&EPA
United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park NC 27711
EMB Report 79-CKO-17
September 1979
Air
Iron and Steel
(Coke Oven Battery Stack)
Emission Test Report
Jones & Laughlin
Steel Corporation
Pittsburgh, Pennsylvania
-------
BY-PRODUCT COKE PLANT
BATTERY P4
Jones and Laughlin Steel Corporation
Pittsburgh, Pennsylvania
Prepared for the
U.S. Environmental Protection Agency
Emission Measurement Branch
Research Triangle Park, North Carolina 27711
Prepared and compiled by
Clayton Environmental Consultants, Inc,
25711 Southfield Road
Southfield, Michigan 48075
EMB REPORT NO. 79-CKO-17
Work Assignment 14
Contract No. 68-02-2817
-------
INTRODUCTION TO REPORT
Two firms prepared this report under contract
to the U. S. Environmental Protection Agency. The
report is presented in two sections. Section I
was prepared by Clayton Environmental Consultants,
Inc., Southfield, Michigan and includes test results
for particulate, carbon monoxide, carbon dioxide,
oxygen, and benzene, as well as visible emission data.
Section II was(prepared by TRW Energy Systems Group,
Durham, North Carolina, and contains benzo(a)pyrene
(B(a)P) sampling data only, and immediately follows
Appendix I of the Clayton report.
TRW did all B(a)P sampling for this study and
four other studies conducted for this coke oven
battery stack series. At the time the initial work
assignments were issued, TRW was the only contractor
suitably equipped for B(a)P sampling.
-------
TABLE OF CONTENTS
Page
SECTION I - CLAYTON REPORT
1.0 Introduction 1
2.0 Summary and Discussion of Results 3
3.0 Process Description and Operation 22
4.0 Location of Sampling Points 28
5.0 Sampling and Analytical Procedures 30
APPENDICES
A. Project Participants
B. Field Data Sheets
B-l. Particulate Test Data Sheets
B-2. Sampling Summary Data
B-3. Visible Emissions Data Sheets
B-4. Summary of Visible Emissions
C. Sample Weights
C-l. Particulate Weight by Fraction
C-2. Sulfate Weight by Fraction
D. Gas Chromatograph Data
E. Carbon Monoxide Data
E-l. Continuous Sampling Charts
E-2. Opacity and Carbon Monoxide
Correlation Data
-------
TABLE OF CONTENTS (CONTINUED)
F. Detailed Summary of Sampling and
Analytical Procedures
F-l. Method 110 -
Determination of Benzene from
Stationary Sources
F-2. Method 10 -
Determination of Carbon Monoxide
Emissions from Stationary Sources
G. Example Calculations
o
H. Calibration Data
I. Process Data
SECTION II - TRW REPORT
-------
SECTION I - CLAYTON REPORT
-------
LIST OF FIGURES
Figure Page
1.1. Process Layout - Waste Heat 2
Flue Duct
2.1. Relationship of CO Concentrations, 12
Stack Opacity, and Duct Temperature;
Particulate Test 1
2.2. Relationship of CO Concentrations, 13
Stack Opacity, and Duct Temperature;
Particulate Test 2
2.3. Relationship of CO concentrations, 14
Stack Opacity, and Duct Temperature;
Particulate Test 3
4.1. Sampling Location - Waste Heat Duct 29
5.1. Particulate Sampling Train 32
5.2. Sampling Train for Continuous 37
Monitoring of Carbon Monoxide
5.3. Integrated Bag Sampling Train for 39
Benzene and Orsat analyses
-------
LIST OF TABLES
Table Page
2.1. Particulate Concentrations and Emission 4
Rates, P4 Battery Waste Heat Duct
2.2. Sulfate Concentrations and Emission 6
Rates, P4 Battery Waste Heat Duct
2.3. Sulfate as a Percent of Particulate, 7
by Weight
2.4. Benzene Concentrations and Emission 9
Rates
2.5. Exhaust Gas Composition, P4 Battery 11
Waste Heat Duct
2.6. Summary of Correlation Results 17
2.7. CO and Opacity Correlation Considering 20
Oven Charge Times
3.1. Plant Design and Operation Record 23
3.2. Record of Dusting and Pressure Rise 24
in Oven during Dusting (in
-------
1.0 INTRODUCTION
The U.S. Environmental Protection Agency (EPA)
retained Clayton Environmental Consultants, Inc. to
determine various gas and particulate emissions from
the P4 coke oven battery stack at the Jones and
Laughlin Steel Corporation, Pittsburgh Coke Works,
in Pittsburgh, Pennsylvania. The results of this
study will be used in research and development
efforts for supporting New Source Performance
Standards for coke oven battery stacks in the iron
and steel industry. This study was commissioned as
EMB Project No. 79-CKO-17, Contract No. 68-02-2817,
Work Assignment 14.
The testing program included the following:
(1) triplicate samples of particulate matter;
(2) integrated bag samples for benzene and
Orsat analyses;
(3) continuous carbon monoxide measurement during
the particulate runs (EPA Method 10, NDIR
analyzer);
(4) sulfate analysis of the particulate samples; and,
(5) visible emission observations recorded for the
duration of each particulate sample run.
Auxiliary data included exhaust gas velocities,
temperatures, and flowrates as determined from the tra-
verses. Figure 1.1 presents a schematic of the process/
control system layout as tested.
-------
Plan view
Coke oven
battery
P4
42'
Elevation view
Sampling location
- .60 '-
Underground rectangular waste heat duct
P4
Battery
stack
12.5
wffl.
Figure 1.1. Process layout - waste heat flue duct.
-------
2.0 SUMMARY AND DISCUSSION OF RESULTS
PARTICULATE EMISSIONS
Results of the particulate emission study are pre-
sented in Tables 2.1, 2.2, and 2.3. Tables 2.1 and
2.2 present the concentrations and emission rates of
filterable and total particulate and sulfate, respec-
tively. Concentrations are expressed as grains per
dry standard cubic foot (gr/dscf) and milligrams per
dry standard cubic meter (mg/dscm). Emission rates
are expressed as pounds per hour (Ib/hr) and kilograms
per hour (kg/hr). Table 2.3 presents sulfate as a
percent by weight of filterable and total particulate.
Averages are presented for each sample run in all three
tables.
