EPA/600/R-97/024
March 1997
TESTING THE PERFORMANCE OF REAL-TIME
INCINERATOR EMISSION MONITORS
By
S. B. Ghorishi, W. E. Whitworth, Jr., C. G. Goldman, and L. R. Waterland
Acurex Environmental Corporation
Incineration Research Facility
Jefferson, Arkansas 72079
EPA Contract 68-C4-0044
Work Assignments 0-4 and 1-1
Project Officer: R. C. Thurnau
Work Assignment Manager: M. K. Richards
National Risk Management Research Laboratory
Cincinnati, Ohio 45268
NATIONAL RISK MANAGEMENT RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before com/
lflEPOHT#A/600/R-97/024
2.
I
4. TITLE ANDSUBTITLE
Testing the Performance of Real-Time Incinerator
Emission Monitors
5 wdDmi
6. PERFORMING ORGANIZATION CODE
7. authobisi
Ghorishi, Vlhitworth, Goldman, and Waterland
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Acurex Environmental Corporation
10. PROGRAM ELEMENT NO.
Incineration Research Facility
Jefferson, Arkansas 72079
11. CONTRACT/GRANT NO.
68-C4-0044
12. SPONSORING AGENCY NAME AND ADDRESS
National Risk Management Research Laboratory
13. TYPE OF REPORT AND PERIOD COVERED
Project Report
Office of Research and Development - USEPA
26 W. Martin Luther King Drive
Cincinnati - OH 45?6fi
14. SPONSORING AGENCY CODE
EPA/600/14
15. SUPPLEMENTARY NOTES
Marta K. Richards (513) 569-7692
is. abstract jn a rcccntly completed test program at the U.S. EPA Incineration Research Facility (IRF), 10 prototype or
developing continuous emission monitors (CEMs) for measuring trace metal or trace organic species concentrations were tested.
Of the 10 CEMs tested, four measured incinerator flue gas concentrations of several specific volatile organic compounds
(VOCs), one measured total parlicuiate-bound polynuclcar aromatic hydrocarbon (PAH) concentrations, two measured flue
gns concentrations of several (up to 14) trace metals, and three measured mercury concentrations. While the testing consistec
of obtaining quantitative measurement data on the four measures of CEM performance checked in a relative accuracy test audit
(RATA) as described in 40 CFR 60 Appendix F — relative accuracy (RA), calibration drift (CD), zero drill (ZD), and response
time the primary project objective focused on the RA measurement. The RA measurement was achieved by comparing the
monitored analvte concentration reported by the CEM to the concentration determined by the EPA Reference Method (RM;
for the analvte. Four series of tests were performed, each simultaneously testing up to three monitors measuring the same oi
similar analyte type. Each test series consisted of performing triplicate RM measurements at each of three target flue gas
monitored analyle concentrations while the tested CEMs were in operation. Thus, each scries gave nine RM, CEM comparisons
in total. All measurements were taken in the wet scrubber exit Hue gas from the pilot-scale rotary kiln incineration systen*
(RKS) at the IRF. The test program was performed in August and September 1995.
The test program results clearly showed the prototype nature of most approaches tested, and the clear need foi
further development. Mercury CEMs will require the least development effort to reach commercial status and are nearly there
However, the approaches tested for multi-metals and VOC determinations will likely require further development. Given this
need, the importance of continuing test programs of the scope and scale of this one cannot be overemphasized.
1 7.
KEY WOROS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b. IDENTIFIERS/OPEN ENOED TERMS
c. COSATI Field/Group
Monitors
Emissions
Measurement
Flue Gas
Combustion
Organic Compounds
Meta1s
18. DISTRIBUTION STATEMENT
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES . or
195
20. SECURITY CLASS iThis page)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EOITION u obsolete
-------
NOTICE
The U.S. Environmental Protection Agency through its Office of Research and
Development partially funded the research described here under Contract No. 68-C4-0044 to
Acurex Environmental Corporation. It has been subjected to the Agency's peer and administrative
review arid has been approved for publiction as an EPA document. Mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
ii
-------
FOREWORD
The U.S. Environmental Protection Agency is charged by Congress with protecting the
Nation's land, air, and water resources. Under a mandate of national environmental laws, the
Agency strives to formulate and implement actions leading to a compatible balance between
human activities and the ability of natural systems to support and nurture life. To meet this
mandate, EPA's research program is providing data and technical support for solving
environmental problems today and building a science knowledge base necessary to manage our
ecological resources wisely, understand how pollutants affect our health, and prevent or reduce
environmental risks in the future.
The National Risk Management Research Laboratory is the Agency's center for investigation
of technological and management approaches for reducing risks from threats to human health and
the environment. The focus of the Laboratory's research program is on methods for the
prevention and control of pollution to air, land, water and subsurface resources; protection of
water quality in public water systems; remediation of contaminated sites and ground water; and
prevention and control of indoor air pollution. The goal of this research effort is to catalyze
development and implementation of innovative, cost-effective environmental technologies; develop
scientific and engineering information needed by EPA to support regulatory and policy decisions;
and provide technical support and information transfer to ensure effective implementation of
environmental regulations and strategies.
This publication has been produced as part of the Laboratory's strategic long-term
research plan. It is published and made available by EPA's Office of Research and Development
to assist the user community and to link researchers with their clients.
E. Timothy Oppelt, Director
National Risk Management Research Laboratory
iii
-------
ABSTRACT
In a recently completed test program at the U.S. EPA Incineration Research Facility
(IRF), 10 prototype or developing continuous emission monitors (CEMs) for measuring trace
metal or trace organic species concentrations were tested. Of the 10 CEMs tested, four measured
incinerator flue gas concentrations of several specific volatile organic compounds (VOCs), one
measured total particulate-bound polynuclear aromatic hydrocarbon (PAH) concentrations, two
measured flue gas concentrations of several (up to 14) trace metals, and three measured mercury
concentrations. While the testing consisted of obtaining quantitative measurement data on the four
measures ofCEM performance checked in a relative accuracy test audit (RATA) as described in
40 CFR 60 Appendix F — relative accuracy (RA), calibration drift (CD), zero drift (ZD), and
response time -- the primary project objective focused on the RA measurement. The RA
measurement was achieved by comparing the monitored analyte concentration reported by the
CEM to the concentration determined by the EPA Reference Method (RM) for the analyte. Four
series of tests were performed, each simultaneously testing up to three monitors measuring the
same or similar analyte type. Each test series consisted of performing triplicate RM measurements
at each of three target flue gas monitored analyte concentrations while the tested CEMs were in
operation. Thus, each series gave nine RM, CEM comparisons in total. All measurements were
taken in the wet scrubber exit flue gas from the pilot-scale rotary kiln incineration system (RKS)
at the IRF. The test program was performed in August and September 1995.
The test program results clearly showed the prototype nature of most approaches tested,
and the clear need for further development. Mercury CEMs will require the least development
effort to reach commercial status and are nearly there. However, the approaches tested for multi-
metals and VOC determinations will likely require 1 to 3 years of further development. Given this
need, the importance of continuing test programs of the scope and scale of this one cannot be
overemphasized.
This report was submitted in fulfillment of Contract No. 68-C4-0044 by Acurex
Environmental Corporation under the partial sponsorship of the U.S. Environmental Protection
Agency. This report covers a period of December 20, 1994 to September 30, 1995, and work was
completed as of September 30, 1996.
iv
-------
CONTENTS
Section Page
NOTICE ii
FOREWORD iii
ABSTRACT iv
FIGURES vii
TABLES ix
LIST OF ACRONYMS xiii
1 INTRODUCTION 1
1.1 BACKGROUND 1
1.2 TEST PROGRAM OBJECTIVES 3
1.3 PROGRAM PARTICIPANT RESPONSIBILITIES 5
1.4 REPORT OUTLINE 5
2 FACILITY DESCRIPTION, TEST PROGRAM DESIGN, AND TEST
CONDITIONS 7
2.1 TEST FACILITY 7
2.2 TEST PROGRAM DESCRIPTION 10
2.2.1 Test Waste Feed 13
2.2.2 Multi-Metals and Mercury CEMs Tests 15
2.2.3 VOC and SVOC CEMs Tests 17
2.3 TEST CONDITIONS 18
3 SAMPLING AND ANALYSIS PROCEDURES 28
3.1 SAMPLING PROCEDURES 28
3.1.1 Feed Sampling 30
3.1.2 Method 0030 Train Sampling 30
3.1.3 Method 0010 Train Sampling 31
3.1.4 Method 29 Train Sampling 31
3.1.5 Method 23 Train Sampling 31
3.1.6 Method 0050 Train Sampling 31
3.1.7 Combustion Gas CEM Measurements 32
v
-------
CONTENTS (continued)
Section Page
3.2 ANALYTICAL PROCEDURES 32
3.2.1 Method 0010 Train Sample Analyses by GC/MS 32
3.2.2 Method 0030 Train Sample Analyses by GC/FID 32
3.2.3 Method 29 Train Sample Analyses by ICP and AAS 36
3.2.4 Noncritical Analyses 36
4 TEST RESULTS 38
4.1 VOCCEM TESTS 38
4.1.1 Test Series 1 — ORNL and EcoLogic VOC CEMs 39
4.1.2 Test Series 4 — EPA/APPCD and MSP VOC CEMs 54
4.2 SVOC (PAH) CEM TESTS 68
4.3 MULTI-METALS CEM TESTS 75
4.4 MERCURY CEM TESTS 87
4.5 RELATIVE ACCURACY DISCUSSION 94
5 CONCLUSIONS 96
6 QUALITY ASSURANCE 101
6.1 VOC ANALYSES 101
6.2 SVOC ANALYSES 110
6.3 TRACE METAL ANALYSES Ill
6.4 TECHNICAL SYSTEM REVIEWS 121
7 DEVELOPER CLAIMS 123
7.1 ORNL VOC CEM DEVELOPER COMMENTS 125
7.2 ECOLOGIC VOC CEM DEVELOPER COMMENTS 131
7.3 ECOCHEM SVOC CEM DEVELOPER COMMENTS 141
7.4 SNL MULTI-METALS CEM DEVELOPER COMMENTS 149
7.5 ECOCHEM Hg CEM DEVELOPER COMMENTS 153
REFERENCES 161
APPENDIX A — DRAFT PS FOR MULTI-METALS CEMs 163
vi
-------
FIGURES
Number Page
1 Schematic of the IRF rotary kiln incineration system 8
2 Typical test day schedule 12
3 Test sampling locations 29
4 Generalized CEM gas flow schematic 34
5 Flue gas VOC concentrations reported by the ORNL CEM for the low VOC
concentration tests 49
6 Flue gas VOC concentrations reported by the ORNL CEM for the
intermediate VOC concentration tests 50
7 Flue gas VOC concentrations reported by the ORNL CEM for the high
VOC concentration tests 51
8 Flue gas VOC concentrations reported by the EPA/APPCD CEM for the
low VOC concentration tests 62
9 Flue gas VOC concentrations reported by the EPA/APPCD CEM for the
intermediate VOC concentration tests 63
10 Flue gas VOC concentrations reported by the EPA/APPCD CEM for the
high VOC concentration tests 64
11 Flue gas VOC concentrations reported by the MSP CEM for the low VOC
concentration tests 65
12 Flue gas total PAH concentrations reported by the EcoChem CEM for the
low (top) and intermediate (bottom) SVOC concentration tests 73
13 Flue gas metals concentrations reported by the Metorex CEM for the low
metals concentration tests 83
14 Rue gas metals concentrations reported by the Metorex CEM for the
intermediate metals concentration tests 84
vii
-------
FIGURES (continued)
Number Page
15 Flue gas metals concentrations reported by the Metorex CEM for the high
metals concentration tests 85
16 Flue gas mercury concentrations reported by the EcoChem CEM for the low
(top), intermediate (middle), and high (bottom) mercury concentration tests .... 91
17 Flue gas mercury concentrations reported by the Perkin-Elmer CEM for the
low (top), intermediate (middle), and high (bottom) mercury concentration
tests 92
18 Flue gas mercury concentrations reported by the Senova CEM for the high
mercury concentration test 93
19 Histogram of VOC surrogate recoveries from Method 0030 samples 107
20 Histogram of SVOC surrogate recoveries from Method 0010 samples 113
viii
-------
TABLES
Number Page
1 Participants in the CEM test program 4
2 Design characteristics of the IRF rotary kiln incineration system 9
3 CEM test program summary 14
4 Test trace metals and target flue gas concentrations 16
5 Concentrated aqueous spike solution composition 16
6 VOCs spiked into flue gas 17
7 Kiln operating data for the CEMs tests 19
8 Afterburner operating data for the CEMs tests 21
9 Air pollution control system 23
10 Combustion gas CEM data 25
11 Continuous emission monitors available and locations monitored 33
12 Analysis procedures 35
13 SVOCs quantitated in Method 0010 train samples 35
14 Trace metals quantitated in Method 29 train samples 36
15 Measured flue gas concentrations for the tests of the ORNL and EcoLogic
CEMs at the low VOC concentration 39
16 Measured flue gas concentrations for the tests of the ORNL and EcoLogic
CEMs at the intermediate VOC concentration 40
17 Measured flue gas concentrations for the tests of the ORNL and EcoLogic
CEMs at the high VOC concentration 40
ix
-------
TABLES (continued)
Number Page
18 Comparison of measured and target concentrations for the ORNL and
EcoLogic CEM tests 42
19 Relative accuracies of the ORNL and EcoLogic CEMs 43
20 ORNL CEM bias and precision estimates for 1,2-dichloroethane 45
21 ORNL CEM bias and precision estimates for benzene, chlorobenzene,
chloroform, 1,1-diehloroethene, tetrachloroethene, toluene, and
1,1,1-trichloroethane 46
22 ORNL CEM bias and precision estimates for carbon tetrachloride and
trichloroethenc 47
23 EcoLogic CEM bias and precision estimates for benzene and
tetrachloroethene 47
24 EcoLogic CEM bias and precision estimates for carbon tetrachloride,
chlorobenzene, chloroform, 1,2-dichloroethane, 1,1-dichloroethene, toluene,
1,1,1-trichloroethane, and trichloroethene 48
25 Calibration and zero drift for the low VOC concentration test day 52
26 Calibration and zero drift for the intermediate VOC concentration test day 53
27 Calibration and zero drift for the high VOC concentration test day 53
28 Estimated CEM detection limits 54
29 Measured flue gas concentrations for the tests of the EPA/APPCD and MSP
CEMs at the low VOC concentration 55
30 Measured flue gas concentrations for the tests of the EPA/APPCD and MSP
CEMs at the intermediate VOC concentration 55
31 Measured flue gas concentrations for the tests of the EPA/APPCD and MSP
CEMs at the high VOC concentration 56
32 Comparison of measured and target concentrations for the EPA/APPCD and
MSP CEM tests 57
33 Relative accuracies of the EPA/APPCD and MSP CEMs 58
x
-------
TABLES (continued)
Number Page
34 EPA/APPCD CEM bias and precision estimates for carbon tetrachloride,
chlorobenzene, chloroform, tetrachloroethene, toluene, 1,1,1-trichloroethane,
and trichloroethene 60
35 EPA/APPCD CEM bias and precision estimates for benzene, 1,2-
dichloroethane, and 1,1-dichloroethene 61
36 EPA/APPCD pre- and post-test zero measurements 67
37 EPA/APPCD pre- and post-test calibration measurements and corresponding
calibration drift 67
38 Zero drift for the EPA VOC CEM 69
39 Measured flue gas concentrations for the EcoChem PAH CEM tests 70
40 Comparison of measured and target concentrations for the EcoChem PAH
CEM tests 71
41 Measured flue gas concentrations for the tests of the SNL and Metorex
CEMs at the low metals concentrations 76
42 Measured flue gas concentrations for the tests of the SNL and Metorex
CEMs at the intermediate metals concentrations 77
43 Measured flue gas concentrations for the tests of the SNL and Metorex
CEMs at the high metals concentrations 78
44 Comparison of measured and target concentrations for the multi-metals
CEM tests 79
45 Relative accuracies of the SNL and Metorex CEMs 80
46 Metorex CEM bias and precision estimates for antimony, arsenic, barium,
cadmium, chromium, cobalt, lead, nickel, selenium, and thallium 81
47 Metorex CEM bias and precision estimates for manganese 82
48 SNL CEM bias and precision estimates for antimony, barium, and lead 82
49 Detection limits reported by SNL and Metorex 87
50 Measured flue gas concentrations and RAs for the mercury CEM tests 89
xi
-------
TABLES (continued)
Number Page
51 Comparison of measured and target mercury concentrations for the mercury
CEM tests 89
52 Bias and precision estimates for the three mercury CEMs tested 90
53 Measured mercury CEM ZDs and CDs 94
54 Precision, accuracy and completeness QAOs for critical measurements 102
55 Results of the VOC analyses of Method 0030 field blank samples 103
56 Comparison of RM and field blank data 104
57 VOC recoveries from matrix spike samples 106
58 VOST audit cylinder analysis results 107
59 VOC measurement MDLs: objectives and achieved 108
60 Comparison of measured and target concentrations for the low concentration
scoping tests 109
61 SVOC recoveries from matrix spike samples 112
62 SVOC SRM analysis results 113
63 SVOC measurement MDLs: objectives and achieved 114
64 Method 29 field blank analysis results: multi-metal CEM tests 115
65 Method 29 field blank analysis results: mercury CEM tests 117
66 Trace metal recoveries from matrix spike samples 119
67 Mercury recoveries from matrix spike samples 120
68 Trace metal SRM analysis results 120
69 Trace metal measurement MDLs: objectives and achieved 121
xii
-------
LIST OF ACRONYMS
ANOVA Analysis of variance
APCS Air pollution control system
APPCD EPA's Air Pollution Prevention and Control Division of NRMRL
ASME The American Society of Mechanical Engineers
BIF Boiler and industrial furnace
CBD The Commerce Business Daily
CD Calibration drift
CEM Continuous emission monitor
CFR The Code of Federal Regulations
CVAAS Cold vapor atomic absorption spectroscopy
DL Detection limit
DOE The U.S. Department of Energy
DRE Destruction and removal efficiency
DSITMS Direct Sampling Ion Trap Mass Spectrometry
EPA The U.S. Environmental Protection Agency
ERDA The Education, Research, and Development Association of Georgia
Universities
FID Flame ionization detector
FTIR Fourier transform infrared spectroscopy
GC/FID Gas chromatography/flame ionization detector
xiii
-------
LIST OF ACRONYMS (CONTINUED)
GC/MS
Gas chromatography/mass spectrometry
GFAAS
Graphite furnace atomic absorption spectroscopy
HEPA
High-efficiency particulate air (filter)
HRCG/HRMS
High-resolution gas chromatography/high-resolution mass spectrometry
IC
Ion chromatography
ICP
Inductively coupled argon plasma spectroscopy
IRF
The EPA's Incineration Research Facility in Jefferson, Arkansas
LASS
Laser Spark Spectroscopy
LIBS
Laser-Induced Breakdown Spectroscopy
MDL
Method detection limit
MS
Matrix spike, a spiked sample used to evaluate analytical method recovery
MS
Mass spectrometry
MSD
Matrix spike duplicate, a duplicate MS sample
MSP
Marine Shale Processors
NDIR
Non-dispersive infrared
NIST
National Institute for Standards and Technology
NRMRL
EPA's National Risk Management Research Laboratory
NRCC
The National Research Council of Canada
ORNL
Oak Ridge National Laboratory
OSW
EPA's Office of Solid Waste
OTD
DOE's Office of Technology Development
PAH
Polycyclic aromatic hydrocarbon or polynuclear aromatic hydrocarbon
PAS
Photoelectric aerosol sensor
-------
LIST OF ACRONYMS (CONTINUED)
PCDD Polychlorinated dibenzo-p-dioxins, a class of compounds often referred to as
dioxins
PCDF Polychlorinated dibenzofurans
PIC Product of incomplete combustion
POHC Principal organic hazardous constituent
PQL Practical quantitation limit
PS Performance specification
QA Quality assurance
QAO Quality assurance objective
QAPP Quality assurance project plan
QC Quality control
RA Relative accuracy
RATA Relative accuracy test audit
RCRA The Resource Conservation and Recovery Act
RKS The pilot-scale rotary kiln incineration system at the IRF
RM Reference method
RPD Relative percent difference
RSD Relative standard deviation
SNL Sandia National Laboratories
SRM Standard reference material
SRTC DOE's Savannah River Technology Center
SVOC Semivolatile organic compound
SW-846 The compendium of EPA sampling and analysis methods to be used for
evaluating solid waste, Reference 3
xv
-------
LIST OF ACRONYMS (CONTINUED)
TSR
Quality assurance technical systems review
TUHC
Total unburned hydrocarbon
VOC
Volatile organic compound
VOST
Volatile organic sampling train, a term often used to refer to Method 0030
XRF
X-ray fluorescence
ZD
Zero drift
xvi
-------
SECTION 1
INTRODUCTION
On May 18, 1993, the Administrator of the U.S. Environmental Protection Agency (EPA)
issued the Draft Combustion Strategy, which proposed more stringent emission standards and
changes in the way that permits for waste combustion facilities are handled. More public
involvement in the process was proposed. Because the public's apparent perception of incinerators
is that high concentrations of hazardous compounds are continually being released from the stacks
of the thermal treatment devices, a means by which the "real-time" (defined as ranging from
instantaneous to a within-several-hours time frame) organic and metals emissions can be monitored
would be of great benefit to both regulators and the regulated community. The purpose of this
project was to test instruments that would offer the ability to gain "immediate" knowledge of the
stack emissions, which in turn would provide assurance that the thermal treatment device is
operating correctly or indicate the change of operating conditions needed to adjust stack emissions.
EPA's Office of Solid Waste (OSW) and the Office of Solid Waste and Emergency Response
would like this monitoring capability as a means of responding to, and allaying, the public's fears
by showing that good, safe, and clean combustion practice can be demonstrated.
Several developers have designed monitoring units that they claim will measure various
regulated hazardous compounds using a number of different innovative concepts and technologies.
This project tested the effectiveness, accuracy, and serviceability of potential continuous emission
monitoring systems.
The immediate needs of the thermal treatment community include the capability of
performing "real-time" monitoring of organic compounds and metals as these exit the stack.
Conventional procedures usually involve sample collection over an extended period of time and
then sample analysis at a later time. "Real-time" monitoring, on the other hand, involves the
virtually immediate analysis of trace quantities of pollutants. In this report, developers' monitoring
systems results are compared with simultaneous flue gas sampling using conventional EPA
Reference Method (RM) sampling trains and analytical procedures.
1.1 BACKGROUND
The impetus for developing continuous emission monitors (CEMs) for measuring flue gas
emission concentrations of trace organic compounds has existed since the promulgation of the
initial hazardous waste incinerator performance standards in 1980 (40 CFR 264, Subpart O). One
requirement of these standards was that an incinerator demonstrate the capability of achieving
1
-------
99.99 percent destruction and removal efficiency (DRE) of the principal organic hazardous
constituents (POHCs) in the hazardous waste incinerated. At the time this standard was
promulgated, the desire was to require a continuous measurement of POHC DRE. However, flue
gas POHC concentrations corresponding to the emission rates associated with 99.99 percent DRE
are in the 1 pg/dscm range, or of the order of 0.2 parts per billion on a volume basis (ppbv). The
most sensitive CEM approaches for measuring organic compound concentrations at the time had
detection limits in the parts per million (ppm) range, not nearly sensitive enough.
As a consequence, compliance with the 99.99 percent DRE standard was shown via a trial
burn. Continuous compliance with the standard was to be assured by constraining incinerator
operation to conditions consistent with those tested in the trial burn. This approach, though
workable, was clearly less attractive than showing continuous compliance via an emission rate
measurement. Thus, the incentive for developing sufficiently sensitive organic compound CEMs
arose.
Similarly, the impetus for developing CEMs for measuring flue gas emission
concentrations of trace metals arose with the promulgation of regulations governing the destruction
of hazardous wastes in boilers and industrial furnaces (BIFs) in 1991 (40 CFR 266, Subpart H).
These rules, extended to hazardous waste incinerators during permit revisions and reauthorizations
under the omnibus authority granted permit writers within the Resource Conservation and
Recovery Act (RCRA), limit the emissions of several trace metals from waste combustion devices.
Again, at the time these rules were promulgated, the ultimate desire was to require a continuous
measurement of regulated metal emission rates. However, at the time these rules were
promulgated, no metals CEMs existed. So, again, compliance with the standard has been shown
via a trial burn. Subsequent continuous compliance with the emission rate standards is
demonstrated by limiting the feedrate of each metal to the waste combustor and again constraining
combustion device operation to conditions consistent with those tested in the trial burn. The metal
feedrate limitation, in turn, requires the operator to have detailed knowledge of the metal
concentrations in wastes being fed to the waste combustor. In application, this often requires
numerous feed metal analyses. Compliance with the BIF rules, thus, would be much simplified
if flue gas metals concentrations could be continuously monitored.
Given these incentives, efforts to develop CEM approaches sufficiently sensitive for trace
organic compounds and effective for determining trace metal concentrations in particulate-
containing flue gas proceeded. For trace organic compounds, Fourier transform infrared (FTIR)
spectroscopy and online mass spectrometry approaches have been investigated, among others. For
trace metals, a number of CEM approaches were discussed at the second American Society of
Mechanical Engineers (ASME)/EPA joint workshop on metals emissions from hazardous waste
combustion systems (Reference 1) in 1993. These approaches were sufficiently promising that the
ASME Research Committee on Industrial and Municipal Wastes formed a subcommittee to
specifically track the progress in metals CEMs development.
In concert with the progress of CEMs development. OSW, in anticipation of being able
to require, or at least offer, the option of using continuous monitoring of regulated pollutant
emission rates in the most recent revision to the hazardous waste combustion process regulations,
2
-------
drafted performance specifications (PSs) for metals CEMs. These PSs are in the same form as
PS2 for SO2 and NOx CEMs contained in 40 CFR 60, Appendix B, and are viewed as being a
complement to a CEMs quality assurance (QA) program much like that required by 40 CFR 60,
Appendix F. These draft PSs are included in Appendix A of this report.
As noted above, the development of CEM approaches for both trace metal and trace
organic analyte classes has advanced to the point that several candidate approaches are now in the
prototype instrument stage. Thus, the general objective of this project was to test several prototype
instruments and establish or estimate their effectiveness, reliability, accuracy, and detection limits.
Support for this project came from both the EPA National Risk Management Research Laboratory
(NRMRL), and the U.S. Department of Energy's (DOE's) Office of Technology Development
(OTD) through the Savannah River Technical Center (SRTC) and the Education, Research, and
Development Association of Georgia Universities (ERDA).
To solicit candidate instruments for testing in the project, an announcement was published
in the January 4, 1995, Commerce Business Daily (CBD). Several proposals were received in
response to this announcement. Proposals addressing both trace metals measurement and trace
organic compound measurement were received. The selection of which CEMs to be included in
this project was subsequently done, taking into consideration the recommendations of a program
coordination committee organized by ERDA, with support from SRTC. The selection process
resulted in 11 offerings being identified for testing in this program. These are listed in Table 1
by monitored analyte class. As shown, included in the list of CEMs selected for testing in this
program were one semivolatile organic constituent (SVOC), four volatile organic constituent
(VOC), two multi-metal, and four mercury CEMs.
1.2 TEST PROGRAM OBJECTIVES
The selected approaches were evaluated in this test program using the rotary kiln
incineration system (RKS) at EPA's Incineration Research Facility (IRF), located in Jefferson,
Arkansas. The testing consisted of obtaining quantitative measurement data on four measures of
CEM performance checked in a relative accuracy test audit (RATA) of a CEM as described in 40
CFR Part 60, Appendix F. These measures are:
• Relative accuracy (RA): the absolute mean difference between the concentrations
determined by the CEM and the value determined by the RM, plus the 2.5 percent
error confidence coefficient of a series of tests, divided by the mean of the RM
tests
• Calibration drift (CD): the difference in the CEM output reading from the
established reference value after a stated period of operation during which no
unscheduled maintenance, repair, or adjustment took place; the reference value is
established by a calibration standard which has a concentration of nominally
80 percent or greater of the CEM's full-scale (span) reading capability
• Zero drift (ZD): the CD where the reference value is 0
3
-------
TABLE 1. PARTICIPANTS IN THE CEM TEST PROGRAM
Monitored
analyte
Developer
Approach
SVOCs
EcoChem
Photoionization of aerosol-bound polycyclic
aromatic hydrocarbons (PAHs)
VOCs
EcoLogic
Continuous chemical ionization mass
spectrometry
Marine Shale Processors
Continuous online mass spectrometry
Oak Ridge National
Direct sampling ion trap mass spectrometry
Laboratory
EPA, Air Pollution Prevention
Online gas chromatography with dual-flame
and Control Division
ionization, electron capture detection
Multi-metals
Sandia National Laboratory
Laser-induced plasma spectroscopy
Metorex
Extractive beta gauge particulate monitor with
x-ray fluorescence metals analysis
Mercury
Perkin-EImer
Gold trap amalgamation collection, cold vapor
atomic absorption spectroscopy (CVAAS)
analysis
Euramark
CVAAS
Senova
Noble metal film solid state chemical
microsensor
EcoChem
CVAAS
• Response time: the time interval between the start of a step change in the
concentration of the monitored gas stream and the time when the CEM output
reaches 95 percent of the final value
RA, CD, and ZD calculation procedures are discussed in the draft PS included in
Appendix A of this report.
