
Simple; no
power
Disadvantages
High initial
High initial
cost
cost; more
radiant super-
heater
Burner modifications Rating
Advantages
Disadvantages
Good
Simple; no power
None
Very good
Simple; no
power
None
Source: Reference 9.
E-196
-------
f
TABLE E4-3. REDUCTIONS OF NOx EMISSIONS WITH COMBUSTION MODIFICATIONS
AT VARIOUS BOILER LOADS
PERCENT REDUCTION IN NOx EMISSIONS WITH
Combustion
Modification
(Percent Full Load)
Low
85/105
Excess
60/85
Air
50/60
85/105
Staging
60/85
50/60
Low Excess Air
and Staging
85/105 60/85
50/60
Flue Gas
Recirculation
85/105 60/85 50/60
Possible* Combined
Modifications
85/105 60/85 50/60
Fuel
Fired
Burner
Arrangement
Front Wall
13
24
7
37
30
30
48
42
36
43
42
36
Gas
Horizontally
Opposed
17
15
32
54
35
59
61
48
68
- - 20
73
52
72
Tangential
-
-
-
-
-
-
-
-
-
60 -
66
65 *
-
Average
16
19
26
45
31
52
54
44
52
60 20
64
51
60
Front wall
27
20
28
29
20
20
39
32
21
46 31
50
41
21
Oil
' Horizontally
Opposed
10
16
12
34
34
47
35
44
42
- - -
38
35
55
Tangential
28
22
-
-
17
-
-
45
-
10 13
-
59
-
Average
19
19
18
30
22
34
38
37
37
28 23
47
42
38
'Possible combination of modifications on the boilers tested.
Sources Reference 9.
-------
Two-stage combustion reduces both the temperature and
the availability of oxygen. The fuel is burned once with only
85 to 95 percent of stoichiometric air. The partially combusted
gases are cooled and then burned again at a low temperature with
10 to 15 percent excess air injected through unused upper burners
or through a "NOx-port" located higher in the furnace. All var-
iations of two-stage combustion involve firing some burners fuel-
rich and.others fuel-lean.
Flame temperature can be reduced by injection of steam,
water, or a water-oil emulsion into the combustion zone. The
oxygen concentration is also diluted when steam or water is used.
Reducing air preheat also reduces the flame temperature. These
methods do lower the thermal efficiency of the unit somewhat;
the dilution effect can be enhanced with less loss of efficiency
if a large amount of waste steam is injected instead of the same
weight of water.
Reducing the load or the firing rate reduces N0x emis-
sions as does oversizing the combustion zone. The capital costs
for this unused capacity are high.
In larger furnaces and boilers, fuel dispersion,
primary/secondary air ratios, burner tilt, burner configuration,
and burner spacing are important variables. Changes in burner
arrangement will probably be small in furnaces with only a single
line of burners.
Low excess air firing and off-stoichiometric firing
are the easiest modifications to apply to an existing unit.
Addition of N0x-ports or the addition of flue gas recycle is
frequently a major effort. Retrofit attempts are hampered by
the large number of units and their diversity. The design of a
E-198
-------
particular unit or the lack of space for ductwork may be a
severe restriction. Individual units must be tuned to avoid
the problems of corrosion, reductive decarburization, and deposits
on boiler tubes, flame instability, blowoff, flashback, combustion-
driven oscillations and combustion noise or roar.
4.1.3.2 Internal Combustion Engines
There are several modifications for controlling N0x
emissions from internal combustion engines. The percent N0x
reduction and the limitations for each of these modifications
are given in Table E4-4.
Modification of combustion chamber design now in use in
diesels and in Honda gasoline-powered automobiles may include the
use of a prechamber, turbulence chambers, or "energy cells." A
prechamber provides a type of two-stage combustion with a very
rich first stage and a very lean second stage. This arrangement
is known as a "stratified charge" engine design. "Energy cells"
provide controlled combustion to prevent high-peak pressures and
rough operation. Internal exhaust recirculation, controlling
oxygen availability, special low-NOx combustors, and premixed,
prevaporized and well-stirred external combustors are being
studied for use on turbines.