Table 2.1 indicates the measured filterable concen-
trations of particulate in the waste heat duct ranged
from 0.195 to 0.286 gr/dscf (446 to 655 mg/dscm) and
averaged 0.234 gr/dscf (535 mg/dscm). Concentrations
of total particulate ranged from 0.362 to 0.370 gr/dscf
(829 to 847 mg/dscm) and averaged 0.365 gr/dscf (837
mg/dscm). Emission rates for filterable particulate
ranged from 74.6 to 116 Ib/hr (33.9 to 52.6 kg/hr)
and averaged 89.3 Ib/hr (40.5 kg/hr). Total particulate
emission rates ranged from 123 to 150 Ib/hr (55.6 to
68.1 kg/hr) and averaged 139 Ib/hr (63.0 kg/hr).
-------
TABLE 2.1. PARTICIPATE CONCENTRATIONS AND.EMISSION RATES
P4 BATTERY WASTE HEAT DUCT
Sample
Number
Sample
Date
Stack Gas
Parameters
Flowrate
d scfm
Temp'
F
Concentration
Filterable
gr/d scf
mg/d scm
Total
gr/dscf
mg/dscm
Emission Rate
Filterable
Ib/hr
kg/hr
Total
Ib/hr
kg/hr
1
2
3
5/1/79 46,200 599 0.195 446
5/2/79 39,500 603 0.220 505
5/3/79 47,300 604 0.286 655
Average 44,300 602 0.234 535
0.364 834 77.2 35.0 144 65.4
0.362 829 74,6 33.9 123 55.6
0.370 847 116 52,6 150 68.1
0.365 837 89.3 40.5 139 63.0
Total Particulate = Front half (filterable) + Back half.
Front Half = nozzle, probe, filter, and front half of filter holder.
Back Half = back half of filter holder, flexline, impingers, and connectors
-------
Filterable concentrations of sulfate (Table 2.2)
ranged from 0.100 to 0.141 gr/dscf (228 to 324 mg/dscm)
and averaged 0.117 gr/dscf (268 mg/dscm). Total sulfate
concentrations ranged from 0.228 to 0.242 gr/dscf (522
to 554 mg/dscm) and averaged 0.235 gr/dscf (538 mg/dscm).
Emission rates ranged from 37.2 to 57.3 Ib/hr (16.9 to
26.0 kg/hr) for filterable sulfate and averaged 44.7
Ib/hr (20.3 kg/hr). Total sulfate emission rates ranged
from 79.8 to 95.8 Ib/hr (36.2 to 43.4 kg/hr) and aver-
aged 89.4 Ib/hr (40.5 kg/hr). Table 2.3 indicates
that sulfate as a percent by weight of filterable par-
ticulate ranged from 49.4 to 51.2 percent and averaged
50.1 percent. Sulfate as a percent by weight of total
particulate ranged from 61.6 to 66.4 percent and aver-
aged 64.4 percent.
Sulfate represents the major constituent of the
filterable and total particulate fractions. This
indicates that by reducing the sulfur content of the
coke oven gas, the particulate concentrations should
correspondingly drop. The planned gas desulfurization
is, therefore, a noteworthy addition.
The general reproducibility of the particulate emis-
sion data collected for all three sample runs was good
although magnitudes were somewhat higher than expected.
Previous testing on Battery P4 had indicated an average
-------
TABLE 2.2. SULFATE CONCENTRATIONS AND EMISSION RATES
P4 BATTERY WASTE HEAT DUCT.
Sample
Number
Sample
Date
Stack Gas
Parameters
Flowrate
d s c f m
Temp
F
Concentration
Filterable
gr/dscf
mg/d scm
Total
gr/dscf
mg/dscm
Emission Rate
Filterable
Ib/hr
kg/hr
Total
Ib/hr
kg/hr
1
2
3
5/1/79
5/2/79
5/3/79
Average
46,200
39,500
47,300
44,300
599
603
604
602
0.100
0.110
: 0.141
0.117
22,8
251 .
324
268
0.
0.
o.
0.
242
236
228
235
554
539
522
538
39,
37.
57.
44.
5
2
3
7
1,7.9
16.9
26.0
2.0,. 3
95.8
79.8
92.5
89.4
43.4
36.2
41.9
40.5
Total Particulate = Front half (filterable) + Back half.
Front Half = nozzle, probe, filter, and front half of filter holder.
Back Half = back half of filter holder, flexline, impingers, and connectors
-------
TABLE 2.3. SULFATE AS PERCENT OF PARTICULATE BY WEIGHT
Sample
Number
Filterable
Percent
Total
Percent
51.2
49.8
49.4
66.4
65.1
61.6
Average
50.1
64.4
-------
filterable particulate concentration of 0.0903 gr/dscf,
61.5 percent lower than this test series. These previous
tests were performed at a less than ideal location.
The heat canal was accessed through four randomly spaced
ports. This random spacing did not permit sampling to
be performed in accordance with Method 1 specifications
for sample and velocity traverses of stationary sour-
ces. For this reason, the validity of the data from
the previous testing program may be questionable.
The heat canal, for this study, was accessed through
ports located in the top of a vertical cut-off damper
slot. The location of these ports was in accordance
with Method 1. Since the slot tapered towards the
bottom, the possibility existed of accidentally picking
up material by scraping the nozzle on the walls of the
slot, as the probe was being raised or lowered through
a port. This would have biased the sample values to-
ward the high side. It is therefore unlikely that any
random scraping was influential, since the test results
are consistent and reproducible.
BENZENE EMISSIONS
Results of the benzene analyses are presented in
Table 2.4. Benzene concentrations ranged from 0.6 to
1.6 ppm and averaged 1.1 ppm. Emission rates ranged
*a
An average of 10 runs from a study conducted in April,
1975 by Betz Environmental.
-------
TABLE 2.4. BENZENE CONCENTRATIONS AND EMISSION RATES
Sampling
Location
Sample
Number
Sample
Date
Concentration
ppm
Emission Rate
Tb/hr
kg/hr
- P4
Battery
Stack
5/1/79
5/2/79
5/3/79
0.6
1.6
1.1
0.3
0.6
0..5
0.1
0.3
0.2
Average
1*1
0.5
0.2
-------
from 0.3 to 0.6 Ib/hr (0.1 to 0.3 kg/hr) and averaged
0.5 Ib/hr (0.2 kg/hr). These results show a high degree
of reproducibility, although they are lower than had
been anticipated.