Within the overall objective definition, the primary project objective was to measure the
RA of each CEM tested. Secondary objectives included obtaining measures of CD, ZD, and
response time for each CEM tested. In addition, the detection limit (DL) of each CEM was
estimated. Normally, DL is defined as three times the standard deviation of nine repeated
measurements of a blank sample or low-level (near-blank) sample. However, this rigorous
approach was not employed in this program; DL was instead estimated from observations of
instrument responses to the flue gas analyte concentrations tested.
4
-------
Measuring a CEM's RA requires comparing the monitored analyte concentration reported
by the CEM to the concentration determined by the RM for the analyte. The RM for trace metal
(including mercury) monitors was draft Method 29, the EPA multiple metals method documented
in the BIF rules (Reference 2). The RM for VOCs was Method 0030 with analysis using thermal
desorption, purge and trap by Method 5040 with quantitation using Method 8015 A (Reference 3).
The RM for SVOCs was Method 0010 with analysis by Method 8270B (Reference 3).
13 PROGRAM PARTICIPANT RESPONSIBILITIES
As part of the tested CEM selection process, each instrument developer participating in
the program was required to commit to supplying a complete CEM system for testing, including
any needed sample lines and equipment stands, all instrument calibration materials and equipment,
and system operators. In addition, each developer was required to commit to supplying the
program sponsors with
• A daily report containing all raw and processed data, results of zero and span
checks, and copies of operator logbook pages for each test day
• A developer test report containing
— A technical description of the instrument principles of operation
— A description of the instrument operating and calibration procedures
— Final measured flue gas analyte concentrations averaged over each RM test
period
— Final ZD and CD measurement results for each test day
In return, each developer was offered the opportunity to review the draft of this test report
and, based on this review, to submit a "vendor claims" discussion for inclusion, without
modification, in this final test report. This discussion could include further developer claims
regarding the capabilities of their technology, a self-assessment of the performance of the
technology in the test program, and any comments on the developer's view of the favorable
aspects and/or the shortcomings of the conduct of the test program.
1.4 REPORT OUTLINE
Results of the test program are summarized in this report. Section 2 describes the IRF's
RKS, in which the tests were performed, and discusses details of the test program design, the test
waste feed, and the test incinerator system operating conditions. Section 3 discusses the sampling
and analysis procedures employed during the tests. Section 4 presents the RM and CEM results
and their comparisons, together with the evaluations of RA, CD, ZD, and CEM response times;
this section is the main focus and the heart of this report. Section 5 summarizes the test program
conclusions. Section 6 discusses the quality assurance (QA) aspects of the test program.
5
-------
Section 7 includes each developer's "vendor claims" discussions resulting from review of the draft
test report. Finally, the Appendices provide a complete assembly of the test data from which
additional test program information of interest can be extracted for further study. Included in
Appendix H are each developer's test report and each developer's daily reports.
6
-------
SECTION 2
FACILITY DESCRIPTION, TEST PROGRAM DESIGN,
AND TEST CONDITIONS
The test program consisted of four series of tests. AH tests were conducted in the RKS
at the IRF. Section 2.1 describes the RKS. Section 2.2 provides a description of the test program
design. Section 2.3 discusses actual test conditions in effect during each test performed.
2.1 TEST FACILITY
Figure 1 is a process schematic of the RKS. Table 2 lists the system's design
characteristics. The IRF RKS consists of a primary combustion chamber, a transition section, and
a fired afterburner chamber. The primary combustion chamber, or kiln, is nominally 1.04 m (3 ft,
4.75 in) in diameter and 2.26 m (7 ft, 5 in) long. The afterburner is nominally 0.91 m (3 ft) in
diameter and 3.05 m (10 ft) long. A 4.43-m (14-ft, 6.5-in) long, 0.61-m (2-ft) diameter refractory-
lined afterburner extension follows the afterburner and is in place for flue gas flow conditioning
to allow isokinetic sampling of hot flue gas prior to quenching. Both the kiln and afterburner are
fitted with 590 kW (2.0 MMBtu/hr) auxiliary fuel burners. Natural gas is the auxiliary fuel,
although liquid waste or fuel can also be fired. Typical firing rates are 290 to 440 kW (1.0 to
1.5 MMBtu/hr) to the kiln, and 440 to 490 kW (1.5 to 2.0 MMBtu/hr) to the afterburner.
In normal operation, combustion flue gas exiting the afterburner extension is rapidly
quenched with water sprays to saturation, corresponding to a flue gas temperature of about 1TC
(170°F), then passed through a two-stage primary air pollution control system (APCS) consisting
of a venturi/packed-column scrubber combination as the first stage, followed by a flue gas reheat
and fabric filter (baghouse) as the second stage. The scrubber system is designed to remove
coarse particulate and any acid gas, such as HC1, from the flue gas. The baghouse removes most
of the remaining flue gas particulate. Reheating the flue gas ensures that no moisture will
condense in the baghouse and adversely affect its operation. Downstream of the baghouse, a
backup, secondary APCS, comprised of an activated-carbon adsorber and a high-efficiency
particulate air (HEPA) filter, is in place.
Provision for performing isokinetic flue gas sampling is provided at four locations in the
flue gas path through the system: at the afterburner exit, the scrubber system exit, the baghouse
system exit, and in the stack downstream of the HEPA filter. Access to the flue gas at each
location is via a number of standard 3-in (7.6-cm) flanged sampling ports. At the afterburner exit
location, these ports are located in the afterburner extension noted above. At the other three
7
-------
ID FAN
ATMOSPHERE
PACKED
COLUMN
SCRUBBER
QUENCH
STACK
SECONDARY
BURNER
VENTURI
SCRUBBER
AFTERBURNER
EXTENSION
AIR
SCRUBBER
LIQUOR
RECIRCULATION
ID FAN
NATURAL
GAS,
LIQUID "
FEED
CARBON BED HEPA
ADSORBER FILTER
AFTERBURNER
SOLIDS
FEEDER
TRANSFER
DUCT
MAIN
BURNER
FLUE GAS
REHEATER
BAGHOUSE
ASH
BIN
AIR
NATURAL
GAS,
LIQUID FEED
ROTARY
KILN
ASH HOPPER
REDUNDANT AIR
POLLUTION CONTROL
SYSTEM
ROTARY KILN
INCINERATOR
PRIMARY AIR POLLUTION
CONTROL SYSTEM
Figure 1. Schematic of the IRF rotary kiln incineration system.
-------
TABLE 2. DESIGN CHARACTERISTICS OF THE IRF ROTARY KILN
INCINERATION SYSTEM
Characteristics of the Kiln Main Chamber
Length
Diameter, outside
Diameter, inside
Chamber volume
Construction
Refractory
Rotation
Solids retention time
Burner
Primary fuel
Feed system:
Sludges
Solids
Temperature
2.26 m (7 ft, 5 in)
1.37 m (4 ft, 6 in)
Nominal 1.04 m (3 ft, 4.75 in)
1.90 m3 (67.2 ft3)
0.95-cm (0.375-in) thick cold-rolled steel
18.7-cm (7.375-in) thick high-alumina castable refractory, variable
Clockwise or counterclockwise, 0.2 to 1.5 rpm
1 hr (at 0.2 rpm)
North American Burner rated at 590 kW (2.0 MMBtu/hr) with liquid
Natural gas
Positive displacement pump via water-cooled lance
Moyno pump via front face, water-cooled lance
Metered twin-auger screw feeder of fiberpack ram feeder
1,0I0°C (1,850°F)
Characteristics of the Afterburner Chamber
Length
Diameter, outside
Diameter, inside
Chamber volume
Construction
Refractory
Gas residence time
Burner
Primary fuel
Temperature
3.05 m (10 ft)
1.22 m (4 ft)
0.91 m (3 ft)
1.80 m3 (63.6 ft3)
0.63-cm (0.25-in) thick cold-rolled steel
15.2-cm (6-in) thick high-alumina castable refractory
1.2 to 2.5 s depending on temperature and excess air
North American Burner rated at 590 kW (2.0 MMBtu/hr) with liquid
Natural gas
1,200°C (2,200°F)
Characteristics of the Afterburner Extension
Length, with transition
Diameter, outside
Diameter, inside
Chamber volume
Construction
Refractory
Temperature (max.)
4.43 m (14 ft, 6.5 in)
0.915 m (3 ft)
0.61 m (2 ft)
1.19 m3 (41.9 ft3)
0.63-cm (0.25-in) thick cold-rolled steel
15.2-cm (6-in) thick high-alumina castable refractory
1,200°C (2,200°F)
Characteristics of the Venturi/Packed-Column Scrubber APCS
System capacity, inlet
Pressure drop
Venturi scrubber
Packed column
Liquid flow
Venturi scrubber
Packed column
pH control
107 m3/min (3,773 acfm) at !,2003C (2,200=F) and 101 kPa (14.7 in WC)
7.5 kPa (30 in WC)
1.0 kPa (4 in WC)
77.2 L/min (20.4 gpm) at 60 kPa (10 psig)
116 L/min (30.6 gpm) at 69 kPa (10 psig)
Feedback control by NaOH solution addition
(continued)
9
-------
TABLE 2. (continued)
Characteristics of the Baghouse Collector
System Capacity, inlet gas
flow
Operating temperature
Operating pressure
Diameter
Overall height
Filter elements (bags)
Material
Length
Number
Total filter area
Material of construction
Collector internals
Airlock
Venturi nozzles
Insulation
70 m3/min (2,500 acfm) at 120°C (250°F)
200°C (400°F)
±12.4 Kpa (±50 in WC)
1.8 m (6 ft)
4.2 m (13 ft, 8.375 in)
16 oz. Nomex
1.8 m (6 ft)
69
45 m2 (488 ft2;
304 SS
316 SS
Aluminum
Heat loss less than 8.8 kW (30,000 Btu/hr) at 200°C (400°F)
locations, the ports are located in straight runs of nominally 12-in (30-cm) diameter, circular cross
section, fiberglass ducting. Sampled gas flows at all locations are fully turbulent and at essentially
constant velocity across the duct's diameter. Flow velocities are nominally 7 m/s (20 ft/s).
For this test program, an afterburner exit flue gas partial quench system was installed in
the afterburner extension so that the afterburner exit flue gas temperature could be decreased to
the 360° to 427°C (680° to 800°F) range prior to completely quenching to saturation. This partial
quench capability was needed for flue gas spiking of VOC and SVOC analytes to be monitored,
as discussed in Section 2.2.3. This partial quench system consisted of a water spray nozzle
inserted into one of the first ports in the afterburner extension. The system was designed to ensure
complete evaporation of the water spray occurred prior to entering the full quench device.
All CEMs tested in this program sampled flue gas at the scrubber exit location. The flue
gas at this location is also saturated with water vapor, but has cooled from the nominally 77°C
(170°F) quench exit temperature to about 60°C (140°F). The duct at this location is at a negative
pressure (draft) of nominally 1.5 kPa (6 in WC). For this program, the length of scrubber exit
ductwork accessible for testing had four sets of ports, each set comprised of four individual ports
at 90° increments in the duct circular cross-section.
2.2 TEST PROGRAM DESCRIPTION
Up to three CEMs were tested at the same time during each of the four test series. Thus,
each CEM under simultaneous testing in a test series had access to one of the four sets of ports
at the sampling location. The remaining set(s) of ports were dedicated to RM sampling performed
by the IRF staff.
10
-------
The major portion of each test series consisted of performing three sequential RM
measurements, while the tested CEMs were in operation, at each of three flue gas concentrations
of monitored analytes. Thus, each test series was designed to supply nine sets of parallel RM and
CEM reading data, three at each of three analyte concentrations. These nine sets of parallel RM
and CEM data supported the calculation of each CEM's RA. Thus, up to three RAs were
calculated for each CEM, one at each of the three flue gas concentrations tested. Other test efforts
discussed below supported the measurements of CD, ZD, and response time. To ensure that the
sets of RM/CEM concentration data were indeed parallel and comparable, the developers were
notified of the start and stop times of each RM procedure so that they could report an average
analyte concentration that corresponded directly to the RM measurement period.
Performing one RM measurement of the flue gas constituent concentrations can require
a 2- to 2.5-hour flue gas sampling period, as discussed in Section 3. With contingency for
sampling train filter changes and other sampling procedure delay events, a 3.5-hour time period
was normally allotted to completing one RM test. Thus, the three sequential RM tests were
targeted for completion over a 10.5-hour period of continuous, steady RKS operation at a
nominally constant scrubber exit flue gas monitored analyte concentration. This nominally
10.5-hour period was termed 1 test day. Testing at three different scrubber exit flue gas
concentrations, thus, required 3 test days, referred to as a week of testing, comprising each test
series.
At the beginning of each test day, the RKS was brought to steady operation at the desired
incineration conditions firing natural gas. After the RKS combustion gas CEMs were calibrated
and all RM sampling preparations completed, test waste feed (see Section 2.2.1) was initiated and
steady RKS operation reestablished. During this time, each CEM developer was given the
opportunity to calibrate his instrument. This calibration included zero and span checks. The day's
test, the three sequential RM sampling efforts, began after all CEMs (up to three) had completed
the zero and span checks. The first of the three sequential RM sampling efforts began after the
CEMs being tested had been calibrated, provided that at least 1 hour of waste feeding had elapsed.
At the end of the test day (after completion of the third RM sampling period), up to two
successive step changes (increases or decreases) in flue gas analyte concentrations were induced,
as discussed in Section 4. Measurements of CEM responses to these step changes gave data on
instrument response time.
After these step change/response observation exercises, test waste feed was stopped and
each developer was given the opportunity to check the calibration of his instrument. These post-
test checks yielded the measures of CD and ZD. The RKS continued to operate, firing natural gas,
until the kiln was visually clear of bottom ash, or for 2 hours, whichever time was longer. After
this time period, the RKS was set to an unattended operating condition firing natural gas in
preparation for the next test day.
Figure 2 shows a typical test day schedule consistent with the above discussion.
Each series of tests was completed over a 2-week time period. This allowed sufficient
time for the developers of the CEMs tested in the series (up to three for each series) to complete
11
-------
0400
0700
0800
3 hr
1 hr
RKS Gas CEMs Calibration
RM Train Preparations
Developer CEM Pre-Test Calibration
Start Test Waste Feed
to
a>
6
g
o
UJ
<
1130
1500
3.5 hr,
First RM Test
3.5 hr
Second RM Test
3.5 hr;
Third RM Test
1830
1930
2030
2230
2330
1 hr
1 hr
2 hr
1 hr
¦ First Step Change
¦ Stop Waste Feed
' Second Step Change
Stop Testing
• Developer CEM Post-Test Calibration
• RKS Secured
Figure 2. Typical test day schedule.
12
-------
instrument setup at the IRF, complete troubleshooting and shakedown efforts, then complete the
3 days of testing. The entire first week of the 2-week test series was set aside for developer setup,
troubleshooting, and shakedown. RM testing occurred on Monday, Wednesday, and Friday of the
second week.
The four test series were completed over the 8-week period beginning July 24, 1995, and
ending September 15. Table 3 summarizes the participants and test dates for each test. Of the
11 instruments selected for testing, 10 were able to obtain some data. The Euramark mercury
monitor was damaged in transport to the IRF and repair efforts by the developer were
unsuccessful. The Marine Shale Processor (MSP) VOC monitor malfunctioned during the initial
tests in Test Series 1 and could not be brought into reliable operation. MSP was given the
opportunity to return and participate in Test Series 4, the last test series completed, which they did.
2.2.1 Test Waste Feed
The incinerator feed material for all tests was a synthetic hazardous waste comprised of
an attapulgite clay solid sorbent combined with a mixture of trace metals and VOCs. Attapulgite
clay sorbents have been used as the solid matrix base material in a number of past IRF test
programs (References 4, 5, 6, and 7). These sorbents can be obtained in a variety of sorbent
particle sizes, ranging from a fine powder to nominally 3-mm particle size. A solid sorbent size
that would result in nominally 30-percent particle entrainment and carryover out of the RKS kiln
and afterburner was selected for these tests.
The mixture of organic compounds added to the sorbent base contained 76 percent toluene
by weight, with 12 percent each of chlorobenzene and tetrachloroethene. This mixture was
combined with the clay sorbent in the ratio of 1.0 kg of organic constituent mixture to 2.4 kg of
clay. The resulting organic compound/clay mixture, thus contained nominally 22.4 percent of
toluene and 3.5 percent each of chlorobenzene and tetrachloroethene. Its chlorine content was
nominally 4.1 percent and its heating value nominally 10.7 MJ/kg (4,590 Btu/lb). The mixture
was a free-flowing solid with no free-standing liquid.
This organic liquid mixture/clay sorbent combination was selected because it was very
similar to the mixture used in four past IRF test programs (References 4, 5, 6, and 7). The organic
liquid mixture was the same as the ones previously used, and the clay sorbent was the same
formulation used in past tests, except that the sorbent size for this test program was finer. The
organic/sorbent mixture combination was as used in the past.
The mixture was prepared by charging a bulk rotating mixer with a weighed quantity of
clay sorbent, then sequentially adding weighed quantities of the three VOC components. After
mixing to uniform visual appearance, the mixture was transferred to 55-gaI (208-L) drums for
storage until needed. The drums were appropriately sealed to prevent volatiles loss.
For all tests, the clay/organic mixture was continuously fed to the RKS via a screw feeder
system. The test waste was charged to the screw feeder hopper from the storage drums filled
during feed preparation. After charging, the feed hopper was sealed under a nitrogen blanket
13
-------
TABLE 3. CEM TEST PROGRAM SUMMARY
Test completion dates
Test
Series
CEM tested
Monitored
analyte
Low analyte
concentration
Intermediate
analyte
concentration
High analyte
concentration
1
EcoLogic
Marine Shale Processors2
Oak Ridge National Laboratory
voc
voc
voc
7/31/95
8/2/95
8/4/95
2
Sandia National Laboratory
Metorex
Euramarkb
Multi-metals
Multi-metals
Hg
8/14/95
8/16/95
8/18/95
3
Perkin-Elmer
Senova
EcoChem
Hg
Hg
Hg
9/1/95°
8/2 8/9 5C
8/30/95c
4
EcoChem
EPA/APPCDd
Marine Shale Processors
svoc
voc
voc
9/11/95
9/13/95
9/15/95
"Marine Shale Processors was unable to bring their system into operation. They relumed during the
last test series.
bEuramark was unable to bring their system into operation.
cAt the Hg monitor developers' request, the planned intermediate Hg concentration became the
highest test concentration; the planned low concentration tested on August 28 became the
intermediate concentration; the low concentration, at half the original low concentration, was tested on
September 1.
dEPA's Air Pollution Prevention and Control Division of NRMRL.
system in place on the feeder. This blanket system served to minimize evaporative loss of the
volatile organic components of the test waste, and to prevent any explosive gas mixtures from
developing due to volatile organic evaporation.
For all tests, the target clay/organic mixture feedrate was 68 kg/hr (150 Ib/hr). The target
kiln exit gas temperature was 870"C (1.600°F), and target afterburner exit gas temperature was
1,065°C (1,950°F). These are conditions typical of industrial hazardous waste incinerator
operation. Combustion system temperatures were maintained by controlling the auxiliary fuel
(natural gas) firing rates to each of the system combustion chambers. The kiln rotation rate was
set to give a kiln solids residence time of about 1 hour.
14
-------
For all tests, the RKS scrubber system was operated at its design operating conditions,
with the exception of the venturi scrubber pressure drop. For these tests, the adjustable-throat
venturi was opened to its widest setting so that the venturi pressure drop was at a minimum, with
the corresponding particulate collection efficiency also at a minimum. The desire was to produce
scrubber exit flue gas metal-containing particulate concentrations in the 100 mg/dscm range.
2.2.2 Multi-Metals and Mercury CEMs Tests
The trace metals of interest to this test program are those being considered for regulation
by OSW and other EPA Program Offices. These are listed in Table 4, which also notes the target
scrubber exit flue gas concentrations of each metal for the tests of multi-metals CEMs. For the
mercury CEMs tests in Test Series 3, the low concentration targets were at half the levels noted
in the low concentration column in Table 4. The intermediate concentration targets were at those
noted in the low concentration column in Table 4, and the high concentration targets were those
noted in the intermediate concentration column in Table 4.
Trace metals were added to the RKS, to result in scrubber exit flue gas levels, via two
routes. Both routes used an aqueous spike solution of the metals. The composition of the most
concentrated spike solution used is given in Table 5. This most concentrated solution was added
for the multi-metal CEM test days at the target high flue gas metals target concentration. This
solution was diluted 10-fold and 40-fold for multi-metal CEM test days at the intermediate and
low target concentrations. The concentrated solution was diluted 10-fold, 40-fold, and 80-fold for
the mercury CEM tests at the high, intermediate, and low target concentrations. The entire
complement of metals was retained in the aqueous spike solution for the mercury CEM tests so
that any potential interferences with CEM readings due to the presence of the other metals could
be assessed.
The two routes of metals addition were incorporated into the clay/organic mixture and
atomized into the kiln main burner flame. The solid waste feed route was effected by metering
the aqueous spike solution into the clay/organic liquid mixture at the screw feeder just prior to
feed introduction into the kiln. A gear pump was used to inject the spike solution at a flowrate
of 2 L/hr. The burner flame atomization route was effected by spraying the aqueous spike solution
through the liquid feed nozzle of the kiln dual fuel main burner at a rate of 6 L/hr.
The solution of the VOCs and SVOCs used to establish the target flue gas VOC and
SVOC concentrations for the organic CEM tests (see Section 2.2.3) was also injected into the
partially quenched afterburner extension flue gas at the intermediate injection concentration for the
multi-metal and mercury CEM tests, again to assess any potential interferences.
After the third sequential RM sampling period for a test day was completed, up to two
step changes in flue gas metals concentrations were attempted to assess CEM response time. It
should be noted that the process step changes used to attempt to induce flue gas metals
concentration step changes may not have resulted in actual concentration step changes because of
the presence of the entire incinerator and scrubber system between spike solution addition points
and the CEM measurement location. However, even the somewhat qualitative measure of CEM
15
-------
TABLE 4. TEST TRACE METALS AND TARGET FLUE GAS
CONCENTRATIONS
Target flue gas concentration, ng/dscm
Metal
Low
Intermediate
High
Antimony (Sb)
10
40
400
Arsenic (As)
5
20
200
Barium (Ba)
50
200
2,000
Beryllium (Be)
0.5
2
20
Cadmium (Cd)
5
20
200
Chromium (Cr)
20
80
560
Cobalt (Co)
10
40
400
Lead (Pb)
50
200
2,000
Manganese (Mn)
5
20
200
Mercury (Hg)
20
80
800
Nickel (Ni)
10
40
400
Selenium (Se)
50
200
2,000
Silver (Ag)
5
20
200
Thallium (Tl)
5
20
200
TABLE 5. CONCENTRATED AQUEOUS SPIKE SOLUTION COMPOSITION
Metal concentration,
Compound
Metal
g/L
Compound
concentration, g/L
Antimony (Sb)
0.32
C4H4KO?Sb
0.85
Arsenic (As)
0.18
AS2O3
0.24
Barium (Ba)
6.3
Ba(N03)2
12.0
Beryllium (Be)
0.06
10,000 ppm Be standard
NAa
Cadmium (Cd)
0.075
Cd(N03)2 • 4 H20
0.21
Chromium (Cr)
1.5
Cr(N03)3 • 9 H20
11.2
Cobalt (Co)
1.6
Co(N03)2 • 6 H20
7.78
Lead (Pb)
1.40
Pb(N03)2
2.24
Manganese (Mn)
0.52
Mn(N03)2 • 6 H-)0
2.72
Mercury (Hg)
0.26
Hg(N03)2 • H20
0.44
Nickel (Ni)
1.90
Ni(N03V> • 6 H20
9.41
Selenium (Se)
0.70
SeO>
0.98
Silver (Ag)
0.11
AgNO,
0.17
Thallium (Tl)
0.075
T1(N03)3 • 3 H?0
0.16
aNA = Not applicable.
16
-------
response time possible via this approach was of interest. Process step change procedures and
CEM responses are discussed in the appropriate subsections in Section 4.
2.2.3 VOC and SVOC CEMs Tests
The list of VOCs present in the scrubber exit flue gas for all tests is given in Table 6.
This list contains many of the VOC species currently being considered for regulation. The target
flue gas concentrations of the compounds were in the 1 to 2, 10 to 20, and 160 to 240 pg/dscm
ranges (low, intermediate, and high concentrations) for all VOCs except carbon tetrachloride and
chloroform. For these two VOCs, the target flue gas concentrations were doubled to 2 to 4, 20
to 40, and 320 to 480 pg/dscm. Naphthalene, phenanthrene, and pyrene were the SVOCs
introduced into the flue gas for all tests. The target flue gas concentrations for these were 1 to
2, 10 to 20, and 160 to 240 pg/dscm.
The VOCs and SVOCs were introduced into the flue gas by metering a solution of the
spiking compounds in methanol through a length of fine bore stainless steel tubing into the
afterburner extension at its centerline. As noted in Section 2.1, the afterburner exit flue gas was
partially quenched to a temperature of between 360° and 427°C (680° and 800°F) by a water spray
introduced into the first port (nearest the afterburner proper) of the afterburner extension. The
VOC/SVOC solution was introduced through a port midway along the length of the afterburner
extension. This arrangement afforded time for the flue gas to cool prior to VOC/SVOC solution
injection as well as time for the VOC/SVOC solution to evaporate and fully mix in the flue gas
prior to entering the water quench section of the RKS.
The concentrated organic spiking solution prepared consisted of 0.4 g/L of carbon
tetrachloride and chloroform, 0.2 g/L of each of the other 8 constituents listed in Table 6, and
0.2 g/L of the three SVOCs, all in methanol. This concentrated solution was used for the high
target flue gas VOC and SVOC CEM tests. The solution was metered into the partially quenched
afterburner extension flue gas at a feedrate in the 800 to 1,200 mL/hr range. The concentrated
organic solution was diluted with methanol 10-fold and 100-fold for the intermediate and low
target flue gas concentration tests, respectively, and fed at the same 800 to 1,200 mL/hr feedrate.
Trace metals were also fed into the kiln in the same manner used for the multi-metals and mercury
CEM tests. The multi-metal CEM test intermediate feed solution concentrations were used.
TABLE 6. VOCs SPIKED INTO FLUE GAS
Benzene 1,1-Dichloroethene
Carbon tetrachloride Tetrachloroethene
Chlorobenzene Toluene
Chloroform 1,1,1-Trichloroethane
1,2-Dichloroethane Trichloroethene
17
-------
After the third sequential RM test run in a test day was completed, up to two step changes
in flue gas VOC/SVOC concentrations were attempted to assess CEM response time. On the
intermediate and high target flue gas concentration test days, the first step change was to reduce
the VOC/SVOC solution feedrate by half; the second step change was to reduce the solution
feedrate to off (zero). On the low target flue gas concentration test day only one step change was
performed. This was to decrease the VOC/SVOC solution feedrate to off (zero). As noted in
Section 2.2.2, the step changes in VOC/SVOC solution feedrate may not have resulted in step
changes in scrubber exit flue gas concentration because of the presence of the scrubber system
between the spike solution injection point and the CEM measurement location. However, even
this somewhat qualitative measure of CEM response time was of interest.
2.3 TEST CONDITIONS
The actual kiln operating conditions in effect for each test day are summarized in Table 7.
The data shown in the table for each test day represent the average over three consecutive RM
sampling periods for the day. The actual afterburner operating conditions in effect for each test
day are summarized in Table 8. Table 9 provides a similar summary of APCS operating
conditions for each test day. Combustion gas CEM data are summarized in Table 10, with the
exception of the kiln and afterburner exit gas 0-? levels; these are given in Tables 7 and 8.
The ranges and the averages of the operating condition data presented in Tables 7
through 10 were developed for the periods of RM flue gas sampling, as noted above, using the
data automatically recorded by a personal computer-based data acquisition system. Transcribed
data from the control room logs of the operating parameters, recorded at 15-minute intervals, are
given in Appendix B.
