Research and Development
EPA-600/S4-82-054 August 1982
oEPA
Project Summary
Field Evaluation of Carbon
Monoxide and Hydrogen
Sulfide Continuous Emission
Monitors at an Oil Refinery
Bruce B. Ferguson and Richard E. Lester
An eleven-month field evaluation was
conducted for five hydrogen sulfide and
four carbon monoxide continuous emis-
sion monitors (CEMs). The H2S CEMs
were installed on e fuel gas tine and the
CO CEMs were installed on a stack from
a fluidized bed catalytic cracking unit at
a refinery. Performance specification
testing was routinely performed on the
instruments as they were operated and
maintained in this field environment.
This Project Summary wes developed
by EPA's Environmental Monitoring Sys-
tems Laboratory. Research Triangle
Park, NC. to announce key findings of
the research protect that is fuffy docu-
mented in e separate report of the seme
title (see Project Report ordering infor-
mation at beck).
Introduction
On March 15, 1978, EPA promulgat-
ed New Source Performance Standards
(NSPS) that required petroleum refineries
to continuously monitor the carbon mo-
noxide (CO) emissions from fluid cata-
lytic cracking (FCC) units.1 Refineries
were required also to continuously moni-
tor either the hydrogen sulfide (H2S)
concentration in fuel gas feed lines or the
sulfur dioxide (S02) concentration in the
exhaust of the boiler burning the fuel
gas. However, refineries were not re-
quired to install H2S or CO continuous
emission monitors (CEMs) until perform-
ance specifications for the instruments
were published by the EPA.
In April 1979, the EPA initiated work
to establish these specifications and
also to determine the durability, mainte-
nance requirements, and data validity of
commercially available CO and H2S con-
tinuous emission monitors. To procure
the monitors, 35 vendors of commer-
cially available stack gas monitors for
CO and H2S were asked to recommend
monitors for evaluation. Sixteen did not
respond. A total of ten H2S and thirteen
CO monitors were advanced for consid-
eration. From this list, five H2S and four
CO monitors were selected for evalua-
tion (Table 1). The selection criteria in-
cluded operating principle, engineering
judgment about suitability for use at pe-
troleum refineries, and total cost.
These nine monitors were installed in a
trailer at the Harmon Engineering & Test-
ing facility in Auburn, Alabama and then
transported to a petroleum refinery for
•n eleven-month field evaluation. Here,
the H2S monitors measured the H2S
concentration in a fuel gas at a point
downstream of an amine treater (an H2S
control device) and the CO monitors
measured the CO concentration in the
exhaust gas from a boiler on an FCC unit.
An EPA-designed stack gas conditioning
system removed particulate and water
from the FCC stack gas at the sampling
port before sending the clean, dry gas to
a manifold in the trailer (where it was dis-
tributed to the monitors). A similar mani-
fold was used to distribute fuel gas to the
H2S monitors.
-------
Table 1. Continuous Emission Monitors Evaluated
Instrument Manufacturer Model Number
Abbreviation
Operating Principle
CARBON MONOXIDE ANALYZERS
Ana red, Inc.
Santa Barbara. CA
P.O. Box 3180
Santa Barbara. CA S31 OS
Ecoiyzer
Energetics Science
85 Executive Blvd.
Elmsford. NY 10523
Applied Automation, Inc.
Fbwhuska Road
Bartlesville, OK 74004
Mine Safety Appliances Company
7522 Neade Street
Pittsburgh. PA 15208
HYDROGEN SULFIDE ANALYZERS
Bendix Corporation Environmental
and Process Instruments
P. O. Drawer 831
Lewisburg. WVA 24901
Del Mar Scientific. Inc.
P.O. Box 486
Addison. TX
Houston Atlas. Inc.
9441 Baythorne Drive
Houston, TX 7 704 1
Process Analyzers, Inc.
1101 State Road
Princeton. NJ 08540
Teledyne Analytical Instruments
333 W. Mission Avenue
San Gabriel, CA 91 776
501R
3107/2949
Optichrom 102
LIRA 202
Anarad
Ecoiyzer
AA102
MSA
7770
DM-W
825R/102
32-230
611 DMC0-20X
Bendix
Del Mar
HA!
PAI
Teledyne
Nondispersive Infrared
Detector (NDIRJ
Electrochemical Sensor
Gas Chromatograph/Flame
Ionization Detector IGC/FIDI
Nondispersive Infrared/Luft
Detector (NDIR/Luftl
Gas Chromatograph/Flame Photo-
metric Detector (GC/FPD)
Lead Acetate Impregrated
Paper Tape
Lead Acetate Impregrated
Paper Tape
Ges Chromatograph/Flame Photo-
metric Detector (GC/FPD)
Ultraviolet Absorption
The output from the monitors was re-
corded on an Esterline Angus PD 2064
data logger. Techtran 816 tape cassette
recorder and an Esterline Angus multi-
point recorder. The multipoint recorder
printed the instantaneous monitor value
at 3-minute intervals. The PD 2064 read
the monitors' output at 3-minute inter-
vals and printed the average after ten
readings, except during relative accuracy
tests when it read it every minute and
printed the average of 30 readings. The
PD 2064 also initiated daily zero and
span checks of the monitors with cali-
bration gases and recorded the results.
Periodic tests for relative accuracy,
calibration error and short- and long-term
drift were conducted periodically during
the field evaluation. EPA Method 11
was the reference method for the rela-
tive accuracy tests on the H2S monitors.
EPA Method 10 (NDIRII and a colorimetric
method (teuco crystal violet, LCV) being
evaluated during this study, were the
reference methods for the relative accu-
racy tests on the CO monitors.
The results of the monitor evaluation
are summarized here. A full report de-
scribes the test results, the LCV analyti-
cal method, and the components and
performance of the FCC stack gas condi-
tioning system.
Results and Discussions
Carbon Monoxide CEMs
Initially, the monitor evaluation was
intended to determine the ability of CO
and H2S monitors to meet some tenta-
tive performance specifications similar
to those for NO, and SO2 CEMs. How-
ever, after the study was initiated, EPA
revised Performance Specification 2
-------
Tabta 2. Summary of CO Relative Accuracy Tests at Refinery
Number of
Mean
LCV
Relative Accuracy
Reference
Method
CO2
Cone
CO
Cone
Ecolyzer
Anarad*
MSAt>
Date
Samples
(%J
(ppm)
%Su*
%IC01
%Stdc
%IC0J
% Stdc
%/COJ
9/80
1/81
2/81 (Al
2/81 (B)
5
9
9
9
2
2
12
3
80
404
607«
712
S
14
>100
•
34
17
>100
a
3.6
13
7.7
12.7
24
16
7.6
8.9
2.9
13
4.2
9.7
18.4
16
4.3
6.8
Mean
40
50
9.3
14
7.4
11
• This is the average of the manual method rasufts using the lauco crystal violet method developed in this study.
A Corrected for C02 interference.
c Standard for CO: 500ppm.
dMonitors sampled cylinder gases containing 502 ppm CO. 12% CO2 and different concentrations of NO and SO2 in nitrogen. This
was done to evaluate the performance of the monitors when they were sampling a stack gas with A10, and SO 2 levels representative
of an FCC stack gas. During the 11 •month study, process upsets and refinery equipment malfunction caused the CO levels to exceed
the span range of the monitors and so it was necessary to dilute the stack gas to bring the CO level into the working range of the
monitors during the relative accuracy test. Dilution was obtained by introducing plant instrument air at the probe. This dilution reduced
the NOm and SO 2 levels by a factor of 4 to 6 and permitted the Ecolyzer to obtain an accurate analysis of the stack gas.
• Monitor not operational because of detector failure from high SO2 and NO levels encountered in Test 2/81 (A).
det gas containing 10% C02 in nitrogen
was introduced at the manifold after
each relative accuracy test and the mon-
itor's response (ppm CO/1 % C02) was
used to correct the relative accuracy test
results for the C02 content of the stack
gas As expected, the C02 interference
remained quite steady during the study
(Anarad. 3 ppm/1% C02, MSA 1 ppm/
1% C02).
Table 3 summarizes the results of the
calibration drift (CD) tests. These tests
were performed in conjunction with the
daily zero and span checks. Since the in-
struments were not zeroed each day. the
values reported represent the daily zero
drift values corrected for the zero value
recorded for the start of the 24-hour
period.
The Applied Automation performed
well during the laboratory check-out, but
suffered from an unknown interference
in the FCC gas sample. A valid relative
accuracy test was not completed be-
cause of this and because of the erratic
performance of the instrument. Testing
of the instrument was discontinued after
a factory representative was unable to
find the cause of the interference and
unable to correct the erratic behavior in
the monitor's output.
The Model 2949 N0»/S0R scrubber
supplied with the Ecolyzer was found to
be inadequate for long-term use of FCC
stack gases where S02 levels of 400-
800 ppm and NO, levels of 100-400
ppm are encountered. Because some of
the scrubbers were exhausted within
Table 3. Calibration Drift Test Results for CO Monitors
Calibration Drift* b I%)
Ecolyzer Anarad MSA
Day
Zero
Span
Zero
Span
Zero
Span
Test 1
1
1.4
0.4
1.7
0.1
0.6
0.1
2
0.3
4.4
0.2
2.4
0.9
1.2
3
0.2
1.7
0.2
1.6
1.2
2.0
4
0.6
1.0
1.0
0.2
1.4
0.8
5
0.4
0.7
2.0
0.8
2.5
6.2
6
0.9
0.9
2.1
1.5
3.3
3.0
7
1.2
1.6
3.2
4.5
2.9
0
Test 2
1
0.2
7.2
2.4
0.7
0
0.9
2
0.2
0.1
0.2
4.4
1.7
2.4
3'
0.7
10.8
0.4
3.4
5.4
1.3
4
0.2
0.2
0.4
0.3
1.3
3.6
5
0
1.6
1.1
0.4
2.5
3.7
6
0.1
0.9
0.1
1.9
0.2
1.1
7
0.1
0.3
0.5
1.0
0.1
1.7
Test 3
1
0.1
0.1
0.9
0.6
0.2
1.2
2
0.3
3.4
1.3
1.7
0.2
1.2
3
0
0.4
0
0.7
0.4
0.1
4
0.1
0.9
0.6
1.1
0.7
1.2
5
3.1
2.7
0.4
1.1
1.8
2.8
6
2.2
2.8
0.4
0.1
4.2
1.1
7
0.7
0.9
0.2
0.1
0.9
2.8
• Values in table represent the daily drift as defined by the following equation
[Calibration Gas Concentration - Monitor Reading', x 100%
Monitor Span Value
* Because long-term drift was of primary concern, the instruments were not zeroed daily.
The values in the table have been corrected by the daily zero drift.
c Instrument recalibration was performed this date.
3
-------
24-hours, it was decided to use them
only during relative accuracy tests. Dur-
ing routine unattended operation a 25
cm long x 2,5 cm ID PVC tube packed
with activated charcoal was substituted
for the 2943 scrubber, (This charcoal
tube could not be used for RA tests, be-
cause it caused a 30-minute delay in the
analyzer's response to a change in stack
gas CO concentration.)
Throughout the field study, the FCC
unit malfunctioned at frequent intervals,
which caused the CO level in the stack to
rise above the span range of the moni-
tors (1000 ppm). Such malfunctions oc-
curred during the first two relative accu-
racy tests and it was necessary to add
ambient air to the stack gas at the probe
exit to bring the CO concentration below
1000 ppm. Since this dilution signifi-
cantly reduced the SO, and NO, levels in
the sampled gas. the Ecolyzer performed
better than expected during the test.
(The actual effect of NO, and SO, on the
Ecolyzer can be seen by referring to RA
Test 3 (February 1981 A) in Table 2-a
test in which cylinder gases containing
SO, and NO, at levels normally expected
in an FCC stack gas were used instead of
diluted stack gas.I
The Ecoiyzer failed to complete the
field study because of detector failure.
The detector failed twice during the
seven months the monitor operated. The
monitor also experienced severe drift
throughout this seven-month period.
Because its output was not compatible
with the PD 2064 input requirements.
the MSA instrument was not operational
until July 1980. The monitor performed
well in alt the relative accuracy tests
after its response for C02 in the sampled
gas was corrected It suffered two major
outages (both of which were corrected
by optical realignment) after it passed
the first relative accuracy test. Because
of these outages and the late start up,
the calibration error test could not be
conducted on this monitor. There is no
reason to suspect, however, that this
monitor would not have passed this test.
The monitor suffered from random
short-term drift, but over the period of a
month its drift averaged less than 3 per-
cent of span. The percent uptime of the
monitor was 86 percent.
The Anarad operated without mal-
function from the time it was installed in
the trailer (November 1979) until the
project was completed in April, 1980. It
performed well during the relative accu-
racy and calibration error tests Like the
MSA, it sometimes suffered from ran-
dom short-term drift, but over the period
of a month, its drift averaged less than 3
percent of span.
Hydrogen Sulfide OEMs
The agreement between the monitors
and Method 11 changed radically during
relative accuracy tests. Initially, the
monitors showed a negative bias with
respect to Method 11. but in early Feb-
ruary a positive bias began to be ob-
served. The agreement between the
monitors and Method 11 sometimes
changed drastically from one day to the
next. Tests to determine the cause of
this bias were inconclusive. When two
laboratories Simultaneously measured
the HjS concentration in the fuel gas,
the two laboratories agreed closely but
their results were significantly different
from that of the monitors (Table 4, RA
Test 5). Further, when the fuel gas was
•piked with known amounts of H2S,
both the monitors and the reference
method recovered the spike. Thus, the
cause of the bias remains unknown. Per-
haps it resulted from a combination of
several factors, including the slow re-
sponse of the monitors and sudden
changes of short duration in the H2S
concentration in the fuel gas.
Table 5 summarizes the results of the
calibration drift (COI tests These tests
were performed in conjunction with the
daily zero and span checks. Since the in-
struments were not zeroed each day. the
values reported represent the daily zero
drift values corrected for the zero value
recorded for the start of that 24-houf
period. Specific comments about the
monitors are presented below.
The Teledyne responded to many
other compounds present in the fuel gas.
The interferences were so severe that
field evaluation was discontinued on the
monitor. It was not possible 10 install a
scrubber that would effectively remove
the majority of the interferences before
the instrument failed completely.
The PA I was not operational for more
than two weeks despite being returned
Table 4. Summary of H2S Relative Accuracy Tests at Refinery
Number
of Method 11 Mauve Accuracy
Test
Start
Reference
Method
Mean H^S
Cone
Bendix
Houston Atlas
Del Mar
Tefedyne
Number
Date
Samples
(ppm)
%St(f
% tUfSJ*
% std*
% IH^SI'
% std>
%
%st
25 1
25 9
115
119
2
5/5/80
9
201
39
32
t>
»
52.4
43 0
74
61
3
6/9/81
9
208
5 8.8
47.4
47.1
37.9
45 8
36.3
b
t>
4A
1/20/81
9
190
18.5
16.1
13.8
12.0
28 4
24 7
e
6
4B
1/20/81
9
197
16.4
13 7
30.0
25.1
48 9
40 9
b
t>
5
2/18/81
Lab 1
Lab 2
9
7
34 6
31.7
18.1
22 4
86 4
117
24 4
28.9
116
148
32.0
39.4
153
205
S
t>
£>
b
6
2/25/81<>
6
150
11.1
12.2
e
€
C
c
b
b
7
8
2/2S/81
3/31/81o
8
10
138
114
35 5
8 3
40.2
12 0
c
t
C
6
e
21.5
c
31.1
b
b
b
b
• Standard 165 ppm HjS. IMjSI * Method 11 result
® Monitor not opera timet.
c Monitor operational but not included in test because it could not be brought into calibration.
* Fuel gas spiked with known amount of H^S using Houston Atlas. Inc. Model 601 dituter.
4
-------
Tabb 5. Calibration Drift Test Results for H^S Monitors
Calibration Drift*- * (%)
Bendix Houston Atlas Dal Mar
Number
Zero
Span
Zero
Span
Zero
Span
Test 1
1
0
0.7
0.1
6.3
12.5
21.3
2
0
0.7
0.3
4.5
9.6
30.9
3
0
1.1
0.4
0.3
19.5
52.6
4
0
1.1
0.1
1.0
0.7
101
5
0.1
9.2
0.3
13.9
4.8
44.0
6
0.1
8.1
0.1
13.9
0
3.3
7
0
0.7
6.9
2.9
2.6
8.1
Test 2
1
0
1.1
3.5
10.8
6.6
34.2
2
0
0
1.4
26.0
4.4
26.5
3
0
0.7
4.5
3.5
1.5
2.6
4
0.2
0.4
4.2
3.1
1.5
2.6
5
0.1
2.6
1.0
12.8
1.5
1.8
6
1.0
1.1
0.6
1.4
0
0
7
0.1
0.4
—
11.5
4.0
4.8
Test 3
1
0.1
1.8
0.1
4.9
1.1
7.4
2
1.0
0.7
0.3
1.0
0.4
6.6
3
0.2
0.7
1.7
15.6
1.1
8.5
4
0.2
1.5
5.2
21.5
10.7
6.6
5
0.1
1.5
7.6
—
9.9
18.4
6
4.2
2.9
5.9
4.9
0
6.3
7
4.2
4.0
2.1
5.2
1.5
3.7
Test 4
1
0
0.7
3.1
3.8
0.4
1.1
2
0
0
0.7
6.6
0.7
3.7
3
0
0.7
0.7
0.7
0
0
4
0
0
0
0.7
0.6
4.8
5
0
0
0
6.6
2.0
12.1
6
0
0.4
0
0
—
—
7
—
—
0.3
10.4
—
—
* Values in table represent the daily drift as defined by the following equation
[Calibration Gas Concentration - Monitor Reading] x 100%
Monitor Span Value
b Because long-term drift was of primary concern, the instruments were not zeroed daily.
The values in the table have been corrected by the daily zero drift.
to the manufacturer three times for re-
pairs. A successful relative accuracy
test was never obtained.
The Bendix operated continuously
fror* the initial start-up until final shut-
•r with only four brief outages. No in-
ferences were detected during initial
check-out; a possible negative interfer-
ence in the fuel gas was indicated but
not confirmed. This instrument experi-
enced the least amount of drift of any of
the CEMs.
One problem encountered with this in-
strument resulted from the fact that it
only sampled the fuel gas approximately
once every 3.5 minutes. This feature of
the Bendix makes comparing its output
for 30 minutes (total sample collection
time less the 10 seconds) to the results
of a 30-minute Method 11 sample of
dubious validity. This is particularly true
when one considers the sudden changes
that can occur in the H2S level of the fuel
gas. The Bendix did complete the field
study with only two major outages and
had a percent uptime of 89 percent.
The Houston Atlas suffered frequent
mechanical failure but did complete the
majority of the field testing. Its percen-
tage uptime was 76 percent. The failures
were primarily in the sample dilution sys-
tem which employs a Teflon sliding
block containing a cavity of known
volume. At times, the instrument suf-
fered from excessive drift. Calibration
usually took at least four hours.
The Del Mar operated for the entire
test program without mechanical failure.
However, because rotameters were used
to achieve a dilution of the fuel gas prior
to analysis, its calibration changed when
the density and viscosity of the fuel gas
changed. This fact caused the instru-
ment to have a significant amount of
drift. The percent uptime was 97 per-
cent. the highest of any of the monitors.
Conclusions and
Recommendations
The Anarad and MSA successfully
completed the eleven-month field study
•nd demonstrated that reliable CO moni-
tors with good drift control are available
for use at petroleum refineries Although
CO2 is an interference with these moni-
tors, its effect is small (in terms of the
span range) and correctable. Water
vapor will also interfere with these NDIR
instruments, but it is unlikely that these
instruments would sample a gas stream
that had not already had the moisture
removed.
Overall, the performance of the five
HjS monitors was disappointing. Two of
the five monitors never obtained a valid
sample and a third suffered frequent
malfunction. The absolute agreement
between Method 11 and all monitors
was poor and variable in eight of the ten
relative accuracy tests. Since the source
of this difference is unknown, the use of
HjS monitors for compliance purposes
cannot be recommended at this time nor
can we advance performance specifica-
tions for accuracy. However, these
monitors may be useful for determining
trends and for indicating the monthly
performance of the amine treater. For
example, the average H2S concentration
for all Method 1 1 tests (140 ppm) com-
pares reasonably well with the average
for each monitor: Del Mar (133 ppm).
Houston Atlas (130 ppm) and Bendix
(119 ppm).
References
1. Performance Standards for New Sta-
tionary Sources. Federal Register,
43: 10869, March 1 5. 1978
2. Proposed Revisions to Performance
Specifications 2 and 3, Federal Regis-
ter, 46: 8359, January 26. 1 981.
5
-------
Bruce B. Ferguson and Richard E. Lester ara with Harmon Engineering and
Testing, inc.. Auburn. AL 36830.
William J. Mitchell is the EPA Project Officer (see below).
The complete report entitled "Field Evaluation of Carbon Monoxide end Hydro-
gen Sulfide Continuous Emission Monitors at an Oil Refinery." (Order No.