In the diesel engine, an increase in the specific
gravity of the fuel is accompanied by an increase in NO emis-
sions. N0x emissions from distillate oil-fired gas turbines are
about twice as high as those from natural gas-fired turbines.
In gas turbines, the effect of fuel composition on N0X emissions
may vary with fuel rate and the geometry of the combustion zone,
so each design must be examined separately. Also, limited data
suggest that fuel nitrogen is converted to N0y at a high rate
in gas turbines.
E-199
-------
TABLE E4-4. ENGINE MODIFICATIONS WHICH REDUCE NO EMISSIONS
FROM INTERNAL COMBUSTION ENGINES
DIESELS
SPARK
IGNITION
CAS TURBINES
ENGINE
MODIFICATION
NOx
REDUCTION
(Percent)
PENALTIES &
LIMITATIONS
t
N0X
REDUCTION
(Percent)
PENALTIES &
LIMITATIONS
/
NOx
REDUCTION
(Percent)
PENALTIES 6
LIMITATIONS
Combustion chamber
design
40
Increased first
cost
-
Not applicable
Under
development
Fuel properties
Variable
Availability of
low nitrogen
fuels; higher
operating cost
Variable
Availability of
low nitrogen fuels;
higher operating
costs
Variable
Availability of
low nitrogen fuel
higher operating
costs
Air/fuel ratio
-
Not applicable
30
Backfiring; reduced
power
50
Increased fuel;
decreased power
Exhaust recycle
50
Increased fuel;
decreased power
50
Increased fuel;
decreased power
Under
development
Fuel Injection
tlalng
40
Increased fuel;
decreased power
-
Not applicable
-
Not applicable
Steam or water
Injection
50
Corrosion
problems
50
Corrosion
problems
75
Need delonlzed
water; costly
Variable compression
ratio
Under development
-
Not applicable
—
Not applicable
Source: Referanca 9.
-------
NOx emissions can be controlled in gas engines by
adjustment of the air/fuel ratio because gas engines operate
with premixed homogeneous mixtures. The flame temperature is
reduced with increasing air-fuel ratios, but too lean a mixture
causes backfiring and loss of power and fuel efficiency. Diesel
engines are not amenable to this modification because they do
not have homogeneous mixtures and combustion occurs in fuel-
rich zones. In turbines, the increased air/fuel ratio reduces
N0X emissions but it also decreases the turbine inlet temperature,
causing a severe loss of power and increased fuel consumption.
Exhaust recycle can effectively control N0x emissions
from gasoline engines. However, the modification causes a
decrease in the maximum power and efficiency of the engine.
Additional piping and a control mechanism are needed at a modest
extra cost. Current studies of combustor designs for gas turbines
provide for internal exhaust recycle.
The pressure and temperature in an injected diesel
engine can be varied by changing the timing of fuel injection.
The point in the cycle at which fuel injection is started or the
rate of fuel addition (or both) can be changed. Two-stage fuel
injection is also possible. The effects of these variables on
N0x emissions from stationary engines have not been fully clar-
ified; the changes will probably also affect power output and
thermal efficiency of specific units.
Injecting water or steam with the air into the combus-
tor of an engine reduces N0X emissions, probably because of a
reduction in the temperature of the flame surrounding the burning
drops of atomized fuel. The modification works for diesel, gas
turbine, and spark ignition engines. The amount of reduction
depends on the water- or steam-to-air ratio and the method of
E-201
-------
injection. Deionized water must be used to avoid undesirable
deposits, but it is not always available. Steam may be avail-
able as a by-product from other processes.
The compression ratio in a diesel engine can be changed
by hydraulically raising or lowering the height of the top of
the piston to maintain a preset peak combustion pressure in the
chamber. Control of this pressure avoids the excessively severe
cycles which occur frequently in conventional diesels and which
may account for a disproportionate part of the N0X emissions.
Variable-compression pistons are more costly than conventional
pistons and may require more maintenance.