EXHAUST GAS COMPOSITION
Table 2.5 displays the results of the exhaust gas
composition analysis, Determinations of carbon dioxide,
oxygen, and carbon monoxide contents were made for each
of the three sample runs. Moisture content is also
presented and shows an average of 16.0-percent. The
initial gas composition sample was voided due to a
leaky bag. The results of Sample Nos. 2 and 3 were
averaged and these averages were used for Sample No. 1
determination.
VISIBLE AND CARBON MONOXIDE EMISSIONS
Visible emissions from the coke oven Battery P4
stack were recorded for the duration of each particulate
sample run, except Run No. 3, for which visible
emissions were abbreviated due to inclement weather.
The observations were performed in accordance with EPA
Method 9 by a qualified visible emissions observer. A
graphic summary of opacities is presented in Figures
2.1, 2.2, and 2.3. Additional visible emission data is
10
-------
TABLE 2.5. EXHAUST GAS COMPOSITION - P4 BATTERY WASTE HEAT DUCT
Sample
Number
Moisture
Content
Percent
Exhaust Gas Composition, Dry Basis
Percent
Carbon
Dioxide
Oxygen
Carbon
Monoxide
Nitrogen
and Inerts
1
2
3
Ave rage
16
16
15
16
.3
.0
.6
.0
3
4
3
3
.6a
.0
.1
.6
11. 6a
10.6
12.5
11.6
0.2a
<0. 1
0.3
0.2
84.
85.
84.
84.
6
4
1
7
Sample 1 was voided due to a leaky bag. and an average of Samples 2 and 3
was used.
-------
Charge cycle-ii
Particula te
Run No. 1
1450-1844
1530
1600
1630
1700
1730
1GOO
1830
1900
TIME
Figure 2.1. Relationship of CO concentrations, stack opacity, and duct temperature.
-------
harge Cycle-:
Oven Number •
Particulate
Run No . 2 .'
1045-1407
1450.
Figure 2.2.
Relationship of CO Concentrations, stack opacity, and duct
temperature.
-------
t Charge cycle-
rticula te
Run No. 3
1049-1540
1105 ,
1130
1200
123-0
1300
1335
Figure 2 3 Relationship of CO Concentrations stack opacity, and duct temperature
-------
included in Appendix B. Carbon monoxide concentrations
are also plotted on Figures 2.1, 2.2, and 2.3.
Correlations seemed to exist between the continuously
monitored concentrations of carbon monoxide (ppm), the
actual charging operation of individual coke ovens
(oven charges indicated by a vertical line above the
axis), and opacity; thus, a linear regression and
correlation analysis was performed on the data.
Based on these statistical results, further statistical
analyses of the data would be conducted if such informa-
tion seemed beneficial to a better interpretation of the
data. The following presents the methodology and
results of these statistical analyses.
General Procedures
The time-concentration curves were reduced in the
following manner: The carbon monoxide strip chart
continuous readings were reduced to individual 15-second
average concentrations to correspond with discrete 15-
second opacity readings. A data file was then created
which included time and each corresponding opacity and
CO reading at that time. One such data file was created
for each sample run.
Since subjective observations indicated that CO
concentration peaks were generally preceded within a
few minutes by a rise in opacity, a computer program
was devised which would accommodate and adjust the
data set pairings to any given lag time. The lag time
15
-------
indicates the time, in minutes, before or after (negative
or positive) the opacity for the associated CO reading.
Each data file was then run through a linear regression
program and correlation routine to determine if a valid
relationship existed between the data for any given
lag time. Different lag times (positive and negative)
were used in determining the maximum correlation coeffi-
cient, beginning at whole minute intervals, then reduc-
ing it to quarter-minute intervals. This usually
required five to ten runs per sample.
Several problems necessitated altering the data
inputs to accommodate a more realistic analysis. For
example, steam exiting the quench tower occasionally
obscured the battery stack emissions. Therefore,
for these points in time, there would be associated
CO readings but no opacity readings. Thus, these
data could not be counted as a valid data set. The
number of complete pairs of data available for correla-
tion then, i.e., the number of data sets used, was less
than the total number of pairs first described (above).
Therefore, a new data file was created based on the
optimum time lag, using only complete data sets in
the subsequent analysis.
Results
Table 2.6 presents the results of the correlation
analyses. Each run showed an optimum lag time to produce
a maximum correlation coefficient (r) which varied from
16
-------
TABLE 2.6. SUMMARY OF CORRELATION RESULTS
Sample
Number
Number
of
Data
Sets
% of
Data
Used
Correlation
Coefficient
(r)
Linear Regression
Equation
Lag Time,3
Minutes
810
100°
89.0
86.5
0.23
0.33
0.33
CO = 11.1 o.p + 570
CO = 13.3 op + 505
CO = 14.0 op + 497
2.5
478
100.u
92.1
79.9 .
0.52
0.65
0.75
CO = 16.0 op + 270
CO = 17.8 op + 226
CO = 20.5 op + 176
394
100 u
95.7
85.3
0.55
0.64
0.73
CO = 10.6 op + 298
CO = 11.3 op + 270
CO = 12.6 op + 227
aMinutes from opacity reading to the carbon monoxide reading.
100% of data implies total number of data sets. Other values are percentages
of this number.
-------
2.5 to 5 minutes. Using 100-percent of the data
produced an r value of 0.55 or less for each sample.
This may or may not be relevant, but does suggest that
there could be factors affecting opacity other than
CO and factors affecting CO other than opacity which
may independently affect the results, in addition to
any relationship shown to exist between CO and
opacity.
The computer outputs also included a table
showing the data distribution (Appendix E-2). Using
the distribution and the file of correlated data points^
it may be seen which pairs were most likely affected
by other variables (i.e., the outliers on the distribu-
tion). Without sufficient operations data to interpret
what other variables were affecting the distribution,
it was impossible to logically eliminate outlying
points. Therefore, certain data pairings were eliminated
from further computer runs according to the following
methodology: When CO readings were high (i.e., 500 ppm)
during periods of low opacity (i.e., 0-percent), it was
intuitively apparent that factors affecting CO levels
were probably not affecting opacity at that time.
Therefore, a certain percentage of data sets were
deleted and the correlation analysis rerun using the
same optimum lag time. This reduced the total data
18
-------
sets considered (shown on Table 2.6 as a percentage
of the total) and increased the correlation coefficient
in each case.