PB 82-227 406: Cost: S 12.00. subject to chenge) will be eveileble only from:
Netionel Technical Information Service
5285 Port Royel Roed
Springfield. VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contected at:
Environmental Monitoring Systems Leboretory
U.S. Environmental Protection Agency
Research Triangle Park. NC 27711
6
~ U S GOVERNMENT PRINTING Off ICI: 1982/559 -092/W89
-------
A STUDY TO EVALUATE CARBON MONOXIDE
AND HYDROGEN SULFIDE CONTINUOUS EMISSION
MONITORS AT AN OIL REFINERY
by
Bruce B. Ferguson and
Richard E. Lester
Harmon Engineering & Testing
Auburn, Alabama 36830
and
W.J. Mitchell
Quality Assurance Division
Environmental Monitoring Systems Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
EPA Contract No. eS-OS-S^OS
Prepared For
QUALITY ASSURANCE DIVISION (MD-77)
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Monitoring
Systems Laboratory, U.S. Environmental Protection Agency and has been
approved for publication. Approval does not signify that the contents
necessarily reflect the views and policies of the U.S. Environmental
Protection Agency, nor does mention of trade names or eoumercial
products constitute endorsement or recommendation for use.
1
-------
FOREVARD
Measurement and nonltorlng research efforts are designed to
anticipate potential environmental problems, to support regulatory
actions by developing an in-depth understanding of the nature and
processes that impact health and the ecology, to provide innovative
neans of monitoring compliance with regulations, and to evaluate the
effectiveness of health and environmental protection efforts through
the monitoring of long-term trends. The Environmental Monitoring
Systems Laboratory, Research Triangle Park, North Carolina, has
responsibility for: assessment of environmental monitoring technology
and systems; Implementation of agencywide quality assurance programs
for air pollution measurement systems; and supplying technical support
to other groups in the Agency including the Office of Air, Noise and
Radiation, the Office of Toxic Substances and the Office of
Enforcement.
The following investigation was conducted at the request of the
Office of Air Quality Planning and Standards (OAQPS) to determine the
performance that can be expected from continuous emission monitors
installed at petroleim refineries. The results of this study will be
used by the OAQPS to determine the appropriateness of these monitors
for use at refineries and to determine reasonable performance specifi-
cations for these monitors.
Thomas R. Hauser, Ph.D.
Director
Environmental Monitoring Systems Laboratory
Research Triangle Park, North Carolina
11
-------
ABSTRACT
The U.S. Environmental Protection Agency (EPA) has promulgated
New Source Performance Standards (NSPS) that require petroleum
refineries to continuously monitor the carbon monoxide (CO) emissions
from fluid catalytic cracking (FCC) wits and also to continuously
monitor either the hydrogen sulfide (H_S) concentration in fuel gas
feed lines or the resulting sulfur dioxide (SOp) concentration in the
boiler exhaust. However, refineries are not required to install H^S
or CO continuous emission monitors (CEMs) until performance
specifications have been published by the EPA. Tentative performance
specifications, proposed by EPA after laboratory and short-term field
evaluations, were extensively evaluated in a year-long field
evaluation conducted using five HpS and four CO continuous emission
monitors. The H.S CEMs were installed on a fuel gas line and the CO
CEMs were installed on a stack from a FCC unit at an east coast
refinery. During the evaluation, performance specification testing
was routinely performed on the instruments as the instruments were
operated and maintained in a work environment. The CO CEMs were
generally reliable and able to meet proposed performance
specifications. The H-S CEMs were not able to meet the proposed
relative accuracy criteria but the difference in measured
concentration could not be isolated to the CEMs or the reference
method.
iii
-------
CONTENTS
Disclaimer 1
Forward 11
Abstract ill
Figures v
Tables vi
List of Abbreviations and Symbols vii
1. Introduction 1
2. Summary and Conclusions 9
Carbon Monoxide Manual Method 9
Carbon Monoxide Monitors 9
Hydrogen Sulfide Monitors 10
3. Recommendations 20
Carbon Monoxide Continuous Emission Monitors 20
Hydrogen Sulfide Continuous Emission Monitors 20
4. Description of Equipoent 22
Continuous Emission Monitors 22
Ancillary Equipoent 24
5. Experimental Procedures..... 29
General Procedures 29
Laboratory Evaluation of the Monitors 29
Field Evaluation of the Monitors 32
6. Results and Discussion 35
Manual CO Method Development/Validation 35
Evaluation of Carbon Monoxide Monitors ^0
Evaluation of Hydrogen Sulfide Monitors 48
References.... 59
Appendices
A. Definition of terms
B. Tentative plan for the evaluation of CO and H^S continuous
monitors at refineries 63
C. Vendors response to letter from EPA 71
D. FCC emissions gas sample conditioning system 74
E. Manual method for measuring carbon monoxide in refinery
gases 81
F. Instrument evaluation history 96
iv
-------
FIGURES
Mumber Page
1 Sketch of field evaluation alte 8
2 Sketch of Instrumentation trailer showing instrunent
arrangement 26
3 Plumbing diagram for FCC gas distribution to the CO CEMs.. 27
4 Plumbing diagram for fuel gas distribution to the H?S
CEMs : 28
5 Zero drift trend for Ecolyzer CEM 45
6 Span drift trend for Ecolyzer CEM 45
7 Zero drift trend for MSA CEM 46
8 Span drift trend for MSA CEM 46
9 Zero drift trend for Anarad CEM 47
10 Span drift trend for Anarad CEM 47
11 Zero drift trend for Bendlx CEM 56
12 Span drift trend for Bendlx CEM 56
13 Zero drift trend for Houston Atlas CEM 57
14 Span drift trend for Houston Atlas CEM 57
15 Zero drift trend for Del Mar CEM 58
16 Span drift trend for Del Mar CEM 58
v
-------
TABLES
Humber Page
1 Tentative performance specifications for CO CEMs 5
2 Tentative performance specifications for H^S CEMs 5
3 CO CEMs evaluated 6
4 H^S CEMs evaluated.....a....7
5 Data summary of CO CEMs 13
6 Sunmary of CO relative accuracy tests at refinery 14
7 Calibration drift test results for CO monitors 15
S Effect of COj, SO^ and CO monitors...«««•«««•*1$
9 Data summary of HpS CEMs relative accuracy tests at
refinery 17
10 Summary of H^S relative accuracy tests at refinery 18
11 Calibration drift test results for H^S monitors 19
12 Change in absorbance of CO reagent blank with time at
roc® temperature38
13 Effect of NO and SO^ on leuco crystal violet 38
14 Comparison between LCV and NDIR results on FCC samples... 39
15 Relative accuracy test results on CO monitors 15
16 Relative accuracy test results on H^S monitors..... 52
17 Results of collaborative RA test on HjS monitors 55
vi
-------
LIST OF ABBREVIATIONS AND SYMBOLS
ABBREVIATIONS
AA — Applied Automation
CQ1 — continuous emission monitor
d.c. — direct current
EPA — Environmental Protection Agency
FCC — fluid catalytic cracker
FID — flame Ionization detector
FPD — flame photometric detector
FS — full scale
CC — gas chrooatograph
HAI — Houston Atlas, Incorporated
ID — Inside diameter
IR — infrared
LCV leuco crystal violet
MSA — Mine Safety Appliance
NBS — National Bureau of Standards
NDIR — nondispersive infrared
NSPS — New Source Performance Standards
CM) ~ outside diameter
PAI — Process Analyzers, Incorporated
ppm — parts per million
PST — Performance Specification Test
PVC — polyvinyl chloride
RA — relative accuracy
SYMBOLS
CH^ — methane
CH.SH — methyl mercaptan
CO — carbon monoxide
CO. — carbon dioxide
H-5 ~ hydrogen sulfide
N§
oxides of nitrogen
SO^ — sulfur dioxide
vii
-------
SECTION 1
INTRODUCTION
On March 15, 1978, EPA promulgated New Source Performance Stan-
dards (NSPS) that required petroleun refineries to continuously moni-
tor the carbon monoxide (CO) missions from fluid catalytic cracking
(FCC) units (1). Refineries were required to also continuously
monitor either the hydrogen sulfide (H~S) concentration in fuel gas
feed lines or the resulting sulfur dioxide (SO.) concentration in the
boiler exhaust (1). However, refineries were not required to install
H?S or CO continuous emission monitors (CEMs) until performance
specifications were published by the EPA.
Tables 1 and 2 present the tentative performance specifications
for both CO and H-S monitors that were subsequently proposed by an EPA
contractor after laboratory and short-termed field evaluations. Terms
used in these tables and throughout the report are defined in Appendix
A.
In the laboratory phase, candidate instriments were evaluated to
determine:
• response characteristics
• stability with time, temperature and flow rate
• sensitivity to potential interferences likely to be
present in the sampled gas.
The five 00 and two H~S monitors that performed adequately in the
laboratory were then evaluated for approximately two months at a
petroleum refinery (2,3). Only one CO monitor and one H^S monitor
performed adequately in the field testing. In the case of the CO
monitors, daily calibration checks were mandatory for reliable opera-
tion of all the instrunents, but even with the inclusion of daily
calibration, the contractor questioned the long-term reliability of
the CO monitors. Further, instrument malfunctions, sampling system
malfunctions and data logger malfunctions plagued the field evaluation
of both types of monitors which resulted in a significant amount of
downtime and lost data.
In April 1979, EPA initiated additional work to determine: (1)
Information about long-term Instrunent durability, data validity and
maintenance requirements of commercially-available CO and H^S CEMs at
a refinery; and (2) the validity of the tentative performance specifi-
cations for the instrunents. In addition, a manual (non-instrunental)
method for measuring CO was to be developed and evaluated to serve as
an alternate to EPA Reference Hethod 10.
1
-------
To procure the monitors, EPA contacted vendors of CO and H^S con-
tinuous stack gas sonltors by letter and asked then to submit informa-
tion about the aonltorCs) they thought would be suitable for the
appropriate refinery process, the operating principle and approximate
cost. A copy of that letter and the Project Accomplishment Plan that
accompanied it are included in Appendix B. Appendix C contains the
vendor response to the letter and, if they suggested a monitor, the
nodel number, cost and operating principle.
Of the 35 vendors contacted, 16 did not respond. A total of 10
H~S and 13 CO monitors were recommended for consideration. From this
list, five HpS and four CO monitors were selected for evaluation. The
pertinent inTormation about each instrument is Included in Tables 3
and 4. The selection criteria (described in detail in Appendix B)
involved total cost, operation/detection principle and engineering
judgement about the likelihood the monitor would be suitable for the
application. For example, one company proposed to use a converted NO
monitor for measuring H^S, but did not consider the likelihood thai
organlcs in the fuel gas would interfere with the chemiluminescence
reaction.
After receipt, the monitors were installed In a trailer at the
Harmon Engineering & Testing (HE&T) facility in Auburn, Alabama and
were subjected to checks for: drift, response time, electronic noise
level, Interferences and response variation due to changes in ambient
temperature and sample flow rate. The monitors and trailer were then
transported to a refinery for an 11-month field study in which they
were tested at periodic intervals for relative accuracy, response
time, calibration error and drift.
The refinery at which the field evaluation was conducted had
added a new CO boiler to the FCC unit in early 1979 to recover addi-
tional energy from the FCC exhaust gas and to reduce the CO concentra-
tion in the gas stream. The emission gas from this unit was used for
the CO CEH evaluation.
An EPA-deslgned sample conditioning system (described In Appendix
D) removed the moisture and particulate matter from the FCC gas at the
sampling port. The conditioned gas was transported to the sample
manifold through 200 meters of 0.95 cm ID black, nylon tubing at a
flow rate ranging from 8 to 15 Lpm. This sample conditioning system
was installed In February 1980, and was operated continuously for
eleven weeks before the monitors arrived at the refinery to allow time
to correct any potential design deficiencies (none were found) before
attaching the monitors to it. Three measurements of C0p ®nd CO
concentrations performed during this 11 week period Indicated these
compounds were not affected by the sample conditioning system. In
these checks, inlet and outlet samples were taken in Tedlar bags and
analyzed by NDIR techniques.
2
-------
A sample manifold was used to distribute the conditioned FCC
stack gas to the four CO CEMs. The distribution system was designed
to vent excess gas not required by the instruments for normal opera-
tion. The unit was also equipped with solenoid valves controlled by a
data logger for automatic zero and span checks each day.
The HpS monitors sampled a fuel gas line at a point downstream of
the mine treater used to remove H~S from the fuel gas. IXiring the
project, both nonoethanolanlne ancT diethanolamine were used in the
•mine treater.
The fuel gas was distributed to the five H^S monitors by means of
a six-port sampling manifold that was suppHed continuously with
treated fuel gas. The fuel gas was transported from the sampling
point to the instrimentation trailer through approximately 100 m of
0.63 cm OD stainless steel tubing.
All instruments were located in a 24-foot long, air conditioned
trailer. Figure 1 shows the location of the instrumentation trailer
in relation to"the two sources that were monitored. .
The output from the monitors was simultaneously recorded on an
Esterline Angus Model PD 2064 data logger, Techtran Model 816 cassette
tape recorder and an Esterline Angus multipoint recorder. While
sampling process gas, each monitor's output was read at 3-minute
intervals and the average value for 10 readings was printed by the
data logger and simultaneously recorded by the Techtran Model 816.
The multipoint recorder printed each 3-n>inute reading without
averaging. During relative accuracy testing, the data logger measured
each monitor's output at 1-minute Intervals and reported the average
every 20 or 30 minutes.
At the beginning of the field evaluation (April 1980), the ten-
tative performance tests listed in Tables 1 and 2 were performed.
Field testing performed after May 1980, however, concentrated on re-
lative accuracy and calibration drift tests In response to a major
change in EPA's overall approach to monitor system performance speci-
fications. This change, formally proposed in the Federal Register(U),
involved a drastic simplification on the Performance Specification
Test (PST) Procedure *». Under these proposed PST revisions the only
mandatory tests are relative accuracy and calibration drift. The
other tests that were previously required (5) are now optional.
Five 00 and ten H_S relative accuracy tests were conducted during
the 11-month field evaluation, but not all monitors were operational
In all tests. In addition, the monitors were subjected to daily 15-
mlnute zero and span checks. Except for days when relative accuracy
3
-------
testing was being performed, zero and span calibrations were not
normally adjusted more often than weekly (frequently less than once
per nonth) in order to provide data on the drift characteristics of
each instrunent. Only the gas chrcnatograph (GC) instrinents were
equipped with automatic zero and no Instrument was equipped with
automatic span adjustment.
Relative accuracy tests on the H.S monitors used EPA Method 11
(6) as the reference method. Relative accuracy tests on the CO
monitors were conducted using EPA Method 10 and an alternate method
(described in Appendix E) developed during this study. This alternate
method can be used to check the accuracy of CO continuous monitors
using NDIR as the measurement technique.
4
4
-------
TABLE 1. TENTATIVE PERFORMANCE
SPECIFICATIONS FOR CO CEMS
PARAMETERS
SKC1T1CATION
Kmc*
Calibration Bitot
ItUtlva Accuracy
1.2
Precision
tarpons* TIm
Output VolM
la to Drift. 2 lour**
Z«ro Drift. 24 Boura'
Span Drift. 24 loura*
Interference Equlv.
15Z C0» aa ppm 00
10Z BjO aa ppa CD
Operational Period
0-1000 ppa
«2Z Span
<101 Mean bf.
Value
OZ Spaa
<10 Mlnutaa
<1Z Span
«1Z Span
«2Z Span
<2.SZ Span
<10 ppa
<5 ppa
>168 Soura
Expressed a* sua of absolute aean value plua
•51 confidence Interval in a eerie* of t«ata.
TM* value 1* kit*! on • relative cooparlson
•f tKe aosltore to each other and not to
Method 10. IU term* are defined in Appendix
A.
TABLE 2. TENTATIVE PERFORMANCE
SPECIFICATIONS FOR H2S CEJS
PARAMETERS
SPECIFICATION
Range
Calibration Error'
Ralatlv* Accuracy
Response Tlae (Systea)
Zero Drift. 2 lour*'
Zero Drift. 24 Bout*'
Span Drift. 2 lours'
Span Drift. 24 lours'
Operational Period
0-300 ppa
<51 of Each
~ Calibration
Mixture
<101 Mean Ref.
~ Value
<15 Minute*
<2Z Span
<2Z Span
<2Z Span
«2.SZ Span
>168 Hour*
E*p retted at *u» of absolute sein value plu*
•51 confidence Interval in a aeries of teata.
5
-------
TABLE 3. CO CEMS EVALUATED
INSTRUMENT MANUFACTURER
NOOn. NUMBER
ABBREVIATION
OPERAT1NC PRINCIPAL
Anifi^i Inc.
P. 0. Box 1180
Santa Barbara. CA 93105
501R
Anarad
Nondleperalve Infrared
Detector (NDIR)
Optical Solid Stat* Detector
Ecolyter
Cneritctlca Science
85 Executive Blvd.
Dtvtaton of Becton Olcklneon Cufauj
Elaaford, NT I0J2J
3107/2*4*
Ecelyter
tlectrocHealcal Senear
Applied Autoaatlon, tne.
Pawhuaka Road
Bartlervllle, OK 74004
Optlctiroa 102
AA102
Cae Cfcroaatoftrapti/Plea*
Ion 1tat Ion Detector
(CC/riD)
H!ne Safety Appliance* Cof any
7)22 Neade Street
Pittsburgh, PA 1)208
URA 202
MSA
Nondleperalve Infrared/
Luft Detector (ND1R/L«ft)
-------
TABLE 4. H3S CEMS EVALUATED
INSTRVHOfT MANUFACTURER
MJDEL NUMBER
ABBREVIATION
0PERAT1NC PRINCIPAL
Inllx bnlrtnamtil t fntm Intra
P. 0. Drawer 831
Levlaburg, W. VA 24901
MRt
7770
Bendls
Caa CtiraBatocra^i/rlaaa
Photometric Detector
(GC/FPD)
Dal Mr Scientific, Inc.
P. 0. Bo* 4B6
AMlaon, Ton
DH-tf
Del Nar
Lead Aeetata I*pr*trat*4
Paper Tap*
fcmaton Atlaa, Inc.
9441 Bay thorn* Drift
H ou a ton, TX 77041
8I5R/10J
IUI
Laad Acatata top r*frated
Paper Tap*
Frociat Analyser*, lac.
1101 Stat* Road
Frlneaton, HJ
92-230
PAt
Caa ChrmetosrapfcfPlaee
Photometric Detector
(CC/FPD)
T
-------
CATALYTIC CRACKER
CO BURNER STACK
SAMPLE CONDITIONING SYSTEM
CO SAMPLE LINE
FUEL OAS PIPELINE
FUEL OAS SAMPLE LINE
INSTRUMENTATION
TRAILER
FIGURE 1.
SKETCH OF FIELD EVALUATION SITE
-------
SECTION 2
SUMMARY AND CONCLUSIONS
CARBON MONOXIDE MANUAL METHOD
The analytical portion of the manual CO test method developed In
this study was biased 4 percent high with respect to a Bendlx 8501-
5CA NDIR CO analyzer. The precision associated with a single analysis
was 4.3 percent of the concentration for the range 10 to 1100 ppm.
The precision was approximately 2.5 percent of the mean concentration
for analyses performed in triplicate. This means that two analytical
results on the same bag sample should differ by more than 8.4 percent
only one time in 20 due to chance alone. The method was significantly
affected by SO^ and NO, so these compounds were removed during sample
collection by bubbling the gas through alkaline potassiun perman-
ganate.
Table 5 summarizes the data collected for each CO instrument
during the study. Tables 6 and 7 summarize the relative accuracy and
calibration drift tests, respectively. Sometimes, less than the
desired nine manual method tests were achieved because of leaking
Tedlar bags and process failure. Table 8 shows the effect of C02, N0X
and SOp on each monitor's response. The following paragraphs
summarize the performance of each lnstruoent.
CARBON MONOXIDE MONITORS
Applied Automation Optichrom 102
The monitor performed well in the laboratory checkout, but not in
the field evaluation. A valid relative accuracy test was never
achieved because of the monitor's erratic performance; thus. Its per-
cent uptime was zero.
Ecolyzer 3107
This monitor was equipped with an Energetics Science Model 2949
scrubber to remove NO^ and SO.,. Originally, each scrubber cost $21, ¦
but by June 1980, this price had increased to $45. The scrubber was
found to be inadequate for long-term use on FCC gases. scrubber
failure, which In turn caused monitor detector failure, occurred from
1 to 20 days after Installation (depending on the NO and ^2
concentration encountered). To protect the detector and save cost, a
9
-------
25 on long by 2.5 cm dimeter PVC pipe containing activated charcoal
was.substituted for the Model 29"9 scrubber during routine use and the
Model 2949 scrubber used only during relative accuracy tests. The
response tine of the system with the large charcoal scrubber Installed
was approximately 30 alnutes compared to less than 1 minute when the
Model 2949 scrubber was used. At the S0? and NO concentrations
normally encountered In an FCC stack (500 ppm SO., 20u ppm NO ) even a
new Model 29^9 scrubber was unable to remove ail of the interfering
gases such that the Instrwent could pass a relative accuracy test
(See RA test 2/81 (A) In Table 6). Carbon dioxide at 15 percent by
volvne did not interfere. The monitor drifted significantly over a
period of several days at frequent intervals. The detector was
replaced once during the study and the evaluation stopped after the
detector failed the second time. Detector failure was slso a problem
In the previous study (2). This instrument does not appear to be
suitable to continuously monitor CO concentrations In FCC emissions.