Appendix C contains graphic representations of the flue gas temperature and continuous
emission monitor data for the kiln and afterburner. Appendix C also contains graphic
presentations of the scrubber exit and stack flue gas continuous emission monitor data. These data
plots were based on incinerator system conditions at about 35-second intervals, as recorded on the
RKS data acquisition system. In addition, the duration of each RM flue gas sampling period,
major events, cumulative amounts of waste fed into the incinerator, and cumulative amounts of
ash removed from the incinerator are included in some plots. These data provide the basis of
assembling a complete picture of the actual incinerator operating conditions for each of the tests
performed.
The data in Table 7 show that clay/organic mixture feedrates were generally at, or just
under, the target feedrate of 68 kg/hr (150 Ib/hr). Average kiln and afterburner exit gas
temperatures, given in Tables 7 and 8, were right on respective targets of 871°C (1,600°F) and
1.066°C (1,950°F). Partially quenched afterburner exit flue gas average temperatures, shown in
Table 8, varied some from test day to test day, but were within the target range of 366° to 427°C
(690° to 800°F).
Several items of note appear in the combustion gas CEM data summarized in Table 10.
First, average scrubber exit NOx levels, which ranged from 106 to 152 ppm, were higher than the
18
-------
TABLE 7. KILN OPERATING DATA FOR THE CEMs TESTS
Parameter
Test Series 1: VOC CEMs
Test Series 2: Multi-metals CEMs
Low Intermediate High Low Intermediate High
concentration concentration concentration concentration concentration concentration
7/31/95 8/2/95 8/4/95 8/14/95 8/16/95 8/18/95
Fcedrates
Clay/organic mixture, kg/hr (Ib/hr)
Metals solution, kg/hr (Ib/hr)
Screw feeder
Burner
Average burner natural gas, scm/hr (scfh)
Average combustion air, scm/hr (scfh)
Average total air (including inleakage),
scm/hr (sclh)
Heat input, kW (kBtu/hr)
Natural gas
Clay/organic mixture
Total
Exit gas
Temperature, °C (°F)
Range
Average
02, %
Range
Average
65 (142)
2.6 (5.6)
6.0 (13.1)
18.1 (640)
284 (10,020)
882 (31,140)
188 (640)
191 (652)
859-881
(1,579-1,618)
871 (1,600)
12.2-15.1
14.2
64 (140)
1.9 (4.2)
6.1 (13.5)
14.7 (520)
274 (9,660)
780 (27,330)
152 (520)
188 (643)
861-883
(1,582-1,621)
871 (1,600)
11.9-16.4
13.6
67 (148)
2.1 (4.6)
6.0 (13.2)
13.5 (475)
271 (9,560)
773 (27,540)
139 (475)
199 (679)
379 (1,292) 340 (1,163) 338 (1,154)
863-879
(1,585-1,615)
870 (1,598)
12.0-13.3
12.7
78 (171)
2.1 (4.6)
6.4 (14.0)
15.3 (540)
259 (9,140)
762 (26,770)
158 (540)
230 (785)
388 (1,325)
851-885
(1,563-1,625)
871 (1,600)
9.6-17.5
12.4
63 (139)
2.1 (4.6)
6.5 (14.3)
16.2 (572)
240 (8,480)
734 (25,790)
168 (572)
187 (638)
355 (1,210)
852-894
(1,566-1,642)
871 (1,600)
9.3-14.5
11.2
65 (142)
2.3 (3.8)
6.6 (14.5)
18.4(649)
245 (8,650)
743 (26,110)
190 (649)
191 (652)
381 (1,301)
857-889
(1,575-1,632)
871 (1,600)
10.8-14.1
11.7
(continued)
-------
TABLE 7. (continued)
Test Series 3: Hg CEMs
Test Series 4: SVOC and VOC CEMs
Parameter
Low
concentration
9/1/95
Intermediate
concentration
8/28/95
High
concentration
8/30/95
Low
concentration
9/11/95
Intermediate
concentration
9/13/95
High
concentration
9/15/95
Feedrates
Clay/organic mixture, kg/hr (Ib/hr)
Melals solution, kg/hr (Ib/hr)
Screw feeder
Burner
Average burner natural gas, scm/hr (scfh)
Average combustion air, scm/hr (scfh)
Average total air (including inleakage),
scm/hr (scfh)
Heat input, kW (kBtu/hr)
Natural gas
Clay/organic mixture
Total
Exit gas
Temperature, °C (°F)
Range
Average
o2, %
Range
Average
64 (140)
2.5 (5.5)
6.9 (15.2)
13.6 (481)
275 (9,690)
775 (27,230)
141 (481)
188 (642)
329 (1,123)
865-879
(1,589-1,615)
871 (1,600)
9.6-14.7
12.3
62 (136)
2.1 (4.7)
6.6 (14.6)
13.9 (492)
264 (9,330)
770 (27,050)
144 (492)
184 (629)
328 (1,121)
853-882
(1,567-1,620)
871 (1,600)
12.1-17.1
14.4
62 (136)
2.0 (4.3)
6.5 (14.3)
13.7 (485)
264 (9,320)
760 (26,600)
142 (485)
183 (624)
325 (1,109)
863-881
(1,586-1,617)
871 (1,600)
11.2-15.5
13.1
60 (133)
2.3 (5.0)
6.8 (15.0)
15.4 (543)
287 (10,120)
1,006 (35,530)
159 (543)
179 (610)
338 (1,153)
846-886
(1,554-1,626)
871 (1,600)
13.4-16.2
14.8
62 (136)
2.3 (5.0)
6.3 (13.8)
14.7 (518)
262 (9,240)
868 (30,660)
152 (518)
183 (624)
335 (1,142)
860-888
(1,580-1,630)
871 (1,600)
12.1-15.2
13.4
64 (140)
2.1 (4.5)
6.6 (14.4)
15.3 (541)
261 (9,230)
998 (35,250)
159 (541)
190 (648)
348 (1,189)
862-884
(1,583-1,623)
871 (1,600)
13.0-17.0
14.7
-------
TABLE 8. AFTERBURNER OPERATING DATA FOR THE CEMs TESTS
Test Series 1: VOC CEMs
Test Series 2: Multi-metals CEMs
Parameter
Low
concentration
7/31/95
Intermediate
concentration
8/2/95
High
concentration
8/4/95
Low
concentration
8/14/95
Intermediate
concentration
8/16/95
High
concentration
8/18/95
Feedrates
Average natural gas, scm/hr (scfh)
Average combustion air, scm/hr (scfh)
Organic spike solution, kg/hr (Ib/hr)
Heat input, kW (kBtu/hr)
Natural gas
Afterburner exit gas
Temperature, °C (°F)
Range
Average
o2, %
Range
Average
Afterburner extension exit gas (post
partial quench)
Temperature, °C (°F)
Range
Average
27.3 (963)
282 (9,970)
0.73 (1.6)
282 (963)
1,061-1,075
(1,941-1,967)
1,069 (1,956)
9.0-11.2
10.3
238-437
(460-818)
424 (795)
28.2 (996)
262 (9,260)
0.77 (1.7)
292 (996)
1,060-1,076
(1,940-1,968)
1,068 (1,955)
9.8-16.1
12.4
384-459
(724-859)
417 (783)
26.8 (947)
260 (9,170)
0.77 (1.7)
277 (947)
1,062-1,076
(1,944-1,968)
1,068 (1,955)
9.2-10.5
9.8
404-434
(760-814)
416 (781)
26.8 (947)
243 (8,580)
0.77 (1.7)
277 (947)
1,058-1,073
(1,936-1,963)
1,069 (1,956)
10.0-14.4
12.6
352-451
(665-844)
400 (752)
27.4 (966)
244 (8,602)
0.82 (1.8)
283 (966)
1,062-1,081
(1,943-1,977)
1,069 (1,956)
10.4-12.5
11.3
374-453
(706-847)
416 (780)
27.6 (974)
245 (8,640)
0.77 (1.7)
285 (974)
1,061-1,077
(1,942-1,971)
1,069 (1,956)
10.8-14.5
12.2
372-443
(701-829)
410 (770)
(continued)
-------
TABLE 8. (continued)
Test Series 3: Hg CEMs
Test Series 4: SVOC and VOC CEMs
Parameter
Low
concentration
9/1/95
Intermediate
concentration
8/28/95
High
concentration
8/30/95
Low
concentration
9/11/95
Intermediate
concentration
9/13/95
High
concentration
9/15/95
Feedrates
Average natural gas, scm/hr (scfh)
Average combuslion air, scm/hr (scfh)
Organic spike solution, kg/hr (Ib/hr)
Heat input, kW (kBtu/hr)
Natural gas
Afterburner exit gas
Temperature, °C (°F)
Range
Average
o2, %
Range
Average
Afterburner extension exit gas (post
partial quench)
Temperature, °C (°F)
Range
Average
27.3 (964)
264 (9,320)
0.77 (1.7)
282 (964)
1,061-1,078
(1,942-1,972)
1,069 (1,956)
12.7-14.8
13.6
350-383
(662-722)
383 (722)
29.0 (1,025)
256 (9,060)
0.77 (1.7)
300 (1,025)
1,062-1,075
(1,944-1,967)
1,069 (1,956)
12.5-13.5
12.9
356-383
(673-722)
366 (691)
26.8 (946)
258 (9,100)
0.77 (1.7)
277 (946)
1,062-1,077
(1,943-1,971)
1,069 (1,956)
10.9-13.6
11.8
363-387
(686-728)
377 (711)
27.8 (980)
259 (9,160)
0.73 (1.6)
287 (980)
1,061-1,078
(1,941-1,972)
1,069 (1,956)
11.9-13.0
12.4
356-390
(682-734)
373 (703)
27.3 (965)
259 (9,160)
0.86 (1.9)
283 (965)
1,058-1,079
(1,936-1,975)
1,069 (1,956)
11.3-15.1
12.4
367-410
(692-770)
389 (733)
26.9 (950)
259 (9,160)
0.73 (1.6)
278 (950)
1,060-1,077
(1,940-1,970)
1,069 (1,956)
10.4-12.1
11.4
352-394
(666-742)
375 (707)
-------
TABLE 9. AIR POLLUTION CONTROL SYSTEM
Test Series 1: VOC CEMs
Test Series 2: Multi-metals CEMs
Parameter
Low
concentration
7/31/95
Intermediate
concentration
8/2/95
High
concentration
8/4/95
Low
concentration
8/14/95
Intermediate
concentration
8/16/95
High
concentration
8/18/95
Flowrates, L^min (gpm)
Average quench liquor 82 (22) 83 (22) 79 (21)
Average venturi scrubber liquor 19 (5) 19 (5) 19 (5)
Average packed-column scrubber liquor 132 (35) 114 (30) 114 (30)
Temperatures, °C (°F)
Average scrubber inlet gas 78 (173) 78 (173) 78 (173)
Average scrubber exit gas 74 (166) 74 (166) 73 (164)
Average scrubber liquor 74 (166) 74 (166) 73 (164)
Scrubber liquor pH
Range 6.2-7.0 6.4-6.7 6.5-6.6
Average 6.6 6.5 6.5
79 (21)
19(5)
114 (30)
77 (171)
70 (158)
72 (161)
6.2-6.9
6.5
83 (22)
19(5)
121 (32)
78 (173)
71 (160)
72 (161)
6.5-6.6
6.5
83 (22)
19(5)
114 (30)
78 (173)
71 (160)
72(161)
6.2-6.7
6.5
(continued)
-------
TABLE 9. (continued)
Test Series 3: Hg CEMs Test Series 4: SVOC and VOC CEMs
Low Intermediate High Low Intermediate High
concentration concentration concentration concentration concentration concentration
Parameter 9/1/95 8/28/95 8/30/95 9/11/95 9/13/95 9/15/95
Flowrates, L/min (gpm)
Average quench liquor
Average venturi scrubber liquor
Average packed-column scrubber liquor
Temperatures, °C (°F)
Average scrubber inlet gas
Average scrubber exit gas
Average scrubber liquor
Scrubber liquor pll
Range
Average
79 (21) 83 (22) 79 (21)
26 (7) 11 (3) 34 (9)
114 (30) 117 (31) 121(32)
76 (169) 76 (169) 76 (169)
68 (155) 69 (156) 68 (155)
71(160) 72 (162) 72 (162)
6.4-6.6 6.4-6.6 6.4-6.6
6.5 6.5 6.5
79 (21) 79 (21) 76 (20)
34 (9) 30 (8) 34 (9)
121(32) 121(32) 117 (31)
76 (169) 77 (170) 77 (170)
69 (156) 70 (158) 70 (158)
71 (160) 71 (160) 71 (160)
6.5-6.6 6.4-6.6 6.4-6.6
6.6 6.5 6.5
-------
TABLE 10. COMBUSTION GAS CEM DATA
Test Series 1: VOC CEMs
Test Series 2: Multi-metals CEMs
Low
Intermediate
High
Low
Intermediate
High
concentration
concentration
concentration
concentration
concentration
concentration
Parameter
7/31/95
8/2/95
8/4/95
8/14/95
8/16/95
8/18/95
Afterburner exit
co2, %
Range
6.6-8.1
3.4-7.7
7.3-8.3
4.5-7.5
5.6-7.3
3.2-7.4
Average
7.3
6.0
7.8
5.8
6.8
6.1
NOx, ppm
Range
189-266
67-337
191-242
127-219
123-227
88-170
Average
221
182
203
170
165
132
Scrubber exit
o2, %
Range
13.2-14.2
13.2-14.1
13.2-14.1
12.6-13.8
12.1-13.4
13.0-14.1
Average
13.5
13.6
13.6
13.2
12.8
13.4
CO,, %
Range
3.6-4.9
4.3-5.1
4.7-5.1
4.1-5.6
4.8-5.7
4.1-5.1
Average
4.6
4.7
4.9
5.1
5.3
4.8
CO, ppm
Range
6-87
26-85
45-58
34-103
38-99
37-59
Average
42
53
52
38
54
48
NOx, ppm
Range
69-133
65-327
120-146
91-248
78-220
51-130
Average
115
139
132
141
106
112
TUHC, ppm
Range
111-206
99-284
140-245
14-255
110-208
112-299
Average
174
187
220
193
171
219
Stack
o2, %
Range
12.0-14.3
12.0-12.8
12.1-13.0
12.0-12.8
11.6-12.7
11.9-13.7
Average
12.5
12.4
12.4
12.3
12.2
12.3
CO, ppm
Range
7-108
33-102
46-60
46-107
37-99
36-63
Average
49
63
54
71
54
50
(continued)
-------
TABLE 10. (continued)
Test Series 3: Hg CEMs
Test Series 4: SVOC and VOC CEMs
Low
Intermediate
High
Low
Intermediate
High
concentration
concentration concentration
concentration
concentration
concentration
Parameter
9/1/95
8/28/95
8/30/95
9/11/95
9/13/95
9/15/95
Afterburner exit
n
o
i-j
£
Range
1.4-6.0
5.5-6.5
5.2-7.5
5.4-6.5
4.0-7.1
6.5-7.9
Average
5.2
6.1
6.7
6.0
6.0
7.1
NOx, ppm
Range
29-160
93-215
147-241
85-235
56-147
66-164
Average
131
190
208
115
94
125
Scrubber exit
o2, %
Range
12.6-15.3
12.5-14.1
12.4-14.3
12.2-14.0
13.0-14.1
12.5-14.3
Average
13.3
13.0
13.2
12.7
13.4
13.3
o
u
Range
3.9-5.9
4.4-6.0
2.1-5.9
4.4-5.6
4.1-5.3
4.3-5.5
Average
4.8
5.4
5.1
5.2
4.8
5.0
CO, ppm
Range
35-61
38-57
34-54
39-56
40-56
30-51
Average
45
50
48
50
47
38
NOx, ppm
Range
63-158
77-177
98-255
87-153
90-140
57-140
Average
127
152
144
129
116
116
TUHC, ppm
Range
60-300
74-287
156-268
90-214
161-234
169-223
Average
171
199
207
177
217
197
Stack
02, %
Range
12.7-15.8
12.4-14.3
12.3-14.3
12.4-14.1
11.9-13.7
12.1-13.5
Average
13.4
12.8
12.8
12.7
12.6
12.4
CO, ppm
Range
28-61
35-55
34-54
38-53
43-58
32-49
Average
45
48
48
49
51
41
-------
70 ppm or less usually experienced. The higher NOx levels in these tests were the result of
introducing the trace metals to the system in what was a very concentrated nitrate solution. In
fact, the generally lower scrubber exit NOx levels, compared to afterburner exit levels, are
consistent with scrubber system removal of some flue gas N02- The predominant NOx species
originating from the nitrate metals spike solution would be N02, which is more water-soluble than
the predominant combustion-generated NOx species, NO. During periods of no metals spike
solution feed to the RKS, flue gas NOx levels indeed reverted to typical lower concentrations.
In addition, the average flue gas CO and total unburned hydrocarbon (TUHC) levels
experienced in these tests were also higher than typically seen from the RKS. Average scrubber
exit CO concentrations from these tests ranged from 38 to 54 ppm, compared to typical levels of
10 ppm or less. Average scrubber exit TUHC levels ranged from 171 to 220 ppm, compared to
typical levels of 1 ppm. The atypicaliy high CO and TUHC concentration resulted from the
organic spike solution added to the partially quenched afterburner exit flue gas. The TUHC
consisted of the spiked VOC and SVOC constituents, along with their methanol solvent carrier,
injected into the partially quenched flue gas. The CO was most likely partially decomposed
methanol. Both CO and TUHC concentrations reverted to the much lower typically-experienced
levels during periods when no flue gas organic spike solution was being fed.
It was also observed that scrubber exit flue gas CO levels were quite sensitive to the
afterburner extension exit gas (after partial quenching) temperatures. As this temperature
increased, CO levels increased. At partial quench exit temperature above 482°C (900°F), CO
levels would increase to several hundred ppm.
27
-------
SECTION 3
SAMPLING AND ANALYSIS PROCEDURES
The RKS sampling locations and the scope of the sampling effort are shown in the
process schematic, Figure 3. The sampling effort performed is discussed in Section 3.1, followed
by a discussion of the sample analysis procedures in Section 3.2.
3.1 SAMPLING PROCEDURES
For all tests, the sampling matrix included:
• Obtaining a composite sample of the clay/organic feed to the kiln
• Continuously measuring 02 concentrations in the kiln exit flue gas; 02, C02, and
NOx concentrations in the afterburner exit flue gas; 02, CO, C02, NOv and TUHC
concentrations in the scrubber exit flue gas; and O-, and CO concentrations in the
stack gas
• Sampling the baghouse exit flue gas for polychlorinated dibenzo-p-dioxins and
polychlorinated dibenzofurans (PCDDs/PCDFs) using Method 23 (Reference 2)
• Sampling the stack gas for particulate and HCI using a modification to the Method
0050 (Reference 2)
For all tests, except those evaluating VOC CEMs, the baghouse exit flue gas was also
sampled for VOCs using Method 0030, the volatile organic sampling train (VOST) (Reference 3).
All of these sampling procedures supplied non-critical measurement data.
The critical measurement sampling procedures were:
• For tests of the SVOC CEMs in Test Series 4, sampling the scrubber exit flue gas
for SVOCs using Method 0010 (Reference 3). Three sequential Method 0010
samples were collected each test day except for the third test day. The SVOC CEM
being tested could not be brought into operation on this day, so only one Method
0010 sample was collected.
28
-------
s
fi.
8
s
Ui
FAN
AFTER-
BURNER
KILN
HEPA
FILTER
FLUE GAS
QUENCH
FLUE GAS
REHEAT
CARBON
BED
VENTURI
SCRUBBER
BAGHOUSE
PACKED
COLUMN
SCRUBBER
1 2 3 4 5 6
Continuous monitors
Flue gas
Sampling point
Clay/organic
mixture feed
°2
CO
co2
NOx
Heated
TUHC
Method 0010,
SVOCs
Method
0030, VOCs
Method 29,
trace metals
(including
mercury)
Method 23,
PCDDs/PCDFs
Method 0050,
particulate
and HCI
I. Kiln screw feeder
X
2. Kiln exit Hue gas
X
3. Afterburner exit flue gas
X
X
X
4. Scrubber exit flue gas
X
X
X
X
X
S
V
M
5. Baghouse exit flue gas
S,M
X
6. Slack gas
X
X
X
X = All tests; S = SVOC CEM tests; V = VOC CEM tests; M = multi-metal and mercury CEM tests.
Figure 3. Test sampling locations.
-------
• For tests of VOC CEMs in Test Series 1 and 4, sampling the scrubber exit flue gas
for VOCs using Method 0030 (Reference 3); three sequential Method 0030 samples
were collected each test day.
• For tests of multi-metal CEMs in Test Series 2 and mercury CEMs in Test Series 3,
sampling the scrubber exit flue gas for trace metals or mercury using Method 29
(Reference 2); three sequential Method 29 samples were collected each test day.
Details of the various sampling procedures performed are discussed in the following
subsections.
3.1.1 Feed Sampling
The composite clay/organic mixture feed sample for each test day was needed to support
the calculation of POHC DRE required by the IRF hazardous waste management permit.
Individual feed mixture samples were collected during all feed hopper charging processes at both
the beginning of each test day and the periodic recharging performed at various times during a test
day. Each drum of clay/organic feed mixture used to charge the feed hopper was thief-sampled
at a location near the center of the drum cross section. This sample was collected just before the
drum contents were charged to the screw feeder hopper. Each collected sample was placed in a
separate container which was filled to contain minimum headspace and sealed with its screw cap
closure.
After all drums charged to the feed hopper on a given test day were sampled, the collected
samples from each individual drum were taken to the onsite laboratory. There the individual drum
samples were combined into one composite sample representing the test day's feed clay/organic
mixture. This composite sample was placed in a larger container, also filled to minimum
headspace, which was sealed and placed into refrigerated storage until use for analysis. All feed
sample handling activities were performed in a manner conducive to minimizing volatiles loss.
3.1.2 Method 0030 Train Sampling
Method 0030 VOC sampling was performed for all tests. Each Method 0030 sampling
consisted of collecting four pairs of sorbent Craps over a 2- to 3-hour period. Four trap pairs were
collected as insurance against trap breakage.
On Test Series 2 and 3 test days, one Method 0030 train sample of the baghouse exit flue
gas (consisting of four sequential sorbent trap pairs) was collected over a convenient time period
during the day that minimized conflicts with the critical RM measurements. Each trap pair
collected nominally 20 L of sampled gas over a 20-minute period.
For the VOC CEM test days in Test Series 1 and 4, three sequential Method 0030 train
samples (each consisting of four sequential sorbent trap pairs) were collected over the 6- to 9-hour
period comprising the test day. Each trap pair collected nominally 20 L of sampled gas over a
20-minute period, except for the high target flue gas VOC concentration tests. For the high target
30
-------
flue gas VOC concentration tests, each trap pair collected 10 L of sampled gas over a 20-minute
time period. This lower sample volume was used because evidence of VOC breakthrough to the
second VOST trap was observed during scoping tests at the high target VOC concentration.
3.1.3 Method 0010 Train Sampling
All Method 0010 trains were operated to support the SVOC CEM tests in Test Series 4.
For the first two test days of this series, three Method 0010 train samples were collected per day.
Only one Method 0010 sample was collected on the third test day due to problems with the SVOC
CEM being tested. Because the CEM could not be brought into operation on this test day, there
was no need for parallel RM data, so the second and third planned Method 0010 samples were not
collected. Each Method 0010 train collected nominally 2.8 dscm (100 dscf) of sampled gas over
a 2.5- to 3-hour period.
3.1.4 Method 29 Train Sampling
Three Method 29 trains were operated per day for the three days of the multi-metals CEM
tests in Test Series 2 and the three days of the mercury CEM tests in Test Series 3. Eight of nine
multiple metals trains collected nominally 2.8 dscm (100 dscf) of sampled gas over a 2.5- to 3-
hour period. One of the nine trains developed a leak and pulled ambient air during the last
30 minutes of operation. The volume and sampling time were adjusted appropriately for this train.
All other trains were operated and recovered per the method.
3.1.5 Method 23 Train Sampling
One Method 23 train was operated at the baghouse exit for each test day of this series.
Each Method 23 train collected nominally 2.8 dscm (100 dscf) of sampled gas over a 2.5- to
3-hour period. During each test day the Method 23 sample was collected over a convenient time
period during the day that minimized conflicts with the critical RM measurements.
3.1.6 Method 0050 Train Sampling
One Method 0050 train was operated at the stack for each test day of this series. This
procedure was performed as a requirement of the IRF's hazardous waste management permit, as
noted above. The Method 0050 trains for particulate and HC1 collection included only dilute
caustic-filled impingers (0.1 N NaOH). The dilute sulfuric acid-filled impingers (0.1 N H2S04)
specified in the method were not used. This approach prevents the separate measurement of HC1
and Cl2 and gives only a combined (HCI and Cl2) measurement. However, this conservative
estimate of HCI concentrations satisfies the IRF compliance requirements. Each Method 0050
collected nominally 1.4 dscm (50 dscf) of sampled gas over a 1.0- to 1.5-hour period. During
each test day the Method 0050 was collected over a convenient time period during the test day that
minimized conflicts with the critical RM measurements.
31
-------
3.1.7 Combustion Gas CEM Measurements
During all tests, the IRF's complement of combustion gas CEMs were in continuous
operation. The combustion gas CEMs available at the IRF and the locations that they monitored
during all tests are summarized in Table 11. This monitoring arrangement was employed in all
tests. Figure 4 illustrates the generalized flue gas conditioning and flow distribution system at the
IRF. Four independent systems, such as the one illustrated in Figure 4, were in place so that the
appropriately conditioned sample gas from the four separate locations could be routed to the
monitors listed in Table 11. The combustion gas CEM setup described in Table 11, with
appropriate gas conditioning per Figure 4, was employed throughout this test program.
Combustion gas CEM data were recorded continuously on strip charts and also by an automatic
data acquisition system.
3.2 ANALYTICAL PROCEDURES
Table 12 summarizes the analytes determined in each test program sample, and the
analysis procedures used. Details of the analytical procedures performed are discussed in the
following subsections.
3.2.1 Method 0010 Train Sample Analyses by GC/MS
As indicated in Table 12, all Method 0010 samples were analyzed for the PAH SVOCs,
with sample preparation performed according to the method, and final GC/MS analysis by Method
8270B. The PAH SVOCs listed in Table 13 were quantitated. Table 13 lists several more PAH
compounds than were spiked into the RKS flue gas. The reason for this is that the SVOC CEM
tested gives only a measure of total PAH concentration. Several PAH compounds other than the
three added to the afterburner extension flue gas might have been present in the scrubber exit at
low levels as incomplete combustion byproducts. By quantitating the expanded list of PAH
analytes in Table 13, a better measure of total PAH concentration could be obtained, as the
contribution from incomplete combustion byproducts would be included.
The SVOC analyses were performed in the IRF onsite laboratories.
3.2.2 Method 0030 Train Sample Analyses by GC/FID
As noted in Section 3.1.2, four Method 0030 trap pair samples were collected for each
procedure run. Three of these trap pair samples were analyzed for VOCs by thermal desorption,
purge and trap according to Method 5040A with gas chromatography/flame ionization detector
(GC/FID) analysis by a modified Method 8015A. The fourth trap pair was collected for breakage
contingency, so that the probability that three trap pair analyses could be done would be increased,
given the inevitability of trap breakage. The list of VOC analytes was comprised of the 10
compounds, listed in Table 6, that were spiked into the RKS flue gas.
Method 8015A is specifically documented for the analysis of nonhalogenated volatile
organics (specifically selected oxygenated solvents) in water samples. The modifications to the
32
-------
TABLE 11. CONTINUOUS EMISSION MONITORS AVAILABLE AND LOCATIONS
MONITORED
Monitor
Location Constituent Manufacturer Model
Principle
Range
Kiln exit 02
Afterburner 02
exit
CO-,
Scrubber exit Oo
Stack
CO,
NO,
CO
TUHC
O,
CO
Beckman
755
Rosemount 755
Paramagnetic
Paramagnetic
Horiba
PIR 2000 NDIRa
NOx Thermo Electron 10 AR Chemiluminescent
Beckman
Horiba
755
Paramagnetic
PIR 2000 NDIR
Thermo Electron 10 AR Chemiluminescent
Horiba
VIA 500 NDIR
Thermo 51
Environmental
Rosemount 755
FID0
Paramagnetic
Horiba
VIA 500 NDIR
0-10 percent
0-25 percent
0-100 percent
0-10 percent
0-25 percent
0-100 percent
0-20 percent
0-80 percent
0-75 ppm to
0-10,000 ppm in
multiples of 2
0-10 percent
0-25 percent
0-100 percent
0-20 percent
0-80 percent
0-75 ppm to
0-10,000 ppm in
multiples of 2
0-50 ppm
0-500 ppm
0-10 ppm
0-100 ppm
0-1,000 ppm
0-10,000 ppm
0-10 percent
0-25 percent
0-100 percent
0-50 ppm
0-500 ppm
aNDIR = Nondispersive infrared.
be
FID = Flame ionization detector.