4.1.4 General Refinery Operations
4.1.4.1 Efficient Gasoline Production
To produce the required volumes of gasoline at the
octane levels specified, the refiner must find the optimum
balance among such parameters as:
• reformer severity
• lead level
• MMT (if used) level in unleaded gasoline
• crude unit cut points (straight run versus
naphtha, naphtha versus kerosene, and diesel
versus gas oil)
• FCC conversion and severity
E-202
-------
All of these factors influence emissions, but with the
exception of lead and MMT, the effect on emissions of choosing
one mix of parametric values over another is minimal when com-
pared to the overall refinery balance. The maximum allowable
levels of lead and MMT are regulated by the government. Nearly
all refiners will operate at the maximum allowable levels of lead
and manganese. Levels of these octane improvers less than the
current maximums result in decreased gasoline volumes and increased
fuel gas or fuel oil burning.
4.1.A.2 Alternate Processing of Intermediate Streams
In most refineries there is some excess capacity in
several of the processing units. This excess is required to com-
pensate for swings in crude composition or results from a desire
to have additional processing flexibility.
In some refineries, portions of the olefin streams
from cracking operations may be polymerized or alkylated. In
the alkylation process isobutane is consumed. The choice between
alkylation and polymerization is based on the values of isobutane,
alkylate and polymer gasoline and the availability of isobutane.
The choice between alkylation and polymerization has no signi-
ficant effect on emissions.
Occasionally, portions of the diesel stream are
charged to the catalytic cracker in an attempt to take advantage
of idle cracking capacity and to make more gasoline. While
opinions differ as to how beneficial this option is, it is cer-
tain to increase emissions to some extent. Lower boiling feed-
stocks do not crack as well as higher boiling stocks, but still
must be heated and vaporized in the riser. Since diesel pro-
duces a negligible coke on catalyst, this heat is not recovered
by combustion of coke in the regenerator and must be compensated
E-203
-------
by additional FCC preheater duty. Once again, higher emissions
are accompanied by higher operating costs.
Some refineries have excess catalytic cracking and
hydrocracking capacity. Processing flexibility is obtained by
splitting the gas oil stream between these two units. While
hydrocrackers do not have a flue gas stack, they do have some-
what higher fuel consumptions than FCC units (100,000 to 200,000
Btu per barrel of fresh feed versus 0 to 100,000 Btu per barrel
of fresh feed.)12 Hydrocrackers also consume 1200-2300 SCF of
hydrogen per barrel of fresh feed. 13 Hydrogen manufacture pro-
duces emissions. The net effect on emissions of switching vol-
umes of gas oil from one of these units to the other is not
significant.
In some refineries, hydrotreating may be used as a
substitute for solvent or chemical treating of products. In these
instances the environmental problems associated with disposal of
byproduct streams (such as acid sludge) are avoided.*3
As a general rule, the refiner's actions to achieve
stable operations which satisfy product volumes and qualities at
minimum cost do not conflict with, and in most cases reinforce,
efforts to reduce emissions.
4.1.4.3 Maximizing Overall Efficiency
Steam Generation and Usage--Boiler stack emissions
may be reduced by maximizing efficiency of the overall boiler
system design and operation, which is reflected in reduced fuel
use. Use of high pressure steam is potentially more efficient
than use of low pressure steam. The problem lies in proper
design. One approach is to reduce steam pressure to the level
required by fractionator reboilers, etc., by passing the high
E-204
-------
pressure steam through power generating turbines, and/or through
turbines driving pumps and compressors.
A simple and energy-saving maintenance procedure is to
ensure that all steam traps operate correctly, specifically that
they pass condensate with minimum steam.
Insulation--Fuel usage in boilers and heaters can be
reduced by proper heat insulation and weather protection of
process equipment and piping which operate significantly above
ambient temperatures. (Some equipment is insulated in colder
climates simply to avoid winter freeze.) With increasing fuel
prices, an "economic thickness" calculation will undoubtedly
justify more insulation than in the recent past.
Fractionator Design and Control—The process engineer
is familiar with the theoretical limits of fractionator operation:
(1) Minimum number of trays - total reflux (no products)
(2) Infinite number of trays - minimum reflux ratio
«-
(minimum tower diameter)
When fuel was cheaper, designs often approached "minimum trays"
because of the relatively low cost of heating and cooling, and
the high cost of equipment. Today, energy costs drive designs
toward limit (2), a great number of trays, with reflux ratios
(reflux rate:distillate rate) approaching the tninimum value.