Mine Safety Appliance (MSA) Lira 202
As received, the output from this monitor was not compatible with
the data logger thus, the laboratory check-out tests were not
completed before the trailer was sent to the refinery. At the
refinery the monitor drifted on a daily basis but, over a month, the
zero and span drift frequently averaged out to less than 3 percent. No
Interference was found from SO^ or NO^ and only a small and constant
interference was found from CO^ (1 ppm per 1 percent CO^). Although
the test was not done, water vapor would not be expected to be a
significant lnterferent. The monitor successfully completed the
11-month evaluation with only two failures, both of which were
corrected by an optical realignment.
Anarad 501-R
Except for the time it was in transit to the refinery, this moni-
tor operated continuously without an outage from October, 1979 until
testing was completed in April 1981. Interferences from SO- ®nd NO
were not experienced, but small, constant interferences did result
from COp (3 ppm per 1 percent CO.) and from water (3 ppm per 1 percent
H.O). Since the stack gas was conditioned to yield a dewpolnt of
-20 C, water was not an lnterferent in the field tests. The monitor's
output was usually more stable than any of the other instruments.
HYDROGEN SULFIDE MONITORS
The H^S CEMs were evaluated In the laboratory prior to the field
testing. Table 9 summarizes the data collected for each H^S
instrument during the study. Table 10 summarizes the relative
accuracy testing performed in the field, and Table 11 sunmarizes the
10
-------
calibration drift tests. In some relative accuracy tests, less than
nine reference method sanples were collected because plant upsets
Increased the H^S concentration above the span range of the monitors.
The following paragraphs summarize the performance of each
instrusent.
Bendlx Model 7770
This monitor operated continuously from initial start-up until
final shut-down with only four brief outages. Two outages were caused
by a ruptured diaphragm in the sampling valve, one outage occurred
when an operator accldently shorted a circuit in the heater control
unit and the fourth occurred from an obstruction in the process air
supply.
Interferences were not detected during the initial checkout.
However, during relative accuracy tests, a possible interference from
something in the fuel gas was indicated but could not be confirmed.
Since the monitor sampled the fuel gas once every 3.5 minutes, it was
difficult to conduct a relative accuracy test when the H^S level in
the fuel gas was changing rapidly (as frequently occurred). The
instrument generally performed in a reliable manner throughout the
evaluation.
Process Analyzers Incorporated Model 32-230
This monitor operated for less than 3 days during the laboratory
check-out and was returned to the manufacturer for repairs on three
separate occasions. Mechanical failure, electronics failure and
corrosion of parts prevented the monitor from obtaining a valid
analysis of the fuel gas and thus its percent uptime was zero. The
instrusent does not appear suitable for use in this application.
Teledyne Model 611 DMCQ-20X
This monitor was received approximately 3 months later than
scheduled, which prevented a complete laboratory checkout of the
monitor. A valid analysis of the fuel gas was never obtained because
of interference from diethanolaraine and sulfur compounds such as
mercaptans and carbonyl sulfide. The molecular sieve scrubber
originally supplied with the monitor could not compensate for these
interferences. By the time Teledyne supplied an improved scrubber,
the monitor had ceased to operate; thus Its percent uptime was zero.
The instrument does not appear to be suitable for use in this
application.
11
-------
Houston Atlas Model B25B/102
Although this monitor suffered frequent mechanical failure, it
did complete most of the field testing program. The nest frequent
cause of failure was the gas dilution system. This is the same
problem that affected the Houston Atlas instrument evaluated in a
previous study(3). This monitor required a minimum of 4 hours for a
major calibration and, at times, was subject to severe drifting. Same
of the operational problems encountered were due to operator error and
the corrosive environment of the instrumentation trailer. Some data
were lost because the person conducting the daily checks failed to
replace the lead acetate tape in a timely manner.
Pel War Scientific Model DH-W
This monitor operated for the entire test program without
mechanical failure. However, because rotameters were used to achieve
a 1:10 dilution of the fuel gas, its calibration changed when density
and viscosity of the fuel gas changed. Sudden changes in the gas vis-
cosity occurred during some of the relative accuracy tests; thus the
agreement between Method 11 and the monitor varied drastically during
some tests. A bias also seemed to exist between the monitor and the
manual method that could not be explained by viscosity changes. Sane
data were lost because the person conducting the daily checks failed
to replace the lead acetate tape in a timely manner.
12
-------
TABLE 5. DATA SUMMARY OF CO CEMS
SPECIFICATIONS
PROPOSED IN
REF. 2
ANARAD
ECOLYZER
MSA
AA102
LABORATORY DATA
Precision (Z FS)
<1
<0.1
<0.1
<0.1
<0.1
Span Noise (Z FS)
<1
<0.1
<0.6
<0.2
<0.1
Zero Noise (Z FS)
a
0.5
0.4
0.1
0.9
tUO Interference (as pp* CO)
<4
32.4
n/a
b
3.0
15Z CO2 Interference (as ppn CO)
<10
28.2
<0.1
b
<0.1
Response Time (seconds)
£600
7.5
53
4.8
138
FIELD DATA
Zero Drift, 24 Hr. (Z FS)
<2
2.5
9.2
2.2
0.4
Zero Drift, 2 Hr. (Z FS)
<1
0.89
1.3
1.3
0.42
Span Drift, 24 Hr. (Z FS)
<2.5
4.2
53.8
3.1
4.3
Calibration Error (Z FS)
<2
- high (Z FS)
a
1.7
5.0
b
2.6C
- medium (Z FS)
a
9.1
13.0
b
6.9C
- low (Z FS)
a
4.7
4.6
b
8.4C
Field Response Time, 90Z/95Z FS
(seconds)
<600
13/16
45/47
11/13
242/242
Instrument Operational Time (*)
a
100
57
86
0
* Not specified.
^ Data not complete due to Instrument malfunction.
c
Result of one test.
Time the Instrument was producing useable data during the 11-month field evaluation without
considering the absolute accuracy of the data. I.e., without correcting for the days the
monitor was out of calibration. However, if the monitor was known to be malfunctioning and
not just out of calibration, the monitor was considered to be off line.
-------
TABLE 6. SUMMARY OF CO RELATIVE ACCURACY TESTS AT REFINERY
DATE
NUMBER OF
REFERENCE
METHOD
SAMPLES
. MEAN
CO
CONC
(*)
LCV*
CO
CONC
(ppm)
RELATIVE ACCURACY
ECOLYZER
ANARADb
MSAb
* Std"
X [CO)
I Std"
I ICO)
Z Std"
* ICOJ
9/80
5
2
80
5
34
3.8
24
2.9
18.4
1/81
9
2
404
14
17
13
16
13
16
2/81
(A)
9
12
507d
>100
>100
7.7
7.6
4.2
4.3
2/81
(B)
9
3
712
e
e
12.7
8.9
9.7
6.8
Mean
40
50
9.3
14
7.4
11
a This is the avirage of the manual method results using the lsuco crystal violet nethod developed
in this study.
b Corrected for CO^ interference.
c Standard for CO: 500 ppm.
^ Monitors sampled cylinder gases containing 502 ppm CO, 12Z CO. and different concentrations
of NO and SO. in nitrogen. This was done to evaluate the performance of the monitors when
they were sampling a stack gas with N0X and SO. levels representative of an FCC stack gas.
During the 11-month study, process upsets and refinery equipment malfunction caused the
CO levels to exceed the span range of the monitors and so It was necessary to dilute the
stack gas to bring the CO level into the working range of the monitors during the relative
accuracy test. Dilution was obtained by introducing plant instrument air at the probe.
This dilution reduced the NO and SO^ levels by a factor of 4 to 6 and permitted the
Ecolyzer to obtain an accurate analysis of the stack gas.
e Monitor not operational because of detector failure from high SO. and NO levels encountered
in Test 2/81(A).
-------
TABLE 7. CALIBRATION DRIFT TEST RESULTS FOR CO MONITORS
CALIBRATION DRIFT*' b (*)
DAY ECOLYZER ANARAD MSA
Zero Span Zero Span Zero Span
Test 1
1 1.4 0.A 1.7 0.1 0.6 0.1
2 0.3 4.4 0.2 2.4 0.9 1.2
3 0.2 1.7 0.2 1-6 1.2 2.0
4 0.6 1.0 1.0 0.2 1.4 0.8
5 0.4 0.7 2.0 0.8 2.5 6.2
6 0.9 0.9 2.1 1.5 3.3 3.0
7 1.2 1.6 3.2 4.5 2.9 0
Test 2
1 "0.2 7.2 2.4 0.7 0 0.9
2 0.2 0.1 0.2 4.4 1.7 2.4
3C 0.7 10.8 0.4 3.4 5.4 1.3
4 0.2 0.2 0.4 0.3 1.3 3.6
5 0 1.6 1.1 0.4 2.5 3.7
6 0.1 0.9 0.1 1.9 0.2 1.1
7 0.1 0.3 0.5 1.0 0.1 1.7
Test 3
1 0.1 0.1 0.9 0.6 0.2 1.2
2 0.3 3.4 1.3 1.7 0.2 1.2
3 0 0.4 0 0.7 0.4 0.1
4 0.1 0.9 0.6 1.1 0.7 1.2
5 3.1 2.7 0.4 1.1 1.8 2.8
6 2.2 2.8 0.4 0.1 4.2 1.1
7 0.7 0.9 0.2 0.1 0.9 2.8
* Values In table represent the daily drift as defined by the
following equation
jCallbration Gas Concentration - Monitor Reading! * 1002
Monitor Span Value
b
Because long-term drift was of primary concern, the Instruments
were not zeroed dally. The values in the table have been
corrected by the dally zero drift.
15
-------
TABLE 8. EFFECT OF COj, NOx, SOj OH CO MTTNITORS
DATE
(ppm)
Ecolyzer
Anarad
MSA
Ecolyzer
Anarad
MSA
6/5/80
iooc
165
149
117
165
113
105
9/20/80
100C
141
139
112
148
101
100
11/18/80
iooc
b
152
113
b
116
101
1/20/81
iooc
206
140
107
206
104
95
2/17/81
500d
1680
544
504
1680
508
492
2/17/81
500*
1430
538
512
1430
502
500
2/17/81
500f
1660
523
504
1660
487
492
a
b
c
d
e
f
COj Interference: Ecolyrer - none; Anarad - 3 ppm/lZ COj* MSA - 1 ppm/IX COj-
Monitor not operational because of detector failure.
Cylinder Analysis: 12Z CO
400 ppm NO; 0 ppm SO2; balance Nj.
0 ppm NO; 700 ppm SOj; balance Nj.
200 ppm NO; 500 ppm SOj. balance Nj«
Cylinder Analysis:
Cylinder Analysis:
Cylinder Analysis:
-------
TABLE 9. DATA SUMMARY OF HjS CEMS
LABORATORY DATA
Precision (X FS)
Span Noise (X FS)
Zero Noise (X FS)
100 X CH$ Interference (X FS)
40X CHa in H2 Interference (X FS)
200 ppm CH3SH Interference (X FS)
Response Time (sec)
FIELD DATA
Zero Drift, 24 Hr. (X FS)
Zero Drift, 2 Hr. (X FS)
Span Drift, 24 Hr. (X FS)
Span Drift, 2 Hr. (X FS)
Calibration Error (X FS)
- high (X FS)
- medium (X FS)
- low (X FS)
Field Response Time, 90X/95X FS (sec)
Instrument Operational Time (X)
SPECIFICATIONS
PROPOSED IN
REFERENCE
3 BENDIX DEL MAR HAI FAI TELEDTNE
a
0.3
4.3
2.8
b
0.6
a
0.4
1.1
0.9
b
0.6
a
0.1
2.1
0.5
b
2.4
a
<0.1
<0.1
-0.2
b
<1
a
<0.1
<0.1
-0.4
b
<1
a
<0.1
<0.1
<0.1
b
82
<900
210
306
630
35
<2
0.04
5.2
1.6
b
b
"<2
0.05
2.1
1.9
b
6.1
<2.5
1.7
11.7
13.4
b
b
<2
0.6
3.4
13.1
b
2.1
<5
a
0.7
9.2C
14c
b
9.3®
a
6.2
32c
20c
b
30c
a
8.0
17c
40C
b
47
<900
230/230
304/328
480/583
b
b
a
89
97
76
0
0
a Not specified.
^ Data not complete due to Instrument malfunction.
c Result of one test only.
^ Time the instrument was producing useable data during 11-month field evaluation without considering
the absolute accuracy of the data, i.e., without correcting for days the monitor was out of cali-
bration. However, if the monitor was known to be malfunctioning and not Just out of calibration,
the monitor wns condidered to be off line.
-------
' TABLE 10. SUMMARY OF H,S RELATIVE ACCURACY TESTS AT REFINERY
NUMBER
OF METHOD 11 RELATIVE ACCURACY
TEST
START
REFERENCE
METHOD
MEAN H2S
CONC
BENDIX
HOUSTON ATLAS _
DEL MAR TELEDTNE
HO.
DATE
SAMPLES
(ppm)
Z std"
x (H2sr
7. std
i IH2S]
* std*
I [H2sr * std" X [HjS
1
4/27/80
9
159
28.9
29.9
b
b'
25.1
25.9 115 119
2
5/5/80
9
201
39
32
b
b
52.4
43.0 74 61
3
6/9/81
9
208
58.8
47.4
47.1
37.9
45.8
36.3 b b
4A
1/20/81
9
190
18.5
16.1
13.8
12.0
28.4
24.7 b b
4B
1/20/81
9
197
16.4
13.7
30.0
25.1
48.9
40.9 b b
3
2/18/81
Lab 1
Lab 2
9
7
34.6
31.7
18.1
22.4
86.4
117
24.4
28.9
116
148
32.0
39.4
153 b b
205 b b
6
2/25/81^
6
150
11.1
12.2
6
c
c
c b b
7
2/25/81
8
138
33.5
40.2
c
c
c
c b b
8
3/31/81 **
10
114
8.3
12.0
b
b
21.5
31.1 b b
* Standard - 165 ppm H-S. Span [H-S] - Method 11 result.
b ^
Monitor not operational.
g
Monitor operational but not Included In test because It could not be brought Into calibration.
^ Fuel gas spiked with knovn amount of l^S uolng Houston Atlas* Inc. Model 601 dlluter.
-------
TABLE 11. CALIBRATION DRIFT TEST RESULTS FOR B2S MONITORS
TEST
CALIBRATION DRIFT®* b
(X)
It UMBER
BENDIX
HOUSTON
ATLAS
DEL MAR
Zero
Span
Zero
Span
Zero
Span
Test 1
1
0
0.7
0.1
6.3
12.5
21.3
2
0
0.7
0.3
4.5
9.6
30.9
3
0
1.1
0.4
0.3
19.5
52.6
4
0
1.1
0.1
1.0
0.7
101
5
0.1
9.2
0.3
13.9
4.8
44.0
6
0.1
8.1
0.1
13.9
0
3.3
7
0
0.7
6.9
2.9
2.6
8.1
Test 2
1
0
1.1
3.5
10.8
6.6
34.2
2
0
0
1.4
26.0
4.4
26.5
3
0
0.7
4.5
3.5
1.5
2.6
4
0.2
0.4
4.2
3.1
1.5
2.6
5
0.1
2.6
1.0
12.8
1.5
1.8
6
1.0
1.1
0.6
1.4
0
0
7
0.1
0,4
—
11.5
4.0
4.8
Test 3
1
0.1
1.8
0.1
4.9
1.1
7.4
2
1.0
0.7
0.3
1.0
0.4
6.6
3
0.2
0.7
1.7
15.6
1.1
8.5
4
0.2
1.5
5.2
21.5
10.7
6.6
5
0.1
1.5
7.6
-
9.9
18.4
6
4.2
2.9
5.9
4.9
0
6.3
7
4.2
4.0
2.1
5.2
1.5
3.7
Test 4 .
-
1
0
0.7
3.1
3.8
0.4
1.1
2
0
0
0.7
6.6
0.7
3.7
3
0
0.7
0.7
0.7
0
0
4
0
0
0
0.7
0.6
4.8
5
0
0
0
6.6
2.0
12.1
6
0
0.4
0
0
-
-
7
0.3
10.4
"¦ Value* la table represent the dally drift •» defined kj tha
fallowing equation
Monitor teadlm! *
1001
{Calibration Cai
Concentration -
Monitor Span Valua
* Because J|-tem drift wa» of jrtmary concern, tH« Instructs
Mn eot itiMl dally, fh# »aluea i® th* tail* Have fcaao
corrected fcj tb« daily »n» drift.
19
-------
SECTION 3
RECOMMENDATIONS
CARBON MONOXIDE CONTINUOUS EMISSION MONITORS
The Anarad and MSA monitors successfully completed the 11-month
study and demonstrated that monitors with good drift control are
available and suitable for the measurement of CO concentrations in FCC
emission gas. Based on the results of this study, the following
specifications are advanced for CO monitors.
• Calibration Drift. The CEMs calibration must not drift
or deviate from the reference value of the calibration
gas by more than 5 percent of the established span value
of 1,000 ppra over each 24-hour operating period.
• Relative accuracy (RA). The RA of the CEMs must be no
greater than 20 percent of the mean value of the
reference method CRM) test data in terms of the units of
the emission standard or 15 percent of the applicable
standard, whichever Is greater, as calculated using
Equation 2-4 In Reference 4 and the manual CO method
results as the reference value.
The correlation among the reference method and CEM data, the
number of RM tests and the calculations should be the sane as those
given in Section 7 of Performance Specification 2(4).
Because of the possibility of leaks when sampling with Tedlar
bags, the number of samples taken for a relative accuracy test should
be at least twelve with the option of discarding the results from any
three if it appears that leakage has occurred. The relative accuracy
testing should be done while the CO concentration In the emissions is
varying less than 10 percent over the duration of the testing.
HYDROGEN SULFIDE CONTINUOUS EMISSION MONITORS
The performance of the H_S monitors was disappointing. Only three
of the five monitors were suitable for field use and only one monitor
(Bendix) had an uptime in excess of 85 percent. The other two moni-
tors that passed the test suffered severe drift problems. The Bendix
20
-------
sampled a total of 17 tines each hour, but each sample duration was
less one second. Thus, the use of the integrated Method 11 sampling
approach for determining the accuracy of this monitor is questionable
unless the fuel gas H^S concentration can be held constant. The
absolute agreement between Method 11 and all the monitors was poor and
variable in eight of the ten relative accuracy tests as shown in Table
10. Since the cause or causes of this difference could not be
identified, the use of H^S monitors for compliance purposes cannot be
recommended at this time.
The monitors may be useful, however, for determining trends and
for indicating amine treater performance, as shown by the overall
agreement between the average for all Method 11 tests and that for
each monitor (i.e.. Method 11-140 ppm, Bendix-119 ppm, Houston Atlas-
130 ppm and Del Mar-133 ppo). (Some Method 11 tests used in calcula-
ting the above averages were not reported in the text of this report
because they represented cases where plant malfunction or Instrument
failure caused a relative accuracy test to be aborted before an ade-
quate number of samples had been obtained).
Due to the apparent bias between the Reference Method 11 tests
and the H^S CEM data, an interference to the reference method tests
is suspected. Additional evaluation of the reference method should be
performed before performance specifications are developed and
installation of monitors is required.
21
-------
SECTION «
DESCRIPTION OF EQUIPMENT
CONTINUOUS EMISSION MONITORS
Carbon Monoxide Monitors
Applied Automation Optichrom 102—
The Model 102 Is a gas chromatograph that uses a flame Ionization
detector (FID) to detect nethane catalytically produced front carbon
nonoxide. The oven, valves and controls are located in an air purged
nodule and the programmer is in a separate cabinet. A sample loop is
used to sample an exact volixne of gas trtiich is Injected onto the
column by a multlport valve. Two colunns in series are used for the
separation - when the CO has eluted through the first column, it is
backflushed to remove the heavier components while additional separa-
tion takes place on the second colixnn. A catalytic methanator con-
verts the CO to methane (in the presence of hydrogen carrier gas)
which is detected by the FID.
Automatic zeroing is accomplished via an auto zero control on the
detector immediately before the CO peak reaches the detector. Span-
ning Is accomplished by a series of attenuation switches and a fine
attenuation potentiometer.
Ecolyzer 3107—
The Ecolyzer Model 3107 monitor utilizes an electrochemical
sensor to measure CO in the ranges of 0 to 1000 ppm and 0 to 500 ppm.
The unit utilizes a sampling pump and a by-pass to vent unused sample
around the sensor. To maintain constant sample hunldity, a salt water
humidifer is located upstream of the detector. An absorbent cartridge
is provided for the removal of SO., and NO from the sample gas.
Sample pressure oust not exceed 2.5 cm water at the sample pump inlet.