33
-------
FILTER
PUMP
FILTER
HIGH
BAY
HEATED
FILTERS 1
CHILLED WATER
IMPING ER
CONDENSATE
REMOVALE
J
L__/00(JL*_ SAMPLE
SAMPLE
PORT
STORE ROOM
PUMP
n:
AIR COOLED
COIL
PORT
calibration
GAS
IS
CONTROL
ROOM
VENT
HEATED
SAMPLE LINE
n
ixK
VENT
PERMA PURE
DRYER
SAMPLE GAS MANIFOLD
HEATED
TUNC
MONITOR
TIT
ROTA-
ROTA-
ROTA-
ROTA-
ROTA-
METER
METER
METER
METER
METER
O2 CO CO2 SPARp
MONITOR MONITOR MONITOR MONITOR
Figure 4. Generalized CEM gas flow schematic.
34
-------
TABLE 12. ANALYSIS PROCEDURES
Sample Analytes Analysis method Number of analyses
Scrubber exit flue gas
Method 0010 train
SVOCs
Extraction of train samples by Method
3510B/3540B, GC/MS analysis by Method
8270B1
3/test day for the SVOC CEM
test series
Scrubber exit flue gas
Method 0030 train
VOCs
Thermal desorption, purge and trap of sorbent
traps by Method 5040A, GC/FID analysis by
Method 8015Aa,b
3 trap pairs/run, 3 runs/test
day for each VOC CEM test
series
Baghouse exit flue gas
Method 0030 train
VOCs
Thermal desorption, purge and trap of sorbent
traps by Method 5040A, GC/FID analysis by
Method 8015Aa'b
3 trap pairs/test day for the
multi-metals and mercury
CEM test series
Scrubber exit flue gas
Method 29 train
Trace metals0
Microwave-assisted digestion of train samples per
Method 29d; ICP or GFAAS analysis by Method
601 OA, 7041, or 7841a
3/test day for the multi-metals
CEM test series
Mercury
Preparation of train samples per Method 29^,
CVAAS analysis by Method 7470A3
3/test day for the multi-metals
and mercury CEM test series
Baghouse exit flue gas
Method 23 train
PCDDs/PCDFs
Extraction and HRGC/HRMS analysis of train
samples by Method 23d
1 /test day for each test series
Stack gas
Method
-------
method for use in this project included adapting the method to the analysis of Method 0030 train
samples, and validating the method for the VOC analytes in Table 6. Method details are given
in the project QAPP (Reference 8). The method was selected for use in this project, instead of
a GC/MS method such as the adaptation of Method 8240A included in Method 5040A, because
it has an extended dynamic range and it is routinely performed at lower project cost in the IRF
onsite laboratories, where the analyses were done.
3.2.3 Method 29 Train Sample Analyses by ICP and AAS
Method 29 train samples collected during the multi-metal CEM tests in Test Series 2 were
analyzed for all 14 spiked trace metals. Method 29 train samples collected during the mercury
CEM tests in Test Series 3 were analyzed for mercury only. Sample preparation for trace metal
analyses was performed according to Method 29 with final analysis by inductively coupled plasma
spectroscopy (ICP) or graphite furnace atomic absorption spectroscopy (GFAAS) per the methods
listed in Table 14. Sample preparation for mercury analyses was according to Method 29, with
final cold vapor atomic absorption spectroscopy (CVAAS) analysis by Method 7470A.
The Method 29 train sample analyses were performed by Triangle Laboratories in
Research Triangle Park, North Carolina.
3.2.4 Noncritical Analyses
Flue gas HCI levels were determined by analyzing the combined Method 0050 train
impinger solutions for chloride via ion chromatography (IC) according to Method 9057.
Method 23 samples were analyzed for PCDDs and PCDFs. Sample preparation and high-
resolution gas chromatography/high-resolution mass spectrometry (HRGC/HRMS) were performed
according to Method 23.
TABLE 14. TRACE METALS QUANTITATED IN METHOD 29
TRAIN SAMPLES
Analysis
Analysis
Trace metal
method
Trace metal
method
Antimony (Sb)
7041 (GFAAS)
Lead (Pb)
601 OA (ICP)
Arsenic (As)
601 OA (ICP)
Manganese (Mn)
601 OA (ICP)
Barium (Ba)
601 OA (ICP)
Nickel (Ni)
601 OA (ICP)
Beryllium (Be)
601 OA (ICP)
Selenium (Se)
601 OA (ICP)
Cadmium (Cd)
601 OA (ICP)
Silver (Ag)
601 OA (ICP)
Chromium (Cr)
601 OA (ICP)
Thallium (Tl)
7841 (GFAAS)
Cobalt (Co)
601 OA (ICP)
36
-------
The clay/organic mixture feed samples were analyzed for the three feed VOCs. Sample
preparation and introduction were by the high level soil procedure of Method 5030A. Analyses
were by the modified Method 8015A discussed in Section 3.2.2.
The chloride analyses were performed in the IRF onsite laboratories. The PCDD/PCDF
analyses were performed by Triangle Laboratories in Research Triangle Park, North Carolina.
37
-------
SECTION 4
TEST RESULTS
The results of the critical measurements of the test program are discussed in this section.
As noted in Section 1, the critical measurements of the test program consisted of obtaining
quantitative measurement data on four measurements of CEM performance checked in a RATA
of a CEM as described in 40 CFR 60, Appendix F. These measures are RA, CD, ZD, and
response time.
The primary objective of the test program was to measure the RA of each CEM tested.
Secondary objectives included obtaining CD, ZD, and response time (if possible) data for each
CEM. In addition, the DL of each CEM was estimated. The results of the test program are
discussed in the following subsections by analyte class. Thus, Section 4.1 presents the RM and
CEM results for the VOC CEM tests. Section 4.2 discusses the SVOC (PAH) comparisons.
Section 4.3 presents the RM and CEM results for the multi-metal CEM tests. Section 4.4 presents
the RM and CEM results for the mercury CEM tests.
For the reader interested in studying the reference method analytical results in more detail,
the analytical laboratory reports are given in Appendix E. The detailed reference method data
reduction worksheets are presented in Appendix F. The CEM developers' daily and final reports,
as received, are included in Appendix H.
4.1 VOC CEM TESTS
VOC CEM testing was performed during Test Series 1 and 4 of the test program (see
Table 3). The participating CEM developers during Test Series 1 were Oak Ridge National
Laboratory (ORNL) and EcoLogic International, Inc. The EPA's Air Pollution Prevention and
Control Division (APPCD) of NRMRL and Marine Shale Processors (MSP) participated during
Test Series 4.
As noted in Section 2.2, each day of VOC CEM testing consisted of performing three
sequential RM measurements while the tested CEMs were in operation. Each week of testing
comprising a test series consisted of three test days at each of three different VOC concentrations
(low, intermediate, and high). The results of Test Series 1 and 4 are discussed separately below.
Specifically, Section 4.1.1 presents the test results for the ORNL and EcoLogic CEMs tested in
Test Series 1, and Section 4.1.2 discusses the test results for the EPA APPCD and MSP CEMs
tested in Series 4. In both sections, CEM measurement results are compared to RM measurement
38
-------
results for all 10 VOC target analytes. In proceeding through this discussion, the reader should
keep in mind the fact that the 1,1-dichIoroethene concentrations arising from the RM
measurements need to be treated with caution. The quantitation of 1,1-dichloroethene in RM
samples was problematic, as discussed in Section 6.1. The conclusion from the Section 6.1
discussion is that reported RM concentrations greater than 30 jig/dscm are most likely
representative of actual flue gas concentrations present. However, reported RM concentrations less
than this need to be viewed with caution.
4.1.1 Test Series 1 — ORNL and EcoLogic VOC CEMs
Detailed descriptions of the principles of operation of the ORNL and EcoLogic CEMs are
included in their respective final reports (Appendices H-1-1 and H-l-2). Briefly, the ORNL CEM
uses a direct sampling ion trap mass spectrometry (DSITMS) technique with the collection of flue
gas VOCs on a sorbent trap, and analysis of the collected material by thermal desorption into the
DSITMS instrument.
The EcoLogic CEM, developed by V&F Analyse-und Messtechnik of Absam, Austria,
and marketed in North America as the Airsense 500 Continuous Chemical Ionization Mass
Spectrometer, uses a real-time mass spectrometry technique. The Airsense 500 produces a well-
defined ion beam that carries a precise internal energy level (the ionization potential) such that the
interaction of this low-energy ion beam with neutral molecules in the sample gas produces defined
ion products with minimal or no fragmentation.
Tables 15 through 17 present the results of the three sequential RM measurements, along
with the ORNL and EcoLogic CEM results, for each of the three VOC concentrations tested in
TABLE 15. MEASURED FLUE GAS CONCENTRATIONS FOR THE TESTS OF THE
ORNL AND ECOLOGIC CEMS AT THE LOW VOC CONCENTRATION
Concentration, fig/dscm
Reference Method 1 Reference Method 2 Reference Method 3
Compound RM ORNL EcoLogic RM ORNL EcoLogic RM ORNL EcoLogic
Benzene
32.4
1.3
NO3
41.9
1.6
NO
59.6
1.6
NO
Carbon tetrachloride
31.2
<0.4
NO
34.2
<0.4
NO
38.0
0.92
NO
Chlorobenzene
55.6
0.76
NO
49.2
1.2
NO
86.7
5.6
NO
Chloroform
40.8
0.56
NO
47.3
0.4
NO
41.6
3.6
NO
1,2-Dichloroethane
2.4
2.5
NO
3.3
1.5
NO
2.6
6.8
NO
1,1-Dichloroethene
86.4
3.2
NO
35.6
<0.4
NO
16.9
11.0
NO
Tetrachloroethene
89.9
1.7
NO
73.9
1.3
NO
126
4.3
NO
Toluene
352
9.2
NO
316
9.2
NO
462
16
NO
1,1.1 -Trichloroethone
2.5
<0.4
NO
4.6
<0.4
NO
6.4
<0.4
NO
Trichloroethene
69
<0.4
NO
5.9
<0.4
NO
3.9
<0.4
NO
aNO = CEM not operational.
39
-------
TABLE 16. MEASURED FLUE GAS CONCENTRATIONS FOR THE TESTS OF
THE ORNL AND ECOLOGIC CEMS AT THE INTERMEDIATE VOC
CONCENTRATION
Concentration, ng/dscm
Reference Method 1 Reference Method 2 Reference Method 3
Compound RM ORNL EcoLogic RM ORNL EcoLogic RM ORNL EcoLogic
Benzene
32.7
12.0
97
28.7
<2.3
820
36.4
5.5
870
Carbon tetrachloride
46.9
10.1
7.9
41.7
3.8
24
58.5
6.4
16
Chlorobenzene
59.8
25.8
81
46.3
7.2
81
74.3
27.6
98
Chloroform
57.1
23.9
140
56.7
9.6
230
66.3
18.4
170
1,2-Dichloroethanc
17.4
43.3
210
12.3
16.6
340
15.3
29.5
290
1,1-Dichloroethcne
24.0
55.3
320
20.5
26.7
430
14.3
40.5
370
Tetrachioroethene
81.4
11.1
120
64.1
2.5
770
101
7.4
710
Toluene
342
147
210
218
71.8
120
413
103
250
1,1,1-Tnchloroethane
13.8
<2.3
800
13.3
<2.3
910
12.9
<2.3
840
Trichloroethene
19.4
4.6
420
18.6
0.9
770
20.4
1.8
510
TABLE 17. MEASURED FLUE GAS CONCENTRATIONS FOR THE TESTS OF THE
ORNL AND ECOLOGIC CEMS AT THE HIGH VOC CONCENTRATION
Concentration, pg/dscm
Reference Method 1 Reference Method 2 Reference Method 3
Compound RM ORNL EcoLogic RM ORNL EcoLogic RM ORNL EcoLogic
Benzene
102
36.8
140
89.1
50.7
160
91.3
28.6
190
Carbon tetrachloride
423
101
380
409
119
350
446
76.4
360
Chlorobenzene
337
138
250
299
170
270
269
97.6
280
Chloroform
417
101
330
411
168
350
413
91.2
390
1,2-Dichloroethane
184
88.4
450
174
114
500
183
6.3
540
1,1-Dichloroethene
116
38.7
350
140
44.2
440
162
35.9
480
Tetrachioroethene
429
62.6
690
374
61.7
740
324
37.8
690
Toluene
1,760
847
1.300
1.393
921
1200
1.024
460
760
1,1,1-Trichloroethane
175
24.9
170
164
37.8
190
182
20.3
210
Trichloroethene
189
15.7
340
176
19.3
300
185
14.7
360
40
-------
Test Series 1. (The reader is reminded of the caution to suspect the four RM measurements of
1,1-dichloroethene concentration less than 30 fig/dscm.) The EcoLogic CEM data for the first day
of testing at the low VOC concentration were not reported in EcoLogic's test report due to
operator error that resulted in inflated and incorrect CEM readings. A more detailed explanation
can be found in the EcoLogic test report, included in Appendix H-l-2. In addition, during the low
and intermediate VOC concentration test days, the concentrations of several analytes were below
the reporting limit of the ORNL CEM. These results are noted with "less than" signs (<) in
Tables 15 and 16. MSP was an original participant in Test Series 1. However, during this series,
they were unable to bring their instrument into operation. Because access space was available for
another CEM in Test Series 4, MSP was afforded the opportunity to return and participate in that
series.
Qualitative review of the RM measurement data in Tables 15 through 17 shows that the
measured flue gas concentrations for the low target concentration test (Table 15) were substantially
greater than the target concentrations outlined in Section 2.2.3 for many, if not most, of the VOC
analytes. The data in Table 18 substantiate this observation. The table shows a comparison of the
average RM concentration for each test day to the nominal target concentration for the test day.
The data in Table 18 show that, with the exception of 1,2-dichloroethane, 1,1,1-trichloroethane,
and trichloroethene, the flue gas VOC concentrations from the RM measurement were roughly an
order of magnitude or more higher than the target levels. The measured toluene concentration was
two orders of magnitude greater than its target. In fact, again with the exception of 1,2-
dichloroethane, 1,1,1-trichloroethane, and trichloroethene, the measured concentrations for the low
target concentration test were quite comparable to the corresponding concentrations for the
intermediate concentration test.
All of the VOCs present at much higher than target levels for the low concentration test
are commonly encountered products of incomplete combustion (PICs) from the incineration of
chlorinated organics. However, the IRF experience has been that these compounds are present in
the scrubber exit flue gas at concentrations in the range of less than 1 to about 10 |jg/dscm during
the incineration of a feed mixture similar to the one used for these tests. And, while toluene is
perhaps the most common PIC from organic waste incineration, measured flue gas concentrations,
in the absence of spiking as performed in these tests, are rarely more than a few tens of jjg/dscm,
based on extensive experience at the IRF.
Because the VOC analysis method used does not give confirmation that a given
chromatographic peak is indeed one of the target VOCs, the method may be subject to positive
interferences for other flue gas unburned or partially burned hydrocarbons. Thus, one could
speculate that the higher than expected VOC concentrations measured for the low concentration
test were the result of positive interferences from other organic compounds that coeluted in the
retention time window of the analysis method used. However, the chromatograms for the low
concentration test Method 0030 sample analyses were not unusual and showed no evidence of
interferences or coeluting compounds. Further, the data from Test Series 4 discussed in
Section 4.1.2 and scoping test data discussed in Section 6.1 also suggest that no positive method
interferences were active. In addition, a plausible explanation for why VOC concentrations were
so much higher than expected is discussed in Section 6.1. Thus, there is no reason to believe the
41
-------
TABLE 18. COMPARISON OF MEASURED AND TARGET CONCENTRATIONS FOR THE ORNL AND
ECOLOGIC CEM TESTS
Concentration, pg/dscm
Compound
Low concentration test
Intermediate concentration test
High concentration test
Average
RM
Nominal
target
Ratio,
RM/target
Average
RM
Nominal
target
Ratio,
RM/target
Average
RM
Nominal
target
Ratio,
RM/target
Benzene
44.6
2.0
22.3
32.6
20.0
1.6
94
200
0.5
Carbon tetrachloride
34.5
4.0
8.6
49.0
40.0
1.2
426
400
1.1
Chlorobcnzenc
63.8
2.0
31.9
60.1
20.0
3.0
302
200
1.5
Chloroform
43.2
4.0
10.8
60.0
40.0
1.5
414
400
1.0
l,2-Dich!oroethane
2.2
2.0
1.1
15.0
20.0
0.8
180
200
0.9
1,1-Dichlorocthene
46.3
2.0
23.2
19.6
20.0
1.0
139
200
0.7
Tctrachloroethene
96.6
2.0
48.3
82.2
20.0
4.1
376
200
1.9
Toluene
377
2.0
189
324
20.0
16.2
1390
200
7.0
1,1,1 -Trichloroethane
4.5
2.0
2.3
13.3
20.0
0.7
174
200
0.9
Trichlorocthene
5.6
2.0
2.8
19.5
20.0
1.0
183 ,
200
0.9
-------
RM measurements for the low concentration test of the ORNL CEM do not reflect actual VOC
concentrations in the scrubber exit flue gas.
The data in Table 18 show that, with the exception of toluene, measured VOC
concentrations for the intermediate and high concentration test days were more in line with the
nominal target concentrations. The measured flue gas toluene concentration was uniformly
substantially greater than the target for all tests.
RAs, as defined in the draft PS given in Appendix A, were calculated using the RM and
CEM data in Tables 15 through 17 for all three concentration test days. These are summarized
in Table 19. For the cases noted above where the concentrations measured by the ORNL CEM
were less than the reporting limit, no RA calculation was performed when two or more of the three
CEM concentrations on a test day were less than the reporting limit. In cases for which only one
of the three CEM concentrations for the test day was less than the reporting limit, the RA was
calculated by using the reporting limit as the CEM concentration. The data in Table 19 show that
the calculated RAs for the ORNL CEM ranged from 123 to 305 percent at the low test
concentration, with an average of 196 percent over the seven compounds reported. ORNL CEM
RAs were improved at the intermediate test concentration, at 113 to 278 percent, with an average
of 154 percent over the nine compounds reported. Further improvement is seen at the high test
TABLE 19. RELATIVE ACCURACIES OF THE ORNL AND ECOLOGIC CEMs
RA, %
ORNL CEM EcoLogic CEM
Test concentration Test concentration
Compound
Low
Intermediate
High
Low
Intermediate
High
Benzene
173
119
98
NC
5,020
154
Carbon tetrachloride
NCa
129
100
NC
135
27
Chlorobenzene
164
93
84
NC
74
52
Chloroform
123
105
97
NC
396
33
1,2-Dichloroethane
305
278
144
NC
2.890
239
l.l-Dichloroethene
299
277
115
NC
2,520
283
Tetrachloroethene
162
142
113
NC
1.640
128
Toluene
145
131
88
NC
65
47
1,1,1 -Trichloroethane
NC
NC
110
NC
7,320
36
Trichloroethene
NC
113
103
NC
5,140
116
Average13
196
154
105
NC
2,520
112
Medianb
164
129
100, 103
NC
1,640, 2,520
52, 116
aNC = Not calculated.
''Average and median excludes NC entries.
43
-------
concentration, with an RA range of 84 to 144 percent, and an average of 105 percent over all
10 compounds reported. In fact, the RA for all VOCs reported uniformly improved as the test
concentration increased.
Because EcoLogic did not report CEM concentrations for the low concentration test day,
no RA calculation was possible. For the intermediate test concentration, the RAs of the EcoLogic
CEM ranged from 65 to 7,320 percent, with an average of 2,520 percent. Much improved
performance was seen at the high test concentration, for which the RA ranged from 27 to
283 percent and averaged 112 percent. As seen for the ORNL CEM, the RAs of the EcoLogic
CEM for all 10 VOCs reported were improved at the high test concentration compared to the
intermediate concentration.
A further measure of CEM performance, beyond the calculated RA which is of eventual
regulatory significance, results if the CEM bias and precision are evaluated. The CEM bias is
defined as the average percent difference between each CEM/RM measurement pair. Specifically:
Bias {%) = 100
^CEM concentration - RM concentration ^
RM concentration
(1)
If the biases of individual CEM/RM measurement pairs are similar regardless of test concentration,
an average bias can be calculated using all, up to nine, CEM/RM data pairs. In such instances the
RA, as defined in Appendix A, represents a worst-case measure of CEM bias. In addition to
calculating the average bias, the standard deviation of the bias over the number of CEM/RM data
pairs can also be calculated. This standard deviation is a measure of CEM precision.
The following procedure was used to assess bias and precision. For each CEM/RM
measurement pair, the ratio of the CEM concentration to the RM concentration, referred to as the
recovery, was calculated. Thus, the recovery, r, in percent is given by:
. 100 • CEM concentration
r (%) = -
RM concentration
(2)
= 100 + bias (%)
Following this approach, up to nine recovery values were calculated for each CEM for each
analyte. These nine recoveries are comprised of three measurement pairs for each of three analyte
concentrations tested. Next, an ANOVA (ANalysis Of VArience) test was performed to evaluate
whether the mean recoveries achieved for each analyte test concentration were statistically the
same. If the mean recoveries for all three test concentrations were statistically equal, data for all
three test concentrations (up to nine recoveries for each analyte) were pooled for the mean bias
and precision evaluation. If the mean recoveries for each test concentration were statistically
different, separate mean bias and precision evaluations were done for each test concentration.
44
-------
The treatment of non-detects in the bias and precision evaluation was analogous to that
adopted in the RA calculation. Specifically, if two or more of the three CEM measurements at
a given test concentration were non-detect, that test concentration was removed from consideration,
both in the ANOVA test and in the mean bias and precision evaluation. If only one of the three
CEM measurements at a given test concentration was non-detect, the reporting limit was used as
the CEM concentration, and the data were considered both in the ANOVA test and the mean bias
and precision evaluation.
For the ORNL CEM, measurement data sufficient to define mean recoveries for all three
test concentrations were reported for 7 of the 10 VOC analytes. However, the ANOVA test
confirmed that the mean recoveries were statistically equal for all three test concentrations for only
one analyte of these seven, 1,2-dichloroethane.
Table 20 summarizes the 1,2-dichloroethane recovery data for the ORNL CEM. Also
noted in the table are the mean recovery over the nine measurements, the standard deviation of
this mean, and the 95-percent confidence interval. Recalling that bias (percent) is recovery
(percent) -100, the data in Table 20 indicate that the bias of the ORNL CEM for
1,2-dichloroethane was 23 ±72 percent at the 95-percent confidence level, and the standard
deviation of the recovery was 93 percent. Individual test concentration level mean bias,
confidence interval, and precision estimates for the ORNL CEM in measuring the other six VOC
analytes with full (nine CEM/RM pairs) measurement data sets are given in Table 21. The table
also includes the estimates for 1,1,1-trichloroethane at the high test concentration, the only level
at which this analyte was detected by the ORNL CEM.
TABLE 20. ORNL CEM BIAS AND PRECISION ESTIMATES
FOR 1,2-DICHLOROETHANE
Reference
Recovery,
Test concentration
Method number
%
Low
RM1
104
RM2
45
RM3
262
Intermediate
RM1
249
RM2
135
RM3
193
High
RM1
48
RM2
66
RM3
3
Mean recovery
123
Standard deviation
93
95-percent confidence interval (±)
72
45
-------
TABLE 21. ORNL CEM BIAS AND PRECISION ESTIMATES FOR BENZENE,
CHLOROBENZENE, CHLOROFORM, 1,1 -DICHLOROETHENE,
TETRACHLOROETHENE, TOLUENE, AND 1,1,1 -TRICHLOROETHANE
Compound
Benzene
Mean bias, %
Standard deviation, %
95-percent confidence interval, %
Chlorobenzene
Mean bias, %
Standard deviation, %
95-percent confidence interval, %
Chloroform
Mean bias, %
Standard deviation, %
95-percent confidence interval, %
1,1-DichJoroethene
Mean bias, %
Standard deviation, %
95-percent confidence interval, %
Tetrachloroethene
Mean bias, %
Standard deviation, %
95-percent confidence interval, %
Toluene
Mean bias, %
Standard deviation, %
95-percent confidence interval, %
1,1,1-Trichloroethane
Mean bias, %
Standard deviation, %
95-percent confidence interval, %
a— Denotes not calculated.
Test concentration
Low Intermediate High
-96.5 -80 -59
0.7 15 14
1.7 37 34
-96.6 -68 -55
2.7 15 11
6.7 36 27
-96.4 -71 -71
4.4 12 10
10.9 31 26
-77 115 -71
36 78 6
90 193 15
-97.6 -91.7 -86
0.9 4.9 2
2.2 12.2 6
-97.0 -66 -47
0.4 9 11
1.0 22 28
-84
6
15
46
-------
For the remaining two VOC analytes, carbon tetrachloride and trichloroethene, full sets
of ORNL CEM measurement data were obtained only for the intermediate and high analyte
concentration tests. The ANOVA test confirmed that the mean recoveries at these two
concentrations were statistically equal. The mean bias and precision evaluations using the six
pooled recoveries achieved over these two concentration test days are summarized in Table 22.
EcoLogic did not report CEM concentration measurement data for the low concentration
test day. Complete sets of data for all 10 VOC analytes were reported for the intermediate and
high concentration test days. However, the ANOVA test confirmed that the mean recoveries were
statistically equal for the two concentrations only for benzene and tetrachloroethene. Table 23
summarizes the mean bias and precision estimates for these two compounds, calculated from the
six pooled CEM/RM measurement pairs at the two concentrations.
The ANOVA test indicated that the mean recoveries for the EcoLogic CEM were
statistically different between the two concentration levels for the eight other VOC analytes.
Individual mean bias and precision estimates at each test concentration are summarized in Table 24
for these eight compounds.
Figures 5 through 7 show the flue gas VOC concentrations reported by the ORNL CEM
versus time for the low, intermediate, and high concentration test days, respectively. Data to allow
assembling corresponding plots for the EcoLogic CEM were not available. However, the
EcoLogic final report (Appendix H-l-2) contains similar plots prepared by EcoLogic. Figures 5
through 7 show that the ORNL instrument reported two to three concentration measurements per
hour. The RM sampling time periods are also shown in each figure.
TABLE 22. ORNL CEM BIAS AND PRECISION ESTIMATES FOR CARBON
TETRACHLORIDE AND TRICHLOROETHENE
Parameter
Carbon tetrachloride
Trichloroethene
Mean bias, %
-81
-89
Standard deviation, %
8
7
95 percent confidence interval, %
8
7
TABLE 23. ECOLOGIC CEM BIAS AND PRECISION ESTIMATES
FOR BENZENE AND TETRACHLOROETHENE
Parameter
Benzene
Tetrachloroethene
Mean bias, %
912
337
Standard deviation, %
1260
430
95-percent confidence interval, %
1320
450
47
-------
TABLE 24. ECOLOGIC CEM BIAS AND PRECISION ESTIMATES FOR
CARBON TETRACHLORIDE, CHLOROBENZENE, CHLORO-
FORM, 1,2-DI CHLOROETH ANE, 1,1-DICHLOROETHENE,
TOLUENE, 1,1,1-TRICHLOROETHANE, AND TRICHLORO-
ETHENE
Test concentration
Compound
Intermediate
High
Carbon tetrachloride
Mean bias, %
-66
-15
Standard deviation, %
21
5
95-percent confidence interval, %
52
11
Chlorobenzene
Mean bias, %
47
-10
Standard deviation, %
24
15
95-percent confidence interval, %
59
37
Chloroform
Mean bias, %
202
-14
Standard deviation, %
90
8
95-percent confidence interval, %
222
19
1,2-Dichloroethane
Mean bias, %
1860
176
Standard deviation, %
780
27
95-percent confidence interval, %
1940
68
1,1 -Dichloroethene
Mean bias, %
1910
204
Standard deviation, %
632
9
95-percent confidence interval, %
1570
23
Toluene
Mean bias, %
-41
-22
Standard deviation, %
3
7
95-percent confidence interval, %
8
17
1,1,1 -T richloroethane
Mean bias, %
6280
10
Standard deviation, %
534
11
95-percent confidence interval, %
1330
27
Trichloroethene
Mean bias, %
2830
82
Standard deviation, %
1060
12
95-percent confidence interval, %
2630
30
48
-------
10
CO
£•
c
o
to 6
»_
c
0)
o
c
O 4
o
(/>
cc
U)
CD o
10
RM1
RM2
v'V
- f\
i.