In Samples 2 and 3, the r value was 0.7 using about
80-percent of the data. The r value from Sample 1 was
extremely low (i.e., 0.23) using 100-percent of the data,
Upon examination of the data distribution it was dis-
covered that 72-percent of the opacity readings were
5-percent, showing very little relationship at all
with CO readings.
In the next step,an additional variable, oven
charge time, was introduced into the data file to
further investigate.the relationship of CO and opacity.
The times of oven charging were marked with a
"C" within the original data file. The program was
structured to utilize various intervals around the
charge period. The correlation analysis was then
rerun based on the optimum time lag as determined from
the first set of runs (Table 2.6).
The results are presented in Table 2.7. Optimum
charge intervals varied from 0.25 to 2.75 minutes.
Sample 3 was run at two intervals surrounding the charg-
ing activity, examining the paired CO and opacity data
at both 15-seconds and 60-seconds before and after the
charge. Samples 1 and 2 were based on 1.0 and 2.75
19
-------
TABLE 2.7. CO AND OPACITY CORRELATION CONSIDERING OVEN CHARGE TIMES
Charge Number % of
Sample interval, of Data
Number Minutes Data used
Sets
90 100b
94.4
1 1
90.0
76.7
to .
o ' —
112 100b
2 2.75
79.5
59 100b
1
84.8
20 100b
0.25
85.0
Correlation
Coefficient
(r)
0.48
0.47
0.61
0.74
0.75
0.92
0.78
0.88
0.90
0.93
"\ £1 O
Linear Regression m. -
77 ^j Time,3
Equation '
Minutes
CO =
CO =
CO =
CO =
CO =
CO =
CO =
CO =
CO =
CO =
21,
22.
29.
27.
24.
30.
12.
13.
15.
16.
0
6
2
6
9
"•
0
3
6
2
0
op -
op -
op •
op -
op -
op -
op -
Op H
op -^
op -
H 461
1- 454
2.5
f- 377
f 324
1- 160
5
h 19.3
h 320
h 264
h 301
h 270
aMinutes from opacity reading to the carbon monoxide reading,
100% of data implies total number of data sets. Other values are percentages of this
number.
-------
minute intervals surrounding the charge, respectively.
Samples 2 and 3 (at the 1.0 minute interval) had correla-
tion coefficients of 0.75 and 0.78, respectively.
By reducing total data pairs to 79.5 and 84.8-percent,
respectively, the same samples had respective correla-
tion coefficients of 0.92 and 0.88. Sample 1 was again
lowest with a coefficient of 0.48 based on 100-percent
of the data, and 0.74 using 76.7-percent of the data.
Based on the results of Sample Nos. 2 and 3, for
80 percent of the time there is good correlation between
opacity and carbon monoxide during the charge intervals.
Additional information would be needed to validate this
relationship, such as the occurrence of extraneous factors
which influence either opacity or carbon monoxide, when
these factors occur, and the magnitude of these influences.
With this information, any given charge interval could be
assessed for its independence from other factors, such as
topside explosions and fires, which could then be isolated
and rationally removed from the carbon monoxide and opacity
data. The data used in the correlation analysis would then
be more qualified and the correlation coefficient would be
more meaningful. With this approach to the analysis, the
correlation is both quantitative and qualitative. Based on
the number of data pairs used, opacity and carbon .monoxide
can be related quantitatively over a percentage of time and
the resulting correlation coefficient would indicate the
quality of the relationship.
21
-------
3.0 PROCESS DESCRIPTION AND OPERATION
(supplied by Midwest Research Institute)
There are five coke oven batteries at the Jones
and Laughlin Pittsburgh Works, designated as PI, P2, P3S,
P3N, and P4. The Jones and Laughlin plant was selected
for testing of battery stack emissions because the plant
uses the silica dusting maintenance technique, coupled
with spray patching and troweling. This technique was
used on several of the batteries, including P4, which
was selected for testing because it provided the best
testing location: in the underground rectangular duct
that carries the flue gases from the waste heat canal
to the battery stack.
Battery P4 is a 79-oven Koppers underjet battery,
underfired with undesulfurized coke oven gas. A gas
desulfurization unit is nearing completion but was not
in operation during the testing. The P4 battery was
originally started up in 1953. It underwent a hot end-
flue rehabilitation in 1976 and was placed back in opera-
tion in early 1977. The battery was operating on a 17-hr
coking time during the testing. Other information about
the battery is shown in Table 3.1.
Use of silica dusting on battery P4 was begun in early
1978 and has continued since then. Each oven has been
"dusted" at least once and some as much as four times.
The dates when each oven was dusted are shown in Table
3.2 along with the pressure rise in the oven, in inches of
22
-------
TABLE 3.1 PLANT DESIGN AND OPERATION RECORD
Date Monday, April 30,
1979
Plant Name Jones and LauKhJLin
Plant Location Pittsburgh, Pennsylvania
Battery No. P-4
Name of Plant Contact Mr. Jim Saunders, Assistant Superintendent.
By-Products Department
Type of Ovens and Designer Koppers - Underlet . .
Date Built Started up in 1953
Date of Last Rehabilitation Early 1977
Type of Last Rehabilitation Hot end flue rehabilitation
Number of Ovens Total 79 In Service 79, except 1 for dusting
Size of Ovens Height 13 ft , Width 17 in. . , Length 40 ft
Type of coke produced Metallurgical .
Normal coking time (hr) l.Z_
Coal charged per oven (tons) 16 to 16.5 Produce about 11.0 tons of coke
Reversal period (min) 30 min
Nozzle decarbonization method Part of flue gas is recirculated to decarbonize
Is flue gas recirculated? (See above)
Type of fuel gas COG Heating value ____________ Btu/scf
Is fuel gas desulfurized? No. ' - -
Note use of stage charging, preheated coal, etc. Stage charging
Stack height and top diameter 225 ft 12.5 ft ID at bottom, 9.5 ft ID at top
Test location (stack or(waste heat canal)) 8 by 8 ft (provide sketch)
Control method used Patching and silica dusting
Fuel gas analysis Coal analysis
Component Vo1.% Component Vol.%
C02 Ash
111. S ;.
00 HO
CO VH
H,
Cfl
23
-------
JL-42768
TABLE 3.2.