Mine Safety Appliances Lira 202—
The Lira 202 has two IR lamps, a "Luft-type" infrared detector
and gold-lined sample and reference cells. One lamp passes through
the sample cell and the other lamp through the reference cell. The
emergent radiation from both cells is directed to a single detector
cell. As the gas in the detector absorbs radiation, its temperature
and pressure Increase. An expansion of the gas in the detector causes
the membrane of a condenser microphone to distort. This distortion is
converted to an electrical signal which is amplified to produce an
output signal. The entire analyzer is kept at a constant temperature
by a thermostatically controlled heater and blower installed in the
case.
22
-------
Ar.arad 501R—
This NDIR analyzer uses unlined Plexiglass sample and reference
cells. The monitor consists of a single IR source, parabolic mirrors,
a chopper, an optical solid state detector and an output nodule that
can be separated from the rest of the monitor.
Hydrogen Sulfide Monitors
Bendix 7770—
The Model 7770 gas chromatograph monitors using an FPD. A
sample block is used to inject a volime of sample gas onto the first
colunn. After the H^S has eluted through the first colunn, the colunn
is backflushed to remove the heavier components. Further separation
is attained in the second coluon prior to H^S detection. Zeroing is
automatic during the running program, but a zero offset control is
also provided inside the programmer. The instrument cycle time is 210
seconds. Clean, dry air is used as the carrier gas.
Process Analyzers Incorporated 32-230—
The Model 32-230 monitor is a gas chromatograph equipped with a
flame photometric detector (FPD) that is sensitive to sulfur. A
volune of sample is injected into the analytical column from a sample
loop. After the H^S sample has eluted past the first colunn, the
column is backflushed to remove heavier compounds. The H^S continues
through the second colunn. At the proper time for the H^S component
to elude through the colunn, the flame photometric sensing circuit is
activated to detect the H?S. Zeroing is automatic and occurs
Immediately before the elution of the peak of interest. Span is
accomplished by adjust- Ing an attenuation potentiometer inside the
case.
Teledyne 611 DMC0-20X—
This dual beam monitor utilizes ultraviolet absorption to quan-
tify H^S concentration in the sample. A 12-inch-long optical cell has
continuous sample flow. On one end of the sample cell is an ultra-
violet source and on the other end a detector. A rotating chopper
with two filters 180° apart is located between the cell and the
detector. One of the filters passes only a known absorption
wavelength for H2S, the other a wavelength at which H^S does not
absorb (reference beam). The wavelengths are not specified.
Synchronizers and electric circuitry subtract the non-H^S absorption
(reference beam) front the total absorption.
Houston Atlas Model 825R/102—
The Model 825R/102 monitor operates on the principle of lead
acetate Impregnated paper tape reacting to change color in the
presence of H~S. A cadmium sulfide photocell Is used as a detector.
The photocell output feeds a preamplifier. The output of the
preamplifier feeds a low-pass filter Which differentiates the signal
with respect to time. The resultant differentiated output provides a
DC signal that has a peak amplitude directly proportional to the H^S
concentration.
23
-------
To obtain a sample In the concentration range of measurement by
lead acetate tape, the nonitor Is equipped with a.sliding block dilu-
ter. This dlluter nixes a measured volune of sample tilth a stream of
diluent gas. The amount of dilution Is adjusted by varying the Injec-
tion frequency of the sample. Zero and span adjustments are made by
adjusting potentiometers Inside the case.
Del-Mar Scientific OM-W—
The DM-W uses a lead acetate impregnated paper tape that Is
exposed to the gas stream as It moves past an aperture. The H.S In
the gas stream causes a black precipitate (lead sulfide) to form on
the tape. Color development is monitored by a photocell that measures
light reflected off the tape.
Because the DK-W normally operates In the 0 to 50 ppm H-S range,
Del Mar supplied a dilution system with the monitor. TTiis system was
comprised of two rotameters, a mixing chamber and a back pressure
regulator. The fuel gas sample was diluted with nitrogen at the
recommended ratio of 1:10 by adjusting needle valves on the
rotameters.
ANCILLARY EQUIPMENT
All monitors and data aquisltion equipment were Installed in a
2'J-foot-long trailer equipped with heating and air conditioning.
Figure 2 shows the location of all equipment within the mobile labora-
tory.
The 00 source was a stack on a FCC unit. The CO boiler stack gas
contained: 150. to 300 ppm NO , 200 to 600 ppm SO., some acid mist,
50 to 300 mg/m particulate n?atter, 10 to 14 percent H_0, 9 to 11
percent C0p, and 20 to 10,000 ppm CO. The temperature at the sampling
point was approximately 320 C (600 F).
The H^S source was a fuel gas pipeline located approximately 120
meters from the trailer. Stainless steel tubing (0.63 cm OD) was used
to connect the gas line to the trailer H^S distribution system. The
sampling point was downsteam of an amine treater. When the monitors
were first Installed in April 1980, the treater was using dlethanol-
amine for the removal of HgS, but in March 1981, the treater was
refurbished and mon6ethanolamine replaced the diethanolamlne. The H-S
level in the fuel line ranged from 5 to over 1,000 ppm during this
study.
Fluid Catalytic Cracker Emission Sampler/Conditioner
An EPA-designed sample conditioning system was located at the
sampling port 60 meters above ground level on the CO boiler stack.
The dried (dewpoint -20°C) and filtered sample was transported 200
meters to the trailer by using unheated, 9.5 nn ID black, nylon
tubing. A detailed description of the sample conditioning system and
its per- forroance is included in Appendix D.
24
-------
Sample Distribution Systems
In the trailer, the gas samples from the sources were distributed
to the monitors by using the distribution (manifold) systems shown In
Figures 3 and 4. The water colinn was Included on the 00 distribution
system to ensure that constant pressure was maintained In the system.
In addition, a pressure reducer was included to ensure that the Ecoly-
xer inlet pressure was maintained below 2.5 cm water. Each distri-
bution systems was fitted with solenoid valves on the span gas and
nitrogen gas lines to facilitate automatic dally zero and span checks
by the data aquisitlon system.
Data Acquisition System/Automatic Zero/Span System
Each monitor was connected to the Esterllne Angus Model PD-2064
data logger. The PD-2064 converted the 4 to 20 mA output of each
monitor to 25 to 125 mV and provided approximately a -25 mV offset.
Each instrument then showed a response on the PD-2064 of 0 to 100 mV.
The PD-2064 is able to monitor all 16 channels continuously or to scan
them at any desired frequency. In addition, the system can scan
selected channels at any desired interval and average the readings for
each channel after the desired number of scans. During the field
testing, each monitor's output was scanned every 3 minutes except when
relative accuracy and calibration error tests were conducted. In
those cases scan times of 1 minute were usually used. After ten
scans, the readings were averaged and the results were printed on the
paper tape. The data and time of printing were also recorded on the
tape.
Every 24 hr, the data logger automatically accomplished a
zero/span check of the monitors. At a set time the data logger closed
the contacts leading to a solenoid valve in the sample distribution
system. This valve shut off the sample gas and opened the nitrogen
(zero) gas. After 15 minutes the data logger took a single reading on
each channel and printed the value. The first set of contacts then
opened and a second set closed to allow span gas to enter. After
15 minutes the data logger again read each channel and printed the
value. The second set of contacts then opened and the original
program of scanning and printing resuned.
EPA Reference Method 11 was used throughout as published. Equip-
ment listings are published in the Method.The manual method for CO
analysis is presented in Appendix E. All equipment is described
therein.
25
-------
Oft
MISC.
OASES
MSA
STANDAM) OASES
OOO*
CO BOX
FIGURE 2. SKETCH OP INSTRUMENTATION TRAILER SHOWING INSTRUMENT ARRANGEMENT.
-------
to to TO
AN/>n*0 MSA Af>0 *UtO.
tO TO
reotrtfn VfMT
N
Mat to
*0
SHut-orr
now
MM
rth-
w
TJ
-«*-
CONSTANT
putssone
FLASK
XJl
FICURE 3. PLUMBING DIAGRAM FOR FCC CAS DISTRIBUTION TO THE CO CEMS
-------
to
KL MAN
TO
PAI
SPARC >*
MMPve .
IN
n no oas w
[NO
FIGURE 4. PLUMBING DIAGRAM FOR FUEL GAS DISTRIBUTION TO THE HjS CEMS
-------
SECTION 5
EXPERIMENTAL PROCEDURES
GENERAL PROCEDURES
Calibration Gases
Calibration gases used for this evaluation were Certified Master
Gases supplied by Scott Specialty Gases, Plunsteadville, PA. Calibra-
tion gases were selected to be 15$ 50, and 90 percent of the instru-
ments' span range with nitrogen being the balance. The gases were
analyzed and certified to be *2 percent of the stated values. Method
11 was used to verify the H-S gas concentrations and a Bendix Model
8501 5CA NDIR CO monitor calibrated with NBS standards, was used to
verify the CO gas concentrations.
Calibration Procedures
Each instrument was calibrated according to the manufacturer's
instructions by using the appropriate sample distribution system to
introduce the calibration gas. Calibration curves were generated for
each monitor.
Throughout this evaluation, all Instrumental response readings
were obtained from the digital millivolt output on the Esterline Angus
PD-2064 data logger. Span and zero adjustments were made on the
lnstrunents to produce a proper millivolt response on the PD-2064 data
logger.
LABORATORY EVALUATION OF THE MONITORS
The monitors were evaluated in the laboratory with the procedures
described in the Federal Register(5). The following parameters which
are defined in Appendix A, were evaluated:
• precision
• noise
• response times (rise and fall)
• H-0 and CO- interference for CO monitors
• CHjj and CH^SH interference for HgS monitors
• flow and t&operature variation
29
-------
Precision (Repeatability)
Precision is the standard deviation about the sean of repeated
measurements on the tase gas concentration (2). In this test,
¦easureoent of the selected gas (e.g., aid-span) Mas Interrupted
alternately by the introduction of a higher and a lower gas
concentration. Six stable readings of the selected gas were obtained
In this Banner and the precision was calculated as follows:
Output Holse
Instrument noise is a short-term variation in instrument output
not caused by changes in output concentration. This value Is
"expressed in concentration units as the standard deviation about the
¦ean (2).
The test procedure involved allowing the instrument to stabilize
on the gas standard (either zero or span) and then taking 25 readings
within a 60 minute period by using a digital voltmeter. These
readings (expressed in concentration units) were entered into the
following equation:
Where: S * Instrunent Noise (ppa)
r± s Instrinent Response for the i reading (ppa)
« *
Instrument Response Tine
Rise tine is the time interval between the initial instrument
response and 95$ of the final response after a step increase in input
gas concentration. The test procedure involved changing the input
froo zero gas to a high-range span gas and determining the time
required to reach 951 of the high-range span gas concentration.
(1)
Where: P * Precision
P. * Instrument response (ppa) for the i
on the selected gas.
th
measurement
(2)
30
-------
Fall time Is the time difference between the Initial response and
95} of the final response after a step decrease In input gas
concentration. The test procedure involved changing the input fran a
high-range span gas to zero gas and determining the time required to
reach 5% of the span value.
To obtain the proper rise and fall times, the "dead volume" in
the sample lines was minimized by placing a three-way stopcock as near
as possible to each instrument's sample port. The sample gas was
switched from zero to span by using this three-way stopcock. The
chart recorder was used at its fastest speed during the test. The
response tines were calculated from the strip chart. Three rise times
and three fall times were calculated and the results of each set were
averaged. The lnstrusent response reported in Tables 5 and 9 is the
larger of these two values.
CO- and' H_0 Interference for CO Monitors
2 2
Carbon dioxide Interference was determined by using a Scott blend
of 10 percent C0? in nitrogen. The response to the introduction of
the 10 percent CO^ is expressed in ppm as an equivalent CO
concentration.
Water interference was determined by adding water vapor while the
instrunent was sampling nitrogen (zero gas). The response to the
added water vapor was expressed in ppm as an equivalent CO
concentration. Water vapor was added by passing nitrogen through 8
flask containing distilled water. The flask was heated sufficiently
to introduce the desired amount of water vapor into the flowing gas
without causing condensation in either the sample lines or the instru-
sent. The water vapor generator was calibrated using EPA Method 4.
Methane and Methyl Mercaptan Interference for H^S Monitors
Methane interference was determined using 99 percent methane.
Each instrument was first zeroed and calibrated according to the
manufacturer's instructions. The methane was then introduced and the
response was recorded as ppm HgS. This same procedure was repeated
using 40 percent hydrogen in methane and 200 ppm methyl mercaptan in
methane. The interference was expressed in ppm as an equivalent
concentration.
Variations In Response Due to Changes in Temperature and Sample Flow
Rate
Variations in instrument output caused by short-term changes in
ambient temperature were measured using zero and span gases. The test
procedure involved allowing the instrument to stabilize at a given
room temperature, recording the response to a gas, changing the
temperature *10oC and again recording the stabilized response. This
test was performed using the trailer's heating and air conditioning
system to control the room temperature.
31
-------
Variations due to flow rate changes were studied using rotameters
calibrated with a soap-bubble flowmeter. The span gas and the flow
rate were varied from approximately 20 to 500 percent of the reconmen-
-------
t0.975
* Student's t-factor.
function
of n; i.e.
n
*0.975
n
*0.975
2
12.706
e
2.365
3
«.303
9
2.306
4
3*162
10
2.262
5
2.776
11
2.228
6
2.571
12
2.201
7
2.447
14
2.160
15
2.145
The 2-hr.span drift was calculated in an analogous manner, except
that the span reading was corrected for any zero drift that occurred
in the 2-hr interval.
24-Hour Drift (Zero and Span)
The data logger automatically introduced zero and span gases into
the sample distribution systems daily and the instrunent response was
automatically recorded by the data logger. Seven consecutive values
were obtained over a one week period to establish the 24-hr drift for
each monitor. Zero and span were adjusted only at 24-hr intervals on
an as needed basis. Equations 3 and 4 were used to calculate the
24-hr zero and span drift values.
Calibration Error Test
The calibration error test was performed by first calibrating
each instrunent, then alternately introducing each calibration and
zero gas until fifteen readings were obtained. The difference between
the instrument's response and the actual concentration of each gas was
then calculated using equations analogous to 3 and 4. (In Equation 3.
the calibration gas concentration would be the denominator and not the
span value of the instrument).
Calibration Drift
Data for this test were taken from the dally zero and span
checks. Zero and span adjustments were made at weekly or longer
intervals.
Relative Accuracy
The relative accuracy of the CO monitors was determined using the
leuco crystal violet wet chemical method described in Appendix E. The
accuracy of the H2S monitors was determined using EPA Reference Method
11 (6). The relative accuracy of the monitors was calculated as
described in Reference 4, i.e.:
33
-------
ltd I + CI] x 1002 (5)
Relative Accur.cy (I) • ^ T,1|]e o{ Eefermc, H.thcKl
where
d ¦ Algebraic mem of the difference between
the reference method value and the aonitor
value
CI » 95Z confidence Interval
3A
-------
SECTION 6
RESULTS AND DISCUSSION
MANUAL CO METHOD DEVELOPMENT/VALIDATION
EPA Reference Method 10 specifies collecting the emission gas
sample in a Tedlar bag and analyzing for CO by NDIR (7). Since NDIR
Is the measurement principle used In most CO CEMs, EPA desired to have
a non-NDIR manual method for checking the accuracy of NDIR-equipped
CEMs. Four wet chemical CO methods previously described in the
literature (8-12) were evaluated In this study.
The first method evaluated (8) Involves the aqueous reduction of
Pd(II) by CO followed by the addition of KI to yield a red, tetralodo-
palladium (II) complex, which is then measured spectrophotometrically
to determine the amount of Pd(II) reduced. The amount of CO present in
the gas sample Is then calculated on the basis that two moles of CO
reduce one mole of Pd(II). Although this method performed adequately
in the laboratory, it was found unsuitable for gas samples that
contained even 1 ppm NO. For example, the following analytical
results (ppm CO) were obtained on three Tedlar bag samples that
contained 266 ppm CO and the following levels of NO in nitrogen: 225
ppra NO (27 ppm); 2 ppm NO (186 ppm); and 1 ppm NO (223 ppm). The
method suffered from: interference by S02 and organlcs; sample
instability; poor precision below 100 ppm u) and had a nonlinear
absorption curve. Because of the slowness of the reduction of Pd(II)
ion by CO, the method also required shaking the sample bulb for 2-hr
before adding the KI. For these reasons, work on this method was
terminated.
Three other methods were then evaluated. These methods used
spectrophotometry and the reduction of Pd(II) ion by CO in the
presence of gun arable (9). phosphomolybdlc acid/acetone (10) and
ieuco crystal violet/potassium lodate (11, 12). The gum arable method
suffered from poor precision, sample Instability, and interference
from low levels of NO and S0^ and required a long reaction time (2 hr
minimum). The molybdenum blue reaction also suffered sample stability
problems, had poor precision for CO concentrations near the level of
the standard (500 ppm) and required a sample reaction time of 1 hr at
60 C ~ 1°C. Further, in both the gun arable and phosphomolybdlc
methods, the reaction product Is a colloid, which means that a
strongly absorbing solution cannot be diluted to bring it into the
linear range of the calibration curve.
35
-------
The last method evaluated, the leuco crystal violet (LCV) method,
yields a soluble complex that permits sample dilution, has good
precision In the range of 0 to 1000 ppm, and is accurate In the
presence of low levels of SOp and NO «5 ppm). Some other favorable
features are:
• the shaking reaction time can be as short as 15 to 20
minutes for CO concentrations from 50 to 1,000 ppm;
• the absorbance Increases linearly with increased shaking
reaction time up to 2 hours;
• the stock reagents are stable and easily prepared;
• the calibration curve is linear in the absorbance range
0 to 1.7;
• the complex that is formed absorbs in the visible region
of the spectrum (587 nm) more than 200 no from where the
reagents absorb. Also, the blank is initially
negligible (Table 12).
Since the LCV method suffers interference from SO- and NO at the
concentrations encountered in FCC stack gases (Table i3), the effec-
tiveness of various scrubbing solutions in reducing these compounds to
less than 5 ppm (without also removing CO) was studied as a function
of flow rate, scrubber volune, concentration and pH, gas volune, and
SO- and NO concentration. The following scrubbing solutions were
evaluated: 6% percent HpO.; 2 percent KMn0^/2 percent NaOH; 4 percent
KMnO^M percent NaOH; 2.% percent KMn0^/1 percent HN0_; and 2 percent
K2Cr!;07/1 percent HNO^ 3
The best scrubbing system is a flow rate of 0.2 to 0.3 L/min
through three Creenburg-Shiith impingers. The first two impingers
contain M00 mL of 4 percent KMnO^/5 percent NaOH and the third 250 mL
of this solution. This system can reduce the SO^ and NO levels to
less than 5 ppm for a 50-llter gas sample containing 15* cbg, 700 ppm
SO^ and 500 ppm NO. Since the Implnger system also removes CO^, an
appropriate volune correction is required when calculating the CO
concentration originally present in the gas. A Fyrite analyzer is
suitable for measuring the C0^ present in the stack gas.
For field sampling, the original single-sample, ambient air
method (12) was modified to allow the analysis of many samples in a
short period of time under field conditions. Sample stability,
shaking time, shaker type, sample volune, and volume of reagent were
36
-------
some * of the parameters studied to simplify the method as much as
possible and at the same tine to optimize its precision and accuracy
at stack CO concentrations.
The method was validated at the petroleum refinery in four field
tests in which the samples were passed through the scrubber solution
and collected in 10-liter Tedlar bags. These bag samples were re-
turned to the laboratory and analyzed in triplicate by the LCV method
and by NDIR on • Bendix Model 8501-5CA calibrated against NBS
certified CO standards. From forty-one samples collected and analyzed
in this manner, it was determined that the manual method had a consis-
tent 4 percent positive bias with respect to the NDIR. The precision
of the method was determined to be 2.5 percent of the mean concentra-
tion for a sample that contained between 15 and 1000 ppm CO when the
sample was analyzed in triplicate. The Tedlar bag samples were stable
for at least two weeks. Representative analytical results are pre-
sented in Table 14 and the actual method is described in Appendix E.
Since the largest source of error was found to be leaks around
the valve of the Tedlar bags, all bags should be checked for leaks
before use, either by submerging filled bags in water or by pulling a
vacuum ( 25 cm H^O) on the bag and seeing if it will hold the vacuum
for at least 4 hours.
37
-------
TABLE 12. CHANGE IK ABSORBANCE OF CO REAGENT
BLANK WITH TIME AT ROOM TEMPERATURE
TIME sncz
nUA*ATI» AlSOMAMCt
(•is)
IBIS STUUT
IXTERTNCE 11
0
0.015
0.020
20
0.013
SO
0.040
43
0.020
•0
0.030
100
0.030
120
0.07
1*0
0.12
230
0.080
280
0.063
320
0.089
MO
0.100
0.21
TABLE 13.
EFFECT OF NO AND
CRYSTAL VIOLET
S02 ON LUECO
KAKTLZ
¦UMBEfc
w.
-------
TABLE 14.