/
. V
-•—•-
12
14
RM3
16
Time of day
Step
change
A
\-
v
\
- — \
18
Benzene
B
Carbon tetrachloride
•&
TetracWoroetHene
• A
1,1,1 -Tricolor oethane
ThcMorocthone
20
30
CO
E 25
O)
20 -
nJ
I—
® 15
O
c
o
o
to 10
ctf
CD
CD
if 5
0 —
10
RM1
RM2
RM3
12
14 16
Time of day
18
Cfc'oro benzene
Step change
r •
f
¦ _
¦
~ ~ - 4 j
¦ • -
*
¦
¦
Vs
¦
r!\
s
^ I
•
• "* V
"\ ^ ^
y. ¦*-,
'.sr.
y
V
: \ '
•« ^ ;
: 1 !
. * +-M—d
I
1 i
Chloroform
f> •
1,2-Otchloroelh arte
tj.
1,1 -Oichloroethene
Toluene
20
Figure 5. Flue gas VOC concentrations reported by the ORNL CEM
for the low VOC concentration tests.
49
-------
60
o
!«
D)
C
o
""5 30
+-*
c
0)
a
c
o 20
o
CO
CO
CD
0
LL
10
RM1
k
J * V'
:
RM2
10
12
RM3
Second
step
change
First -
step
change
/
i~r
t
•AH. .
¦i
! ¦ Yt ,
'-/¦ [7^-TV
• .1. i \.
¦4
K.
14
16
Time of day
18
Benzene
Cuban tetrachloride
-=>
TetracHoroethefra
A •
1,1,1 -Trichloroethane
•
TricfilOfocthene
20
250
m
E
O)
3.
200
c
.9 I
CO 150
-4—'
c 1
CD
O
C I
O 100 '
o
cn
cc
U)
^ 50
RM1 RM2
RM3
-m First
step change
— Second
step
change
Jl ¦ . ¦ ' \ '
\ /
¦¦¦ a / \\ i'_ A •
V —
£_ V:; - 7
, ¦»
; ^
! f-\ \
i i »¦
i\
10
12 14 16
Time of day
18
Chkxobenzene
Chloroform
-&
1,2 OicWorocthane
1,1 Oichforoefriene
4
Toluene
20
Figure 6. Flue gas VOC concentrations reported by the ORNL CEM
for the intermediate VOC concentration tests.
50
-------
200
RM2
RM3
RM1
Caffoon tetrachloride
-o
Telraeiilofoettiene
i— First
step change
1,1.1-Tricftlaroethane
Triehloroelhefie
,Second
/ step
r change
100
LL
8 10 12 14 16 18
Time of day
1,400 -
CO
75)
c
o
V-»
ca
+—»
c
0)
o
c
o
o
CO
O)
CD
ID
1,200
1,000 h
!
800 r -
600
400 !
i
200
RMI
RM2 RM3
"~V,
v
,»V
First
step change
Second
l
10
12 14
Time of day
-V
16
i step
' change
Chlorobenzene
Chloroform
<-•
1,2-Oichloroethane
1,1-Oichlo»oelhene
«
Toluene
18
Figure 7. Flue gas VOC concentrations reported by the ORNL CEM
for the high VOC concentration tests.
51
-------
As noted above, the secondary objectives of the test program were to evaluate CD, ZD,
response time, and DL of the CEMs. Response times were originally planned to be assessed by
noting CEM response to step changes in flue gas VOC concentrations at the end of each test day
(see Section 2.2.3). The step change effected on the test day at the low VOC concentration was
the shut-off of spike solution feed. The time of this step change is shown in Figure 5 for the
ORNL CEM. The data in the figure show that the first concentration measurement reported by
the ORNL CEM after the step change was at or higher than the concentration just preceding the
step change for six of the nine VOC constituents detected. This suggests that the step change in
VOC spiking solution feedrate did not cause a corresponding step change in scrubber exit flue gas
concentrations. The presence of the scrubber system between the spike solution injection point
and the CEM measurement location evidently caused a step change injection rate to be manifested
as a gradual change in scrubber exit flue gas concentration. The ORNL CEM data in Figure 5
suggest that about an hour was needed for scrubber exit flue gas VOC concentrations to decrease
significantly.
For test days at the intermediate and high VOC concentrations, the first step change was
to reduce the VOC spiking solution feedrate by half. The second step change was to shut off
spike solution feed. Again, these times are indicated in Figures 6 and 7 for the ORNL CEM. The
data in these figures support the observation that about an hour was required for a step change in
VOC spike feed to be fully expressed at the scrubber exit.
Tables 25 through 28 summarize the CD, ZD, and DL measured and reported by ORNL
and EcoLogic for their respective CEMs. The ORNL final report (Appendix H-l-1) mentions the
fact that the reporting limits used in Tables 15 and 16 for compounds not quantitated is not the
TABLE 25. CALIBRATION AND ZERO DRIFT FOR THE LOW
VOC CONCENTRATION TEST DAY
ORNL, %
EcoLogic, %
Compound
CD
ZD
CD
ZD
Benzene
-11
36
NOa
NO
Carbon tetrachloride
40
36
NO
NO
Chlorobenzene
-1.6
42
NO
NO
Chloroform
-28
195
NO
NO
1,2-Dichloroethane
-13
11
NO
NO
1,1-Dichloroethene
71
3.4
NO
NO
Tetrachloroethene
37
23
NO
NO
Toluene
71
23
NO
NO
1,1,1-Trichloroethane
-49
181
NO
NO
Trichloroethene
36
29
NO
NO
aNO = Not operational.
52
-------
TABLE 26. CALIBRATION AND ZERO DRIFT FOR THE
INTERMEDIATE VOC CONCENTRATION
TEST DAY
O
r
%
EcoLogic, %
Compound
CDa
ZD
CD
ZD
Benzene
40, 68
15
9.45
47.5
Carbon tetrachloride
0.6, -15
1.3
12.6
13.8
Chlorobenzene
-0.2, 28
-1.0
9.19
9.44
Chloroform
-40, 222
-13
8.4
9.57
1,2-Dichloroethane
-5.2, -7.3
6.2
10.2
16.7
1,1-Dichloroethene
-12,-29
10
10.9
14.6
Tetrachloroethene
-6.9, 22
0.1
5.13
3.40
Toluene
5.4, 42
5.5
9.82
5.8
1,1,1 -Trichloroethane
-0.8, 318
-18
9.61
2.41
Trichloroethenc
-1.9, 13
1.5
6.17
1.46
aMid-day calibration was performed; the first value corresponds to
the mid-day calibration, the second to the end-of-day calibration.
TABLE 27. CALIBRATION AND ZERO DRIFT FOR THE
HIGH VOC CONCENTRATION TEST DAY
ORNL, % EcoLogic, %
Compound
CD
ZD
CD
ZD
Benzene
-54
6.2
17.0
18.2
Carbon tetrachloride
-64
-1.3
17.0
53.7
Chlorobenzene
-60
-1.8
12.2
71.1
Chloroform
-100
-18
14.7
18.4
1,2-Dichloroethane
-61
0.2
17.8
63.7
1,1-Dichloroethene
-100
-0.3
17.2
36.0
Tetrachloroethene
-67
-2.6
8.8
15.7
Toluene
-64
-0.7
16.3
48.2
1,1,1 -T richloroethane
-66
-12
17.7
35.5
Trichloroethene
-65
-2.1
22.9
48.6
53
-------
TABLE 28. ESTIMATED CEM DETECTION LIMITS
DL, ng/dscm
Compound
ORNL
EcoLogic
Benzene
0.16
2
Carbon tetrachloride
0.16
20
Chlorobenzene
0.16
8
Chloroform
0.16
5
1,2-Dichloroethane
0.16
3
1,1-Dichloroethene
0.16
2
Tetrachloroethene
0.16
50
Toluene
0.16
2
1,1,1-Trichloroethane
0.16
20
Trichloroethene
0.16
40
instrument detection limit, but is, rather, the lowest VOC concentration used in the calibration
curve for that test day.
4.1.2 Test Series 4 — EPA/APPCD and MSP VOC CEMs
The EPA/APPCD CEM tested was an online gas chromatograph (GC) analytical system
consisting of a heated sample delivery component, a sample concentrating component, and an
analysis component. The sample concentrating component consists of a sorbent trap that collects
the flue gas VOC analytes over an adjustable time period. The sorbent trap is subsequently
thermally desorbed into the GC analytical component. The analytical component uses both a flame
ionization detector (FID) and an electron capture detector (ECD) for VOC quantitation.
The MSP final report (Appendix H-l-3) does not provide a system description. However,
information provided in their proposal to participate in the test program describes their system as
consisting of a heated flue gas extraction and sampling system and a process mass spectrometer
(MS) with dual detectors. The MS employs an electron impact ionizer with a high-speed, high-
resolution quadrupole for mass separation.
Tables 29 through 31 present the results of the three sequential RM measurements, along
with the EPA/APPCD and MSP CEM results, for each of the three VOC concentrations tested in
Test Series 4. (The reader is again reminded of the caution to suspect the low and intermediate
concentration RM measurement data for 1,1-dichloroethene as the reported concentrations are less
than 30 jjg/dscm.) EPA/APPCD concentrations are given to two decimal places in the tables
because they were reported as such in the EPA/APPCD final report (Appendix H-l-4). The tables
indicate that, out of the nine sampling periods, MSP obtained data for only two. As described in
54
-------
TABLE 29. MEASURED FLUE GAS CONCENTRATIONS FOR THE TESTS
OF THE EPA/APPCD AND MSP CEMS AT THE LOW VOC
CONCENTRATION
Concentration, pg/dscm
Reference Method 1 Reference Method 2 Reference Method 3
EPA/ EPA/ EPA/
Compound
RM
APPCD
MSP
RM
APPCD
MSP
RM
APPCD
MSP
Benzene
8.2
21.31
NOa
5.9
29.93
795
8.4
22.21
707
Carbon tetrachloride
13.9
9.68
NO
11.9
5.99
118
13.3
4.99
126
Chlorobenzene
21.6
18.56
NO
20.8
29.16
143
16.0
18.8
60.2
Chloroform
15.8
16.95
NO
18.4
14.89
3,439
15.8
9.25
1,515
1,2-Dichloroethane
1.8
43.21
NO
1.5
39.18
73.7
1.6
30.33
78.5
1,1-Dichloroethene
2.0
76.34
NO
6.3
90.23
322
5.1
84.49
271
Tetrachloroethene
32.7
15.65
NO
26.9
31.7
124
20.6
8.56
107
Toluene
160.9
131.92
NO
149.4
221.38
1,308
97.1
57.85
814
1.1,1-Trichloroethane
2.1
2.21
NO
1.9
2.22
3.6
1.8
3.53
4.1
Trichloroethene
2.6
2.18
NO
2.9
2.71
3,022
3.0
1.74
1,602
aNO = Nol operational.
TABLE 30. MEASURED FLUE GAS CONCENTRATIONS FOR THE TESTS OF
THE EPA/APPCD AND MSP CEMS AT THE INTERMEDIATE VOC
CONCENTRATION
Concentration, fig/dscm
Reference Method 1 Reference Method 2 Reference Method 3
EPA/ EPA/ EPA/
Compound
RM
APPCD
MSP
RM
APPCD
MSP
RM
APPCD
MSP
Ben/cnc
33.9
35.8
NO3
32.9
42.41
NO
33.6
50.34
NO
Carbon tetrachloride
53.5
31.45
NO
57.8
37.58
NO
64.1
40.55
NO
Chlorobenzene
29.5
24.86
NO
64.6
54.15
NO
75.0
40.73
NO
Chloroform
43.2
26.33
NO
62.1
31.61
NO
63.5
43.31
NO
1,2-Dichloroethane
20.7
27.97
NO
18.1
34.9
NO
16.6
55.71
NO
1,1-Diehloroethene
14.2
47.14
NO
11.3
54.48
NO
9.9
101.19
NO
Tetrachloroethene
39.3
22.87
NO
92.7
58.32
NO
96.8
30.3
NO
Toluene
143
90.22
NO
498.5
306.1
NO
551.8
163.48
NO
1,1,1 -Trichloroethane
17.3
13.68
NO
16.2
14.33
NO
17.4
14.54
NO
Trichloroethene
22.9
16.04
NO
20.6
16.63
NO
19.1
15.74
NO
aNO = Not operational.
55
-------
TABLE 31. MEASURED FLUE GAS CONCENTRATIONS FOR THE TESTS
OF THE EPA/APPCD AND MSP CEMS AT THE HIGH VOC
CONCENTRATION
Concentration, fig/dscm
Reference Method 1 Reference Method 2 Reference Method 3
EPA/ EPA/ EPA/
Compound
RM
APPCD
MSP
RM
APPCD
MSP
RM
APPCD
MSP
Benzene
102.6
96.33
NOa
129.5
98.73
NO
117.7
88.66
NO
Carbon tetrachloride
222.5
209.55
NO
266.2
205.53
NO
283.7
135.72
NO
Chlorobenzene
104.8
119.78
NO
146.5
113.78
NO
127.2
126.99
NO
Chloroform
229.4
190.9
NO
243.8
199.06
NO
241.1
178.42
NO
1,2-DichIoroethane
93.7
95.44
NO
121.2
107.96
NO
114.9
90.81
NO
1,1-Dichloroethcne
65.9
113.67
NO
65.5
124.81
NO
71.9
144.06
NO
Tetrachloroethene
112.8
150.65
NO
162.6
161.66
NO
132.2
131.62
NO
Toluene
176.6
191.23
NO
445.4
213.8
NO
261.5
217.23
NO
1,1, l-Trichloroethane
97.8
92.89
NO
106.8
87.31
NO
103.5
57.93
NO
Trichloroethene
98.2
98.46
NO
114.3
91.81
NO
113.1
70.04
NO
aNO = Not operational.
their final report (Appendix H-l-3), the MSF CEM failed during the first test day. The failure was
caused by moisture condensation in the instrument's sampling system which, in turn, caused
pressure fluctuations at the MS inlet. These pressure fluctuations persisted throughout the rest of
the test series. The MSP team was unable to effect system repairs in time to participate in any
subsequent testing.
It also bears noting that the MSP final report gives flue gas analyte concentrations for the
two RM test periods reported in units of parts per billion (ppb), without mention of the associated
gas conditions. Verbal communication with MSP staff confirmed that the reported ppb were on
a dry basis. The unit conversions to (jg/dscm were performed by the IRF staff.
Table 32 presents a comparison of the flue gas VOC concentrations as determined by the
RM to the target concentrations outlined in Section 2.2.3. Thus, Table 32 is the Test Series 4
analog to Table 18 for Test Series 1. Comparing the data in Table 32 to the data in Table 18
shows that the measured flue gas concentrations for the low concentration test for Test Series 4
were greater than corresponding targets, as was the case for Test Series 1. However, the measured
concentrations for Test Series 4 were uniformly much closer lo their respective targets.
As in Test Series 1. the compounds measured at the highest concentrations in Test
Series 4 were toluene, tetrachloroethene, and chlorobenzene. However, except for these three
compounds, measured concentrations in Test Series 4 were uniformly within a factor of about 1
56
-------
TABLE 32. COMPARISON OF MEASURED AND TARGET CONCENTRATIONS FOR THE EPA/APPCD
AND MSP CEM TESTS
Concentration, fig/dscm
Low concentration test Intermediate concentration test High concentration test
Compound
Average
RM
Nominal
target
Ratio,
RM/target
Average
RM
Nominal
target
Ratio,
RM/target
Average
RM
Nominal
target
Ratio,
RM/target
Benzene
7.5
2.0
3.8
33.5
20.0
1.7
117
200
0.6
Carbon tetrachloride
13.0
4.0
3.3
58.5
40.0
1.5
258
400
0.6
Chloroben/.ene
19.5
2.0
9.8
56.4
20.0
2.8
126
200
0.6
Chloroform
16.7
4.0
4.2
56.3
40.0
1.4
238
400
0.6
1,2-Dichloroclhane
1.6
2.0
0.8
18.5
20.0
0.9
110
200
0.5
1,1-Dichlorocthene
4.5
2.0
2.3
11.8
20.0
0.6
67.8
200
0.3
Tetrachloroethene
26.7
2.0
13.4
76.3
20.0
3.8
136
200
0.7
Toluene
136
2.0
67.9
398
20.0
19.9
295
200
1.5
1,1,1 -Trichloroethane
2.9
2.0
1.5
17.0
20.0
0.9
103
200
0.5
Trichloroethenc
2.8
2.0
1.4
20.9
20.0
1.0
109
200
0.5
-------
to 4 of their respective targets. As also seen in Test Series 1, measured concentrations at the
intermediate test concentration were uniformly closer to their respective targets. In fact, again with
the exception of toluene, tetrachloroethene, and chlorobenzene, the measured concentrations for
the intermediate concentration test were within a factor of 1.7 of the respective targets. Measured
concentrations for the high concentration test were 50 to 70 percent of their respective targets, with
the exception of the measured toluene concentration.
RAs corresponding to the RM/CEM concentration data given in Tables 29 through 31 are
summarized in Table 33. Because the MSP system only operated on the first test day at the low
VOC concentration, only this one set of RAs is noted in the table. However, even this RA
calculation is not strictly appropriate because it is based on only two pairs of RM and CEM
measurements. All EPA CEM performance specifications require a minimum of three pairs of RM
and CEM measurements for an RA calculation.
The data in Table 33 show that the RAs for the EPA CEM ranged from 71 to
3,190 percent, and averaged 638 percent, for the low test concentration. The relatively high
average RA was driven by the two very high RAs for 1,2-dichloroethane and 1,1-dichloroethene,
however. And, although the 1,1-dichloroethene RM measurement needs to be viewed with
caution, another reason the EPA CEM appears to have compared poorly in comparison is
TABLE 33. RELATIVE ACCURACIES OF THE EPA/APPCD AND MSP CEMs
RA, %
EPA/APPCD
MSD
Test concentration
Test concentration
Compound
Low
Intermediate
High
Low
Benzene
429
83
48
18,400
Carbon tetrachloride
86
45
95
1,200
Chlorobenzene
87
98
53
3,150
Chloroform
76
71
34
85,800
1,2-Dichloroethane
3,190
334
40
6,740
1,1-Dichloroethene
2,040
1,130
133
10,700
Tetrachloroethene
137
134
50
673
Toluene
113
158
138
3,040
1,1,1-Trichloroethane
150
29
73
314
Trichloroethene
71
45
70
384.000
Average
638
213
73
51,400
Median
113, 137
83. 98
53, 70
3,150, 6,740
58
-------
discussed below. The median RA for the low concentration test, at a much improved 113 to
137 percent, removes the dominant influence of the two compounds for which the CEM did
poorly. The RAs for the EPA CEM were improved at the intermediate test concentration, ranging
from 29 to 1,130 percent and averaging 213 percent. Poor performance in quantitating 1,2-
dichloroethane and 1,1-dichloroethene compared to the RM again accounts largely for the high
average RA. Again, the median RA, at 83 to 98 percent, better reflects the mean performance of
the CEM by removing the dominant influence of the RAs for the two VOCs poorly quantitated.
Further improved performance of the EPA CEM was seen at the high test concentration, with an
RA range from 34 to 133 percent and an average RA of 73 percent. In fact, at the high test
concentration, the RAs for two compounds poorly quantitated at the low and intermediate test
concentrations are more in line with those calculated for the other eight compounds. For this
reason, the median RA, at 53 to 70 percent, is comparable to the average RA.
The calculated RAs based on the two available CEM/RM measurement pairs for the MSP
CEM were quite large, ranging from 314 to 384,000 percent and averaging 51,400 percent. Even
the median RAs for the MSP CEM, at 3,150 to 6,740 percent, were quite high.
EPA/APPCD reported a full set of CEM/RM data; CEM results were reported for each
RM sampling period completed. The ANOVA test of the achieved recoveries indicated that the
mean recoveries were statistically the same over all three test concentrations for seven of the VOC
analytes. Table 34 summarizes the mean bias and precision estimates for the EPA/APPCD CEM,
based on the nine pooled CEM/RM measured pairs for these seven VOCs. The data in Table 34
show that the bias for the EPA/APPCD CEM over the seven compounds ranged from
0.2 ±31 percent to -37 ±13 percent, both at the 95-percent confidence level. The corresponding
standard deviation of the mean CEM recovery ranged from 14 to 40 percent.
Table 35 summarizes the EPA/APPCD CEM bias and precision estimates for the other
three VOCs, based on the evaluation of individual test concentration data. These three compounds,
for which the ANOVA test indicated that the mean recoveries were not the same at all three
concentration levels tested, are the three compounds having the highest RAs for the low
concentration test — benzene, 1,2-dichloroethane, and 1,1-dichloroethene. The data in Table 35
confirm this observation. Clearly, the CEM bias and precision for these compounds were quite
poor at the low concentration. Both improved significantly at the intermediate concentration for
all three compounds, with even further improvement at the high concentration.
Because MSP reported concentration data for only two RM sampling periods over the
entire week of Test Series 4 testing, and because reported CEM concentrations were uniformly
much greater than the corresponding RM concentration, no further statistical evaluation of the MSP
data was deemed warranted.
Figures 8 through 10 show the flue gas VOC concentrations reported by the EPA CEM
for the low, intermediate, and high concentration test days, respectively. Figure 11 is a
corresponding plot for the MSP CEM on the low concentration test day, the only day the MSP
CEM was operational. Figures 8 through 10 show that the EPA instrument reported about one
concentration measurement per hour for the low and intermediate concentration tests, and one to
59
-------
TABLE 34. EPA/APPCD CEM BIAS AND PRECISION ESTIMATES FOR CARBON TETRACHLORIDE,
CHLOROBENZENE, CHLOROFORM, TETRACLOROETHENE, TOLUENE, 1,1,1-TRICHLORO-
ETHANE, AND TRICHLOROETHENE
Parameter
Carbon
tetra-
chloride
Chloro-
benzene
Chloro-
form
Tetra-
chloro-
ethene
Toluene
1,1,1-
Trichloro-
ethane
Trichloro-
ethene
Mean bias, %
-37
-0.2
-30
-23
-24
0.2
-21
Standard deviation, %
17
21
19
36
35
40
14
95-percent confidence interval, %
13
16
14
28
27
31
11
-------
TABLE 35. EPA/APPCD CEM BIAS AND PRECISION ESTIMATES
FOR BENZENE, 1,2-DICHLOROETHANE, AND
1,1-DICHLOROETHENE
Test concentration
Compound
Low
Intermediate
High
Benzene
Mean bias, %
244
28
-18
Standard deviation %
142
22
11
95-percent confidence interval, %
352
55
26
1,2-Dichloroe thane
Mean bias, %
2200
121
-10
Standard deviation %
368
103
11
95-percent confidence interval, %
914
256
28
1,1-Dichloroethene
Mean bias, %
2200
512
88
Standard deviation %
1320
363
14
95-percent confidence interval, %
3270
902
35
two per hour for the high concentration test. The MSP CEM gave essentially continuous
concentration readings, though, as noted in Table 29, these readings were, for the most part, much
greater than the corresponding RM measurement. RM measurement periods are indicated in each
of Figures 8 through 11.
Secondary test program objectives were to evaluate CD, ZD, response time, and DL of
the CEMs. The MSP final report (Appendix H-l-3) does not offer any information regarding these
performance measures parameters. Figures 9 and 10 illustrate the qualitative responses of the EPA
CEM to the step changes induced. No EPA nor MSP CEM data were taken after the step change
on the low VOC concentration test day, as indicated in Figures 8 and 11. The step changes
induced for this test series were the same as for Test Series 1, as discussed in Section 4.1.1. As
the EPA CEM data point frequency was about one per hour, the EPA data shown in Figures 9
and 10 support the observation noted in Section 4.1.1 that about an hour was needed for scrubber
exit flue gas VOC concentrations to fully respond to a scrubber inlet concentration step change.
Interestingly, the data in Figure 9 show that the concentration of all 10 VOCs measured
by the EPA instrument decreased after the first step change on the intermediate concentration test
day; most decreased to about half their pre-step-change levels. Recall that the first step change
was a halving of the VOC spike solution feedrate. Further concentration decreases were measured
after the second step change (stopping VOC spike solution feed) for 8 of 10 constituents. Similar
observations can be seen in the high concentration test data in Figure 10. Decreases in the
61
-------
70
60
50
O)
3.
c
o
CO Af\
*— 40
c
0
o
c
o
o
c/>
CO 20
U)
0
3
n: 10
30
500
CO
E
CT)
400
c
o
2
300
c
0
o
c
o
200
o
cn
cd
U)
0
13
100
Ll_
RM1
RM2
RM3
Step —
change
¦ - - ¦>,
Benzene
,-^-y
Carbon tetrachloride
o
Tetrachloroettiane
a
1,1,1-Trichloroettiane
Trichloroethene
10
12 14
Time of day
16
18
RM1
RM2
RM3
Step
change
Chlorobenzene
Chloroform |
1,2-Dichloroethane j
1,1-Dichloroethene
«
Toluene
10
12 14
Time of day
16
18
Figure 8. Flue gas VOC concentrations reported by the EPA/APPCD CEM
for the low VOC concentration tests.
62
-------
120
rt
.E 100
o>
3.
O 80 h
2 I
S 60 I -
g '
o
o
C/5
cd
CD
1,2-Dichloroethane
u
1,1-Dichloroethene
—
Toluene
20
Figure 9.
Flue gas VOC concentrations reported by the EPA/APPCD CEM
for the intermediate VOC concentration tests.
63
-------
300
CO
E
250
D)
C
o
200
<-
150
a>
o
c
o
o
03
100
05
CD
CD
3
50
LL
0
300 r
CO
E:
250
c
o
200
cd
v_
C
150 -
Carbon tetrachloride
-Q-
Tetrachtoroettiane
A
1,1,1-Trichloroethane
—
Trichlocoethene
10 12 14
Time of day
16
18
RMl RM2 RM3
•• ^ s* - -X
/ *
First step
change
\
Second
step
change
Chlorobenzene
Chloroform
1,2-0ichloroethane
A •
1,1 -Dichioroethene
Toluene
s •
\
10 12 14
Time of day
16
18
Figure 10. Flue gas VOC concentrations reported by the EPA/APPCD CEM
for the high VOC concentration tests.
64
-------
5,000
«
O) 4,000
c
o
(0 3,000
i—
c
(D
O
o 2,000
o
(/)
(0
CD
0) 1.000
LL
RM1
10
!&! i'Si
lit iij:
in »;
rtt ftr-
"i
if), |H
I'll! II!
i'tt': : i'r.
1'it:. : ii||
M : :Ji'iJ
!W
!;!'• •'
k ~< m i
i'i'- v,! ?•
RM2
r ]
V:i •£ 'i
".1
12 14
Time of day
RM3
Step -
change
r::
! !
I • 11'
k t * b !
16
Carbon Tetrachloride
Chlorobenzene
1,2-Dichloroethane
1,1-Dichloroethene
Tetrachloroethene
1.1,1-Trichloroethane
18
Figure 11. Flue gas VOC concentrations reported by the MSP CEM
for the low VOC concentration tests.
65
-------
measured concentrations after the first step change were seen for 9 of 10 constituents, again most
by roughly half. Further decreases were measured for all 10 constituents after the second step
change.
EPA/APPCD reported zero and span checks as pre- and post-test quality assurance
measurements (see Appendix H-l-4). These data are summarized in Tables 36 and 37. Table 36
gives the pre- and post-test zero measurements obtained by the collection of a nitrogen/system
blank sample. Table 37 summarizes the pre- and post-test system calibrations.
The data in Table 36 show that very high levels of 1,1-dichloroethene were present in
both the pre- and post-test system blanks for the low concentration test day. Elevated levels of
benzene and 1,2-dichloroethane were also present. Elevated levels of benzene persisted in both
pre- and post-test system blanks for the intermediate and high concentration test days, as well as
elevated levels of 1,1-dichloroethene in post-test blanks, and 1,2-dichloroethane in one of the two
blank checks. EPA did not blank-correct their measurement data, so the EPA CEM concentrations
noted in Tables 29 through 31 were not adjusted for the high blank levels for the three compounds
noted above. Had blank correction been performed, the RAs for the low concentration test for
these three compounds would have measurably improved, and the RAs for the intermediate
concentration test would have been somewhat better. In addition, the ANOVA test would likely
have indicated that mean recoveries for these three compounds were statistically equal, and CEM
bias and precision estimates would have been much improved.
The EPA calibration procedure consisted of injecting a 5-mL sample of calibration gas
directly into the analytical unit. The quantities of target analytes contained in this 5-mL sample
are given in the column headed "Spiked amount" in Table 37. The pre- and post-test columns in
the table show the analyzed amounts for each calibration. CD can be calculated directly from the
pre- and post-test analyzed amounts as:
x-ti ,ct\ Post-test amount - Pre-test amount
CD (%) = x 100 (-3)
Pre-test amount
CDs so calculated are given in Table 37.