RECORD OF DUSTING AND PRESSURE RISE IN
OVEN DURING DUSTING (IN. H20)
1
6/15/78-7..8
6/19/78-8.4
9/9/78-7.7
5/17/73-6.1
8/9/73-8.0
3/12/78-7.9
fi/24/78-8.6
6/1/78-5.0
7/12/78-8.0
5/7/78-7.4
11/26/78-4.0
6/10/78-5.0
7/26/78-7.4
5/25/78-5.0
R/lfi/7R-fi.n
6/6/78-4.5 10/20/78-7.8
7/-W7fi-fi.n
6/25/78-4.0
6/23/78-7.3
7/18/78-7.4
3/14/78-4.1
7/25/78-8.4
6/27/78 11/11/78-6.5
10/30/78-8.8
6/20/78-3.1
9/20/78-7,4
6/26/78-4.2
7/20/78-5.0
6/12/78-6.3
6/14/78-8.5
6/3/78-4.2
4/2/78-5.4 7/21/78-6.5
7/11/78-5.5
3/3/78-3.9
7/-n/7H_a.?
2/7/78-1.0 i
11/17/78-7.7
, 3/30/78-5.3
"'/10/78-7,?
4/15/78-5.0
J./23/78-6.7
4/28/78-4.0 10/8/78-7.0
7/7/7R-7.R 11/1/78-7.7
3/1/78-4.2
.5/8/78-3.9
10/27/78-7.4
7/26/78-5.0
8/27778-7.8
2
3/19/78-5.0
.5/16/78-4.2
11/5/78-6.0
5/11/78-4.9
8/31/78-4.0
3/6/78-4.9
7/30/78-6.4
4/14/78-6.1
3/22/78-5.4 12/2/78-6.4
7/6/7R.5.R
4/26/78-5.5
11/14/7R-6.0
5/5/78-6.0
8/6/7R-6.4
5/21/78-6.1 3/27/79-2.8
8/l/7«_7_6
5/10/78-4.2
6/30/78-6.0
8/25/78-4.3 8/19/78-7.4
7/4/7R-S.O
4/18/78-6.0 11/24/78-7.4
9/22/78-3.0 4/4/79. 7. R
4/5/78-6.3
7/16/78-6.2
5/20/78-6.1
11/13/78-7.2
5/4/78-7.5
5/26/78-6.5
11/V7R_7.fi
5/13/78-5.0
__7/5/7fi-5.4
5/18/78-6.2
6/4/78-5.0
10/7R/7R-6.8
4/28/78-7.4 3/28/79-4.0
7/14/78-8\7
6/8/78-4.3
7/Q/7R-6.0
4/21/78-8.0 11/8/78-9.7
Q/n/7R_s.n
6/6/78-3.9
11/18/78-5.0
5/1/78-7.0
5/31/78-5.0
9/6/78-7. n
6/12/78-5.0 4/30/79-4.0
RM/7R-7.2
3/19/78-5.0
3
3/3/78-4.2
8/ll/78-7n6
2/10/78-2.9
11/16/78-6.0
5/26/78-4.0
5/22/78-5.0 4/1/79-5.1
7/1/78-7.7
4/19/78-6.4
8/14/78-6-0
4/26/78-3.0
7/13/78-6 0
5/15/78-5.0
6/22/78-7.6
6/13/78-6.6
6/17/78-8.1
8/29/78-8.4
6/27/78-6.8
5/14/78-5.2
6/18/78-5.1
6/24/78-7.6
1/2/79-4.5
6/11/78-8.4
6/25/78-7.2 11/7/78-7.0
in/i4/7s_R,n
6/22/78-7.2
10/1R/7R-7.0
6/8/78-4.0
4/8/78-5.9 4/3/79-3.8
7/74/7R-Q.O
3/2/78-5.7
10/7/7R-7.0
3/16/78-3.7
1 1/??/7fi-^,n
3/24/78-4.1.
3/8/78-4.3
4/4/78-6.0
5/2/78-5.0
24
-------
water, that is measured during the dusting. This table
shows only the dates when patching and dusting occurred,
but not the dates when additional patching without
dusting may have been carried out.
During 1978, the ovens were patched and dusted
more frequently than in 1979, but on a random basis.
Recently Jones and Laughlin has instituted a systematic
maintenance technique utilizing computerized maintenance
record keeping that includes observation of whether
the stack "smoked" or did not "smoke" after charging
of specific ovens. The computerized records show the
type and date of any oven maintenance and the date
the stack was observed and the number of minutes that
the stack smoked after a specific oven was charged.
These data are now used to select ovens which should
be dusted, rather than the random basis on which ovens
were previously selected for dusting. Jones and
Laughlin personnel believe this computerized information
system should enable them to achieve much better control
of battery stack emissions.
The actual patching and dusting operation usually
requires at least 24 hours in which an oven is out of
service. During the first 8 hours, spray patching and
troweling of the end flues is carried out by a three-
man crew, on both the coke side and pusher side of
the oven. During the next shift, a three-man crew "dusts"
25
-------
the oven with all doors and lids closed and sealed.
This takes 2 to 3 hours and uses about 100 Ib of the
silica dust. The dust is hand-fed into a small hopper
that is air aspirated to disperse the dust and carry
it into the oven through a special charging lid. Pressure
rise in the oven is monitored during this operation
and usually increases from 3 to 8 in. of water.
No "dusting" was carried out during the testing
because of possible interference with the particulate
tests. However, a dusting operation on one oven was
observed on the day preceding the first test. It
was observed that silica dust leaked out the chuck
door at first. The chuck door was tightened, after
which the dust began leaking out the coke-side door.
We were told that this was an abnormal example of the
dusting operation but it does demonstrate some of the
problems that can occur when using this technique.
Overall, the condition of the walls in the P4 ovens
appeared to be very good, with few visible cracks in any
of the ovens.
During each test day process operating data was
obtained at approximately 1-hr intervals, and the time
that each oven was pushed and leveled was recorded when-
ever possible. Copies of circular charts showing process
data were obtained, along with coke and coal analyses done
by Jones and Laughlin, and fuel gas analysis. Also,
26
-------
some flue inspections were made and noted along with
other observations or information. All of these
records are shown in 'Appendix I.
Jones and Laughlin personnel cooperated with the
test team during the testing, and appreciation is
expressed for the help provided by Mr. Jim Saunders,
Assistant Superintendent, and Mr. Ken Kobus, Heater
Foreman.