COMPARISON BETWEEN LCV AND NDIR RESULTS
ON FCC SAMPLES
TEST
LCV METHOD RESULTS (ppa CO)
NDIR (ppm CO)
DATE
1
2
3
AV
BENDIX 8501-5CA
SEPTEMBER
149
1980
178
159
163
167
121
120
129
123
115
31.1
31.0
30.1
31.0
34.0
- 78.6
91.1
76.4
82.0
75.0
26.5
28.0
27.8
27.0
26.0
25.2
24.0
26.9
25.0
22.0
28.8
28.0
29.3
29.0
27.0
16.7
17.6
16.5
17.0
19.0
33.0
30.2
33.2
32.0
28.0
JANUARY
1981
1094
1002
1111
1069
1005
356
365
321
347
315
546
592
569
575
176
184
182
180
178
181
184
211
192
175
158
145
153
152
155
188
182
189
186
181
81
87
78
82
62
102
114
110
109
98
519
482
536
512
539
258
248
267
258
235
-------
EVALUATION OF CARBON MONOXIDE MONITORS
Five relative accuracy tests were conducted on the CO monitors at
the refinery. The results of these tests are summarized In Table 6
and the results of each run are presented In Table 15. In these
tests, the samples for the nanual method were passed through
Greenburg-aslth lmplngers containing alkaline KMnO^ solution (to
remove N0X, SO^ and organlcs) and then collected In Tedlar bags. All
samples were collected from the distribution manifold to ensure that a
sample-conditioning system malfunction did not contribute significant
error to the monitor validation procedure.
Tests to verify that CO was not lost in the sample conditioning
system were conducted before each relative accuracy test and at other
times between February 1980 and April 1981. Until September 1980,
these checks consisted of simultaneously collecting stack gas samples
In Tedlar bags at the manifold (distribution box) and at the probe,
and analyzing the bags by NDIR for CO and CO^ at the EPA laboratory in
Research Triangle Park, North Carolina. In September 1980, a Bendix
Model 8501-5CA NDIR CO- analyzer was installed in the trailer to con-
tinuously measure the cO^ concentration in the gas leaving the mani-
fold. After this, the sample conditioning/sample transport system was
checked monthly and also before each relative accuracy test by
analyzing the stack gas for CO^ using a Fyrlte analyzer and by
comparing the result to the Bendix 8501-5CA monitor reading. If the
Bendix C0_ monitor value differed by more than 0.5 percent frcxn the
Fyrlte value, remedial action was taken before the relative accuracy
test was Initiated.
The operation of the FCC unit during this 11-month study was
erratic due to FCC failure and process upsets. The CO levels varied
between 20 ppm and 10,000 ppm CO over a period of several days;
frequently, the particulate emissions were much higher than the 50 to
100 mg/m normally present. Each time a relative accuracy test was
conducted, the CO concentration in the stack exceeded the 1,000 ppm
span range of the monitors, so plant instrument air was introducted at
the probe to bring the CO concentration into the operating range of
the monitors. This dilution also reduced the NO^ and SO^ concentra-
tions In the sample gas by factors of 4 to 10.
During each relative accuracy test, a cylinder gas containing 100
ppm CO, 12 percent CO-, 500 ppm S0_, and 200 ppm NO in N2 was intro-
ducted at the manifold for 30 min and the monitors' output was
recorded while the Tedlar bag sample was taken for analysis by the
manual method. This served three purposes. First, It showed how the
monitors would perform when sampling a stack gas representative of FCC
exhaust gases. (Recall, It was necessary to dilute the stack gas with
air to bring the CO level Into the range of the monitors.) Second, it
served as a control sample to establish how constant a monitor's
response was to the same sample over the length of the study. Third,
it provided a sample of known CO concentration so that the accuracy of
the manual method (sampling and analysis) could be established.
40
-------
taring each of the 30-minute relative accuracy tests, the monitor
output was read at 1-alnute Intervals. The first test, conducted in
June 1980, was unsatisfactory because the manual method being
evaluated (palladlun/potasslixn Iodide) suffered sample stability
problems and interference lYoo NO. Thus, the only way to evaluate
¦onltor accuracy in this test (Table 15) is to compare one nonltor to
another (Runs 1-9) or to compare their response to the 100 ppm 00
cylinder (Run 10).
However, since the other four relative accuracy tests were done
using the LCV method, satisfactory manual method results were
obtained. Ihe fourth and fifth relative accuracy tests were performed
during the same week. In the fourth test, three gas cylinders
containing 500 ppm 00, 12 percent COp, and different levels of SO^ and
NO were employed to see how the monitors would perform If sampling an
undiluted FCC stack gas. As expected, the Ecolyzer detector failed
during the test from exposure to N0„ and SO- which were not effec-
tively removed by the Model 29"9 scrubber, tn the first three rela-
tive accuracy tests, the Anarad and HSA monitors were quite stable,
but, by the last two tests, they somtlmes drifted up to 5 percent of
span over several hours, but at other times remained stable for a day.
This drift sometimes occurred only in the span and at other times only
in the zero setting. No explanation is available for the difficulty.
As mentioned earlier, the only problem encountered with the LCV
manual method was the tendency of the Tedlar bag to leak at the point
where the valve enters the bag. TMs was a significant problem only
in the fourth relative accuracy test where new sample bags were used
and a small leak was found at the metal washer that served as a gasket
between the valve body and the bag. This leak only occurred when the
sample was withdrawn from the bag and for this reason was not detected
when the bags were leak-checked by inflating them before use.
Two calibration error tests were performed: the first in Hay 1980
and the other in June 1980. The results are presented in Table 5.
The interference of C0~ on the monitors remained constant throughout
the study.
Figures 5 through 10 show the long-term zero and span drifts in
the CO monitors. These data, which were collected from May 1980 to
January 1981, were used to determine the calibration drift of the
instruments. (Recall that: (1) only the Applied Automation had auto-
matic zero and that none of the instruments had automatic span
correction; (2) the zero and span of the monitors were checked every
24 hours; and (3) except when relative accuracy tests were being
conducted, the zero and span of the monitors were never adjusted at
intervals shorter than 1 week.)
41
-------
Applied Automation Optlchroa Model 102
The Instrument Is comprised of two nodules, a programmer and a
GC-oven assembly. It was purchased without vendor Installation and
Interconnecting wiring was not furnished. lhe instruction aanual was
furnished three weeks after the instrument was received; hence, the
nonltor was not operational as received. Extensive time was required
to complete the instrument wiring and Install the circuit boards.
After the monitor was wired and Installed, numerous startup problems
were encountered. After a serviceman was sent by Applied Automation,
it was determined that wire of too small a gauge had been used for the
Interconnecting wiring. Nunerous other difficulties were encountered
In obtaining the correct output signal.
To adjust the span, a lengthy procedure requiring a strip chart
recorder and many adjustments is necessary. After the initial span
adjustment, a trial-and-error method is used for fine tuning, al-
though it Is not specified In the Instruction manual.
The response time is limited by the cycle time of the analytical
program. The GC completes a cycle every 105 seconds. Varying the
flow or pressure causes no response variance because a sampling loop
vented at atmospheric pressure is used. An oven maintains a constant
temperature for the chromatographic columns. Since the sample value
Is ambient temperature or in a thermostatically controlled area, it
was not affected by changes In ambient temperature or sample gas flow
rate.
In June 1960, a negative concentration was sometimes measured by
the instrument. After careful examination of the chrooatogram, this
problem was related to a negative signal dip that occurred immediately
before the CO peak eluted. Several trips to the refinery were made by
HE&T personnel to repair the instrument with no success. In November
1960, Applied Automation performed a service call. After two days of
work, the serviceman still could not correct the problem and further
testing of this lnstrunent was discontinued.
Ecolyzer 3107
This instrument performed without failure during the laboratory
testing. Special provisions were necessary to ensure that a sample
pressure of no more than 2.5 en water pressure was present at the
sample inlet port. The span and zero adjustments were not responsive
and thus a alow drift in zero and span followed any adjustment In
these controls.
Sample flow variations from 0.5 to 1.5 times the recommended rate
caused response variations of 10 percent reduction and 3 percent
increase, respectively. In varying the ambient temperature from 10 C
to 30 C no change In response was observed.
42
-------
For removal of NO and SO- from the sample stream, a gas scrubber
cartridge (filter Model 2919) was supplied. This was located Inside
the case and required removal of the sensor for replacement. Because
of the extremely short lifetime of these filters and the difficulty In
replacement, the filters were relocated outside of the case during the
field evaluation. Normal life expectancy of these filters in a stack
gas containing 500 ppm SO^ and 300 ppm N0X is 1 to 5 days. These
filters cost approximately $45 (a considerable operating expense).
Filter failure caused damage to the electrochemical sensor (requiring
replacement at considerable cost).
MSA Lira 202
The MSA Lira 202 operated as received. After less than a week of
operation, the power supply and output signal circuit boards failed.
After the boards were replaced, satisfactory output in the range of
4 to 20 mA could not be obtained, despite several replacements of each
of these circuit boards.
In June 1980, an MSA serviceman corrected the problem. Instead
of using the 4 to 20 mA output on the Instrixnent, he used an Isolated
external converter to convert the normal 0 to 100 mV output to 4-20
mA. After this repair, the monitor operated well throughout the
evaluation. Occasional optical realignment was required to maintain
calibration. Excessive drift was sometimes observed.
Tests for response variation due to sample flow rate and ambient
temperature changes were not peformed due to the late arrival of the
Instrument.
Anarad 501R
This NDIR monitor performed without any difficulty during the
evaluation. No special provisions were necessary for the sample gas
except that there had to be free discharge from the sample exit port.
Sample flow variation from 0.3 to 5 times the recommended rate caused
response variations of less than 1 percent. Varying the ambient
temperature from 10°C to 30°C did not affect the response. Carbon
dioxide and water did give a slight positive Interferences. Overall,
the Instrument proved to be extremely accurate and reliable.
43
-------
TABLE 15. RELATIVE ACCURACY TEST
RESULTS ON CO MONITORS
>n MB
am
m,
at'
to mim<
T|» (M* »)•
irv
i
9
*
fi
«*
»"«' I »)
l
»
»
Mf
IM
in
I
>
t
SM
III
MS
I
>
%
IM
9M
it*
«
»
»
aw
910
Mt
»
i
t
«u
Ml
41?
•
i
»
m
9*
M?
»
i
t
IN
90*
9*0
•
i
t
IM
J?S
MO
•
»
t
911
Ml
n*
M
u
t
Ml
IU
MS
» t..1 » Mm. *01
ij
i
M?
u?
IM
IM
ij
*
W
M
IM
ij
»
12
U
•0
n
»!
t
is
»
9?
9)
ti
t
0
AO
M
4)
•J
*
1)
M
M
4)
»•
p
9M
Ml
Ml
M0
•»
—
m
M
M
90
i!
1
HM
MM
MM
MS
£
9
9*0
9?J
MS
M4
<
1
111
lis
IM
in
i
t
If}
9)4
t»
MS
£
U
M?
9D»
9M
ts
t
It*
to*
Ml
its
»•
9
ttt
MS
49*
in
w
«¦»
—
4M
490
49*
4M
•"""»*¦ •!>
1*
U
SSI
un
9M
MT
I*
U
91)
mo
M?
M?
<
1)
90)
mo
•M
91*
?
a
«n
1100
*M
910
li
ii
1)7
UM
Ml
910
•i
u
49?
IMI
90S
9M
u
901
m*
Ml
90S
•J
u
M)
MM
IM
Ml
•*
u
915
Ml*
91*
M?
—
90?
IMI
91*
M?
¦» »»« » "¦» •!>
if
9
m
J
ns
?1S
t
9
9k)
J
MS
ns
tl
9
Ml
J
M0
m
»'
9
•00
9
ns
IM
**
9
Mi
J
M2
m
9
m
j
Mf
m
*
9
«M
J
no
ns
•*
9
MI
j
7)4
m
•*
9
Ml
J
994
VM
m
9
no
no
ImuJu tmr 6D} >»rfn—i.
tfv HIM IMI M( «Hi.
M»4: tco Vf» BD; MO Kj. MB WD, til— Bj.
•uck |»* liUti# 1:4 «ttl> toirwat «lr.
lufk lllaiiM 1110 vltli tetnn tu.
Iwi1ni My
Ctltmdmt m* NO *«¦ 0D. MC V. MC Ky IB C0J; »*1m»
fe*Uar NBltf •m M MiUrautf. IV) yp 00 M4i(< nd M0 «*».
44
-------
A • 4**• am MUmi
m k^t
1
•
Aft
- M
_
*^fra
IC tt«
FIGURE 5. ZERO DRIFT TREND FOR ECOLYZER CEM.
MM
¦»e
MO
•to
MO
liM< I |M MlutM
HI
* * Me in* (ii)M m fet«
Ihhi.
—
\A,
W
*
V
•j
I
¦i
\
«m mi «tt •» mi »«•
FIGURE 6. SPAN DRIFT TREND FOR ECOLYZER CEM.
45
-------
1
A » teci MM fsllwe
» iau
1 • *mmt
-
-0
*"°
•
A,
*A-~
1
/
V
"^1
*"-\J
m «d m >n arr m* hc «• m w
FIGURE 7. ZERO DRIFT TREND FOR MSA CEM.
-
li« hiuim
|
1
ulikrttlA |«< t«t.
A • JUu ih" UUm
• ny 1«h't .
.
A
i ~
V
¦'l
1
I
»»
i
*
.
Ifn
h
HI ^
M «U M *C« ¦>> HI »!• MM
FIGURE 8. SPAN DRIFT TREND FOR MSA CEM.
46
-------
la am* *ill «w
mm H—ir.
k
1*.
* • tm
m
%
mc
wm
•
i •
'!
i
i
U
|
1
*=7
«M m mm «« «cr ¦* «( Ml
FIGURE 9. ZERO DRIFT TREND FOR ANARAD CEM.
mm
|»«>»< 1IM tnafiratat
... mm.
•
*
fl
4«Im
i • Awt
|)l lla*
JyMf f]«
• a*r» fi
a latnr
t*
(*•«
H«n m
-s
y
r
- ?
t
ri
¦
ml
IT
n
-------
EVALUATION OF HYDROGEN SULFIDE MONITORS
The results of the nine relative accuracy tests are summarized In
Table 10. The individual test results are presented in Tables 16 and
17. From an examination of the data, the following observations are
Bade:
1. When two laboratories simultaneously conducted a rela-
tive accuracy test of the nonitors, very good agreenent
was obtained between the laboratories, but their results
differed significantly from the nonitors' results (Table
17, RA Test 5).
2. A monitor could compare well one day but poorly the next
(Table 16, RA Test MA, 48). The overall agreement
between Method 11 and the monitors, as measured by the
relative accuracy tests, was quite variable over the
length of the study (Table 16). This variability was
also noted in the previous short-term study (3).
3. Generally, the Bendix and Houston Atlas monitors
measured H^S concentrations lower than Method 11, but on
one occasion (RA Test 6, Table 16) the Bendix measured
HjS levels higher than Method 11. Experiments to
determine the cause of these differences were
Inconclusive. Vhen the fuel gas was spiked with known
amounts of H^S upstream of the manifold, the spike was
adequately recovered by both the monitor (Bendix) and
the Method 11 procedure. During these tests, (February
and March 1981) the Houston Atlas monitor was not
operational and the Del Mar was drifting severely, so
only the Bendix results were reliable. The two
laboratory collaborative test (Table 17) shows that the
difference in concentrations was not due to laboratory
bias in Method 11. Both Method 11 and the monitors
agreed well when analyzing H^S in nitrogen cylinder
gases.
H. When the eighteen, 30-minute tests (Table 16, RA Test 4A
and 4B) were grouped to yield nine 1-hour tests, the
relative accuracy of the monitors with respect to Method
11 did not improve significantly. This indicates that
increasing the sampling time to 1 hr does not
significantly improve the results obtained.
Figures 11 through 16 show the long-term zero and span drift ob-
served in the HpS monitors. These data, which were collected from
May 1980 to January 1981, were used to determine the calibration drift
of the instrunents. (Recall that: (1) only the Bendix and PAI had
automatic zero and that none of the instruments had automatic span
48
-------
correction; (2) the tero and span of the monitors was checked every 24
hr; and (3) except when relative accuracy tests were being done, the
zero and span of the monitors were never adjusted at Intervals shorter
than one week).
Bendlx 7770
This Instrument Is equipped with numerous safety devices Includ-
ing automatic shutdown and start up If certain hazardous conditions
exist In the analyzer. A molecular sieve was added to the carrier air
and hydrogen supply to ensure contamination free gases. The Bendlx
operated 89 percent of the time during field evaluation. Host of the
downtime was attributable to two pneixnatlc. valve failures; the down-
time was lengthened due to personnel not being on-site to replace the
part.
The flow variation and vnbient temperature change tests were not
applicable to this instriment because It used a heated sampling block
to inject a constant and known volune of sample Into the chromotagraph
column, regardless of the sample flow rate. The cycle time of the
Instrixnent was 210 sec. For this reason, the time from injection to
100 percent response was used in the laboratory evaluation for
response time rather than from first response to 95 percent of the
final response.
The results of the precision, drift, noise and calibration error
tests were excellent. Relative accuracy testing showed a consistant
difference between the monitor and Method 11 for the fuel gas. The
agreement was always good for H^S in nitrogen. An interference in
either the monitor or the reference method was indicated.
Process Analyzers Incorporated 32-230
This instrument never functioned properly. Many hours were spent
troubleshooting circuit boards to attempt to find the cause for the
lack of a proper output signal. The schematic wiring diagrams
furnished In the operating and maintenance manual were not in agree-
ment with the actual wiring. The unit was returned to the manufac-
turer for repair. After more than 4 weeks, it was returned to HE&T
and installed In the test trailer on the final day of the first field
evaluation. Before the next scheduled field evaluation, the monitor
•gain failed due to a worn sample valve.
The output of the monitor was not the 4 to 20 mA specified in the
equipment order; and thus the monitor was not compatible with the data
aquisltlon system. Thus, Its output was recorded on the narrow
recorder supplied with the monitor. After several breakdowns, the
monitor was shut off and Its evaluation was discontinued.
49
-------
Teledyne 611 DHC0-2QX
This instrianent required alignment of the chopper by using an
oscilloscope prior to operation. Severe Instability caused by any
¦ovement or vibration necessitated.that laboratory testing be done at
night when wind or personnel would not cause Bovenent of the
evaluation trailer. Significant interferences were caused by nethyl
¦ercaptan and other compounds found in fuel gas. The nolecular sieve
scrubber provided with the Instrument could not correct the
interference. An alternate scrubber was eventually furnished but
complete failure of -the instrument had already occurred. Because of
¦ultiple factors preventing acceptable operation of this nonltor
testing was discontinued before • valid test was achieved (See
Appendix F).
Houston Atlas 825R/102
The instrument performed properly as received. The Instrument's
gas dilution system was not affected by sample flow rate or pressure
changes, but it was affected by ambient temperature changes. That is.
Increasing the ambient temperature decreased the response In accor-
dance with the ideal gas law. This affected the accuracy of the dilu-
tion.
Because of a wiring defect, the instrunent did not operate during
the first field test and was returned to the manufacturer. It was
rapidly repaired and returned to the refinery.
Extremely large variance in the dally span values (10 to 30
percent) during the first 8 months of operation resulted from water
condensation in the sample vent line. Increasing the downward slope
of the vent line eliminated the problem.
The pneivnatlc actuator for the dilution system failed twice
during the field evaluation causing monitor downtime. No specific
interferences were detected but the general agreement between Method
11 and the monitor was not within the desired 10-percent range. Each
time the lead acetate tape was changed, the monitor required recali-
bration. This required a mlnlmun of 2 hr due to the instrument's slow
response to calibration adjustment. A roll of tape lasted
approximately 14 days.
Del Mar Scientific DH-W
The instrument operated as recieved, but before continuous
operation was possible, the tension on the tape take up reel belt had
to be increased by shortening the belt. The gas dilution system
supplied by the vendor was Inadequate. It was comprised of two rota-
meters (one for sample and one for dilution gas), a mixing chamber and
a back pressure regulator. Slight changes in sample gas viscosity and
50
-------
changes in both sample and dilution gas flow rate and pressure caused
drastic calibration shifts in the instrunent. Large volwes of
dilution gas (10 L/nin) were specified by the nanufacturer. At this
rate, a cylinder of nitrogen would last 1 day. Thus, to conserve gas,
both sample and dilution flows were reduced to 15 percent of the
reconmended rate. Adverse effects were not noted by this reduction in
flow.
Variations in sample pressure greater than 1 psl caused large
variances in the sample flow to the dilution system, which in turn
caused poor results in precision, noise and drift evaluations, the
major problem with the instrunent appeared to be the sample gas
dilution system.
51
-------
TABLE 16. RELATIVE ACCURACY TEST RESULTS ON HjS MONITORS
TZST HQ)
MTTHOD 11
DITTMENCE ll-MoMteT>
fnrrm
am
t£5Ll.TS
MKDXX
•0UST0N ATLAS
on. MAX
(PP» %S)
<»P«)
(PP»)
(ppm)
KA T«»t 1
(4/27/80)
1
161
43
37
-197
2
119
S2
25
-248
3
165
47
4
- 77
4
172
21
37
-141
3
166
22
39
-138
6
167
32
40
•138
7
166
44
28
-147
•
156
3fl
34
-158
9
2£3
11
45
-153
Itoao
160
39
32
-155
KA T«»t 2
(5/3/80)
1
157
22
-6
-143
2
189
52
23
-104
3
173
50
25
-120
4
. 222
44
18
- 67
S
219
29
128
- 69
6
226
35
43
- 64
7
251
75
•6
- 35
•
247
•8
117
- 43
9
230
26
1
-158
Mud
202
47
48
- 89
U T«*t 3
(6/9/80)
1
262
99
76
•0
2
269
*6
67
87
i
226
73
50
49
4
245
91
74
68
S
199
78
61
•6
6
233
102
•6
33
7
219
•2
78
30
S
163
62
47
37
~*
28
28
28
28
Hub
205
79
63
55
.