Further economy can be realized in selected fractionation sys-
tems by "thermal coupling,"1" computerized composition control,15
and generalized system approaches.16 One author states that a
10 percent decrease in distillation energy consumption in the
USA would conserve 0.2 Quads (1.0 Quad = 1015 Btu) per year based
E-205
-------
upon 1976 figures. The saving is equivalent to 100,000 bbl of
oil per day. 1 7
In general, efficient energy utilization in the refin-
ery is in accord with environmental goals.
4.2 Alternative Fuels16
Refiners traditionally have had a choice of natural
gas, refinery gas, or distillate or residual fuel oils as fuel
for refinery boilers and process heaters. The present trend is
toward the heavier oils, both for economic reasons and because
lighter oils are being reserved for smaller consumers with fewer
emissions controls. This trend, however, results in higher
emissions from refinery boilers and heaters.
Emissions factors for the use of natural gas and fuel
oils in industrial boilers are found in Tables E4-5 and E4-6.
According to these factors, particulate emissions from natural
gas and No. 2 oil are independent of sulfur content, but increase
with sulfur content for No. 4 oil and heavier oil (see foot-
note C of Table E4-6).
Sulfur dioxide emissions are, of course, directly
related to sulfur content of the fuel gas or oil. Nitrogen
oxide emissions result from oxidation of atmospheric and fuel
bound nitrogen; the former increasing with combustion temperature,
and the latter increasing according to the square of the bound
nitrogen content. Oxygen partial pressure and point flame tem-
perature are strong determinants of N0X yield.
Table E4-7 presents comparative emissions for natural
gas and fuel oils for 106 Btu heat release. The AP-42-based
E-206
-------
TABLE E4-5. EMISSION FACTORS FOR NATURAL GAS COMBUSTION
Industrial Process
Boiler
Pollutant
lb/106ft3
kg /106m3
Particulates
5-15
80 - 240
Sulfur Oxides (S02)a
0.6
9.6
Carbon Monoxide
17
272
Hydrocarbons (as CHi.)
3
48
Nitrogen Oxides (N02)
(120 - 230)b
(1920 - 3680)b
aBased on an average sulfur content of natural gas of 2,000 g/106 std ft3
(4,600 g/106 N m3).
^This represents a typical range for many industrial boilers. For large
industrial units (100 MM Btu/hr), the N0X factors presented for power
plants are used.
Source: Reference 18.
E-207
-------
TABLE E4-6. EMISSION FACTORS FOR FUEL OIL
COMBUSTION IN INDUSTRIAL BOILERS3
Grade of Oil
Residual oil
lb/10i gal kg/103 liter
Disti
lb/103 gal"
late oil
Pollutant
kg/10 liter
Particulate ^
Sulfur dioxide
Sulfur trioxide
0
Carbon monoxide
Hydrocarbons
(total, as CHO
Nitrogen oxides
(total, as NO2)
c
157S
2S
5
1
60
g
c
19S
0.25S
0.63
0.12
7.58
2
142S
2S
5
1
22
0.25
17S
0.25S
0.63
0.12
2.8
Industrial boilers can be classified, roughly, according to their gross
(higher) heat input rate as >15 x 10 , but <250 x 106Btu/hr
^Particulate is defined as that material collected by EPA Method 5 (front
half catch).
Particulate emission factors for residual oil combustion are best described,
on the average, as a function of fuel oil grade and sulfur content, as shown
below.
Grade 6 oil: lb/103 gal = 10(S) + 3
Where: S is the percentage (by weight) of sulfur in the oil
Grade 5 oil: 10 lb/103 gal
Grade 4 oil: 7 lb/103 gal
^S is the percentage (by weight) of sulfur in the oil.
Carbon monoxide emissions may increase by a factor of 10 to 100 if a unit is
improperly operated or not well maintained.