27
-------
4.0 LOCATION OF SAMPLING POINTS
The underground waste heat duct sampling location
was accessed through a vertical cut-off damper slot.j
located approximately 42-feet downstream from the
battery waste heat flue centerline and 60-feet upstream
of the battery stack. The duct itself is located
4 - feet 3 - inches below ground level with an 8 x 8-foot
(nominal) cross-section. This particular location was
selected since there was no feasible method to install
ports in the concrete battery stack. The damper slot
measured approximately three inches across at the
ground level elevation and tapered to one-inch across
at the underground duct entry point.
The six sampling ports on top of the damper slot
were spaced 16 inches on center. Each vertical traverse
consisted of five sampling points; six were originally
proposed but, due to an irregular build-up of deposit
on the bottom of the duct, the sixth point was eliminated,
Velocity pressures and temperatures were measured at
each of the 30 sampling points. Figure 4.1 is a diagram
of the sampling location showing each of the traverse
points and their respective distances from the duct walls
28
-------
80"
NJ
Points
1
2
3
4
5
Vertical Distance
Inside Duct
feet
0.7
2.1
3.5
4.9
6.3
Dep
centimeters
21.3
64.0
106.7
149.4
192.0
- '.
f *
* * ' -
t *
^ ,
' *
f '
• '
,
4
•;•
, i
*
• '•
A B C |
=a I" -v. ;
LJ 1
Slot '
, • ' • • i ' " • * * 'i ' ' , ••'* •' • ''
'* 't • > J * * f'
!
S3 jg— T -7r| C=
i E F
3"
i
'Vr''. \''>! •'•/'','. ''-'' -t'r
I
2
» • • • " • •
3
4
*
5
.•:.;•.'.'.'.' '/.'-'/ '.. •;.'.'-.-''••.'"-.•::.•
--.'.
' ,
'
''
'
. , '
'
• ;
. -
- •
Ground level
Avg 5'U1
Section view
Avg 6 '11'
8'0'
Figure 4.1. Sampling location - waste heat duct
-------
5.0 SAMPLING AND ANALYTICAL PROCEDURES
PARTICULATE EMISSIONS
Triplicate two and one-half hour particulate sam-
ples were extracted isokinetically for five minutes at
each of 30 sampling points in the waste heat duct.
During each test, the probe, Pitot tube, and thermo-
couple assembly were moved to each sampling point,
the velocity pressure and temperature of the exhaust
gas were measured^ and isokinetic sampling flowrates
were adjusted accordingly^ An orifice-type meter was
used to indicate instantaneous flowrates. \
Proper nozzle alignment with the flue gas stream
was maintained throughout the test by clamping the probe
assembly at the top of each port. As a result of this
vertical support system, there was no unusual difficulty
in measuring the actual gas velocity pressures. The
impinger assembly was moved as required to gain access
to each vertical entry port. Some interference was
anticipated from the push ram (on the pusher car) passing
above the sampling location; however, proper synchroniza-
tion of the push ram operation with the probe position
eliminated this potential problem. All field data
sheets are included in Appendix B.
30
-------
The sampling train was checked for leaks before
and after each sample run in accordance with the require
ment that the initial leak rate shall not exceed 0.02
O
ft-'/min at 15 inches of mercury vacuum and the final
leak rate shall not exceed 0.02 ft^/min at the greatest
vacuum incurred during the test.
A modified EPA Method 5 sampling train was used
(Figure 5.1). The sampling train consisted of a
gharp, tapered, stainless steel sampling nozzle; a
14-foot stainless steel probe assembly (instead of
glass); a heated preweighed 110-mm glass-fiber filter;
flexible Teflon^ tubing leading to two Greenburg-
Smith impingers, the first modified, the second
standard, each containing 100-ml of distilled water;
an empty modified Greenburg-Smith impinger; a modified
Greenburg-Smith impinger containing approximately 400
grams of silica gel; a leakless pump with vacuum gauge;
a calibrated dry gas meter equipped with bimetallic
inlet and outlet thermometers; and, a calibrated orifice-
type flowmeter that was connected to a O-to-10-inch
range inclined (water gauge) manometer. As a result of
the unusual configuration of the underground sampling
location, it was impossible to utilize a 14-foot glass
probe for sample extraction, due to obvious difficulties
with flexing and fracturing.
31
-------
Stainless
steel probe
Heated 110-mm
Type A glass-
fiber filter
~]
S-type
tube
Pitot
Braided Teflon
tubing
Inclined
manometer
Orifice
Inclined
manometer
if
0
r
oro
-5
a
O/J
oO
100-ml
.distilled
water
y;: -| |J
Dry 400g
trap silica
gel
Thermometers
Main Vacuum
valve gauge
Vacuum
pump
Figure 5.1. Particulate sampling train.
-------
The impinger train was immersed in an ice bath
to maintain the temperature in the last impinger at
70F or less. All of the sampling train glassware was
connected by ground glass joints, sealed with stopcock
grease, and clamped to prevent leakage. A calibrated
S-type Pitot tube was connected to the sampling probe
and velocity pressures were read on the inclined
manometer. An iron-constantan (I/C) thermocouple,
attached to the Pitot-probe assembly, was connected to
a calibrated pyrometer. During the course of testing, the
filter temperature was kept at 250 + 25F.
Following each sample run, the entire sampling
train was transferred to a sheltered clean-up area.
The probe and nozzle assembly., and the front half of the
filter holder were initially rinsed and brushed with
water, then acetone. The two rinsings were collected
(Si
in separate glass sample bottles with Teflon^ -lined
caps. The glass-fiber filter was returned to its original
petri dish and sealed. The volumes of the impinger
contents were measured and volume increases recorded.
The solutions were placed in glass sample bottles with
Teflon^-lined caps. The impingers were first rinsed
with water, then acetone. The water rinsings were placed
in the same sample bottle as the impinger solutions,
33
-------
and the acetone rinsings were placed in a separate
glass sample bottle. The flexline and back-half of
the glass filter holder were brushed and rinsed
with water, then acetone, and the rinsings placed in
the respective sample bottles. The silica gel was
weighed to determine the weight gain (as condensate).