(Continued)
\
52
-------
TABLE 16. RELATIVE ACCURACY TEST RESULTS ON B^S M3NITORS
BATE
tiSUlTS
ItKDlX
HOUSTON ATLAS
BEL MAX
(pp« KjS)
tfl*>
Cm)
(**¦)
1A test 4*
(1/20/81)^
S
222
SI
c
§9
1
in
24
e
72
1
190
24
16
- 5
•
1*6
20
16
0
*
171
4
2
- 7
10
179
11
3
3
11
its
24
28
10
12
182
14
17
-95
13
lii
11
11
-70
Hiis
1»0
22
IS
1
KA Tcct 48
(1/20/81)*
•
14
174
1
2
- 97
IS
177
S
8
- 92
1*
203
25
18
- 31
n
209
24
28
. 28
18
1*8
15
30
- 33
19
208
29
54
- 31
20
198
21
4?
- 33
21
206
27
57
- 43
22
203
27
56
-108
Hkaq
197
19
33
- 56
KA T.ft 6 (2/25/81)'
1
285
22
c
c
2
151
16
c
c
3
71
-1
c
c
4
97
6
c
c
3
128
0
e
c
•
MI
IS
c
c
litan -
ISO
9
—
•
ia r«*t ?
(2/25/81)
l
131
17
c
c
2
281
48
e
c
J
§2
20
c
c
4
145
43
c
c
s
203
*4
c
c
*
238
§7
e
c
1
17
27
e
c
•
12
Ik
e
c
Main
137
U
-
*
(Continued)
53
-------
TABLE 16. RELATIVE ACCURACY TEST RESULTS' OK B?S MDKITORS
IXSI U© METHOD 11 PITTEREKCE (Method 11-Wonlter) (Mm IhS)
BfcTE ttSUITS ID© IX 1OUST0H AT1AS PtL KM.
P« %s) (ppm) (rpm)
U Teat t (3/31/81)
1
64
13
e
27
2
116
6
e
42
S
171
16
c
70
*
70
1«
e
10
S
US
0
e
26
6
153
-»
c
-2
I
212
-1
c
t
s
77
15
e
4
V
78
12
c
-1
10
Mean
84
- 114
18
8
c
9
19
* Fuel |ii aaaple collected In a Tedlar Uk*ft at tUi tlM atkd aaalyaad
fcy CC eo* day later showed 1«»* than 2 pjw H,S.
%
90-aisute rail.
c Monitor set operat local.
* Fuel ^a« Mia iked vlth knows amount of t^S for all ai* rrna.
54
-------
TABLE 17. RESULTS OF COLLABORATIVE RA TEST ON HjS MONITORS
DIFFERENCE (Method 11-Monltor)
(pp« W»S)
BEHDPt WOUSTOW ATLAS DTL HAK
LAB 1 LAB 2 LAB 1 LAB 2 LAB 1 LAB I
M Test 5 (2/18/81)
1
3
-19
-10
-2?
—
2
8 27
-11 -12
-4,
-8
-3
-*
3
1 4)
-3! -39
-33
-41
-36
-43
4
3
-26 —
-38
-35
—
5
1 3)
-23 -33
-37
-43
-43
-51
6
3 29
-29 -33
-40
-44
-52
-56
7
2 23
-24 -31
-34
-41
-4)
-52
8
1 32
-28 -27
-37
-3#
-52
-51
9
9 33
-32 -32
-39
-35
-64
-60
Mean
3 32
-25 -30
-30
-36
-40
-45
Ml HUD 11*
TtfC AND (PP" "»S>
DATE LAB 1 LAB 2
a The reference aatM miI^ni were ptrfonH ilMltmeuily by two different laboratories at 11 Icing
totally separate equipment, reagenta* and standard*.
-------
FIGURE 11. ZERO DRIFT TREND FOR BENDIX CEM.
biM lla* i»4UAim MlttrWMi
fNtntniM.
i • l*i:r«ni NmUm
1 • b
-------
A • Am
10 Ml tlllH.
!
fx*
•
-
I
1
I'
m
L
1 W* 9
1
\K
A
X.
V
mm
1
FIGURE 13. ZERO DRIFT TREND FOR HOUSTON ATLAS CEM.
i)
la
I!
cjJ Uri
Ifatlfl
ft
I
ill
ii
i /
1
r
1
!i|T(i
A~
t
r
r»
i
•\\\
i
h
liil
m m
«u M «n art ¦( tm m
FIGURE 14. SPAN DRIFT TREND FOR HOUSTON ATLAS CEM.
57
-------
4 • «*« fallv* •
iiu I««r.
i • Itr*
mrtummt.
1
a
1}
1
h/\
fm
lv
J\'
L>
\
•
r
¦
FIGURE 15. ZERO DRIFT TREND FOR DEL MAR CEM.
I* • _
t* ite fm-
B • ••• e«pe^»tr»ti
d*|ti %» Mt pp»-
• 0
III
FIGURE 16. SPAN DRIFT TREND FROM DEL MAR CEM.
56
-------
REFERENCES
1. Standards of Performance for New Stationary Sources.
Petroleum Refineries. Federal Register, 43:
10866-10873, March 15. 1978.
2. Repp, H. Evaluation of Continuous Monitors for Carbon
Monoxide in Stationary Sources. EPA 600/2-77-063, U.S.
Environmental Protection Agency, Research Mangle Park,
MC, 1977. 155 pp.
3. Malnes, G.D., and V.C. Kelly. Determining Laboratory
and Field Performance Characteristics of H.S Monitoring
Systems as Applied to Petroleum Refinery Fuel Gas Lines.
Draft Final Report, EPA Contract 68-02-2707, Scott
Environmental Technology, San Bernandlno, CA, 1978.
4. Standards of Performance for New Stationary Sources.
Proposed Revisions to General Provisions and Additions
to Appendix A, and Reproposal of Revisions to Appendix
B. Federal Register, 46: 8352, January 26, 1981.
5. Standards of Performance for New Stationary Sources.
Appendix B. Performance Specifications Federal Register,
40: 46250, 46271, October 6, 1975.
6. Determination of Hydrogen Sulfide Emissions from New
Stationary Sources. Petroleum Refineries. Federal
Register, 43: 1495, January 10, 1978.
7. Determination of Carbon Monoxide Emissions from New
Stationary Sources. Petroleum Refineries. Federal
Register. 39: 9319-9323, March 8, 1974.
8. Allen, T.H., and V.J. Root. Colorlmetrlc Determination
of Carbon Monoxide in Air by An Improved Palladium
Chloride Method. J. Biol. Chem., 216: 309-17, 1955.
9. Anonymous, Determination of Low Concentrations of Carbon
Monoxide. J. Soc. Chem. Ind., 57: 79-82, 1938.
59
-------
10. Polls, B.D., L.B. Berger, H.H. Schrenk, Colorimetric
Determination of Low Concentrations of' CO by Use of a
Palladium Chloride - Phosphomolybdlc Acid-Acetone
Reagent, Publication No. 3785, U.S. Dept. of Interior,
Bureau of Nines, Report of Investigations, November
19^4. 16 pp.
11. Veins, R.E. Colorimetrlc Methods for the Determination
of CO in Air. Thesis. Kansas State University,
Manhattan, Kansas 1973- 32 pp.
12. Lambert, J.L., and R.E. Veins. Induced Colorimetrlc
Method for Carbon Monoxide Anal. Chem., 46: 929-930
1971. ""
60
-------
APPENDIX A
DEFINITION OF TERMS
Calibration Drift - The difference in the monitor's output
readings from the established reference value after a stated
period of operation during which no unscheduled maintenance,
repair, or adjustment has taken place.
Calibration Error - The difference between the pollutant
concentration Indicated by the continuous monitoring system
and the known concentration of the test gas mixture.
Interference Equivalent — Positive or negative response
caused by a substance other than the one being measured.
Operational Period - A mlnimun period of time over which a
measurement system Is expected to operate within certain
performance specifications without unscheduled maintenance.
Output Noise - Spontaneous, short duration deviations in the
analyzer output that are not caused by input concentration
changes. Noise is determined as the standard deviation
about the mean expressed as a percentage of full scale.
Precision - Variation about the mean of repeated measure-
ments of the same pollutant concentration, expressed as one
standard deviation about the mean.
Range - The mlnimun and maxlmuo measurement levels.
Relative Accuracy - The degree of correctness with Which the
continuous monitoring system yields the value of gas
concentration of a sample relative to the value given by a
defined reference method or the emission standard.
Span Drift - The change in the continuous monitoring sys-
tem's output over a stated period of normal and continuous
operation when the pollutant concentration at the time of
measurement is the same known upscale value.
61
-------
System Response Time - The time interval between a step
change In pollutant concentration at the input to the
¦onitoring system and the tine at which 951 of the
corresponding value is displayed on the system data
recorder.
Zero Drift - The change in the continuous monitoring system
output over a stated period of time of normal and continuous
operation when the pollutant concentration at the time of
the measurement is xero.
62
-------
APPENDIX B
TENTATIVE PUN FOR THE EVALUATION OF CO
AND H^S CONTINUOUS MONITORS AT REFINERIES
by
William J. Mitchell
Quality Assurance Division
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
APRIL 1979
63
-------
i % UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
t VW 5 ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
V^Wri RESEARCH TRIANGLE PARK
^ NORTH CAROLINA 27711
April 26, 1979
Dear
He are Initiating a program to evaluate the long-term reliability of
continuous monitors for measuring H-S in fuel gas feed lines at petroleum
refineries and for measuring CO emissions from fluid cat cracker stacks.
On March 15, 1978, EPA promulgated a regulation that required petroleum
refineries to install such monitors, but only after EPA developed
performance specifications for these monitors. The objective of our
program 1s to evaluate tentative specifications recommended for these
monitors by an EPA contractor.
The program will Involve Installing CO and H~S stack gas monitors at
a petroleum refinery and operating and maintaining these monitors for
approximately one year. The 1979 Pollution Engineering Yearbook and
Product Guide indicates that you may have one or more monitors that could
be used in this study. If this is correct, we would appreciate receiving
Information about your monitor as an aid 1n planning our program. For your
convenience we have enclosed an attachment with this letter that outlines
the information that we need.
For your Information and comment, I have attached a copy of our
tentative plan. We anticipate that the H-S continuous monitors will be
monitoring a stack gas that will be comprised primarily of methane and
hydrogen and contain 30-500 ppm H2S, 10-500 ppm mercaptans and some SO2.
The CO monitors will be monitoring a flue gas that will have the following
character: 25-800 ppm CO, 8-15% CO^J 200-600 ppm SOg* ?00 ppm NO, 8-12*
HgO, and particulate concentrations between 25-150 mg/m .
If you wish to submit Information about your monltor(s) or need
further information, please write to me at the address above or call
me at (919) 541-2769.
Sincerely yours,
William J. Mitchell, Ph.D., Chemist
Source Branch
Quality Assurance Division (MD-77)
Attachments
64
-------
TENTATIVE PLAN FOR THE EVALUATION OF CO
AND H2S CONTINUOUS MONITORS AT REFINERIES
I. BACKGROUND
On March 15, 1978, EPA promulgated New Source Performance Stan-
dards that required petroleun refineries to continuously monitor the
CO emissions from fluid catalytic crackers and the H-S levels in fuel
gas feed lines. However, at the time the regulations were promul-
gated, EPA did not have performance specifications available for the
CO and H«S continuous monitors. Therefore, the refineries are not
required To install the monitors uitil these performance specifica-
tions are developed.
Tentative performance specifications for both CO and H2S monitors
have now been advanced based on laboratory and field evaluations done
by Scott Environmental Technology. In the laboratory phase of Scott*s
program, candidate instrunents were evaluated to determine response
characteristics, stability with time, temperature and flow rate, and
sensitivity to potential interferences'likely present in the stack
gas. Instruments that performed adequately in the laboratory were
then evaluated at a petroleum refinery to establish their field
performance.
Five CO monitors and two H^S monitors were evaluated at the
petroleuo refinery during field trials that lasted approximately 55
days. Only one CO monitor and one H^S monitor performed adequately.
In the case of the CO monitors, daily calibration checks were manda-
tory for reliable operation of the instruments, but even with the
Inclusion of dally calibration, Scott questioned the long term
reliability of these monitors. Instrument malfunctions, sampling
system malfunctions and data logger malfunctions plagued these field
evaluations.
Based on these studies, Scott proposed tentative performance spe-
cifications for both CO and H-S monitors. The proposed CO monitor
specifications are summarized in Table B-1. The proposed HgS monitor
specifications are the same as those described in EPA Performance
Specification 2 - "Specifications and Test Procedures for SO^ ®nd NO^
Continuous Monitoring Systems in Stationary Sources."
65
»
-------
TABLE B-1
RECOMMENDED PERFORMANCE SPECIFICATIONS FOR
CONTINUOUS MONITORS OF CARBON MONOXIDE AS
APPLICABLE TO PETROLEUM REFINERIES
PARAMETERS
SPECIFICATION
Range
0-1000 ppm
Calibration Error
<2f Span
Relative Accuracy1
_O0f Mean Ref. Value
Precision
Of Span
Respone Time (System)
<10 Minutes
Output Uoise
<1% Span
Zero Drift, 2 Hours1
Of Span
Zero Drift, 2H Hours1
£2X Span
Span Drift, 2U Hours1
<2.5% Span
Interference Equiv. 15% CO^
<10 ppo
Interference Equiv. 10J H^O
<5 ppm
Operational Period
166 Hours
Expressed as sun of absolute mean value plus 95X confidence
interval in a series of tests. This value is based on •
relative comparison of the monitors to each other and not to
Method 10.
II. PROPOSED PUN
A. Objective
Establish the long-tern operational performance (durability/
reliability/accuracy) of CO and H^S monitors when the monitors are
installed and maintained as directed by the instrument manufacturer/
vendor and from this data determine what are reasonable and useful
performance specificalons for those monitors.
66
-------
B. Duration of Project
The field testing will last approximately one year. The labora-
tory testing preceding the field study vlll last one to three months
depending on the problems encountered and the Instruments selected for
evaluatlon.
C. Site Selection Criteria
The CO monitors will be Installed at a fluid cat cracker and the
H.S monitors will be Installed on a fuel gas feed line preferably
equipped with an amine treater for removal of HjS. The actual
selection of the test slte(s) will be made using the following
criteria:
1. Attitude of plant management toward program.
2. Accessibility of site to EPA and EPA contractor
personnel for installation, calibration and
maintenance of equipment.
3. Scheduled plant shut-downs for maintenance,
production changes, etc.
4. Availability of a room or trailer that is suitable
for Installing continuous monitors, i.e., one that
can be maintained at a constant temperature and
humidity.
5. Availability of a stable, continuous supply of
electrical power.
D. Equipment Procurement
1. Continuous Monitors
The continuous monitors will be selected based primarily on
engineering Judgment about their technical reliability and durability,
maintenance requirements, data recording requirements, and avail-
ability of spare parts. Specific criteria for monitor selection
Include the following:
a) Willingness of monitor vendor to cooperate with EPA,
e.g., supply the monitoring system requested
including all pre-delivery instrument check-outs
requested.
b) Cost of the monitoring system to the government In
relation to other monitors with similar operating
principles.
c) Delivery time involved in obtaining the monitor.
67
-------
d) Anticipated cost of maintenance, calibration and
repair of the Instrument including availability of
spare parts, ease of on-site repair and availability
of service personnel for aajor equipment repair.
e) Sampling conditions required by the aonltor. I.e.,
temperature, stack gas character, humidity, flow
rate, etc., and availability of a suitable sample
conditioning system for obtaining the required
sampling conditions.
f) Existence of a similar system on other petroleum re-
fineries and the demonstrated performance of the
system.
g) Availability and cost of training EPA contractor per-
sonnel in the operation and maintenance of the moni-
toring system.
2. Stack Gas to Monitor Conditioning System
The actual sampling system required to bring the stack gas to the
monitors cannot be determined until the field site(s) and the candi-
date monitors are selected. If a commercially-available system exists
that can be obtained at reasonable cost, such a system will be bought.
However, if necessary, a site-specific sample conditioning system will
be designed and installed by EPA. The system installed will meet or
exceed requirements of every monitor that will be used in the evalua-
tion.
3. Data Recording System
It Is anticipated that strip chart recorders will be used to
record all data generated by the monitors. Data loggers and magnetic
tapes will likely not be used since these devices have not demon-
strated long-term operational stability. Similarly, manual calibra-
tion and span checks will be an Integral part of the program rather
than relying simply on automatic controllers.
E. Trailer for Housing the Monitors
Monitors require a well-controlled environment for operational
stability. A trailer or similar facility that has temperature and
humidity controls will be used to house the monitors.
F. Determination of Performance Specifications
and Operational Characteristics
The performance of the CO monitors will be compared to the ten-
tative performance specifications now being advanced by QAQPS as
Performance Specification 4. The performance of HjS monitors will be
compared to the most recent specifications for SO^ and NO^ monitors.
68
-------
Relative accuracy performance testa for 00 will be conducted using EPA
Method 10 unless ESRL is able to supply a wet chhemlcal test nethod
for 00. Hie relative accuracy tests for H-S will be conducted using
EPA Nethod 11. Initially, the number of samples specified in perfor-
mance Method 2 for SO, will be used for the H-S and 00 relative
accuracy tests. Span checks and drift checks will oe done daily using
calibration gas mixtures that correspond to OS, 251, SOS, 751 and 100X
of span. However, not all span gases will be used everyday.
Described below is the tentative schedule for these tests. This
schedule stay change based on final selection of nonitors and test
sites:
Parameter
Relative Accuracy Test
Zero/Span Checks
0, 501, 1001 span check
0, 251, 501, 751, 100* span cheek
Time Interval
1 week
1, 2, *, 8, 12 Bonths
dally
weekly
Response Time/Noise Check weekly
Interference of other gases monthly
C. Tentative Schedule for Accomplishment of Study
Task
Obtain permission to presurvey
tentative sites
Complete site visits and select
test sites
Order monitors/sampling
systems, etc.
Procure necessary supplies for
reference method tests
Complete laboratory check-out
of monitors
Install/check-out nonitors
Complete field studies
Completion Date
May 15, 1979
June 15, 1979
July 15, 1981
August 15, 1979
November 15, 1979
January 15, 1980
January 15, 1981
69
-------
III. SPECIAL MOTES
A. Each monitor will be subjected to tAiatever lab tests are
necessary to determine if it will work in a field situation.
These lab tests nay be done by the vendor before shipping the
equipment to the EPA contractor.
B. Tentatively, we plan to have the person or persons res-
ponsible for maintaining the monitors receive monitor-specific
training from the instrunent vendor either on-site or at the
vendor*s facility.
C. Based on cost and anticipated delivery tine, spares of
those components most likely to fail (phototubes, scrubber
coitions, switches) will be stored at the test site.
D. If possible, two identical instruments may be operated
at the site to yield an estimate of instrument precision. The
method of standard additions may also be employed on occasion to
check the reliability of the monitors. In this case, a calibra-
tion gas would be Injected into the sample stream to see if the
expected Increase in response was obtained.
70
-------
APPENDS C
WNDORS RESPONSE TO LETTER FROM EPA
¦BCATIVt
¦o usrann
lURiSt 00 |,|
raorosro wjtkuwwi
cost/of oat tuc niMcim
AfflM AuKm*turn, lac
fawfcyalfc* Im4
•artlaavllla, OK MOO*
*¦
CO: Optlchrne 102
¦if: Optlehro* 102
CO:
»»7tS/€C/nD
»»j,ooo/cc/rr»
Ittn tcolep C**».
r. e. in M>»t
Imim, t* >»M
n
CO: MOO
CO:
$»oo/u
•acharacb hitranli
1100 U»Kom ttraat
¦mctala View. CA 9MM3
XX .
feaaattea l»<»«trlaa
r. 0. fca Ml
tye«a. CO §01*0
CO: P I030A
¦tS: r IOJOA
CO:
>lt:
~4500/CC/TC
IMOC/CC/TTD
Bactaam Imetr*Mita, tee.
hwiM Inatruaanta (IftiiM
MOO Marker Uvtf.
FnUartaa. CA HtM
CO: MV-U
Rtt: tSl
CO:
•»$:
MNOmiB
»»000 /«*•! luaiAa aceaca
•*a
Mtitak, tec.
23' Fut naa Avenue
Cutr»|t, HA 021 J*
t.l. BuTont BtHraovr* 4 C*.
100? Market Street
VUiU|lw, PC ItTOt
(Hr|ti let Science, tee.
IS laacutlva »lw<.