^Hydrocarbon emissions are generally negligible unless unit is improperly
operated or not well maintained, in which case emissions may increase by
several orders of magnitude.
g
Nitrogen oxides emissions from residual oil combustion in industrial boilers
are strongly dependent on the fuel nitrogen content and can be estimated
more accurately by the following empirical relationship:
lb NO2/IO3 gal = 22 + 400 (N)2
Where: N is the percentage, by weight, of nitrogen in the oil. Note: For
residual oils having high (>0.5%, by weight) nitrogen contents, one should
use 120 lb NO2/IO3 gal as an emission factor.
Source: Reference 19
E-20S
-------
TABLE E4-7. REFINERY FUEL EMISSIONS
AT EQUIVALENT HEAT RELEASE
FUEL TYPE
NATURAL GAS
(a)
#2 F.O.
(b)
#6 F.O
(b)
Gross Heating Value
Specific Gravity
Amount Equivalent to
106 Btu heat release 988 ft3
Emissions lb per 10
Btu:
S02: NSPS
SO2: (av)
Particulates (av)
CO
HC
NOx (av)
Sulfur Wt% )
(oils only) i
1012 Btu/ft3(c)
0.55/Air
2.66 x 10
5.9 x 10"*4
9.9 x 10"3
1.68 x 10"2
3.0 x 10"3
0.17
-3(e)
19,460 Btu/lb(d) 18,200 Btu/lb(d)
0.856/water
7.20 gal
0.306
0.014
0.036
7.2 x 10*3
0.16
0.3
(typical)
0.972/water
6.78 gal
(d)
2.13
0.16
0.034
6.8 x 10~3
0.41
2.0
(high)
(f)
(a)
Reference 20
^Reference 21
Cc)
Accepted value for pure methane
^Reference 22
(e)„„„„ . . , r ^ 0.1 gr H2S w 64 lb SO2 1 lb
NSPS emission llndt of S02 - 10 ftiKnat. gas x 34 lb H;S x TOOOl?
- 2.69 x 10"6 ¦ .lb f2
ft'1 nat. gas
^Residual fuel oil sulfur levels range from 0.3 Wt. percent (N.Y. City)
to 2.0 percent (Midwest) according to sales data as regularly published
in Oil and Gas Journal.
E-209
-------
"average" natural gas contains only about 22 percent of the
sulfur equivalent allowed by NSPS (0.10 grain/10 ft3 gas)
(Column 1). Fuel oils containing 0.3-2.0 weight percent sul-
fur, which represent the broad range of sulfur levels in fuels,
emit from 500 to 3,600 times as much S0X as the "average"
natural gas. Particulates from No. 2 oil are estimated to be
only 1.4 times that of natural gas, but No. 6 fuel oil emits
about 16 times as much.
Carbon monoxide and hydrocarbon emissions at identical
heat release are about two times higher for fuel oils than for
natural gas; nitrogen oxides emissions for gas and No. 2 fuel
oil are about equal and smaller than No. 6 fuel oil N0X emis-
sions by a factor of 2.6.
Over all, the substitution of fuel oils for natural
gas without instituting additional controls will have a detri-
mental impact on the environment.
A.3 Hydroprocessing as Feedstock Pretreatment
Hydroprocessing includes those processes in which
hydrogen is combined with a feedstock and passed over a catalyst
at elevated temperature and pressure. There are three categories
of hydroprocessing, depending on the severity of the process:
(1) hydrocracking, in which 50 percent or more of the feed is
reduced in molecular weight; (2) hydrorefining, in which 10 per-
cent or less of the feed is reduced in molecular weight; and
(3) hydrotreating, in which essentially no reduction in molecu-
lar weight occurs.
Though all three types of processes are used to clean
feedstocks, the main purpose of hydrocracking is the making of
new products; therefore it is not discussed further in this
E-210
-------
section. Hydroprocessing of products (such as lube oils) is
also not discussed here, even though such treatment may contrib-
ute to reduced emissions when the products are used. Hydrorefin-
ing and hydrotreating of feedstocks to be subjected to further
processing are included in this section.