Calculating ZDs from the data in Table 36 is not as straightforward, however. The blank
check consists of collecting a volume of nitrogen zero gas that is nominally the size of a flue gas
volume sampled during operation, and the result of a blank check is a gas concentration (fjg/m3)
that can be directly compared to a measured flue gas concentration. However, the calibration
consists of directly injecting a small quantity of high-concentration calibration gas. The calibration
gas analyte concentrations are far greater than any expected measured flue gas concentrations.
66
-------
TABLE 36. EPA/APPCD PRE- AND POST-TEST ZERO MEASUREMENTS
Concentration, jig/m3
Low concentration Intermediate High
test concentration test concentration test
Compound Pre-test Post-test Pre-test Post-test Pre-test Post-test
Benzene
3.94
17.60
3.61
6.57
3.11
5.74
Carbon tetrachloride
NDa
0.39
0.04
0.86
0.30
1.04
Chlorobenzene
1.33
3.13
0.69
0.67
0.65
7.86
Chloroform
ND
3.96
0.56
2.54
0.96
2.46
1,2-Dichloroethane
3.63
33.20
ND
11.11
5.04
ND
1-Dichloroethene
50.89
108.73
0.45
4.57
0.77
7.23
Tetrachloroethene
ND
0.51
0.24
1.20
1.28
5.55
Toluene
ND
3.27
0.97
3.40
1.24
6.30
1,1,1-Trichloroethane
ND
0.71
0.66
1.28
0.89
0.63
Trichloroethene
ND
0.81
0.42
1.00
1.08
0.86
aND = Not detected.
TABLE 37. EPA/APPCD PRE- AND POST-TEST CALIBRATION MEASURE-
MENTS AND CORRESPONDING CALIBRATION DRIFT
Measured amount, ng
Compound
Spiked
amount,
ng
Low
concentration test
Pre- Post- CD,
test test %
Intermediate
concentration test
Pre- Post- CD,
test test %
High
concentration test
Pre- Post- CD,
test test %
Benzene 32.27
Carbon tetrachloride 40.48
Chlorobenzene 46.04
Chloroform 49.0?
1,2-Dichloroethane 55.11
1,1 -Dichloroethene 40.05
Tetrachloroethenc 68.51
Toluene 37.87
1,1,1-Trichloroethane 63.86
Trichloroethene 54.01
30.33 25.63 -15.5
36.0(3 39.16 8.6
23.85 24.26 1.7
39.92 35.11 -12.0
32.85 27.21 -17.2
45.56 55.13 21.0
42.20 26.76 -366
31.95 20.78 -35.0
45.86 41.43 -9.7
44.96 30.61 -31.9
24.22 29.36 21.2
44.52 77.32 73.7
27.71 25.50 -8.0
40.97 45.28 10.5
28.21 36 91 30.8
38.90 42.34 8.8
40.77 49.97 17.7
20.49 29.19 42.5
48.08 53.78 11.9
43.30 46.66 7.8
27.40 33.11 20.8
59.42 67.72 14.0
35.44 38.83 9.6
48.43 49.04 1.3
36.46 38.88 66
43.11 39.41 -8.6
66.21 69.27 4.6
32.24 36.85 14.3
54.28 54.76 0.9
47.60 56.14 17.9
67
-------
The definition of zero drift is:
~r. Post-test concentration - Pre-test concentration ,AS
ZD = x 100 (4)
Reference concentration
Normally, the reference concentration is the concentration of the most concentrated calibration
standard. However, because the procedures used for zero and calibration checking are different,
and, because the analyte concentrations in the calibration gas are much greater than actual flue gas
concentrations, the choice of what reference concentration to use is not readily apparent.
During test operation, the volume of flue gas sampled for each analytical determination
is adjusted to bring the analyte quantity in the flue gas sample into the calibrated range of the EPA
instrument. Thus, a larger volume of flue gas was sampled for each determination on the low
concentration test day than on the intermediate concentration test day, which, in turn, was a larger
volume than that sampled on the high concentration test day. However, the quantity of analyte
(ng) contained in the volume of gas sampled each test day was in the range of the quantity of
analyte in the 5-mL calibration gas sample.
Data in the EPA final report show that each flue gas concentration measurement on a
given test day was derived from nominally the same sampled flue gas volume. Thus, on the low
concentration test day, each measurement was derived from a flue gas sample of about 0.76 L, on
the intermediate concentration test day from a sample of about 0.47 L, and on the high
concentration test day from a sample of about 0.27 L. Given this, a logical choice for the
reference concentration to use in calculating ZD would be the pre-test calibration analyzed amount
(ng) divided by the nominal flue gas volume sampled on each test day.
Table 38 summarizes the calculated ZDs on each test day on this basis. For each test day,
the table notes the reference concentration for each analyte as well as the calculated ZD. Each
analyte's reference concentration is the analyte's pre-test measured amount from Table 37 divided
by each day's nominal flue gas sample volume, noted above. ZDs arc calculated from equation
(4) using these reference concentrations and the pre- and post-test zero concentration given in
Table 36.
4.2 SVOC (PAH) CEM TESTS
The SVOC CEM tests were performed during Test Series 4 of the test program
simultaneously with the second set of VOC CEMs. The EcoChem Technologies, Inc., total
particulate-bound PAH CEM was the only SVOC CEM tested in the program.
Like all other test series, each day of SVOC CEM testing consisted of performing three
sequential RM measurements, while the tested CEM was in operation, and the test series consisted
of three test days at each of three different SVOC concentrations: low, intermediate, and high.
Due to problems in the flue gas conditioning (moisture removal) system, the EcoChem CEM could
not be brought into operation on the last day of the testing at the high SVOC concentration.
68
-------
TABLE 38. ZERO DRIFT FOR THE EPA VOC CEM
Low
concentration test
Intermediate
concentration test
High
concentration test
Compound
Reference
concentration,
Mg/m3
ZD, %
Reference
concentration,
jig/m3
ZD, %
Reference
concentration,
Mg/m3
ZD,
%
Benzene
39.91
34.2
51.53
5.7
101.48
2.6
Carbon tetrachloride
47.45
0.8
94.72
0.9
220.07 -
0.3
Chlorobenzene
31.38
5.7
58.96
-0.03
131.26
5.5
Chloroform
52.53
7.5
87.17
2.3
179.37
0.8
1,2-Dichloroethane
43.22
68.4
60.02
18.5
135.04
-3.7
1,1-Dichloroethene
59.95
96.5
82.77
5.0
159.67
4.0
Tetrachloroethene
55.53
0.9
86.74
1.1
245.22
1.7
Toluene
42.04
7.8
43.60
5.6
119.41
4.2
1,1,1 -Trichloroethane
60.34
1.2
102.30
0.6
201.04
-0.1
Trichloroethene
59.16
1.4
92.13
0.6
176.30
-0.1
Therefore, because EcoChem was the only SVOC CEM participant in the test program, RM
measurements were stopped after the first RM of this test day.
EcoChem terms their PAH CEM the PAS lOOOe Photoelectric Aerosol Sensor. The
photoelectric aerosol sensor (PAS) works on the principle of photoionization of aerosol-bound
PAHs. Sample gas is first passed through an electrofilter to remove charged aerosols. Then, an
ultraviolet light lamp ionizes the electrically neutral flow of aerosols. The wavelength of the light
is chosen such that only PAH-coated aerosol particles are ionized, while gas molecules and non-
carbon aerosol particles remain neutral. The aerosol particles that have PAH molecules adsorbed
on their surfaces then emit electrons, which are subsequently removed by an electric field. The
remaining positively charged particles are collected on a filter inside an electrometer, where the
charge is measured and a signal proportional to the total particulate-bound PAH concentration is
produced. The analyzer does not differentiate between various PAH species; instead, the analyzer
signal is a measure of total PAH adsorbed on the flue gas particulate.
Table 39 presents the results of the three sequential RM measurements performed each
test day, and compares these to the EcoChem CEM results for the test days at the low and
intermediate SVOC concentrations. The data from the single RM measurement performed on the
high SVOC concentration test day is also presented in the table, although the EcoChem CEM was
not operational on this test day. No data were obtained during the first RM period on the
intermediate concentration test day because the EcoChem CEM was not in operation, again due
to problems with the flue gas moisture removal system.
69
-------
TABLE 39. MEASURED FLUE GAS CONCENTRATIONS FOR THE ECOCHEM
PAH CEM TESTS
Concentration, pg/dscm
Reference
Reference
Reference
Test
Method 1
Method 2
Method 3
RA, %
Low concentration test
Naphthalene
1.7
1.8
1.7
Phenanthrene
1.3
1.2
1.3
Pyrene
1.0
0.8
0.9
Total PAHa
4.0
3.8
3.9
EcoChem CEM
6.9
14.8
15.5
527
Intermediate concentration test
Naphthalene
17.5
10.9
15.8
Phenanthrene
15.7
10.1
15.3
Pyrene
9.1
19.6
9.7
Total PAH
42.3
40.6
40.8
EcoChem CEM
NOb
33.2
39.0
99
High concentration test
Naphthalene
97.0
NPC
NP
Phenanthrene
91.4
NP
NP
Pyrene
68.2
NP
NP
Total PAH
256.6
NP
NP
EcoChem CEM
NO
NO
NO
NCd
aActual total may have been larger due to possible contributions from PAHs not spiked
and not detected at MDLs comparable to the measured concentrations of spiked PAHs.
bNO = Not operational.
CNP = Not performed.
dNC = Not calculated.
70
-------
Table 40 compares the measured flue gas concentrations to the target levels discussed in
Section 2.2.3. As indicated, actual measured concentrations were in the range of 45 to 87 percent
of the target levels.
Table 39 notes RM concentrations only for the three PAH compounds spiked into the
afterburner extension flue gas. The entire list of compounds in Table 13 was sought in the
analyses of RM samples. However, none of the non-spiked compounds was detected in any
sample at MDLs ranging from 0.38 to 1.25 pg/dscm.
This latter point deserves some discussion. As outlined in Section 6.2, the method
detection limits (MDLs) achieved for the spiked SVOC analytes were 0.91 pg/dscm for
naphthalene, 0.67 |ig/dscm for phenanthrene. and 0.63 )ig/dscm for pyrene. All three MDLs
achieved met the MDL objective for the measurement of 1 (jg/dscm. Furthermore, the flue gas
concentrations measured by the RM for the low target concentration test day ranged from factors
of 1.3 greater than the MDL for pyrene to 2.0 greater than the MDL for naphthalene. These
factors are in the range of the factor of 2 deemed acceptable in the approved QAPP (Reference 8)
for the project.
The MDL objectives for the other 14 SVOC analytes were also 1 jig/dscm. The
discussion in Section 6.2 shows that these objectives were, for the most part, also met. One may
question whether these MDL objectives were appropriate in light of both the target and measured
concentrations of the three spiked PAH analytes. For example, if all 14 non-spiked analytes were
present in the flue gas at concentrations just under their respective MDLs achieved (see
Section 6.2), their contribution to the total PAH concentration could be as high as 13.1 jjg/dscm.
This compares to the measured total PAH from the three spiked analytes of 3.8 to 4.0 ng/dscm.
However, this possibility was acknowledged during the planning efforts for the test program, as
documented in the approved QAPP for the program (Reference 8). The desire to challenge the
PAH CEM with a low spiked PAH concentration was given priority over the inevitable ambiguity
in the ability to measure the total flue gas PAH concentrations given the MDL capabilities of the
RM.
TABLE 40. COMPARISON OF MEASURED AND TARGET CONCENTRATIONS
FOR THE ECOCHEM PAH CEM TESTS
Concentration, pg/dscm
Low concentration test Intermediate concentration test
Compound
Average
R\1
Nominal
target
Ratio,
RM/target
Average
RM
Nominal
target
Ratio,
RM/target
Naphthalene
1.7
2.0
0.87
14.7
20.0
0.74
Phenanthrene
1.3
2.0
0.64
13.7
20.0
0.69
Pyrene
0.9
2.0
0.45
12.8
20.0
0.64
71
-------
Nevertheless, the total PAH concentrations listed in Table 39 for the low concentration
test, and, to a lesser though still significant degree, for the intermediate concentration test, more
defensibly represent lower bounds for the actual total PAH concentrations potentially present. The
fact that the concentrations reported by the EcoChem CEM were uniformly greater than the
corresponding totals noted in Table 39 suggests that perhaps other PAH compounds were present
in the flue gas at levels which, while below MDLs, may still have been sufficiently high to have
had a measurable effect on the total RM PAH concentration reported.
Table 39 gives the calculated RA of the EcoChem PAH CEM for the two test days the
CEM was in operation using the RM concentrations reported in the table. As was done for the
MSP VOC CEM discussed in Section 4.1.2, an RA was calculated for the intermediate
concentration test day using only the two pairs of RM and CEM data, despite the universal PS
requirement that at least three pairs of data be used in a true RA determination.
Table 39 indicates that the RAs for the EcoChem CEM were 527 and 99 percent. As was
seen in the VOC CEM tests, the RA at the higher test flue gas concentration was improved in
comparison to the lower test concentration. However, the calculated RA for the low concentration
test may portray an overly pessimistic view of the EcoChem CEM's capabilities, given the
possibility that actual flue gas total PAH concentrations may have been higher than measured by
the RM, as discussed above.
The ANOVA test confirmed that the mean CEM recoveries for the two test concentrations
having CEM data were statistically equivalent. Pooling all five CEM/RM measurement pairs
shows that the EcoChem PAH CEM had a mean bias of 127 ±193 percent at the 95-percent
confidence level, and that the standard deviation of the mean recovery was 156 percent.
Figure 12 shows the measured PAH concentration over the duration of each test day for
the low concentration test (top) and the intermediate concentration test (bottom). The periods of
RM sampling are also shown in the figure. In addition, as the EcoChem PAH CEM gives a single
reading, total PAH, the position of the line denoting each RM sampling period along the y-axis
of the plots corresponds to the RM measurement concentration.
The data for the low concentration test in Figure 12 show that the EcoChem CEM reading
was in the range of the RM measured concentration for the first RM sample. However, the CEM
reading roughly doubled near the end of the first RM, and remained at this high level for the other
two PM sampling periods. Nevertheless, the second and third RM concentration measurements
were essentially the same as the first RM measurement. The data for the intermediate
concentration test in Figure 12 show that the mean CEM reading remained relatively constant over
the second and third RM sampling periods, and at a level in the range of the RM measurements.
In addition to measuring CEM RA, secondary test program objectives were to evaluate
CEM CD, ZD, response time, and DL. Response time was to have been evaluated by the
introduction of step changes in monitored flue gas analyte concentrations. As the EcoChem PAH
CEM was tested during Test Series 4, along with the EPA/APPCD and MSP VOC CEMs, the
attempts to introduce scrubber exit flue gas step changes were as discussed in Section 4.1.2.
72
-------
20
CO
75)
=1.
C 15 h
o J
2 i
c i
CD I
O '
£T 10 r-
o I
O I
X
< i
Q_ 1
CO 5 .
CO 19 ,
CD I
CD
13
Time of day
RMI
20
100
CO
E
CT>
80
c
o
CO
-+~*
c
60
a>
o
cz
o
o
X
40
<
Q.
CO
¦
ca
CD
20
a>
•
=3
UL
KM1
10
First step
change
Second
change
y*#1'
12 14 16
Time of day
20
Figure 12. Flue gas total PAH concentrations reported by the EcoChem CEM for the
low (top) and intermediate (bottom) SVOC concentration tests.
73
-------
However, as also noted in Section 4.1.2, the step changes in organic spiking solution feedrate
induced apparently did not cause corresponding step changes in monitored flue gas concentrations.
Nevertheless, the data in Figure 12 for the low concentration test show a relatively rapid dropoff
in the CEM reading beginning about 10 minutes after the scrubber inlet step change, which was
terminating SVOC spike solution feed. The drop off leveled about 30 minutes after the step
change. A second period of rapidly decreasing CEM reading began about an hour after the step
change.
In contrast, the data for the intermediate concentration test in Figure 12 show that the
CEM reading abruptly increased within a few minutes of the first step change. Interestingly, this
first step change was a decrease in SVOC spiking solution feedrate by half. Only about an hour
after the first step change did the CEM reading decrease substantially. This was right at the time
of the second step change for this test, the stopping of spike solution feed. While the EcoChem
PAH CEM response time could not be well-evaluated in the manner planned, an estimate of CEM
response time can be obtained from the data in Figure 12. These data suggest that the EcoChem
PAH CEM has a response time on the order of a minute or less.
The Figure 12 plots, as well as additional data and plots given in the EcoChem final
report included in Appendix H-2-1, show the potential usefulness of the EcoChem PAH CEM as
a combustion process operation monitor. The data and plots in Appendix H-2-1 especially show
that the CEM signal was quite sensitive to transients in the RKS operation.
EcoChem measured ZD on the two test days the CEM was in operation, and reported
these ZDs in their final report as -0.001 pico amp (pA) on the low concentration test day and 0
on the intermediate concentration test day. CD was not measured because no calibration standard
for particulate-bound PAH is available. EcoChem translated CEM signals into monitored PAH
concentrations using a calibration curve developed over many past tests of the instrument. This
calibration curve is linear and has a slope of 1 to 3 ng/m3 PAH per pA of instrument signal. A
slope of 1 |ig/m3/pA was used by EcoChem for these tests. EcoChem notes that site-specific CEM
calibration can be performed, but this requires repeated RM measurements from the particular
source application at varying flue gas PAH concentrations. This effort-intensive exercise was not
performed as part of this test program.
The absence of a daily calibration check raises the question of what reference
concentration should be used in calculating the normalized (percent) ZD using the equation
presented in Section 4.1.2. Normally the reference concentration is the concentration of the
highest calibration standard used. In the absence of such a standard, it was decided to use the
mean flue gas concentration measured on a given test day. On this basis, the ZD of -0.001 pA
on the low concentration test day corresponds to a ZD of -0.012 percent. The ZD on the
intermediate concentration test day remains 0 percent.
The EcoChem specifications for the PAS 1000c CEM (see Appendix H-2-1) note that the
most sensitive instrument range is 0 to 2 pA output. With an estimated of DL of 2 percent of
span, the EcoChem PAH CEM would have a DL in the 0.04 (ig/m3 range.
74
-------
4.3 MULTI-METALS CEM TESTS
The multi-metals CEM tests were performed during Test Series 2. The participating
multi-metals CEM developers were Sandia National Laboratories (SNL) and Metorex Incorporated.
Like all other test series, each day of multi-metals CEM testing consisted of performing three
sequential RM measurements, while the tested CEMs were in operation, and the test series
consisted of three test days at each of three different multi-metals concentrations: low,
intermediate, and high.
Detailed descriptions of the principles of operation of both the SNL and Metorex CEMs
are included in their respective final reports (Appendices H-3-1 and H-3-2). Briefly, the SNL
CEM uses an optical technique known as Laser Spark Spectroscopy (LASS), which has also been
referred to as Laser-Induced Breakdown Spectroscopy (LIBS). In LASS, a high-energy pulsed
laser is focused on a small volume in the flue gas stream. The high energy density in the focal
region generates a laser-induced plasma, or "laser spark," in which particles and molecules are
decomposed into excited atoms and ions. Metal concentrations in the laser plasma volume can
then be measured via atomic emission spectroscopy.
The prototype Metorex CEM consists of a flue gas sampling module and a sample
analysis module. These modules perform two different but complementary functions. The
sampling module aspirates the flue gas through a heated dilution probe and directs the gas flow
through a quartz or Teflon membrane filter. The total volume of the gas passing through the filter
is measured. The analysis module then uses x-ray fluorescence (XRF) analysis to determine the
concentrations of metals collected in the particulate on the filter.
Tables 41 through 43 summarize the results of the three sequential RM measurements
performed each test day, and compare these to the SNL and Metorex CEM measurements.
Table 41 summarizes the low concentration test day results, Table 42 the intermediate
concentration test day results, and Table 43 the high concentration test day results. As indicated
in the tables, the Metorex instrument did not measure beryllium or mercury. The SNL CEM did
not detect any of the test trace metals on the low concentration test day; only arsenic, barium, and
lead were reported on the intermediate concentration test day; and only antimony, arsenic, barium,
and lead for one or more RM periods were reported on the high concentration test day.
Table 44 compares the flue gas metals concentrations, as measured by the RM, to the
respective target concentrations listed in Table 4. As indicated, the measured concentrations for
the low target concentration were within a factor of about 5 of the target concentration, with the
ratio of the measured concentration to the target concentration ranging from 20 to 550 percent.
Corresponding ratios ranged from 25 to 145 percent for the intermediate target concentration test,
and, excluding silver, from 18 to 111 percent for the high target concentration test. Flue gas silver
concentrations measured by the RM are highly suspect. As discussed in Section 6.3, spike
recoveries from QA samples were quite poor for silver.
The RAs corresponding to the measurement pair data in Tables 41 through 43 are
summarized in Table 45. Neither beryllium nor mercury is included in Table 45 because neither
75
-------
TABLE 41. MEASURED FLUE GAS CONCENTRATIONS FOR THE TESTS
OF THE SNL AND METOREX CEMS AT THE LOW METALS
CONCENTRATIONS
Concentration, pg/dscm
Reference Method 1 Reference Method 2 Reference Method 3
Metal
RM
SNL
Metorex
RM
SNL
Metorex
RM
SNL
Metorex
Antimony (Sb)
4.5
NDa
ND
5.1
ND
ND
4.5
ND
5.13
Arsenic (As)
4.4
ND
3.65
3.8
ND
0.83
3.6
ND
1.19
Barium (Ba)
11.7
ND
ND
15.8
ND
ND
18.6
ND
6.23
Beryllium (Be)
0.1
ND
NMb
0.1
ND
NM
0.1
ND
NM
Cadmium (Cd)
9.7
ND
2.63
12.1
ND
ND
13.2
ND
10.02
Chromium (Cr)
22.3
ND
2.49
23.5
ND
0.56
28.0
ND
22.29
Cobalt (Co)
7.8
ND
12.11
7.1
ND
ND
7.1
ND
14.68
Lead (Pb)
101
ND
11.51
85.6
ND
9.06
110
ND
12.36
Manganese (Mn)
21.8
ND
5.89
29.2
ND
ND
31.6
ND
19.43
Mercury (Hg)
14.9
ND
NM
17.2
ND
NM
11.2
ND
NM
Nickel (Ni)
39.6
ND
27.52
29.1
ND
6.15
42.4
ND
21.87
Selenium (Se)
11.4
ND
1.51
12.3
ND
1.47
12.3
ND
3.62
Silver (Ag)c
2.9
ND
0.98
4.8
ND
0.91
4.5
ND
0.99
Thallium (Tl)
1.1
ND
ND
1.5
ND
ND
1.7
ND
ND
aND = Not detected.
*>NM = Not measured.
CRM data for silver not reliable due to low spike recoveries.
76
-------
TABLE 42. MEASURED FLUE GAS CONCENTRATIONS FOR THE TESTS
OF THE SNL AND METOREX CEMS AT THE INTERMEDIATE
METALS CONCENTRATIONS
Concentration, jig/dscm
Reference Method 1 Reference Method 2 Reference Method 3
Metal
RM
SNL
Metorex
RM
SNL
Metorex
RM
SNL
Metorex
Antimony (Sb)
11.0
NDa
39.73
11.6
ND
22.00
9.5
ND
8.49
Arsenic (As)
11.1
63
11.92
10.8
42
0.74
8.7
115
6.61
Barium (Ba)
78.0
251
44.37
80.0
199
11.87
49.2
463
9.92
Beryllium (Be)
0.6
ND
NMb
0.6
ND
NM
0.4
ND
NM
Cadmium (Cd)
14.0
ND
7.22
15.0
ND
26.93
14.2
ND
10.09
Chromium (Cr)
54.7
ND
56.48
59.5
ND
72.51
50.3
ND
25.68
Cobalt (Co)
32.3
ND
14.72
33.9
ND
20.79
27.4
ND
9.79
Lead (Pb)
141
144
107.07
141
93
51.80
136
106
40.86
Manganese (Mn)
24.2
ND
61.4
24.6
ND
55.58
18.2
ND
31.25
Mercury (Hg)
54.3
ND
NM
83.7
ND
NM
75.3
ND
NM
Nickel (Ni)
59.9
ND
26.48
61.2
ND
21.26
52.6
ND
15.76
Selenium (Se)
43.2
ND
29.34
54.5
ND
21.27
53.2
ND
18.12
Silver (Ag)c
5.0
ND
20.90
7.9
ND
12.77
6.9
ND
6.05
Thallium (Tl)
11.1
ND
12.96
11.2
ND
4.49
12.4
ND
2.72
aND = Not detected.
hNM = Not measured.
CRM data for silver not reliable due to low spike recoveries.
77
-------
TABLE 43. MEASURED FLUE GAS CONCENTRATIONS FOR THE TESTS
OF THE SNL AND METOREX CEMS AT THE HIGH METALS
CONCENTRATIONS
Concentration, pg/dscm
Reference Method 1 Reference Method 2 Reference Method 3
Metal
RM
SNL
Metorex
RM
SNL
Metorex
RM
SNL
Metorex
Antimony (Sb)
114
233
27.32
75.7
186
18.35
43.5
131
6.59
Arsenic (As)
82.2
75
21.82
64.8
86
13.28
54.8
65
4.68
Barium (Ba)
331
650
207.37
484.3
ND
111.07
285
ND
27.29
Beryllium (Be)
12.1
NDa
NMb
6.8
ND
NM
4.0
ND
NM
Cadmium (Cd)
88.0
ND
33.58
60.9
ND
31.73
88.7
ND
22.18
Chromium (Cr)
425
ND
129.33
299
ND
91.85
241
ND
34.70
Cobalt (Co)
357
ND
100.16
229
ND
67.47
248
ND
37.62
Lead (Pb)
1.650
ND
297.20
1,082
ND
282.71
2,176
54
167.04
Manganese (Mn)
179
ND
52.89
89.9
ND
35.25
95.6
ND
16.56
Mercury (Hg)
625
ND
NM
451
ND
NM
581
ND
NM
Nickel (Ni)
550
ND
160.10
347
ND
111.42
429
ND
67.89
Selenium (Se)
421
ND
102.52
399
ND
96.70
383
ND
39.92
Silver (Ag)c
5.0
ND
33.60
8.1
ND
25.60
7.1
ND
14.88
Thallium (Tl)
114
ND
32.29
94.3
ND
28.10
113
ND
16.84
aND = Not detected.
bNM = Not measured.
CRM data for silver not reliable due to low spike recoveries.
78
-------
TABLE 44. COMPARISON OF MEASURED AND TARGET CONCENTRATIONS FOR THE MULTI-METALS
CEM TESTS
Concentration, (ig/dscm
Metal
Low concentration test
Intermediate concentration test
High concentration test
Average
RM
Nominal
target
Ratio,
RM/target
Average
RM
Nominal
target
Ratio,
RM/target
Average
RM
Nominal
target
Ratio,
RM/target
Antimony (Sb)
4.7
10
0.47
10.7
40
0.27
77.7
400
0.19
Arsenic (As)
3.9
5
0.78
10.2
20
0.51
67.3
200
0.34
Barium (Ba)
15.4
50
0.31
69.1
200
0.35
367
2000
0.18
Beryllium (Be)
0.1
1
0.20
0.5
2
0.25
7.6
20
0.38
Cadmium (Cd)
11.7
5
2.34
14.4
20
0.72
79.2
200
0.40
Chromium (Cr)
24.6
20
1.23
58.2
80
0.73
322
560
0.57
Cobalt (Co)
7.3
10
0.73
31.2
40
0.78
278
400
0.70
Lead (Pb)
98.9
50
1.98
139
200
0.70
1636
2000
0.82
Manganese (Mn)
27.5
5
5.50
22.3
20
1.12
122
200
0.61
Mercury (Hg)
14.4
20
0.72
71.1
80
0.89
552
800
0.69
Nickel (Ni)
37.0
10
3.70
57.9
40
1.45
442
400
1.11
Selenium (Se)
12.0
50
0.24
50.3
200
0.25
401
2000
0.20
Silver (Ag)
4.1
5
0.82
6.6
20
0.33
6.7
200
0.03
Thallium (Tl)
1.4
5
0.28
11.6
20
0.58
107
200
0.54
-------
TABLE 45. RELATIVE ACCURACIES OF THE SNL AND METOREX CEMs
RA, %
SNL
Metorex
Test concentration
Test concentration
Metal
Low
Intermediate
High
Low
Intermediate
High
Antimony (Sb)
NCa
NC
188
NC
467
158
Arsenic (As)
NC
1,560
65
125
174
101
Barium (Ba)
NC
905
NC
NC
135
153
Cadmium (Cd)
NC
NC
NC
89
177
123
Chromium (Cr)
NC
NC
NC
158
94
113
Cobalt (Co)
NC
NC
NC
236
72
118
Lead (Pb)
NC
64
NC
115
112
177
Manganese (Mn)
NC
NC
NC
116
261
146
Nickel (Ni)
NC
NC
NC
88
77
121
Selenium (Se)
NC
NC
NC
104
113
93
Thallium (Tl)
NC
NC
NC
NC
171
111
Average*1
—
843
127
129
168
129
Median^
—
905
65, 188
116
135
121
aNC = Not calculated.