Uaafaitf. Wt 1052)
IX COi Iftatru. Mt ap*ctfla4 || («»i BpaelltaO
1,1: M0
•,$. IM00/W
n CO: ¦cclyaer >107 CO: SJIOO/elettreck
71
-------
APPENDIX C
VENDORS RESPONSE TO LETTER FROM EPA
nCATlVt
•0
Usrmii
unnn
Co b(i norotrc wmuwwT
corr/ofouTi»c niRCiru
IrWKtVMI
(n*lranaantal FrWactl tlv.
fp«»Uf 11m Im4
tiwtaii. r* m:i n
Ictltur l«t<, lac.
•2 J 7 Av«.
Ckatavorth. O VlStl IS
bhrl latartacfc C*if.
It Koata] (oad
Prlacataa. U MM XI
Inlitl-Fwkirl
icltniiflc tdiiriwnti >i«.
1*01 California A*«au«
Pale Alto. CA **H
¦ertka luitrianti, lac.
1021 Bury** Ava. —
Irrlaa. CA «7I*
¦button M)t(, lac.
•4* 1 brtkem Dri«a
¦euatoe, 11 7704)
lafrarad Mvitfiti
thatarn Dlvlalaa
r. 0. lot M9
lut< karkara, CA UI02
l|i
¦l>
(2SI-102M
722**102
703
¦ at: 17200Ac
¦at: »12.0(»/FhAe»
co: ims/iroa
bur tttwrcti Uba< lac.
SOf M. Fairfax a*«.
Lot An|«laa. CA *00M
t*«da t korthruv Ca^nqr
(tartytm fUt
¦ortfe Valaa. FA IMM
Wtloy l«Wor«torl««, lac.
"15 Clactreelc trln
¦ . tprlagflaid. U 22111
Mm Safety Appllaaca Co.
tOO Fann Caetar Blvd.
Flttakurgh. PA 1J2JS
Monitor Laka, lac.
101(0 Scrlppa taaeii Blvd.
•as Blago, CA (2111
hoem taalyttri, lac.
1101 Stata toad
Prlacataa, MJ MJ40
Wit on tojr Ccf mij
flow Control Blv.
10} Ivjlaad la*d
toylMMd, PA 1(174
00: Lira 202
60: Laft datactor
¦(S: Not apaclfjad
CO: >*-222-*
III: 32-2)0
CO: M400Arc XI
CO: $7000/KB a
¦at: cc/rro
CO: M100/CC/TC
¦ts: $*»oo/cc/ftd
T»l«dr*«. lac.
Analytical Inttnawnta Mi.
>11 V. Hi•aloe Drlva
!•> Cakrial. CA (17F*
CO: 9300
¦»S: ill BHCO-20X
CO: (MOO/WII
¦at: (4700/UV
f^ai-ae flactraa Carp.
¦n*iroe**ntal Inatrxa
101 katli (tract
¦ofklatoa. HA C17U
ata M«.
72
-------
APPENDIX C
VENDORS RESPONSE TO LETTER FROM EPA
WBtkTVn
•0 '
uamtt
U tro»u
co «»t rtoroso wrmmPT
eon/ofn*T»c nmctni
lm«i bt.
btinnut IMiUb
AWO Trtcvr Laix
tattt, sx nm
t«Marck m*4
U4.
mi - Mth *«•., n
Cal|«i7 Uu
¦»! Nar (clmtf It
f. 0. k« *M
Mlm, IX 71001
-------
APPENDIX D
FCC EMISSIONS GAS SAMPLE CONDITIONING SYSTEM
The stack gas was conditioned at the stack to remove particulate,
acid nlst and noisture. The gas conditioning system was housed in a
¦etal case 55 cm by 76 cm by 33 cm. The system (Figure D-l) consisted
of the following parts:
1. Probe. SS 202, length 130 an, OD 5.1 cm, ID 4.94 cm,
slots 2 to 4 an wide cut in the 40 cm nearest the tip
and covered with Balston lype 20/80-A microfiber filter
(5.1 cm ID by 45 cm long), glass wool plug in probe tip;
and steel plate welded 5 cm from the end to attach probe
to port flange and probe extension (Figure D2).
2. Probe Extension. Alunlnum pipe, 28 cm long, 10 cm OD,
one end welded to alunlnun plate (to attach extension to
probe) and other end externally threaded to accommodate
pipe cap. Each side contained a piece of alunlnum
tubing 2.5 cm OD by 8 cm long to allow sample to pass
from probe into the gas conditioning system and to allow
a stack sample to be withdrawn Into a Tedlar bag. The
bottom of the extension contained a piece of alunlnum
tubing 1.6 cm OD by 5 cm long for attaching a 60 cm
U-shaped drain. This drain continuously removed
condensate from the probe extension while maintaining a
water seal to prevent ambient air from entering the
system.
3. Balston 20/80-A Filter Housing (SS304) with type 200-80
Grade D and Type 200-35 SS filter support core filter.
Gas passed from inside to outside to coalesce water and
remove small particulates. A U-trap was attached to the
bottom to allow continual draining of the condensate.
4. Balston 97S6 Filter Housing (SS316) with 050-05CH micro-
fiber filters (1.2 an ID by 3.1 cm long). One of these
filters was added to the system December 5, 1980, to
reduce plugging of the Perma Pure dryer; the second was
added January 9« 1981.
5. Perma Pure Model PD-1000-24S (200 tubes, 60 cm long).
Two connected in series to remove moisture. A pressure
regulator was used to maintain dryer purge air at a flow
of 17 L/mln at 2 to 3 psl*
74
-------
6. Pump. ADI Model 19320-T dual-stage (with Carpenter 20
heads to reduce corrosion). Teflon-coated dlaphran and
VIton valve gasket and discs. The single-stage pump
used originally would be adequate for most systems.
7. Tubing. SS 316 between probe extension and Balston 97S6
filter, polypropylene between 97S6 filter and pump.
8. Sample Line. 200 m long, 0.95 on ID by 1.2 cm OD, black
nylon, unheated.
9. Balston 20/80-A Filter Housing with Type 20/80-A Grade
D, nlcroflber filter. Located immediately In front of
the trailer to remove fine particulate.
The overall performance of the gas conditioning system was very
good. Problems and changes that occurred between February 11, 1960
and April 28, 1981, are sunmarized in Table D-1. In general, the
following comments are noted:
1. Polypropylene tubing was found to be superior to
stainless steel tubing. The latter reacted with the
stack gas to yield a fine particulate that collected at
bends and elbows in the conditioning system.
2. The Ealston 20/80-A filter was difficult to disassemble
in the field due to its large diameter and the lack of
large wrenches.
3. The U-shaped drains worked well, but care should be
taken to prevent freezing; either by wrapping the drain
with heating tape or by using heat radiating from the
stack.
U. From February 18, 1981, to April 28, 1981, the system
worked without failure or maintenance at a sampling rate
of 5 to 8 L/mln. When disassembled on April 28, 1981,
the Perma Pure dryer inlet contained no particulate, the
Balston 20/80-A filters were unbroken, the pump inlet
dlaphram contained • reddish syrupllke liquid (pH less
than one) that absorbed water rapidly upon exposure to
the ambient air. An anion analysis showed that the only
anion present was sulfate.
5. The system described was constructed to provide a sample
flow rate of 6 to 15 L/min to the trailer with a minimum
system response time. For most applications, a flow
rate of 1 L/min would be sufficient; thus one PD-750-24
inch (100-tube) Perma Pure dryer might be adequate for
75
-------
¦ost applications. This would result in a considerable
savings. In addition, the probe extension and the
Balston 20/80-A filter in the sample case could be
replaced with smaller units without adversely affecting
the sample conditioning system. Probe filter integrity
was maintained as long as the slotted side of the probe
was facing downstream. When the probe was rotated 90
so that only half the slotted side of the probe was
facing downstream, a hole about one cm in diameter
formed in the filter at a point about 35 cm from the
tip. A glass wool plug in the probe also helped keep
the probe clear.
76
-------
SAMPLE WOK IN-8TACK FILTER
BAL8TON TYPE WRITER
2 STAGE "ADr
OIAPHRAM PUMP
PURMA-PURE
GAS DRYERS
WATER TRAP ft DRAIN
FIGURE D-l. SCHEMATIC DRAWING OF SAMPLE CONDITIONING SYSTEM.
-------
FIGURE D-2. PROBE, PROBE EXTENSION AND BALSTON 20/80-A FILTER.
78
-------
TABLE D-l. MAINTENANCE/REPAIR OF STACK GAS CONDITIONING ST STEM OF FCCU
A
*
A
•
•
m
S
to*
k
»*
V
•
•
c
I
i
«
I
•
A
0
1
m
&
•
4.
a
•*
•
•«
s.
9
3
«
«
c
Ik
-s
I
*
4
8
«k
3
C
*
\
m
*
I
«
8
1
o
t
1*5
Ktmrr-.
t/u
•
•
11
lintw. >i »*< «t*tt«4 il* (mntrm trftk »1aaa aaal »1af la tip. Hit t$-IH Ml«(. Ma* (I^M M
•CI-MJIO' vllh fat rKltlXlln lyllM It wltl flair.
tm
1
•
«
IT
Ho nilrWH by fl* fray pe»*r it w* lulrt. Inlm< tt tit Ml«| Wtnw fft trf pap wit ft fatyvraaytaaa. Qi
•t aatlat »*4 1* itKt n m ky UmI na»ltar r«*4 M iw. SfilM InI-(IikM at 1 Inch Nf. hiHr
ft: HI (M); kilw S04, pi) J.J, Cwtmiti $'»**.
vt%
1
•
1
•
II
tiaaul mm mtnttrtlw lyttn kacawaa nlw M l-tn* 1» U nltM iwlmHi. Irali taa IwH^ at in
a«lat*4 f«b* to* «M iiwiiiI «>•«• ml »'¦» ti im «((Kt aa H«tltaliiii»i at trail** to KtaaNHi Maatla a»lto»
rtailnamti.
trm
t
It
•
M
IMH Ml«| iwtrlctlal (la Into «<¦». tiatto< Ml flltor. Ml M aaall H*tlt*l«to fcailt. tat «M
nt kcIiiM.
%n
II
l»
1/
Hut CO «»7 kl|k. Pat la I t lllitla ay*l«a to krlaf CO lata apaa ranyt af aaalto*. Ollatlaa ayataa raaa< koat
1*0.
rnt
M
IJ
II
II
It
Nt italaa paaka nimln aailiH • train auto prata.
SytlM 0*. ImtalM tOj aaaltar 1* trallar.
vn
•
i»
¦
If
h»lm4 iaU»i< M
-------
TABLE D-I. MAINTENANCE/REPAIR OF STACK GAS COKDITIOfllffC ST STEM Of PCCU
WW
ftetMtr
i»/n
1
i
a
*** l*Ht «**» **— 4mmH. AmIjmIi *m*: 11:1.1 fm ?**:&), m*m HiaiWi, tiRJl^ «•*#*!.
****«?*£ tanpi* flaw *iW air * *
11m
9
-
S
a
a
**a Hlt.r M *)« 1* ft. I»tani. m »artl«tfy »(¦**. «W>Mi t W NM-ff>$S (11 ¦ m Uk»f) «ttft mm
W »»•« «»l tm c 1M) totot). pm* )m4i (JlHSj vttft Uttwiff » toa*. Wl (*•« Mn aa latx r*9 fl«N
an
»
-
a
a
Mi* >• pm» ri«ni tm fiitar tatwaa r-a»-i m*m t» f+n »**¦*«¦ •* **•
Htl
!/*
*
*
a
V-#at« M *«to tmtaMlan twMlnN f»»«« twKmel*. tralnatf. (ivlaratf «m fitter. •1nnirtly rmm * 1-10-tl itlnHnf wdMrtt to #111 tfttm m»1Im IIh. (Mar
f«*i# »l nl^il M «•"»< *cl«| <4 1*01, ArHnf <>r wn ural. fef1at*4 l-»-| «ni »•»•* fllltrt. t*M-l
wl, mm with inn «»1«a. »-00-I e«vtr*4 »lth m »a»4tr. tm fit tor tmiH »)Oi Iwdr. n»taca«
flttar
tm
•
*
a
~«*a fin*r M tola, n#1ac*4 an* ratatrt paka m «lattai il* fatlof Immmtnm. M flnt «aal la fata tip. Ml
Men a» »n»a flittf an nit m« ar frata, *• (fiat 11 twcM tto »rata •¦(Ml** t» ta »ra»l«» iMItlnit
muring.
tm
•
Nt (at* wnr apt* to «WI»# ttotl'c *a««Bt M to yrtiam frMtla* *f M tartti* top* m prata
aitwntan WW rt W >»l, tato to wUw*. )« Mm Km
•lt» a t* WDfr/W W. M t«« fttt flltrrt In Nrln. M tolltan 49 C flltor Awnatraaa tf pap an* hllwH
¦tik tMtiwy nln to mini fla» ta • tpa.
VN
•
f
IftUm rtllt fanrtlaalat. frato «lw. 01, ttoct ta, CI, aMWi UL
Ml
i
Syttaa Hilt fwtUanlf*. C*, tlart IK, aaMftW IS. to—ill (Tit*.
-------
APPENDIX E
MANUAL METHOD FOR MEASURING
CARBON MONOXIDE IN REFINERY CASES
1. PRINCIPLE AND APPLICABILITY
1.1 Principle
An Integrated sample Is extracted from the gas stream, passed
through hydrogen peroxide and alkaline permanganate solutions and
collected In a Tedlar bag. The carbon monoxide (CO) concentration In
the sample Is measured spectrophotometrlcally using the reaction of CO
with a palladium chloride/leuco crystal violet/potassium lodate
solution.
1.2 Applicability
This method is applicable for determining 00 emissions from
stationary sources.
2. RANGE AND SENSITIVITY
2.1 Range. As written, the method applies to gas samples that
contain 20 to 1000 ppa CO. Samples containing In excess of 1000 ppra
CO can be analyzed by reducing either the gas volune or the shaking
time. Samples containing less than 20 ppo can be analyzed by
increasing the volume of sample reacted with the palladium
chloride/leuco crystal violet/potassium lodate solution or by
Increasing the shaking time or frequency. However, if such changes
are made in the procedure, the linearity of the absorption curve must
be checked under these conditions.
2.2 Sensitivity. Sensitivity depends on shaking time, shaking
frequency, gas vol me and shape of reaction vessel and cannot be
specified absolutely. As written, the sensitivity of the method Is
approximately 10 ppo CO.
2.3 Interfering Agents. Sulfur oxides, nitric oxide, and other acid
gases which Interfere with the reaction, are removed from the gas
sample during sample collection. These gases are removed by passing
the Sctnpled gas through a 4 percent potassium permangate/5-percent
sodium hydroxide solution. Carbon dioxide does not interfere with the
reaction, but, because it is removed by the scrubbing solution, its
concentration must be measured and an appropriate volune correction
¦ade.
81
-------
4. PRECISION, ACCURACY AND STABILITY
4.1 Precision. The estimated intralaboratory standard deviation of
the method is 2.5 percent of the Bean for gas samples analyzed in
triplicate. This estimate, which applies to the concentration range
20 to 1000 ppm, was determined from 22 samples oollected at a
petroleum refinery. The Interlabortory precision has not been
established.
4.2 Accuracy, Cta the average, the manual method results were biased
4 percent high for 22 samples analyzed by an NDIR calibrated with NBS
standards. The manual method was biased 5 percent high when used to
analyze certified calibration gas mixtures that contained SO^, MO,
C02% >nd 100 to 500 ppa CO In nitrogen.
4.3 Stability. The individual components of the colorimetric reagent
are stable for at least one month, but the colorimetric reagent must
be used within 3 hours after preparation to avoid excessive blank
correction. For optimum accuracy the samples must be reacted and
analyzed no later than 3 hours after the colorimetric reagent has been
prepared.
5. APPARATUS
5.1 Sampling (Figure E-1)
5.1.1 Probe. Stainless steel, sheathed Pyrex glass or equiva-
lent, equipped with a glass wool plug to remove particulate matter.
5.1.2 Impinger. Three Greenburg-Snlth impingers connected in
series with leak-free connections.
5.1.3 Pump. Leak-free pump, Metal Bellows Model 110 or equiva-
lent with stainless steel and Teflon parts to yield a flow rate of 0.2
to 0.4 L/min.
5.1.4 Surge Tank. Installed between the pump and the rate meter
to eliminate the pulsation effect of the pimp on the rate meter.
5.1.5 Rate Meter. Calibrated rotameter, or equivalent, to
measure flow rates between 0 to 0.4 L/mln.
5.1.6 Flexible Bag. Tedlar, or equivalent, with a capacity of 10
liters. Bag must be leak-free.
5.1.7 Valve. Needle valve, or equivalent, to adjust flow rate.
5.1.8 Fyrite Analyzer, or equivalent, to measure CO^ concentra-
tion to within 4-0.5 percent accuracy.
82
-------
5.2 Analytical
5.2.1 Spectrophotometer. Single or double beam to measure
•bsorbance at 569 no. Slit width should not exceed 20 nm.
5.2.2 Vacuum Gauge. U-tube nanometer, 1 meter (36 In.), with
1-nm divisions, or other gauge capable of measuring pressure to within
x +2.5 nm Hg (0.10) in. Hg).
5.2.3 Pump. Capable of evacuating the gas reaction bulb to a
pressure equal to or less than 75 nm Hg (3 in. Hg) absolute, equipped
with coarse and fine flow control valves.
5.2.4 Barometer. Mercury, aneroid, or other barometer capable of
measuring atmospheric pressure to within 2.5 on Hg (0.1 in Hg). In
many cases, the barometric reading may be obtained from a nearby
national weather service station, in which case the station value
(which is the absolute barometric pressure) must be requested. An
adjustment for elevation differences between the weather station and
sampling point must then be made at a rate of mlnlus 2.5 nm Hg (0.1
in. Hg) per 30 m (100 ft) elevation increase, or vice versa for
elevation decrease.
5.2.5 Reaction Bulbs. Pyrex glass, 100-125 mL with Teflon
stopcock, leak-free at 650 an Hg. Designed so that 10 mL of the
colorimetrlc reagent can be added and removed easily and accurately
(Figure E-2). Conmerclally available gas sample bulbs such as Supelco
#2-2161 and Alltech #7012 can also be used.
5.2.6 Volumetric Pipettes. Class A, 4 mL and 10 mL and 1 mL
graduated pipette.
5.2.7 Volumetric Flasks. 100 mL
• 5.2.8 Graduate Cylinder. 1000 aL.
5.2.9 Shaker Table. Reciprocatlng-stroke type such as Eberback
Corp. Model 6015. A rocking arm or rotary-motion type shaker may also
be used. The shaker must be large enough to accommodate at least six
gas sample bulbs simultaneously. It may be necessary to construct a
table top extrusion for most commercial shakers to provide sufficient
space for six bulbs (Figure E-3).
5.2.10 Spectrophotometer cells. 1 cm pathlength.
83
-------
6.0 REAGENTS
6.1 Sampling
6.1.1 Alkaline Permanganate Solution. Prepare by dissolving 40
grams of ACS reagent grade sodiun hydroxide and 50 grams of ACS
reagent grade potassium permanganate In 1 liter of distilled water.
This is sufficient for removing HO and SO- from 50 liters of gas
containing 15J CO^.
6.2 Analysis
6.2.1 Stock Solutions
6.2.1.1 Potassium lodate. ACS reagent grade, or equivalent.
6.2.1.2 SodluB Chloride. ACS reagent grade, or equivalent.
6.2.1.3 Palladium Chloride. ACS reagent grade, or
equivalent.
6.2.1.4 Sod1ud Monohydrogen Phosphate Keptahydrate
(Na^HPO^.TH^O). ACS reagent grade, or equivalent.
6.2.1.5 Leuco Crystal Violet CH,4',U" methylidynetris (N,N-
dlmethylanillne)). Eastman Kodak Company Stock No. 33651.
6.2.1.6 Phosphoric Acid (851). ACS reagent grade, or
equivalent.
6.2.2 Working Solutions
6.2.2.1 Sodium Monohydrogen Phosphate (0.1M). Dissolve
2.68 grams In 100 bL of distilled, deionized water. This solution is
stable Indefinitely.
6.2.2.2 Sodium Tetrachloropalladate(II) Solution (0.005M).
Dissolve 0.0887 grams of palladiun chloride and 0.0595 grams of sodium
chloride In 50 *>L distilled, deionized water and dilute to 100 ml.
This solution Is stable for at least one month. If a brown preclpate
forms It can be dissolved by adding a few crystals of sodium chloride.
6.2.2.3 Leuco Crystal Violet Solution. Dissolve 0.0256
grams leuco crystal violet in 80 »L water containing 0.25
-------
€.2.2.5 Colorimetric Solution. Pipet 4.0 nL each of potas-
sium lodate solution (6.2.2.4), leuco crystal violet solution
(6.2.2.3) and sodium tetrachloropalladate (II) solution (6.2.2.2) Into
a 100 mL voluoetric flask. Pipet 0.6 mL sodium monohydrogen phospate
solution (6.2.2.1) Into the flask and dilute to voluoe. This solution
should be used within 3 hours of preparation to minimize the contribu-
tion of the blank and the sample absorbance. This is sufficient
vol use to analyze three stack gas samples in triplicate.