Thus, five fractions were collected for each
particulate sample:
(1) water rinsings of probe and nozzle assembly,
and front-half of the filter holder;
(2) acetone rinsings of probe and nozzle assembly,
and front-half of the filter holder;
(3) 110-mm type A glass-fiber filter;
(4) impinger contents and water rinsings of back-half
of filter holder, flexline, and impingers; and
(5) acetone rinsings of back-half of filter
holder, flexline, and impingers.
Filterable particulate was the sum of Fractions
1, 2, and 3. Total particulate was the sum of Fractions
1 through 5. The particulate weights by fraction are
presented in Appendix C.
34
-------
In the laboratory, the liquid fractions were
measured volumetrically and placed in tared beakers.
A five milliliter aliquot was taken from each
fraction for sulfate analysis. The water fractions
were then evaporated to residue at 105C and the partic-
ulate weight determined. The acetone fractions were
evaporated at room temperature and weighed until
constant. The filter was desiccated at room temperature
and weighed until constant. All weight determinations
were performed on an analytical balance having a
sensitivity of 0.1 milligrams.
For the determination of sulfates in the liquid
samples, the 5-ml aliquot was brought up to 25-mls
with 80-percent IPA. The filter was also combined
with 80-percent IPA. The acidity was adjusted with
perchloric acid to a pH of between 2.5 and 4.0. Three
to five drops of thorin indicator were then added and
the solution titrated with standardized barium perchlo-
rate. The results are reported as sulfuric acid
(including sulfur trioxide), and as a percent of total
particulate.
CARBON MONOXIDE SAMPLING
A sample of flue gas was drawn through a stainless
steel probe, Teflon® tubing, and then through a
particulate and condensate trap containing a glass
wool plug, to a 3-way valve. This valve was used to
35
-------
divide the gas sample into two streams; one for the
continuous analysis of carbon monoxide and one to
provide an integrated bag sample for the determinations
of benzene content and exhaust gas composition by
the Orsat method.
The gas stream used for carbon monoxide monitoring
was then passed through two modified Greenburg-
Smith impingers, the first containing approximately
250 grams of silica gel and the second containing
approximately 500 grams of Ascarite^, for moisture
and carbon dioxide removal, respectively. Finally,
a leak-free diaphragm pump forced the sample through
a needle valve to a rotameter and the Beckman Model 865
NDIR analyzer. At the sample interface, a flowrate of
approximately 1,5 cfh with a delivery pressure of
10psig was maintained for the duration of the contin-
uous sampling. An analog strip chart recorder was
used to record all instrument outputs. This sampling
system is depicted in Figure 5.2.
The daily calibration sequence included passing
a certified standard zero gas (dry nitrogen) and
a certified standard span gas in concentrations of
9,900 ppm or 29,800 ppm carbon monoxide in nitrogen
through the analyzer. The instrument output was
calibrated for two anticipated ranges, 0-10,000 and
0-30,000 ppm carbon monoxide by adjustment of the
' 36
-------
Stainless steel probe
®
Teflon sampling line
Three-way valve
To integrated
bag sample
J&-"
-------
zero and gain settings for the appropriate signal, as
indicated on a calibration curve. The instrument
operates by the Luft principle as specified in
Reference Method 10. A 4-position valve allows the
introduction of sample gas or any of the required
standard calibration gases.
The actual measured concentrations of CO were
determined by adjusting the recorded strip chart values
with the factory calibration curve for the 16-mm
cell, which had been adjusted to the standard gas
concentrations used in the field.
INTEGRATED BAG SAMPLING (BENZENE AND ORSAT)
An integrated bag sample was withdrawn from the
waste heat duct simultaneously with each particulate
sampling run utilizing a train as depicted in Figure 5.3,
and as described previously under Carbon Monoxide Sampling,
An evacuated Tedlar® bag with a volume of 100 liters
was placed inside an insulated steel drum. The drum
was then gradually evacuated, thereby filling the
(R)
Tedlar bag at a controlled flowrate, using a
rotameter and valve assembly as shown in Figure 5.3.
When the bag was filled, it was removed and transferred
to a laboratory for immediate gas chromatographic (GC)
analysis of the benzene content and later, Orsat
analysis.
38
-------
. Stainless steel sampling line
Teflori©tubing
Stainless
steel
probe
Needle
valve
Dry trap
with glass
wool plug
Vacuum
pump
Insulated steel drum
Figure 5.3. Integrated bag sampling train,
-------
The method used for the determination of benzene
concentrations is in accordance with the EPA gas chromato-
graphic Method 110, "Determination of Benzene from
Stationary Sources", delineated in Appendix F-le
Gas chromatographic field analyses were performed
utilizing an Analytical Instrument Development (AID)
Model 511, portable gas chromatograph with a flame
ionization detector and a 61 x 1/8" stainless steel
column packed with 1.75-percent Bentone and 5-percent
SP1200 on 100/120 mesh Supelcoport. The following
operating conditions were maintained for all analyses:
85C oven, 105C detector, 190F gas sampling loop with 1 ml
capacity, and 16 ml/min zero nitrogen carrier gas.
The samples were analyzed for benzene on the same day
they were collected. Peak areas were measured using
a compensating polar planimeter. The sample chromatograms
had three apparent peaks, which were completely resolved.
Following the GC analyses, each integrated bag
sample was analyzed by the Orsat method for carbon
dioxide, oxygen, and carbon monoxide concentrations,
as specified in EPA-Method 3. These results were
used to calculate the molecular weight of the process gas.
40
-------
Visible Emissions
Visible emissions from the P4 battery stack exhaust
were recorded for the duration of each sample run except
for Sample No. 3, for which visible emissions were
abbreviated due to foul weather. The observations
were performed in accordance with EPA Method 9 by a
qualified visible emissions observer. A summary of
the visible emission data is presented in Appendix B-4.
41
-------
SECTION II - TRW REPORT
-------
Log of B(a)P tests the Product Coke Oven Battery Stack,
Jones and Laughlin Steel Corporation, Pittsburgh,
Pennsylvania plant.
Monday, April 30, 1979
TRW spent the day preparing for the first B(a)P test
which was to be run simultaneously with the Clayton
Environmental Consultant's particulate test. It
was considered that TRW would follow Clayton and use
Clayton's pitot values for TRW's run since TRW did
not have their own pitot of sufficient length.