^Average and median exclude RAs not calculated (NC).
CEM tested measured these two metals. In addition, results for silver are not included in Table 45
because the RM concentrations are suspect.
As was done for the VOC CEMs discussed in Section 4.1, no RA was calculated in cases
where the CEM reported nondetectable concentrations for two or more RM periods on a given test
day. In cases where only one nondetectable level over the three RM periods in a test day was
reported, the metal's reported DL was used for the RA calculation.
The data in Table 45 show that the RAs for the SNL CEM ranged from 64 to
1,560 percent, for the three metals reported on the intermediate concentration test day, and from
65 to 188 percent, for the two metals reported on the high concentration test day. RAs for the
Metorex CEM ranged from 88 to 236 percent, with an average of 129 percent and a median of
116 percent for the low concentration test. Corresponding RAs for the intermediate concentration
test were 72 to 467 percent, with an average of 168 percent and a median of 135 percent.
Corresponding RAs for the high concentration test were 93 to 177 percent, with an average of
129 percent and a median of 121 percent. The RAs for the Metorex CEM were comparable for
each test concentration. No marked RA improvement, as flue gas concentration increased, as
observed for the VOC CEMs, is seen in the Metorex CEM data.
80
-------
The ANOVA test indicated that the mean recoveries of the Metorex CEM were
statistically equal, at all three concentrations tested, for arsenic, cadmium, chromium, cobalt, lead,
nickel, and selenium. Furthermore, the ANVOA test confirmed that the mean recoveries of the
Metorex CEM were the same for the two test concentrations with detectable CEM measurement
data (the intermediate and high concentrations) for antimony, barium, and thallium. Table 46
summarizes the mean bias and precision estimates for these 10 metals, based on the pooled six
(antimony, barium, and thallium) to nine (other seven metals) CEM/RM measurement pairs. The
data in Table 46 show that the absolute mean biases of the Metorex CEM were less than
75 percent, with 95-percent confidence limits generally between a fifth to on the order of the
corresponding absolute bias (except for antimony). With the exception of antimony, the standard
deviations of the mean CEM recovery were, at the most, 66 percent; most were less than
40 percent.
Tabic 47 summarizes the bias and precision estimates for the Metorex CEM for
manganese. Means recoveries for this metal were not the same in the three test concentrations
according to the ANOVA test.
TABLE 46. METOREX CEM BIAS AND PRECISION ESTIMATES FOR
ANTIMONY, ARSENIC, BARIUM, CADMIUM, CHROMIUM,
COBALT, LEAD, NICKEL, SELENIUM, AND THALLIUM
Metal
Parameter3
Mean bias,
%
Standard deviation,
%
95-percent
confidence
interval, %
Antimony (Sb)b
17
137
143
Arsenic (As)
-57
36
28
Barium (Ba)^
-69
23
69
Cadmium (Cd)
-33
47
36
Chromium (Cr)
-51
43
33
Cobalt (Co)
-31
66
51
Lead (Pb)
-75
21
17
Nickel (Ni)
-64
16
13
Selenium (Se)
-72
18
14
Thallium (Tl)b
-58
38
39
aBased on nine pooled CEM/RM measurement pairs except as noted.
bBased on six pooled CEM/RM measurement pairs.
81
-------
TABLE 47. METOREX CEM BIAS AND PRECISION ESTIMATES
FOR MANGANESE
Test concentration
Low Intermediate High
Manganese (Mn)
Mean bias, % -64 117 -71
Standard deviation, % 23 42 11
95-percent confidence interval, % 56 104 27
SNL reported only arsenic's concentration on more than one test day. The ANOVA test
confirmed that the mean arsenic recoveries for the SNL CEM were statistically equal at the two
test concentrations reported. Based on the six pooled CEM/RM measurement pairs for arsenic,
the mean bias was 337 ±496 percent at the 95-percent confidence level. The standard deviation
of the arsenic recoveries was 472 percent. Mean bias and precision estimates for the SNL CEM
for barium and lead, at the intermediate test concentration, and for antimony, at the high test
concentration, are summarized in Table 48.
Figures 13 through 15 show the flue gas metals concentrations reported by the Metorex
CEM for the low, intermediate, and high concentration test days, respectively. Data to allow
assembling corresponding plots for the SNL CEM for the two or three metals detected on the
intermediate and high concentration test days were not available. Figures 13 and 14 show that the
Metorex instrument reported one to two concentration measurements per hour for the low and
intermediate concentration tests; Figure 15 shows this frequency increased at up to three
measurements per hour for the high concentration tests. As in preceding plots, the RM sampling
periods are indicated in the figures.
TABLE 48. SNL CEM BIAS AND PRECISION ESTIMATES FOR ANTLMONY,
BARIUM, AND LEAD
Test
95-percent
concentration
Mean bias, Standard deviation.
confidence interval,
Metal
reported
% %
%
Antimony (Sb)
High
150 49
121
Barium (Ba)
Intermediate
404 380
945
Lead (Pb)
Intermediate
-18 18
45
82
-------
RM1
RM2
RM3
Step -
change
.-4,
As
-
Ba
Cr
£
Pb
- -
Mn
- -
Mi
&
10
12 14 16
Time of day
18
20
22
30
m 50
E
jf 40
O
<0
c
CD
O
c
o
o
520
"S
E
ro 10
CD
/
; /
• /
¦/
RM3
Step
change
V.
. .
Sb
Cd
Co
Se
Ag
¦ * -
n
A
10 12 14 16
Time of day
18
20
22
Figure 13. Flue gas metals concentrations reported by the Metorex CF.M for the
low metals concentration tests.
83
-------
to
500
O)
=i.
c 400
o
15
300 h
d
CD
O
c
o
o
75 200
-4—'
-------
o 500
E
D)
5L
C 400
0
"S
1 300
o
c:
O
o
"S 200
co
4-»
E
(8 100
CD
CD
LL n
RM1
RM2
RM3
•
•
—
First step
change
* •
\ •-
i
i
Second
/ ^
\ *' •'
\ :
\ ! " •
-
/ s,eP
; change
r. .1*
•
r
*' '* \
S
\ A
)L. - - -*
-• " A V- -
\ ~
*
*- .. -¦
¦" ¦-
/ " A
*
~ "
A
A
\
A
»'
li:»
As
Ba
Cr
A
Pb
- -•
Mn
Ni
10 12 14
Time of day
16
18
o 140
E
5 120
c
.2
CO
100
§ 80
c
o
U 60
¦2
0)
E
40 -
CO
CO
D) 20
0
3
RM1
RM2
RM3
•- •
V' *\ *
» * W*
• • v-
y-y*' ^
' ¦- * '-IK-*
V .
V
First step
change
Second
' step
change
Sb
Cd
Co
Se
- -
Ag
- -
n
10 12 14
Time of day
16
18
Figure 15. Flue gas metals concentrations reported by the Metorex CEM for the
high metals concentration tests.
85
-------
Step changes in metal spike addition rate were induced at the end of the test day, as was
done during the VOC and SVOC CEM tests, in an attempt to evaluate CEM response time. On
the low concentration test day, the only step change performed consisted of stopping metal spike
solution feed. On the intermediate concentration test day, the first step change consisted of
substituting the intermediate concentration metals spike solution with the high concentration
solution, and the second step change consisted of stopping spike solution feed. On the high
concentration test day, the first step change was to decrease the metals spike solution feedrate by
half, and the second step change was to stop spike solution feed. However, as for other test series,
this exercise yielded only qualitative response time estimates at best.
The SNL final report does not include metals concentration data for any step change
periods. The Metorex data in Figure 13 shows that no concentration measurements were taken
after the step change on the low concentration test day. The data in Figure 14 show that measured
metals concentrations increased, some quite sharply, after the first step change for 11 of the 12
metals on the intermediate concentration test day reported. Similarly measured concentrations
decreased, some quite sharply, after the second step change. The data in the figure suggest that
the system response time to a step change in feed metals feedrate was shorter than that indicated
for the VOC constituents. Scrubber exit flue gas metals concentrations show much quicker
response to feedrate step changes as measured by the Metorex CEM. Response times on the order
of 15 minutes are suggested by the data in Figure 14.
In contrast, the data in Figure 15 show that the metals concentrations reported by the
Metorex CEM for the high concentration test were not consistently changed following the first step
change. No measurements were taken after the second step change on the high concentration test
day.
Neither SNL nor Metorex included ZD or CD data in their respective final reports.
SNL reported DL values for each of the test metals based on laboratory measurements.
Metorex established DLs for the test metals by sampling a particulate-free gas and calculating
3 times the standard deviation of the measured concentration of this blank sample; these DLs are
summarized in Table 49. Based on the test data in terms of each instrument's ability to detect a
flue concentration measured by the RM, the SNL estimates of their CEM's DL are consistent with
the test data for antimony, arsenic, barium, lead, and thallium. However, the SNL DL estimates
for beryllium, cadmium, chromium, cobalt, manganese, nickel, and selenium seem optimistic. Had
their estimated DLs been achieved for these metals, the SNL CEM should have measured the flue
gas concentrations present on one or more of the test days.
Test data are consistent with the DLs reported by Metorex for most metals. In fact, in
several cases Metorex reported a flue gas concentration at a level lower than the stated DL. The
DLs for barium and cobalt stated by Metorex are perhaps slightly optimistic. The test data suggest
that the DLs for these two metals are closer to about 10 pg/dscm than to the values listed in
Table 49.
86
-------
TABLE 49. DETECTION LIMITS REPORTED BY SNL AND METOREX
DL, jig/dscm
Metal
SNLa
Metorex
Antimony (Sb)
20
6.01
Arsenic (As)
10
2.51
Barium (Ba)
100
6.63
Beryllium (Be)
0.2
NMb
Cadmium (Cd)
1
9.59
Chromium (Cr)
0.4
5.76
Cobalt (Co)
1
3.25
Lead (Pb)
100
2.72
Manganese (Mn)
0.2
5.43
Mercury (Hg)
10
NM
Nickel (Ni)
20
2.80
Selenium (Se)
100
1.19
Silver (Ag)
100
4.57
Thallium (Tl)
100
3.25
aConverted from pg/m3 using test condition temperature, pressure,
and moisture content.
^NM = Not measured.
4.4 MERCURY CEM TESTS
The mercury CEM tests were performed during Test Series 3. The participating mercury
CEM developers were Perkin-EImer, Senova Corporation, and EcoChern Technologies. Euramark
was a fourth mercury CEM developer originally scheduled to participate in the program as well.
However, the Euramark CEM was damaged in transit to the IRF, and repair efforts by the
Euramark team were unsuccessful.
Detailed descriptions of the principles of operation of each CEM tested are included in
each respective developer's final report (Appendices H-4-1, H-4-2, and H-4-3). Briefly, the
EcoChem CEM, termed the Hg-Mat 2 Analyzer by EcoChern, provides a continuous measure of
flue gas total mercury concentration. Sample gas is transported through a heated (200°C) sample
line to the analyzer. Sample gas is first conditioned by passing through two reactors that convert
ionic mercury to elemental mercury and desorb any particle-bound mercury. The mercury
concentration of the sample gas is subsequently measured using CVAAS.
87
-------
The Perkin-Elmer CEM, termed the MERCEM by Perkin-Elmer, also uses a heated
(185°C) sample line to transport sample gas to the analyzer. The sample gas probe for the
MERCEM includes two sintered metal filters for fine particulate removal. The MERCEM also
converts ionic mercury to elemental mercury via reaction with stannous chloride (SnCl2) solution.
However, the MERCEM also includes a gold trap for mercury concentration via amalgamation.
Including the gold trap allows for a lower detection limit. Mercury quantitation is also by
CVAAS.
The Senova CEM differs from the other two in that the conversion of ionic mercury to
elemental mercury is accomplished via a proprietary solid reactor bed, and the mercury
concentration measurement is performed via solid-state noble metal thin-film microsensors.
Mercury in the conditioned sample gas adsorbs onto the noble metal thin film, causing a change
in the film's electrical resistance. This change is directly related to the mercury concentration.
Depending on the size and mercury concentration of the analyzed sample, a number of samples
can be analyzed before the thin film becomes saturated. When saturated, the thin film is thermally
regenerated and a new cycle of gas sample analysis begins. The thin-film microsensor also
requires that moisture and acid gases be removed from the sample gas. This is also done in the
sample conditioning system via a dryer and a solid acid gas sorbent bed. This CEM also features
a provision for including a gold trap for mercury concentration, if needed.
Table 50 summarizes the results of three sequential RM measurements performed on each
mercury CEM test day, and compares these to the corresponding three mercury CEM
measurements. Calculated RAs for each CEM are also given in the table for the three test days,
each representing a different flue gas mercury concentration.
Table 51 compares the flue gas concentration measured by the RM at each test
concentration to the corresponding target concentration discussed in Section 2.2.2. As noted,
actual achieved concentrations were greater than, up to more than twice, target concentrations.
The data in Table 50 indicate several periods during which the Perkin-Elmer and Senova
CEMs were not in operation. The Perkin-Elmer team was unable to obtain reliable results during
the second RM period on the first test day (intermediate concentration) because of a buildup of
a white powder that clogged the probe's sintered metal filter. On the second day of testing (high
concentration), the Perkin-Elmer CEM developed a defect at the sample drain pump and a broken
probe fitting, so no data were obtained for the first two RM periods on this test day. As a result,
no RA could be calculated for this Perkin-Elmer CEM for the high concentration test, and the RA
in Table 50 for the intermediate concentration test is based on only two pairs of RM/CEM
measurements.
In the Senova case, a critical part of the Senova CEM was broken during the CEM's
packing for shipment to the IRF. The time required to locate and secure a replacement part caused
a delay in the Senova team's arrival at the IRF such that the first test day, at the intermediate flue
gas mercury concentration, was missed. The Senova CEM was in operation on the second test
day, at the high flue gas mercury concentration. However, a malfunction that caused unstable
sample gas flow to the analyzer prevented Senova from obtaining valid data on the third and last
88
-------
TABLE 50. MEASURED FLUE GAS CONCENTRATIONS AND RAs FOR THE
MERCURY CEM TESTS
Mercury concentration, jig/dscm
Test
RM
Perkin-Elmer
CEM
Senova
CEM
EcoChem
CEM
Low mercury concentration
RM 1 21
RM 2 16
RM 3 13
RA, %
Intermediate mercury concentration
RM 1 56
RM 2 34
RM 3 40
RA, %
High mercury concentration
RM 1 119
RM 2 94
RM 3 86
RA, %
78
42
11
602
61
NO
125
1,150
NO
NO
405
NC
NOa
NO
NO
NCb
NO
NO
NO
NC
232
116
165
186
22
20
19
60
83
43
56
92
137
81
62
61
aNO = Not operational.
bNC = Not calculated.
TABLE 51. COMPARISON OF MEASURED AND TARGET MERCURY
CONCENTRATIONS FOR THE MERCURY CEM TESTS
Concentration, fig/dscm
Test concentration Average RM Target Ratio, RM/Target
Low 17 10 1.7
Intermediate 43 20 2.2
High 100 80 1.2
89
-------
day of testing at the low flue gas mercury concentration. Thus, the Senova team was able to
obtain only one day of test results.
The data in Table 50 show that the EcoChem CEM had an RA of about 60 percent for
both the low and high concentration tests. The RA at the intermediate concentration was
increased, at 92 percent. The RA of the Perkin-Elmer CEM was 602 percent at the low mercury
concentration, and 1,150 percent (based on two measurement pairs) at the intermediate mercury
concentration. The RA of the Senova CEM was 186 percent at the one test concentration for
which data were collected.
Table 50 indicates that the EcoChem CEM reported a full set (nine CEM/RM pairs) of
test data. The ANOVA test confirmed that the mean EcoChem CEM recoveries were statistically
the same at all three test concentrations. The Perkin-Elmer CEM reported sufficient data to
calculate an RA at two test concentrations. The ANOVA test confirmed that mean recoveries were
the same at the two test concentrations reported by Perkin-Elmer. The Senova CEM reported data
at one test concentration. Table 52 summarizes the bias and precision estimates for the three
mercury CEMs. These estimates are based on all nine pooled CEM/RM data pairs for the
EcoChem CEM, on five pooled data pairs for the Perkin-Elmer CEM, and on the three high
concentration test data pairs for the Senova CEM. The table shows that the mean bias of the
EcoChem CEM was 18 ±20 percent at the 95-percent confidence level, with a standard deviation
of the mean recovery of 26 percent. Corresponding values for the Perkin-Elmer and Senova
CEMs were 128 ±157 percent/126 percent and 70 ±100 percent/40 percent, respectively.
Figures 16 through 18 show the flue gas mercury concentrations reported by the
EcoChem, Perkin-Elmer, and Senova CEMs, respectively. Periods of RM sampling are indicated
in each figure. In addition, as done in Figure 12 for the SVOC CEM, the position of the line
denoting each RM period in Figures 16 through 18 along each figure's y-axis corresponds to the
RM measurement concentration. The figures graphically illustrate each CEM's minute-to-minute
readings and their comparison to corresponding RM results.
TABLE 52. BIAS AND PRECISION ESTIMATES FOR THE
THREE MERCURY CEMs TESTED
Parameter
Perkin-Elmer
CEM
Senova
CEM
EcoChem
CEM
Number of pooled measurement
pairs used
5
3
9
Mean bias, %
128
70
18
Standard deviation, %
126
40
26
95-percent confidence interval, %
157
100
20
90
-------
40
Ui 35
O 30
£ 25
15
RM3
cn
© 10
5
20
e
Time of day
140
cn 1?0
5. i
O IX
80
C 150
O
c
©
u 100
o
Q
05
X
<2 50 ;
©
J3
U_
bm; V
Fint *!<-p ef.angf
V, r.
fi —J .) , RMJ
m,Jkr
J y
Second aep ditngc
TihifdslepcbJinje
to 12
i« 16 18
Time of day
23
Figure 16. Flue gas mercury concentrations reported by the EcoChem CEM for the low
(top), intermediate (middle), and high (bottom) mercury concentration tests.
91
-------
c
O
c
o
u
OJ
X
m
to
Ui
©
_3
LL
160
140
120
100
60
60
40
20
0
VWwv^AV
RM2 I
14 16
Time of day
n
E
O) 60C
a
c
o
500
?
c
400
0)
o
c
o
300
u
D)
X
200
u>
CO
D)
100
~
U-
10
A'
t?
Si*p w
ctiaog?
\!\ , ,
!M1 *' »
/ !/
14 16 18
Time of day
20
22
600
n
E
a>
a
500
c
o
m
400
c
e
Second
ttrp rtiarj-c
Third
Mrp
change
10
12 14 16
Time of day
10
Figure 17. Flue gas mercury concentrations reported by the Perkin-Elmer CEM for the
low (top), intermediate (middle), and high (bottom) mercury concentration tests.
92
-------
1.000
o>
a
c
o
600
C
©
O
C
o
O 4i
O)
X
i/i
O) 20C .
©
D ;
a,VV
- Rntitepcittftgc
,V
'VvTJ
<;v\M
12 14 16
Time of day
- Secowl
siepcfciaie
Third
ittp
dnitfe
20
Figure 18. Flue gas mercury concentrations reported by the Senova
CEM for the high mercury concentration test.
As in all other test series, inducing step changes in monitored flue gas mercury
concentration to evaluate CEM response time was attempted at the end of each test day. On the
first lest day, at the intermediate flue gas mercury concentration, the step change consisted of
stopping metals (including mercury) spike solution feed. On the second test day, at the high flue
gas mercury concentration, the first step change consisted of replacing the high concentration
metals spike solution with the intermediate concentration spike solution. The second step change
on this test day consisted of reducing the intermediate concentration spike solution feedrate by
half. Also on this test day, a third step change, consisting of terminating spike solution feed, was
performed. On the third test day, at the low flue gas mercury concentration, the first, and only,
step change consisted of stopping spike solution feed.
CEM responses to these attempts to induce flue gas mercury concentration step changes
are shown in Figures 16 through 18. The data in Figures 16 (low concentration test) and 17 (low
and intermediate concentration tests) suggest that the system response time (the time between a
feed composition step change and the results of this step change being manifest at the scrubber
exit) was about 10 to 25 minutes. In these three cases, respective CEM readings increased after
the step change occurred for the above time period, then decreased sharply. The effects of
gradually changing scrubber exit flue gas mercury concentrations cannot be separated from
instrument response time effects. Nevertheless, the data in the figures can be used to estimate the
upper bound instrument response time. This upper bound is in the 25- to 50-minute range for the
EcoChem CEM and in the 25- to 35-minute range for the Perkin-Elmer CEM.
Each mercury CEM developer performed pre- and post-test calibrations and zero checking
for their instrument on the days the instrument was in operation, and used these data to calculate
ZD and CD. Reported results are given in Table 53.
93
-------
TABLE 53. MEASURED MERCURY CEM ZDs AND CDs
Instrument drift, %
Test
Perkin-EImer Senova EcoChem
CEM CEM CEM
Low mercury concentration
ZD
CD
Intermediate mercury concentration
0.14
21
NOa
NO
0
0
ZD
CD
0.13
2.8
NO
NO
0
-10
High mercury concentration
ZD
CD
NO
NO
-4.8
4.3
0
-23
aNO = Not operational.
All three mercury CEM developers state the capability to detect flue gas mercury
concentrations down to the 2- to 3-^jg/dscm range. Test data from the low concentration test day
confirm that the EcoChem and Perkin-EImer CEMs can quantitate mercury in the 10- to
20-jjg/dscm range.
4.5 RELATIVE ACCURACY DISCUSSION
The data presented and discussed in the preceding sections show that the measured RAs
of the CEMs tested were often quite large. In fact, even the RAs for the best performing CEMs,
based on qualitative comparisons of CEM readings to RM measurements, were generally no better
than of the order of 100 percent. By comparison, EPA performance specifications (PSs) for
combustion gas CEMs, 02 for example, are generally in the range of 20-percent RA. This raises
questions concerning whether CEMs for mercury, other trace metals, and organic compounds, even
after further development, can potentially meet a 20-percent RA PS.
A partial answer to this question may lie in the statistics of the RA definition. The draft
PS given in Appendix A notes that RA is defined as follows:
\d\ ~ SD
,./7
(5)
RA =
m
94
-------
Where:
The absolute value of the mean difference between the CEM reading and the
corresponding RM measurement
Number of pairs of CEM/RM measurements taken
RM = Mean RM measurement value
SD = Standard deviation of the differences between the CEM reading and the RM
measurement
l0 975 ~ f-statistic at the 2.5 percent error confidence
The RATA of the combustion gas CEM requires that nine CEM/RM measurement pairs
be the minimum taken to perform an RA calculation. For nine measurement pairs, tQ 975 is 2.306
and
'°'975 - 0.769 (6)
While the tests in this program were designed to collect nine pairs of CEM/RM measurements,
the nine pairs were collected over three flue gas concentrations. Thus, each RA calculation
performed used at most three CEM/RM measurement pairs. For three measurements, t0 97j is
4.303 and
f()'975 - 2.484 (7)
Thus, it is apparent that the RAs calculated based on three CEM/RM measurement pairs can be
large, being "penalized" by the much larger tQ 975 lyfn when n is 3 compared to when n is 9.
95
-------
SECTION 5
CONCLUSIONS
In a recently completed test program at the EPA's IRF, 10 CEMs designed to measure
trace metal or trace organic species concentrations were tested. Some of these devices were
prototypes, others were in the developmental stage. Of the 10 CEMs tested, four measured
incinerator flue gas concentrations of several specific VOCs, one measured total particulate-bound
PAH concentrations, two measured flue gas concentrations of several (up to 14) trace metals, and
three measured mercury concentrations.
While the testing consisted of obtaining quantitative measurement data on the four
measures of CEM performance checked in a RATA as described in 40 CFR 60, Appendix F —
RA, CD, ZD, and response time — the primary project objective focused on the RA measurement.
The RA measurement was achieved by comparing the monitored analyte concentrations reported
by the CEM to the concentration determined by the EPA RM for the analyte. EPA draft Method
29 was the RM for trace metals (including mercury); Method 0030, with analysis by Methods 5040
and 8015A, was the RM for VOCs; and Method 0010, with analysis by Method 8270B, was the
RM for PAHs.
Four series of tests were performed, each simultaneously testing up to three monitors
measuring the same or similar analyte types. Each test series consisted of triplicate RM
measurements performed at each of three target flue gas monitored analyte concentrations while
the tested CEMs were in operation. Thus, each series yielded a total of nine RM and CEM
comparisons. All measurements were taken in the wet scrubber exit flue gas from the IRF's pilot-
scale RKS. The test program was performed in August and September, 1995.
The following is a summary of test program conclusions focusing on the measured RAs
of the tested CEMs. The CEM measurement bias and precision were also evaluated. Results of
these evaluations are discussed in Section 4.
The four VOC CEMs tested were:
• A direct sampling ion trap mass spectrometer technique, under development by
ORNL
• The Airsense 500, developed by EcoLogic, Inc., which uses a chemical ionization
mass spectrometer
96
-------
• An online GC technique with dual detectors, an FID and an ECD, developed by
EPA/APPCD
• An electron impact mass spectrometer with dual detectors, developed by Marine
Shale Processors, Inc.
The flue gas VOC analytes were benzene, carbon tetrachloride, ehlorobenzene,
chloroform, 1,2-dichloroethane, 1,1-dichloroethene, tetrachloroethene, toluene, 1,1,1-
trichloroethane, and trichloroethene. The three flue gas concentration ranges tested were 2 to 400,
10 to 500, and 70 to 1,800 |ig/dscm, depending on the individual VOC analyte.
The measured RAs of the ORNL CEM ranged from 123 to 305 percent at the low test
concentration, with an average of 196 percent over the seven compounds reported. ORNL CEM
RAs were improved at the intermediate test concentration, at 113 to 278 percent, with an average
of 154 percent over the nine compounds reported. Still further improvement in this CEM's RAs
were seen at the high test concentration, with an RA range of 84 to 144 percent, and an average
of 105 percent over all 10 compounds reported. In fact, the RAs for all VOCs reported uniformly
improved as the test concentration increased.
The RAs of the EcoLogic CEM ranged from 65 to 7,320 percent, with an average of
2,520 percent at the intermediate test concentration. Much improved performance was seen at the
high test concentration, for which the RAs ranged from 27 to 283 percent and averaged
112 percent. As with the ORNL CEM, the RAs for nine of the 10 VOCs reported were improved
at the high test concentration compared to the RAs at the intermediate test concentration.
EcoLogic did not report CEM concentrations for the low concentration test.
The RAs for the EPA/APPCD CEM ranged from 71 to 3,190 percent, and averaged
638 percent, at the low test concentration. It should be noted, however, that the relatively high
average RA was driven by the two very high RAs for 1,2-dichloroethane and 1,1-dichloroethene.
And, although the RM concentration for 1,1-dichloroethene for the low as well as the intermediate
concentration tests must be viewed with caution due to the problematic nature of its analysis via
the procedure used, the poor RA of the EPA/APPCD CEM for this compound was more likely due
to its presence in the EPA/APPCD CEM system blanks. The median RA for the low concentration
test, at a much improved 113 to 137 percent, removes the dominant influence of the two
aforementioned VOC compounds for which the CEM did poorly. The RAs for the EPA/APPCD
CEM were improved at the intermediate test concentration, ranging from 29 to 1,130 percent and
averaging 213 percent. Poor performance in quantitating 1,2-dichloroethaneand 1,1-dichloroethene
again accounts, in large measure, for the high average RA. As before, the median RA, at 83 to
98 percent, better reflects the mean performance of the CEM by removing the dominant influence
of the RAs for the two poorly quantitated VOC compounds. Further improved performance of the
EPA/APPCD CEM was seen at the high test concentration, with an RA range from 34 to 133
percent and an average RA of 73 percent. In fact, at the high test concentration, the RAs for the
two compounds poorly quantitated at the low and intermediate test concentrations are more in line
with those calculated for the other eight compounds. For this reason, the median RA, at 53 to 70
percent, is comparable to the average RA.
97
-------
Despite having the opportunity to participate in both of the VOC CEM test series, the
MSP CEM was operational for only one test day, that being at the intermediate test concentration
tested during the second series of the VOC CEM tests. Furthermore, the MSP CEM was
operational for only two of the three RM test periods on this test day. The RAs of the MSP CEM,
based on the two available RM/CEM measurement pairs, were quite high, ranging from 314 to
384,000 percent and averaging 51,400 percent. Even the median RAs for the MSP CEM, at 3,150
to 6,740 percent, were quite high.
The SVOC CEM tested was the PAS lOOOe Photoelectric Aerosol Sensor developed by
EcoChem, Inc. This CEM provides a continuous measure of flue gas total particulate-bound PAH.