Some examples of hydroprocessing as feedstock pre-
treatment are:
• hydrodesulfurization of residual feedstock to
be used in fuel oil production or catalytic
cracking;
• hydrodesulfurization of heavy gas oils and
middle distillates to be used in the produc-
tion of jet fuels, diesel fuels, and heating
oils;
• hydrodesulfurization of heavy gas oils to be
used as high-quality catalytic cracking feed;
• hydrodesulfurization and hydrodenitrogenation
of naphtha and straight-run crude distillate
streams to be used as primary feeds for the
isomerization and catalytic reforming units.
These and other uses of hydroprocessing are discussed more
thoroughly in Section 4.4 of Appendix F.
Hydroprocessing removes sulfur- and nitrogen-containing
compounds, the heavy metals, oxygen, and halides. Hydrotreating
also stabilizes unsaturated hydrocarbons by saturating the double
bonds. Typical reactions for these operations are given in
Table E4-8.
E-211
-------
TABLE E4-8. TYPICAL HYDROPROCESSING REACTIONS
Desulfurization:
Mercaptans:
Sulfides:
Disulfides;
Thiophenes:
RSH + H2 RH + H2S
R2S + 2H2 -+¦ 2RH + H2S
(RS)2 + 3H2 -+¦ 2RH + 2H2S
HC — CH
HC CH + 4H2 CfcHl0 + H2S
Denitrogenation:
Pyrrole:
Pyridine:
CfcHfcNH + 4H2 CfcHio + NH3
C5H5N + 5H2 C9Hl2 + NH3
Deoxidation:
Phenol:
Peroxides:
C6Hs0H + H2 C6H6 + H20
C7H1300H + 3H2 C7H16 + 2H20
Dehalogenation:
Chlorides:
RC1 + H2 + RH + HC1
Hydrogenation:
Pentenes:
C3Hio + Ha C3H12
Source: Reference 22.
E-212
-------
Desulfurization requires less severe operating condi-
tions that does nitrogen removal. Sulfur removal is less diffi-
cult for paraffins than naphthenes and aromatics.21
Hydrogen consumption is also less for sulfur removal
than for nitrogen removal: about 70 scf/bbl feed/percent sulfur
compared to about 320 scf/bbl feed/percent nitrogen. About
180 scf of hydrogen are required per barrel of feed per percent
removal of oxygen. Hydrogen consumption increases rapidly if
cracking occurs. About two to ten times the stoichiometric
hydrogen are actually required because of hydrogen dissolving
in the oil.2 2
A number of catalysts have been used for hydroprocess-
ing, including cobalt and molybdenum oxides on alumina, nickel
oxide, nickel thiomolybdate, tungsten and nickel sulfides, and
vanadiun oxide. Cobalt and molybdenum oxides or alurania are
used most commonly, but nickel-cobalt-molybdenum or nickel-
molybdenum compounds or alumina are more efficient if nitrogen
removal is a significant factor.22
Catalyst consumption varies from 0.001 to 0.007 lb/bbl
feed, depending on the severity of the operation, the gravity of
the feed, and the metals content of the feed.22 The deposition
of metals from the feed onto the catalyst can be a major problem,
especially with residual feeds. The catalyst should be capable
of remaining active even with accumulated metals. It should also
be cheap enough to be thrown away after a single cycle, since
deposited metals sometimes make regeneration unattractive.23
The overall impact of hydroprocessing is generally
beneficial for- reducing emissions. Removal of objectionable
materials, besides reducing emissions from subsequent processes,
can significantly reduce catalyst poisoning and equipment
E-213
-------
corrosion and can also increase yields. Hydrodesulfurization
catalytic cracking unit feeds is a very effective method for
reducing sulfur emissions from catalytic cracking catalyst
regeneration.
E-214
-------
4.4 References
1. Stover, R. D. Control of Carbon Monoxide Emissions from FCC
Units by Ultracat Regeneration. In: Ind. Process Ses.
for Poll. Control, Proc. of the AIChE Workshop, 6:80-85, 1975.
2. Rheume, L., et al. Two New Carbon Monoxide Oxidation Catalysts
Get Commercial Tests. Oil & Gas J., 74(21):66-70, 1976.