EPA's Technical Manager, Mr. Frank Clay requested TRW
to try and rig a temporary pitot for the three B(a)P
runs. The pitot was constructed with a 10-foot
stainless pitot and a 6-foot of Teflon tube which was
connected to the metal tube by rubber hose and taped.
Clayton checked the duct and found a dust and dirt build-
up at the bottom which naturally changed the effective
duct size (Figure 1).
TRW cleaned the sampling train after which blanks were
taken from each of the train's components, the same as
after a normal sampling run. After this, all major
components were leak checked and capped ready for the
first test. See Figure 2 for sampling train configuration.
-------
n n n
i i
! i
' '
n n n
y
x
*
x
X
0U5T
X X X * X
xx x x x
X x, X XX
XXX X X
X X X X X
MO DTRT &UILOUC
. Concrete
' OUCT
OUST
x i
X H
X 5
10 THE
-------
FIGURE 3.
FILTER HOLDER
OVEN
NCATF0. TEFLON
-------
Tuesday, May 1, 1979
TRW arrived at site and set up the train for the first
test. During the leak check the heated Teflon hose
collapsed. The heated hose was replaced and sampling
started at 1450. At 1555 it was discovered that the
female inlet to the XAD-2 coil was cracked. It was leak
checked to be airtight but it was expected that it would
start leaking later so it was replaced. While the coil
was being replaced it is thought that the taped tubing
of the pitot developed a leak which affected pitot
reading on the last three points of the remaining ports
and the same points on the first three ports of the second
test.
No problems were encountered except for trying to keep
the gas temperature at the exit of the XAD-coil at 127F.
The temperature tended to fluctuate, and at times, it
was impossible to reach 127F. The final leak check was
.02 at 15" of mercury. The test was completed at 1842.
While recovering the sample it was noticed that the water
in the first impinger was discolored, a mixed tan and
brown color.
Wednesday, May 2, 1979
Second test started at 1043 with no problems, and finished
at 1356. The impinger water was not discolored at the
end of the test.
-------
Thursday, May 3, 1979
Testing started at 1049 and finished 1442. No major
problems were encountered. It was still impossible
to hold the XAD-2 coil's exit gas to 127F but it was
possible to hold in a +20F range of 120-127F.
Discussion of Tests
For results of the tests see Table 1 and 2. Since TRW's
pitot readings were somewhat questionable it was decided
to use both TRW's and Clayton's pitot results to calculate
the duct flow rates.
Due to the electric shock hazard caused by the high
temperatures of the duct burning off the probes insulation
it was decided not to heat the probe. The filter oven
temperature was monitored by taping a thermocouple
to the inlet of the filter. The flex line was heated
using heat tapes controlled by a Variac, and the XAD-2
coil was heated using a water bath system. The temperature
was monitored by taping a thermocouple to the outlet
of the XAD-2 coil. It was possible to maintain the
filter temperature at 248F+ 25 F during '-most of the testing
period. The XAD-2 coil was controlled at 122F+ 18F during
most of the testing period.
-------
TABLE 1
BaP - TEST RESULTS
TEST #
DATE
TRW DATA
•METER VOLUME DSCF
JMETER VOLUME DSCM
DUCT VOLUME ACFM
'DUCT VOLUME DSCFM
'DUCT VOLUME DSCMM
% ISOCKINETIC
CLAYTON'S DATA
DUCT FLOW ACFM
DUCT FLOW DSCFM
;DUCT FLOW DSCMM
! % ISOKINETIC USING
TRW METER VOLUME
AND CLAYTON DUCT FLOW,
BaP-JL-1 .
5/1/79
88.025
2.493
94,856
40,543
1148.3
105
112,607
47,626
1348.8
89
BaP-JL-2
5/2/79
74.032
2.097
79,337
33,498
948.7
108
96,020
40,088
1135.3
91
BaP-JL-3
5/3/79
98.539
2.791
113,243
47,282
1339.0
104
115,503
48,176
1364.4
102
. 3 RUN AVERAGE
86.8695
2.460
95,812
40,442
1145.3
108,043
45,297
1,282.8
-------
TABLE 2
J & L - BaP TEST RESULTS
SAMPLE
BaP-OL-1
BaP-JL-2
BaP-JL-3
3 RUN AVERAGE
vg/FILTER
< 0.002
:<0.002
0.050
< 0.021
yg/PROBE
RINSE
1.6170
0.4329
0.1710
0.7403
yg/HEATED
HOSE
0.4674
0.3150
0.3780
0.3868
jig/XAD-2
<0.001
<0.001
<0.001
<0.001
ug/TOTAL
2.0874
0.7509
0.6100
1.1494
DUCT FLOW
*DSCMM
1345.8
1135.3
1364.4
1282.8
METER VOLUME
DSCM
2.493
2.097
2.791
2.462
yg/DSCM
8.35xl04
3.58X.104
2.18xl04
2.70xl64
gm/HR
~
0.0672
OY0241
0.0177
0.0363
!».
DUCT FLOW
DSCFM
47,626
40,088
48,176
45,296
METER VOLUMS
DSCF
88.025
74.032
98.539
86.940
SAMPLE
BaP-JL-1
BaP-OL-2
BaP-GL-3
3_ RUN AVERAGE
LBS/DSCF
5.21X10"11
2.23X10"11
1.36X10"11
2.93X10"11
LBS/HR
1.48xlO"4
0.53xlO"4
0.39xlO"4
O.SOxlO"4
;LBS/24' HR
DAY.
.3. 5.5x1 0"3
1.27xlO"3
0.94xlO"3
1.92xlO"3
LBS/365
..DAYS
1.29
0.46
0.34
0.70
; ***
/EARS/LB
0.78-^.
2.18
2.94
1.43**
* Based on Clayton Environmental Consultants' pi tot and temperature readings
** lbs/365 days must be used to derive years/1b.
*** Years required to emit one pound
-------
Due to the minute quantities of B(a)P present in the
stack gas relative to the total volume of stack gas,
it is difficult to conceive the concentrations in
perspective. If the average weight in pounds of
B(a)P emijtted per dry standard cubic foot (dscf)
of stack gas is used with the average dscf per minute
stack gas volume, then it is possible to calculate
the number of days of continuous operation needed to
emit one pound of B(a)P. Based on the data in this
report, the time required to emit one pound of B(a)P
from the battery stack is 526 days, or '1.44 years
(1 year, 5 months, 8 days approximately).
------- |