6.2.2.6 Standard Gas Mixtures. Traceable to NBS standards
and containing between 100 and 1000 ppm CO in nitrogen. The calibra-
tion gases shall be certified by the manufacturer to be within ~ 2
percent of the specified concentration.
7.0 PROCEDURE
7.1 Sampling. Evacuate the Tedlar bag completely using a vacuum
pimp. Assemble the apparatus as shown in Figure E-1. Loosely pack
glass wool in the tip of the probe. Place 400 mL alkaline
permanganate solution (6.2.1) in the first two implngers and 250 mL in
the third. Connect the pump to the third impinger and follow this
with the surge tank and the rate meter. Do not connect the Tedlar bag
to the system at this time.
Leak-check the sampling system by plugging the probe Inlet and
observing the rate meter for flow. If flow is Indicated on the rate
meter, do not proceed further until the leak Is found and corrected.
Insert the probe into the stack and draw sample through the system at
300 ~ 50 mL/min for 5 minutes. Connect the evacuated Tedlar bag to
the system, record the time and sample for 30 minutes, or until the
Tedlar bag Is nearly full. Record the sampling time, the barometric
pressure and the ambient temperature. Purge the system as above
Immediately before each sample.
The sampling system above Is adequate for removing sulfur and
nitrogen oxides from 50 liters of stack gas when the concentration of
each is less than 1,000 ppm and the COp concentration is less than
15%. The samples In the Tedlar bag should be stable for at least one
month, if the bag Is leak-free.
7.2 Ancillary Methods
Measure the CO- content in the stack to the nearest 0.5X each
time a CO bag sample is collected. A grab sample analyzed by the
Fyrlte analyzer Is acceptable.
85
-------
7.3 Analysis
Assemble the system shown In Figure E-4 and record the Informa-
tion required in Table E-1 as it is obtained. Pipet 10.0 aL of the
colorimetrie reagent (6.2.2.5) In each gas reaction bulb (5.2.5) and
attach the bulbs to the system. Open the stopcocks on the gas bulbs
but leave the valve on the Tedlar bag closed. Turn on the pump, fully
open the coarse-adjust flow valve and slowly open the fine-adjust
valve until the vacuum is at least 550 no Hg. Now close the coarse
adjust valve and observe the Manometer to be certain that the system
is leak-free. Vait a nlnimuro of two minutes and if the pressure has
decreased less than 1 mn, proceed as described below. If a leak is
present, find and correct it before proceeding further.
Fecord the vacuum pressure to the nearest 0.1 mn Kg and close off
the gas bulb stopcock. Open the Tedlar bag valve and allow the system
to cocne to atmospheric pressure. Close the bag valve, open the pimp
coarse adjust valve and evacuate the system again. Repeat this
fill/excavation procedure at least twice and then close off the pimp
coarse adjust valve, open the Tedlar bag valve and let the system fill
to atmospheric pressure. Open the stopcocks on the gas bulbs and let
the entire system come to atmospheric pressure. Close the gas bulb
stopcocks, remove the bulbs and place them on the shaker table with
their main axis either parallel to or perpendicular to the plane of
the table top.
Record the room temperature and the barometric pressure (nearest
0.1 ion Hg) after each set of gas bulbs is filled. At least one set of
bulbs from a Tedlar bag containing a known concentration of CO in
nitrogen must be used each time a set of samples is shaken. Improved
accuracy will be obtained if two standards are Included each time.
Also, to avoid cross contamination of samples, the bulb filling system
must be purged with ambient air for several minutes between samples.
Shake the samples for 25 minutes if the expected concentration is
less than 600 pprn and for 20 minutes if it is between 600 and 1,000
ppm. Place the contents of each bulb in a labeled test tube or other
suitable vessel.
Measure the absorbance of each sample at 589 nm using water as
the reference; also measure the absorbance of the unreacted
colorimetrie reagent used for that set of samples to serve as a
reagent blank. Reject the analysis if the blank absorbance is greater
than 0.1.
The absorbance curve is linear to an absorbance of 1.8. If the
sample absorbance exceeds this, the sample can be diluted with the
colorimetrie reagent.
86
-------
The reaction between 00 and the colorlnetrlc solution is slow.
For example, unshaken samples set aside for one hour after filling
show no significant absorbance.
8.0 CALIBRATION AND QUALITY ASSURANCE
8.1 Gas Bulb Calibration
Weigh the empty bulb to the nearest 0.1 gram. Fill the bulb with
distilled water and again weigh to the nearest 0.1 gram. Subtract the
tare weight and calculate the volune in liters to three significant
figures using the density of water (at the measurement temperature).
Record the volune on the bulb or, alternately, etch an identification
number on the bulb and record the volune in a notebook.
8.2 Rate Meter Calibration
Assemble the system as shown in Figure E-1 (the Implngers may be
removed) and attach a volume meter to the probe Inlet. Set the rota-
meter at 300 mL/min, record the volune on the volune meter, start the
pump and pull gas through the system for 10 minutes. Record the final
volume on the volixne meter and determine the volune of gas that passed
through the system. Repeat this procedure for at least two other flow
rates between 100 and 500 mL/min and prepare a calibration curve for
the rate meter.
8.3 Spectrophotometer Calibration
The calibration curve is established for each batch of samples
shaken at the same time by running standards along with the stack
samples. For highest accuracy two sets of standards should be run,
but acceptable results have been obtained when only one set of
standards was used per batch of samples.
8.4 Quality Assurance
8.4.1 System Shaking Time
Before field samples are analyzed, the linearity of the
absorption curve as a function of shaking time must be established for
the analytical system. This can be accomplished as follows: Fill a
Tedlar bag with a CO concentration at least as high as the highest
level expected to be encountered In the field samples. Fill at least
six gas bulbs with gas from the bag as described in Section 7.3 and
set them on a shaker. After shaking has been Initiated remove two
bulbs at 15 minutes, two at 30 minutes and two at 60 minutes and
measure the absorbance of each solution. Calculate the absorbance per
87
-------
vol we of gas (SA) according to Equation E-1 (Section 9.2) for each
•ample and deteraine the average SA for all samples with Identical
shaking tines. Plot this sverage SA versus the shaking time to
determine the linearity of the absorption curve. Use this data to
deteraine the shaking time and sample volune required for sample
analysis.
6.4.2 Sample Bag-Leak Checks
Vhlle a bag-leak check Is required subsequent to bag use, it
should also be done before the bag is used for sample collection. The
bag should be leak-checked in the Inflated and deflated condition
according to the following procedures. Connect the bag to a water
nanometer and pressurize the bag to 5 to 10 cm HJ) (2 to 4 inches
H^O). Allow the bag to stand for 60 minutes. Any (Hsplacement in the
water nanometer indicates a leak. Now evacuate the bag with a
leakless pixsp that Is connected on the downstrearo side to a flow
indicating device such as a 0 to 100 nL/min rotameter or an impinger
containing water. When the bag is completely evacuated, no flow
should be evident if the bag is leakless.
9.0 CALCULATIONS
9.1 Abbreviations and Symbols
SA c Sample absorbance per volune of gas in bulb
COB * Concentration of CO in Tedlar bag (ppro, dry basis)
CO . * Concentration of CO In stack (ppra, dry basis)
FCO_aC s Volume fraction of CO,, in stack
By = Moisture content of gas in bag
9.2 Calculation of CO
9.2.1 Calculation of SA
Calculate the sample absorbance per volune of gas (SA) for each
gas bulb (Equation E-1) and record the value In Table E-1. Calculate
the average Sh for each bag sample and compare the three values to the
average. If any single value differs by more than 10 percent from the
average, reject this value and calculate a new average using the two
remaining values.
(E-1)
SA • I Sample Absorbance Corrected for Blank! j Barometric Pressure 1
I Bulb Volume - Volune Keagent J I Sample Pressure In Bulb J
88.
-------
9.2.2 Calculation of 00 Concentration in Bag
9.2.2.1 Single Standard Shaken with Samples
Calculate the CO concentration In the bag using Equations E-2
and E-3* If condensate Is visible in the Tedlar bag, calculate B
using Table E-1 and the temperature and barometric pressure in the
room where the analysis was done. If condensate Is not visible,
calculate Bw using the temperature and barometric pressure at the
sampling site.
2 „ Vapor pressure of water In bag
w Barometric pressure
_|ppm CO In stdj [Average SA sample /„
Bag T ]-Bu j I Average SA .td J
-------
TABLE E-1. MOISTURE CORRECTION
TEMPERATURE VAPOR PRESUR
( C) H^O (nn Hg)
4
6.1
6
7.0
8
8.0
10
9.2
12
10.5
1H
12.0
16
13.6
18
15.5
20
17.5
22
19.8
24
22.4
26
25.2
28
28.3
30
31.8
90
-------
RATE METER
PROSE
SlASsVrOOl
FILTER
AMBIENT
TEMPERATURE
BATH
SURGE TANK
IMPINGERS
RIOIO
CONTAINER
FIGURE E-l. CO SAMPLING TRAIN
-------
(DIMENSIONS IN cm)
II
1J
II
FIGURE E-2. CAS REACTION BULB (0.1 liter)
• 92
-------
FIGURE E-3. ADAPTOR FOR HOLDING GAS BULBS ON SHAKER TABLE
93
-------
MANOMETER
ei
EIRE ADJUST
VALVE
COARSE ADJUST
fVALVE
IAS SHUT-
OFF VALVE
VACUUM
PUMP
T1DLAR
SAMPLEIAO
FIGURE E-4. GAS BULB FILLING SYSTEM
-------
TABLE E=2« DATA REC08B1WG SlffiET FtM SAMPLES AKAXtfZSS W TRIPLICATE
0SKH.B
no./
rrro
WHO
TEK?.
<*ci
owsea
8
TO
MM
CO.
own
TO.
a»
753.. 0BSCB3?
tRBVU
PflOTtAt
poesswe c?
CAS IB mo
to na)
EflraKTrQie
PRESSUBE
Cca BrJ
ocsasra
TIK2
(oln)
G>o
raasms
bateh
OAKTtQ
flbo
©Banwrres
rea gim
CCSZV2
Aba/L
-------
APPENDIX F
INSTRUMENT EVALUATION HISTORY
1/4/80
1/22/80
2/2/80
2/4/80
3/10/80
3/31/80
4/2/80
4/3/80
4/10/80
4/16-4/17/80
4/21/80
4/25/80
6/9/80
7/3/80
9/18/80
11/10-11/12/80
ECOLYZER 3107
10/18/79
4/16-4/17/80
4/21/80
4/27/80
4/28/80
APPLIED AUTOMATION OPTICHROM 102
Instrument received with no Instruction
manual.
Instruction manual received, started wiring
instrunent.
Power supply failed.
New power supply failed.
Incorrect output found (overloaded). Power
supply failed.
Serviceman from Applied Automation arrived to
repair instrument.
Instrument repaired, power supply, output
board and temperature controller replaced.
Output is negative.
Output problem fixed by reversing wiring to
output board and by moving a resistor on it to
another position.
Transported to refinery in trailer.
Instrument turned back on.
Fuse blew in heater circuit.
Zero response varying badly.
Negative peak before CO peak on chranatogram
detected. Chromatograms sent to manufacturer
as aid in troubleshooting.
Condensation in FID caused corrosion to Inter-
nal parts. Monitor shut off until manufac-
turer's service representative call.
Applied Automation serviceman tried to repair
monitor - was unsuccessful. Monitor shut
down.
Instrunent received. Checked out.
bleras found in laboratory checkout.
Transported to refinery in trailer.
Instrument turned on.
Model 2949 filter failed. Replaced.
Model 2949 filter failed. Replaced.
No pro-
96
-------
5/3/80
5/80
9/18/81
1/17/81
2/17/81
ANARAD 501-R
10/2/79
4/16-4/17/80
4/21/80
MSA LIRA 202
11/2/80
1/14/80
2/27/80
3/3/80
3/4/80
3/11/80
3/10-4/3/80
4/15/80
4/16-4/17/80
4/21/80
4/21-5/1/80
Model 2949 filter failed. Replaced.
Large carbon filter added to monitor sample
line (replaces Model 2949 filter) to extend
operation time.
Electrochemical cell going bad. Span control
was unable to span monitor.
Mew electrochemical cell Installed, nonltor
zeroed and spanned.
Model 2949 filter failure caused
electrochemical cell to fall. Monitor removed
from evaluation.
Instrunent received and checked out in labora-
tory without problems.
Transported to refinery in trailer.
Instrunent started up. No problems or outages
occurred before testing was completed In
April 1981.
Instrunent not operational as received.
Put into operation.
Apparent failure of output board. Output had
worked previously (on PD-2064) but output now
less than 4 mA. Called MSA. New output board
ordered.
New output board arrived, installed. Output
still not correct. MSA suggested checking
output by voltage drop across a resistor.
Output seemed to be okay by this method, but
power supply board failed soon after test.
Ordered new power supply. MSA suggested that
output problem could be solved by installing a
matching circuit between the MSA instrunent
and the data logger.
Power supply board arrived.
Electronics expert, R. Burdine, made several
visits to ascertain output problem and built a
circuit to correct problem. His conclusion
was that such a circuit was not needed, but
that output board t/as bad (again). New board
ordered.
New board arrived, not tested (trailer packed
for moving).
Transported to refinery in trailer.
Instrunent started up. Output still bad.
Instrunent off. MSA contacted from Auburn.
97
-------
5/1/80
5/2/80
5/3-5/20
6/9/80
6/9-6/16/80
9/18/80
D. Tlskiewlc of MSA ene to check problem.
According to his meter, output was 4-20 mA.
Mr. Tlskiewlc said that Batching circuit was
needed. MSA-Pittsburgh would find and send
one.
Instrwent off. No work on matching circuit.
Matching circuit received and Installed by
MSA serviceman.
Lab evaluation completed in field.
Span and zero adjustment not possible. Com-
plete optical alignment restored performance.
Instrument continued to operate normally
throughout the remaining part of the evalua-
tion.
TELEDYNE 611 DMC0-20X
2/80
3/13/80
1/16-4/17/80
1/21/80
4/29/80
5/1 /80
5/7/80
9/18/80
11/2/80
Instrument received with no instructions on
how to Install or use molecular sieve canis-
ters.
Initial setup and adjustment. Dimensional
instability noted: very sensitive to even the
slightest change of attitude with respect to
horizontal - even microscopic changes signi-
ficant.
Transported to refinery in trailer.
Instrunent turned on and Interference in fuel
gas noted.
Molecular sieve Installed on sample line. No
effect on sample concentration. Impossible to
zero and span instrument.
Molecular sieve removed.
2-hour drift RA Test spoiled by very high
sample concentration. Instrument readout did
not change before, during or after period of
high concentration.
Monitor shut off due to lack of reliable data
caused by Interference.
Molecular sieve received from Teledyne. When
Installed, no startup was possible because of
optical system failure.
BENDIX 7770
1/3/80
2/14/80
Instrument received.
Connections to gas services made, Instrunent
started up. Several tube fittings inside
leaked and had to be tightened. Initial
adjustment made (oven temp, flows, etc.).
.98
-------
3/11/80
3/21/80
4/1/80
4/16-4/17/80
4/21/80
5/7/80
11/5/80
12/15/80
2/25/81
Correct
success-
Internal plumbing modified to accept separate
sources for valve and heater air. When
Instrument was restarted, oven heater was
Inadvertently shorted causing destruction of
TRIAC in heater control circuit. New TRIACs
ordered.
Parts arrived but proved to be wrong,
parts ordered.
Correct parts received, installed
fully. Instrument restarted.
Transported to refinery in trailer.
Instrument started back up.
2-hour drift RA Test spoiled due to high HpS.
Instruoent system recovered from high H^S in
about 4 hours.
Monitor failure from 2 weeks previous caused
by diaphragm on sample valve. New part
ordered.
Diaphragm replaced. Monitor operating pro-
perly.
Sample valve diaphragm ruptured and repaired.
Instrument operated properly throughout the
remaining field evaluation.
DEL MAR SCIENTIFIC JJH-V
2/10/80
4/16-4/17/80
4/21/80
5/7/80
Monitor received and set-up.
Transported to refinery in trailer.
Instrument turned on. Pressure/flow problems
noted. Sensitivity critically dependent on
flow (which varies with pressure and gas
viscosity).
Very high H?S concentration encountered in
fuel gas at 10:00 am. Analyzer did not re-
cover until next day. Instrument operated
properly throughout study, but gas dilution
system was not- appropriate for fuel gas
dilution.
HOUSTON ATLAS 825/102R
12/14/79
1/20/80
4/16-4/17/80
4/21/80
4/23/80
Instrument received.
Instrument setup.
Transported to refinery in trailer.
Instrument turned on. Output overloads data
logger. Tape does not advance. Ordered new
output/cycle time board.
New board arrived. No effect. Timer relay
ordered.
99
-------
4/28/80
5/1/80
5/5/80
5/8/80
5/8/80
9/11/80
9/18/80
11/2/80
1/6/81
3/31/81
Relay received. No effect.
Abraham Asperlc of HAI arrived to trouble-
shoot. Did not find problem. Took analyzer
back.
Asperlc informed us of wiring mistake to data
logger. Sent lnstriment back to us.
Instrunent arrived with wiring Instructions.
Instrument turned on - works.
Monitor sent to manufacturer for repair.
Monitor returned by manufacturer.
Houston Atlas serviceman found cause of drift
to be condensation in vent line.
Dilution system breakdown. Circuit board in-
stalled backwards by contractor personnel.
New part delivered.
Monitor failure due to corrosion in timer.
PROCESS ANALYZERS INCORPORATED 32-230
12/20/79
1/80-3/80
5/9/80
6/9/80
8/28/80
9/18/80
10/1/80
11/5/80
Instrument received.
Instrument not operational - many circuit
cards exchanged with manufacturer. Instrunent
eventually returned for repair.
Monitor returned from manufacturer allowed to
wannup until next site visit.
Monitor broken down during 1-oonth warmup.
Returned to manufacturer.
Monitor returned to manufacturer.
Monitor returned to field test, 4 to 20 nA
output not functioning.
Monitor stopped operating.
Monitor failure caused by wearout of 10 port
valve. Monitor shut down and removed from
further testing.
100
-------
yreoxuxf irtiirai iiLwi ink rrrurw mrjorr sompittiMyi
1. OEPOOT WO.
a.
a. RECIPIENT'S ACCESSION NO.
a. TITLE AND OuOTITLtE
' FIELD EVALUATION OF CARBON EPOXIDE AND HYDROGEN
SULFIDE CONTINUOUS EMISSION EDITORS AT AN OIL
REFINERY
p. REPORT ©A"?G
3. PERFORMING ORGANISATION CODE
7. aut«orisj
R. IB. tester and 3. 3. Ferguson (Haraon)
W. J. Mitchell (EPA)
p. PERFORMING ORGANISATION REPORT NO.
3. P&RFORMIWG ORGANISATION WAMt AND A0DREDS
Haraon Engineering and Testing
3oa 2207
Auburn industrial Park
Auburn„ AL §5810
16. PROGRAM ELEMENT CaO.
»». gonymacy/GRany evo.
(S8-02-S405
13. SPONSORING AGENCY NAME AND ADDRESS Il3. TYPE Of REPORT AND PERIOD COVERED
Environmental Monitoring Systems Laboratory j Final
Office of Research and Development jia. sponsoring agency code
U.S. Environmental Protection Agency j
Research Triangle Park0 NC 27711 J EPA 600/08
13 SUPPLEMENTARY WOTdS
To be published as on ORD Project Report
18. ADSTRACT
An eleven month field evaluation was done on five hydrogen sulfide and four
carbon monoxide aonitors located at an oil refinery. The hydrogen sulfide aonitors
sampled a fuel gas feed line and the carbon monoxide aonitors sampled the emissions
from a fluid cat cracker (FCC). Two of the four carbon monoxide monitors operated
over the eleven aonth period and showed good agreement with the leuco crystal violet
(LCV) wet chemical method developed for the purpose of checking monitor accuracy.
The LCV method and the special stack gas conditioning system employed to remove mois-
ture and particulate from the FCC stack gas are also described. The gas conditioning
system operated for 14 months without a major failure. None of the five hydrogen sul-
fide aonitors was found acceptable. Two of the five never obtained a valid sample and
the other three did not agree well with the EPA Reference Method 11 during relative
accuracy testing.
17. KIEV WORDS AMD DOCUMENT ANALVBIS
o. OESCRlPTOOS
b. IDE NT If IE RS/OPEN ENDED TERMS
c. COSATi Field/Croup
Stack gas conditioning system
Carbon monoxide aonitors
Hydrogen sulfide aonitors
Fuel gas feed line
Carbon monoxide wet chemical oethod
Leuco crystal violet
EPA Method 111
vrr
•
10 DlSTRIDUTION STATEMENT
RELEASE TO PUBLIC
18. SECURITY CLASS (Thii Report)
UNCLASSIFIED
81. WO. Of PAGES
SO SECURITY CLASS (ThitpaQel
UNCLASSIFIED
33. PRICE
EPA P»o 3220-1 (Qo
------- |