3. Rheume, L., et al. New FCC Catalysts Cut Energy and Increase
Activity. Oil & Gas Journal, 74(20):103-110, 1976.
4. Davis, J. C. FCC Units Get Crack Catalysts. Chem. Eng.,
84 (12):77-79, 1977.
5. American Petroleum Institute, Refining Department, Proceedings
of the 41st Mid-Year Meeting, Am. Pet. Inst. Ref. Dept.,
Washington, D.C. 1976.
6. Hartzell, F. D. and A. W. Chester. FCCU Gets a Catalyst
Promoter. Hydrocarbon Process., 58(7) .-137-140, 1979.
7. Vasalos, I. A., et al. New Cracking Process Controls FCCU
S0x. Oil Gas J., 75(26):141-148, 1977.
8. New Catalyst Cuts FCC Sulfur Oxides 80-90 Percent. Oil Gas
J.,75(13):38-39, 1977.
9. American Petroleum Institute, Division of Refining. Manual
on Disposal of Refinery Wastes, Volume on Atmospheric
Emissions. API Publ. 931. Washington, D.C., 1976.
E-215
-------
10. Iya, K. S. Reduce NOx in Stack Gases. Hydrocarbon Process.,
51(11):163, 1972.
11. Reed, R. D. How to Cut Combustion-Produced N0x-Oil Gas J.,
72(3):63-64, 1974.
12. 1978 Refining Process Handbook, Cracking. Hydrocarbon
Process, 57 (1) : 107-111, 117-120, 1978.
13. Mason, H. F. and R. S. Tolberg. Hydrogen Processing--Clean
Routes to Clean Products. AIChE Symp. Ser., 71(148):222,
1975.
14. Stupin, W. J. Thermally coupled Distillation--A Case History.
Chem. Eng. Prog., 68(10):71, 1972.
15. Smith, D. E.,et al. Distill with Composition Control.
Hydrocarbon Process., 57 (2):99, 1978.
16. Null, H. R., D. E. Steinmeyer, and R. J. Larson. Process
Design for Energy Conservation, Seminar Papers. AIChE
Today Series. AIChE, New York, 1975.
17. Mix, J. J.,et al. Energy Conservation in Distillation.
Chem. Eng. Prog., 74(4):49, 1978.
18. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Monitoring and Data Analysis
Division. Compilation of Air Pollutant Emission Factors.
3rd edition, AP-42. Research Triangle Park, North Carolina,
August, 1977.
E-216
-------
19. Gary, J. J. and G. H. Handwerk. Petroleum Refining, Technology
and Economics. Marcel Dekker, New York, 1975.
20. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Monitoring and Data Analysis
Division. Compilation of Air Pollutant Emission Factors.
3rd edition, AP-42. Research Triangle Park, North Carolina,
April, 1977.
21. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Monitoring and Data Analysis
Division. Compilation of Air Pollutant Emission Factors.
3rd edition, AP-42. Research Triangle Park, North Carolina,
May 1974.
22. Steam, 39th Edition. Babcock and Wilson, 1978, p. 5-17.
23. McKinney, J. D. and J. Stipanovich. Gulf HDS Handles Resids.
Hydrocarbon Process., 50(5):97-102, 1971.
E-211
-------
5.0
CONVERSION FACTORS
To Convert From
To
Multiple By
Btu
kcal
0.252
bbl
A
159.0
gal
i
3.785
ton
kg
907.2
lbs
kg
0.454
cm
in
0.394
ft3
m3
0.0283
psi
kg/cm2
14.223
g/gal
g/A
0.264
Btu/bbl
kcal/ Si
0.0016
kWh/bbl
kWh/ i
0.0063
lb/bbl
kg/i
0.0285
lb/106 Btu
g/Mcal
18.0
grain/ft3
g/m3
2.29
gal/106ft3
£/106m3
133.7
gpm
m3/hr
0.227
lb/1000 gal
mg/Z
119.8
E-218
\
\
-------
TECHNICAL REPORT DATA
(Please read lnurucrions on the rtvene before completing)
1. REPORT NO.
EPA-600/2-80-075d
2.
3 "Ec,pM&f Ac
-------