Three PAH analytes present in the flue gas were naphthalene, phenanthrene, and pyrene. The
three resulting total PAH concentrations tested were about 4, 40, and 260 pg/dscm.
The RAs of the EcoChem CEM were 527 percent at the low test concentration and
99 percent at the intermediate test concentration. This CEM was not operational during the high
concentration test day. The RA measurement at the low concentration may have been adversely
affected by the fact that the MDLs of the RM for other PAH compounds were in the range of the
total PAH concentrations measured. If other PAH compounds were present at concentrations just
under their MDLs in the RM, the actual total PAH concentration may have been larger than
reported by the RM at the low test concentration.
The two multi-metals CEMs tested were:
• The laser spark spectroscopy (LASS) instrument, developed by Sandia National
Laboratories
• A technique involving the collection of a particulate sample on a quartz or Teflon
filter, with subsequent analysis of the collected metal-containing particulate sample
by XRF, developed by Metorex, Inc.
The 14 metal analytes were antimony, arsenic, barium, beryllium, cadmium, chromium,
cobalt, lead, manganese, mercury, nickel, selenium, silver, and thallium. The three flue gas
concentration ranges tested were 0.1 to 110, 0.4 to 140. and 4 to 2,200 pg/dscm, depending on the
specific metal. Silver concentrations, as measured by the RM, were highly suspect, so no
comparison data for silver are reported.
The SNL CEM did not detect any of the test trace metals on the low concentration test
day; only arsenic, barium, and lead were reported on the intermediate concentration test day; and
only antimony, arsenic, barium, and lead, for one or more RM periods, were reported on the high
concentration test day. Corresponding RAs ranged from 64 to 1,560 percent for the three metals
reported on the intermediate concentration test day, and from 65 to 188 percent for two metals
reported on the high concentration test day.
The Metorex CEM did not measure beryllium or mercury. For the other metals, the RAs
of this CEM ranged from 88 to 236 percent, with an average of 129 percent and a median of 116
98
-------
percent for the low concentration test. Corresponding RAs for the intermediate concentration test
were 72 to 467 percent, with an average of 168 percent and a median of 135 percent. For the high
concentration test, corresponding RAs ranged from 93 to 177 percent, with an average of 129
percent and a median of 121 percent. The RAs for the Metorex CEM at each test concentration
were comparable. Unlike the VOC CEMs, no marked RA improvement with increasing flue gas
concentration was seen in the Metorex CEM data.
The three mercury CEMs tested were:
• A CVAAS-based instrument, termed the Hg-Mat 2, and represented in the United
States by EcoChem
• A CVAAS-based instrument, termed the MERCEM, and represented in the United
States by Perkin-Elmer
• A noble metal, thin-film microsensor technology, developed by Senova, Inc.
The three flue gas mercury concentrations tested were 13 to 21, 34 to 56, and 86 to
119 pg/dscm.
The EcoChem CEM demonstrated an RA of about 60 percent in both the low and high
concentration tests. The RA at the intermediate concentration was increased, at 92 percent.
The RA of the Perkin-Elmer CEM was 602 percent in the low concentration test and
1,150 percent in the intermediate concentration test. This CEM was not operational during two
RM sampling periods on the high concentration test day, so no RA calculation was possible.
The RA of the Senova CEM was 186 percent on the high concentration test day, the only
test day on which it was operational.
Test program results indicate that:
• Mercury CEMs are almost to the point of being commercial offerings in the United
States, although operational problems were significant for two of the three
instruments tested. RAs for mercury CEMs of no less than about 100 percent,
based on three measurements, might be the best achievable, however.
• The XRF analysis of a collected particulate sample performed better in yielding
real-time multi-metals concentration data than the LASS technique in these tests.
The LASS CEM experienced detection limit problems such that few concentration
data were obtained. Even with further development, achievable RAs may be no
better than 200 percent for some metals, and no better than 100 percent for any
metal, again based on three CEM/RM measurement pairs.
99
-------
The PAH CEM may see its greatest use as a sensitive indicator of combustion
process operation and efficiency
The EPA/APPCD GC/dual-detector technique performed better in yielding real-time
VOC concentration data than any of the three MS techniques tested, although the
ORNL instrument's performance was comparable. Both the EPA/APPCD and the
ORNL techniques are, at present, real-time, not continuous, and still rely on much
manual operation. Further development work is needed before these instruments
could be considered true CEMs. The other two MS instruments tested offered
continuous concentration measurements, although the MSP instruments suffered
repeated operational problems. All three MS instruments need further verification
testing. Again, the achievable RAs for any of the VOC CEM approaches may be
no better than 100 percent, based on three CEM/RM measurements.
Measured RAs based on three CEM/RM measurement pairs may become
significantly reduced if the PS-required nine measurement pairs are taken
The test program clearly showed the prototype nature of most of the approaches
tested, and the clear need for their further development. As noted above, the
mercury CEMs will require the least development to reach commercial status.
However, the approaches tested for multi-metals and VOC determinations will
likely require one to three years of further development. Given this need, the
importance of continuing test programs of the scope and scale of this one cannot
be overemphasized.
100
-------
SECTION 6
QUALITY ASSURANCE
This test program was carried out as outlined in the "Quality Assurance Project Plan for
Testing the Performance of Real-Time Incinerator Emission Monitors" (Reference 8). The QA
aspects of the program were carried out in accordance with the Quality Assurance Project Plan
(QAPP). Except as noted, all tests were performed in accordance with the procedures documented
in the QAPP.
All RM samples analyzed to obtain data reported in this report were taken at the IRF by
members of the IRF operating staff. All samples were collected and/or recovered in accordance
with the methods appropriate to their eventual analysis. After appropriate preservation, the
samples were relinquished to the custody of the onsite Sample Custodian. The Sample Custodian
subsequently directed the splitting of samples and the transport of these to the appropriate
laboratories for analysis. The sample chain-of-custody procedures described in the QAPP for these
tests were followed. Except as noted no compromise in sample integrity occurred.
For this test program, the critical measurements identified in the QAPP were SVOCs,
VOCs, and trace metals in the flue gas samples.
Numerous QA procedures were followed to assess the data quality of laboratory analytical
measurements performed in this test program. These include blank sample analyses, and matrix
spike (MS) and matrix spike duplicate (MSD) sample analyses. Table 54 summarizes the critical
measurement Quality Assurance Objectives (QAOs) for precision, accuracy, and completeness.
Results of QA procedures performed for the critical laboratory measurements are discussed, by
analyte group in the following subsections.
6.1 VOC ANALYSES
A total of 67 Method 0030 (VOST) samples were analyzed for VOCs by GC/FID using
Method 8015A. Included in this group were five field blanks and eight matrix spikes. All but
one of the 67 VOST samples, or 99 percent, were analyzed within the 42-day method hold time
limit. The Test Series 2 intermediate concentration test VOC field blank sample was broken and
therefore could not be analyzed.
Table 55 summarizes the results of the analyses of the five Method 0030 field blank
sample. The data in the table are in units of jjg/dscm, corresponding to the analysis results (in
101
-------
TABLE 54. PRECISION, ACCURACY, AND COMPLETENESS QAOs FOR CRITICAL MEASUREMENTS
Measurement
parameter
Measurement/analytical
method
Reference
Conditions
Precision,
%
Accuracy,
%
Completeness,
%
SVOCs in Hue
gas samples
Method 0010 sampling, with
collected sample extraction,
concentration, and GC/MS
analysis
SW-846 Methods 0010,
3510B, 3540B, and
8270B
Methylene
chloride
extraction
±50
21-133a
70
VOCs in flue
gas samples
VOST sampling, thermal
desorption purge and trap
GC/FID analysis
SW-846 Methods 0030,
5040A, and 8015A
±50
D-234b.
70
Trace metals
in flue gas
samples
Method 29 sampling, ICP,
CVAAS, and GFAAS
analysis
BIF methods, SW-846
Methods 601 OA,
7470A, and 7841
Microwave
digestion by
BIF methods
±30
60-140
70
aCompound-specific acceptance criteria taken from column 5 of Table 6 of Method 8270B.
hD denotes detected; compound-specific acceptance criteria taken from column 5 of Table 6 of Method 8240B.
-------
TABLE 55. RESULTS OF THE VOC ANALYSES OF METHOD 0030 FIELD BLANK
SAMPLES
Analysis result, pg/dscm
Test Series 1 field blanks
Test Series 4 field blanks
Low
High
Low
Intermediate
High
concentration
concentration
concentration
concentration
concentration
Compound
test
test
test
test
test
Benzene
<0.3
0.5
1.1
<0.3
0.9
Carbon tetrachloride
<0.3
<0.3
4.0
<0.3
1.6
Chlorobenzene
<0.2
<0.2
<0.2
0.6
0.4
Chloroform
<1.4
<1.4
<1.4
<1.4
<1.4
1,2-DichIoroethane
<0.1
<0.1
<0.1
<0.1
<0.1
1,1-Diehloroelhene
4.4
<0.5
10.4
<0.5
<0.5
Tetrachloroethene
<0.3
<0.3
0.4
<0.3
<0.3
Toluene
<0.4
<0.4
<0.4
<0.4
<0.4
1.1,1 -Trichloroethane
<0.4
1.0
<04
<04
<0.4
Trichloroethene
<0.4
<0.4
<0.4
0.5
<0.4
ng/total sample) divided by a hypothetical 20-dsL gas sample. Tabulating data in this form allows
direct comparison to the RM sample concentration data discussed in Section 4.1.
The blank data in Table 55 show that no field blank contained detectable quantities of
three of the VOC analytes — chloroform, 1,2,-dichloroethane, and toluene — at MDLs ranging
from 0.1 to 1.4 jig/dscm. Detectable quantities of the other seven VOC analytes were found in
one or more field blank samples at levels corresponding to up to 1.1 jjg/dscm for benzene,
4.0 (jg/dscm for carbon tetrachloride, 0.6 (jg/dscm for chlorobenzene, 10.4 pg/dscm for 1,1-
dichloroethene, 0.4 jjg/dscm for tctrachloroethcne, 1.0 jig/dscm for 1,1,1-trichloroethane, and
0.5 pg/dscm for trichloroethene.
Table 56 compares the highest field blank sample concentration to the lowest RM
concentration from the data in Section 4.1. As shown, the lowest RM concentration was at least
a factor of 11 greater than the MDL for the three compounds not detected in any field blank
sample. In addition, for four of the compounds detected at measurable levels in at least one field
blank sample, the highest blank concentration measured was at least a factor of 5 smaller than the
lowest RM concentration reported. These four were benzene, chlorobenzene, tetrachloroethene,
and trichloroethene. Of the remaining two compounds, carbon tetrachloride and 1.1,1-
trichloroethene, the ratios of the minimum RM concentration to the maximum blank concentration
were 3.0 and 1.8, respectively.
103
-------
TABLE 56. COMPARISON OF RM AND FIELD BLANK DATA
Concentration, fjg/dscm
Maximum Ratio RM/
Compound Minimum RM field blank field blank
Benzene
Carbon tetrachloride
Chlorobenzene
Chloroform
1,2-Dichloroethane
I, I-Dichloroethene
Tetrachloroethene
Toluene
1,1,1 -Trichloroethane
Trichloroethene
5.9
11.9
16.0
15.8
1.5
2.0
20.6
97.1
1.8
2.6
4.0
0.6
<1.4
<0.1
10.4
0.4
<0.4
1.0
0.5
1.1
>11.3
>15
0.2
51.5
>243
1.8
5.2
5.4
3.0
27.7
The tenth compound not discussed above, 1,1,-dichloroethene, was measured in two of
the five field blank samples at levels corresponding to 4.4 and 10.4 jjg/dscm. The lowest RM
concentration measured was 2.0 pg/dscm.
No blank corrections were applied to any of the RM analysis results. Any such correction
would not have measurably affected test data or conclusions for the seven compounds having
ratios of the minimum RM concentration to the maximum field blank concentration greater than 5.
This also holds true for RM concentrations of carbon tetrachloride greater than about 20 jug/dscm
(5 times the maximum field blank level), which was the case for 14 of the 18 RM measurements.
This holds true, as well, for RM concentrations of 1,1,1-trichloroethane greater than about
5 |ig/dscm, the case for 13 of the 18 RM measurements. For the other four or five measurements,
the reported RM concentration probably should be considered an upper bound to the flue gas level.
The quantitation of 1,1-dichloroethene was problematic throughout the test program. In
fact, as noted in the case narrative for the Method 8015A analyses included in Appendix E-3,
1,1-dichloroethene was not quantitated in the second trap (Tenax/charcoal) of any Method 0030
trap pair because of a high level of chromatographic interferent. Several observations concerning
this interferent include:
The interferent was only seen in the second trap of a trap pair and it was found to
be present in all sample types, including actual RM samples, field blanks, and
matrix spikes
The interferent was not seen in blank traps; it was only seen in traps after being
spiked by flash evaporation with surrogates or standards
104
-------
• Surrogate or standards solutions were analyzed directly through injection into water
and analysis by purge and trap, with no evidence of the interferent
• The interferent was always seen only in the second trap of a trap pair even when
the combined trap pair was spiked by flash evaporation
Because 1,1-dichloroethene was not quantitated in the second trap of a Method 0030 trap
pair, all RM data presented in Section 4.1 were based on the first trap analysis only. Normally,
this would imply that the analysis result would represent a lower bound estimate of the flue gas
concentration because the contribution of any material breaking through to the second trap is not
taken into consideration. However, given the variability of the results of the field blank analyses
for 1,1-dichloroethene, and the fact that blank levels were quite high for some of the field blanks,
even this statement may not hold true.
In summary, based on the problematic character of the 1,1-dichloroethene analyses, the
RM concentrations discussed in Section 4.1 should be viewed with suspicion. Reported
concentrations greater than 30 fjg/dscm (3 times the highest field blank level) are most likely
representative. This was the case for 8 of the IB RM concentrations reported. The other 10 RM
concentrations must be viewed with caution.
GC/FID analytical accuracy for VOCs was assessed by preparing eight matrix spike
VOST traps and measuring the percent recovery. Table 57 summarizes the VOC spike recoveries
from the VOST samples analyzed by GC/FID. The data in Table 57 show that all 80 spike
recovery measurements were within the compound-specific recovery ranges. Therefore, the
accuracy QAO, as measured by spike recovery, met the 70 percent QAO completeness objective.
Measuring method surrogate recoveries was another check on the accuracy of the GC/FID
analyses of test samples. Two method surrogates were used: both 1,1,2-trichloroethane and
4-bromofluorobenzene were spiked onto the first trap of a trap pair; only 1,1,2-trichloroethane was
spiked onto the second trap of a trap pair. Figure 19 is a histogram of the recoveries achieved.
The figure shows that 181 out of 200, or 91 percent, of the surrogate recoveries were within the
accuracy QAO range. Since the completeness QAO was 70 percent for this measurement, the
accuracy QAO, as measured by surrogate recovery, was met.
As a secondary check on VOST accuracy, a certified VOST audit cylinder (RCRA Audit
#729) was requested from the EPA's National Exposure Research Laboratory. The analytical
results of this independent audit are summarized in Table 58. The acceptable recovery objective
for this accuracy check is 50 to 150 percent. All recoveries determined were in this range.
Table 59 shows the method detection limit (MDL) objectives and the achieved values for
the VOC analyses. Achieved values are based on repeated analyses of low-level samples. MDL
objectives were achieved for all of the target analytes on this basis. However, the practical
quantitation limits (PQLs) can be higher than these MDLs for the analytes detected in field blanks.
Analytes detected in field blank samples at levels potentially affecting test program RM analysis
results were carbon tetrachloride, 1,1 -dichloroethene, and 1,1,1 -trichloroethane, as discussed above.
105
-------
TABLE 57. VOC RECOVERIES FROM MATRIX SPIKE SAMPLES
Spike recovery, %
Compound
Master Index Number
11146
11147
11148
11149
11150
11151
11152
11153
QAO
Benzene
98.0
98.2
96.0
96.6
97.2
99.3
96.2
98.7
37-151
Carbon tetrachloride
90.5
119
116
118
116
119
116
120
70-140
Chloroben/.ene
96.2
97.8
95.8
97.4
96.7
97.1
99.1
97.9
37-160
Chloroform
104
107
101
101
102
103
100
103
51-138
1,2-Dichloroethane
98.0
100
98.3
99.4
99.3
98.1
98.5
100
49-155
1,1-Dichloroethene
97.1
99.3
105
97.7
103
99.3
96.5
102
D-234
Tetrachloroethene
97.3
100
96.6
97.3
99.7
97.1
97.3
99.2
64-148
Toluene
96.9
97.7
96.9
98.0
98.6
97.5
97.5
98.5
47-150
1,1,1 -Trichloroethane
95.8
101
99.3
100
100
98.9
99.3
100
52-162
Trichloroethene
97.9
100
97.9
98.7
99.1
112
112
99.3
71-157
-------
50
1,1,2-T rkrfilofoethano
First trap
¦»— OAO Range
4-Brwnofluofobeazen*
First trap
»
40
1,1 ,2-Trichloroelhan*
Second trap
32
30
o 30
25
CT
lZ 20 -
10 -
90 105 121 135 1 50 >150
60
74
<45 45
Surrogate Recovery, %
Figure 19. Histogram of VOC surrogate recoveries from Method 0030 samples.
TABLE 58. VOST AUDIT CYLINDER ANALYSIS RESULTS
Concentration, ppb
Recovery,
Analyte Analyzed EPA value %
Benzene
21.47
20.64
104
Carbon tetrachloride
12.38
11.10
112
Chloroform
50.55
44.99
102
Tetrachloroethene
10.35
10.11
102
107
-------
TABLE 59. VOC MEASUREMENT MDLs: OBJECTIVES AND ACHIEVED
MDL, |ig/dscm
Analyte
Objective
Achieved
Benzene
I
0.334
Carbon tetrachloride
2
0.319
Chlorobenzene
1
0.224
Chloroform
2
1.41
1,2-Dichloroethane
1
0.143
1,1-Dichloroethene
1
0.457
Tetrachloroethene
1
0.299
Toluene
1
0.408
1,1,1 -Trichloroethane
1
0.396
Trichloroethene
1
0.440
The discussion in Section 4.1 noted that measured flue gas VOC concentrations for the
low target concentration test were higher, in many cases much higher, than the target
concentrations. This was particularly true during Test Series 1. As discussed in Section 2.2.3,
VOCs were introduced into the flue gas by spiking a solution of the target analytes in methanol
into afterburner exit flue gas that had been partially quenched to a temperature of between 360°
and 427°C (680° to 800°F). The expectation was that flue gas concentrations of the target VOC
analytes as incomplete combustion products from the incinerator would be small, so that the
scrubber exit flue gas concentrations for the CEMs to measure would be the amount from the
injected spiking solution dispersed into the volume flow of flue gas.
To verify that the low target concentrations could indeed be approximated, some scoping
tests were performed. Specifically, two scoping tests were performed on July 25, 1995. For these
tests, the RKS was operated at the desired test operating condition while being fired with auxiliary
fuel (natural gas) only. For the first scoping test, the scrubber exit flue gas was sampled for the
VOC analytes while no VOC spike solution injection occurred. This is referred to as the blank
burn test. Next, the VOC spike solution was set at the rate expected to result in the low target
flue gas concentrations.
Results of these scoping tests are given in Table 60. Also shown in Table 60 are the
average RM measured concentrations from the low target concentration test in Test Series 1. The
blank burn data shown in Table 60 confirm that background flue gas concentrations were
negligible fractions of the low concentration targets for carbon tetrachloride, 1,2-dichloroethane,
1.1 -dichloroethene, 1,1,1 -trichloroethane, and trichloroethene. Background flue gas chlorobenzene
and tetrachloroethene concentrations were half their low concentration targets. But background
benzene, chloroform, and toluene were at up to 4 times their low concentration targets.
Accordingly, the low concentration scoping test data confirm that injecting the VOC spiking
108
-------
TABLE 60. COMPARISON OF MEASURED AND TARGET CONCENTRATIONS FOR THE LOW
CONCENTRATION SCOPING TESTS
Concentration, pg/dscm
Average Test Scries 1
low concentration test Low concentration scoping test Blank burn test
Compound
Measured
Nominal
target
Ratio,
measured/
target
Measured
Nominal
target
Ratio,
measured/
target
Measured
Nominal
target
Ratio,
measured/
target
Benzene
44.6
2.0
22.3
5.2
2.0
2.6
2.6
2.0
1.3
Carbon tetrachloride
34.5
4.0
8.6
5.1
4.0
1.3
<0.3
4.0
<0.1
Chlorobenzene
63.8
2.0
31.9
2.9
2.0
1.5
1.0
2.0
0.5
Chloroform
43.2
4.0
10.8
17.5
4.0
4.4
7.6
4.0
1.9
1,2-Dichloroethane
2.2
2.0
I.I
2.4
2.0
1.2
0.4
2.0
0.2
1,1-Dichloroethene
46.3
2.0
23.2
5.3
2.0
2.7
0.1
2.0
0.1
Tetrachloroethene
96.6
2.0
48.3
4.4
2.0
2.2
0.9
2.0
0.5
Toluene
377
2.0
189
8.6
2.0
4.3
8.3
2.0
4.1
1,1,1-Trichloroethane
4.5
2.0
2.3
2.2
2.0
1.1
0.2
2.0
0.1
Trichloroethene
5.6
2.0
2.8
4.2
2.0
2.1
<0.4
2.0
<0.2
-------
solution gave flue gas concentrations of carbon tetrachloride, chlorobenzene, 1,2-dichloroethane,
and 1,1,1-trichloroethane quite near their respective targets. Concentrations of benzene,
1,1-dichloroethene, tetrachloroethene, and trichloroethene were 2 to 3 times their targets, while
chloroform and toluene were a little over 4 times their targets. Thus, while target concentrations
were achieved for only 6 of the 10 VOC analytes, it was decided from the scoping test that
achieved concentrations for all were close enough.
That the actual flue gas concentrations measured during the low target concentration test
of the initial test series were so much higher than those achieved in the scoping test for all
analytes except 1,2-dichloroethane, 1,1,1-trichloroethane, and trichloroethene came as quite a
surprise. However, there was one major difference between the scoping test and the actual CEM
tests that may have caused such significant changes. The scoping test was performed with the
RKS firing auxiliary fuel only. No synthetic waste feed was fed. Past experience suggested that
this difference should not have affected the flue gas VOC concentrations. This past experience
was that flue gas VOC concentrations resulting from the incineration of the same synthetic waste
fed during these tests, under the incineration conditions of these tests, would be quite similar to
the blank test levels given in Table 60.
However, in this past experience, no aqueous metal spiking solution atomization through
the kiln main burner was employed. Thus, the key difference between the scoping test and the
actual CEM tests was the combination of synthetic waste feed to the kiln via the screw feeder and
the atomization of an aqueous metals spiking solution through the main burner. With this
combination in place, the formation of significantly higher concentrations of the VOC analytes
seen as PICs would not be unexpected, in retrospect.
One concern arises due to the use of a modified Method 80I5A as the VOC analysis
procedure. Because the method does not give confirmation that a given chromatographic peak is
indeed one of the target VOCs, the method may be subject to positive interferences for other flue
gas unburned or partially burned hydrocarbons. This opens the possibility that the higher than
expected VOC concentrations measured for the low concentration test were the result of positive
interferences from other organic compounds that coeluted in the retention time window of the
analysis method used. However, as noted in Section 4.1, the chromatograms for the low
concentration test Method 0030 sample analyses were not unusual and showed no evidence of
interferences or coeluting compounds. In fact, the chromatograms for all tests, including those at
the high target concentrations, while quite "busy" with a large number of peaks, still showed good
separation of the target analytes' peaks from other responders. Thus, given this and the possible
explanation for actually having higher than expected VOC concentrations noted above, there is no
reason to believe that the RM measurements for the low target concentration tests did not reflect
the actual flue gas concentrations present.
6.2 SVOC ANALYSES
A total of 13 Method 0010 samples were analyzed for SVOCs. Included in this group
were one field blank, one method blank, and two MS/MSD sample sets. All samples were
extracted and analyzed within the required hold times specified by the method — 14 days for
110
-------
extraction and 40 days for extract analysis. Neither the field blank sample nor the method blank
sample (clean sorbent resin) contained detectable amounts of any of the 17 SVOC analytes. MDLs
are discussed below.
SVOC analysis precision and accuracy were assessed by preparing and analyzing
MS/MSD sample sets. Table 61 summarizes the spike recovery data obtained for these MS/MSD
samples. The data in Table 61 show that all 68 spike recovery measurements were within the
compound-specific recovery objective ranges. Thus, the accuracy QAO as measured by spike
recovery was met. The data in Table 61 also show that all 34 relative percent difference (RPD)
measurements were within the precision QAO of 50 percent.
All samples slated for SVOC analyses were spiked with method surrogates prior to
extraction in order to evaluate surrogate recoveries. Figure 20 is a histogram illustrating the
surrogate recoveries achieved. As indicated, 47 out of 48 surrogate recovery measurements, or
98 percent, were within the surrogate-specific recovery range. As the completeness QAO for this
measurement was 70 percent, SVOC measurement accuracy, as assessed by surrogate recovery,
was met.
As an independent check on SVOC accuracy, a certified standard reference material
(SRM) was obtained from the National Research Council of Canada (NRCC). The results of the
analysis of this SRM are given in Table 62. As indicated in the table, of 14 analytes quantitated
in the SRM, 8 analyzed concentrations were in the acceptable range established by the NRCC.
This represents an acceptable fraction of 57 percent, compared to an accuracy completeness
objective of 70 percent. However, in terms of percent recovery of analytes from this SRM, 13 of
14 quantitated recoveries, or 93 percent, fell within the project recovery QAO ranges. Thus, while
the accuracy objective in terms of the project QAO was met, the more restrictive NRCC objectives
were not.
Table 63 lists the MDL objectives and achieved levels for the SVOC measurements. As
shown in Table 63, the MDL objective was achieved for all analytes except anthracene,
benzo(ghi)perylene, dibenz(a,h)anthracene, fluoranthene, and indeno(l,2,3-cd)pyrene. The MDLs
achieved for these compounds were only 0.1 to 0.2 |jg/dscm above the MDL objective of
1 jjg/dscm, however. Exceeding MDL objectives by the small amounts experienced, by itself, had
no effect on test program results. However, the propriety of having MDL objectives representing
significant fractions of target flue gas concentrations is subject to challenge. Still, as discussed
in Section 4.2, the potential problems associated with this occurrence were acknowledged during
test program planning efforts, and were documented in the approved QAPP for the program
(Reference 8).
6.3 TRACE METAL ANALYSES
A total of 140 samples from Method 29 sampling trains were analyzed by ICP and
GFAAS for trace metals (excluding mercury) and by CVAAS for mercury. Included in these
samples were 26 field blanks and 24 MS/MSD samples. The 140 samples represent different
fractions of the nine Method 29 sampling trains and three field blank trains, collected for the
111
-------
TABLE 61. SVOC RECOVERIES FROM MATRIX SPIKE SAMPLES
Set 1 Set 2
Spike recovery, % Spike recovery, %
Compound
QAOa
MS
MSD
RPD, %
MS
MSD
RPD, %
Acenaphthene
47-145
77.8
73.3
6.0
74.7
79.9
6.7
Acenaphthylene
33-145
77.8
72.8
6.6
75.5
81.3
7.4
Anthracene
27-133
83.2
76.6
8.3
74.9
81.5
8.4
Benzo(a)anthracene
33-143
85.2
77.0
10.1
81.6
86.9
6.3
Benzo(b)fluoranthene
24-159
80.3
75.0
6.8
76.6
89.8
15.9
Benzo(k)fluoranthene
11-162
82.1
77.3
6.0
76.7
78.8
2.7
Benzo(ghi)perylene
D-219b
81.7
75.5
7.9
77.7
80.7
3.8
Benzo(a)pyrene
17-163
81.7
75.2
8.3
77.8
80.6
3.5
Chrysene
17-168
84.1
76.0
10.1
80.5
85.1
5.6
Dibenz(a,h)anthracene
D-203
81.6
75.9
7.2
77.4
80.6
4.1
Fluoranthene
26-137
89.0
77.3
14.1
75.5
83.9
10.5
Fluorene
59-121
85.2
78.5
8.2
77.6
85.5
9.7
Indeno( 1,2,3-cd)pyrene
D-171
79.0
74.6
5.7
75.0
78.7
4.8
2-Methylnaphthalene
21-133c
89.0
86.1
3.3
87.2
96.2
9.8
Naphthalene
21-133
75.8
71.0
6.5
69.6
77.0
10.1
Phenanthrene
54-120
82.0
76.8
6.4
77.1
93.7
19.4
Pyrene
52-115
74.6
71.1
4.8
76.5
80.3
4.8
RPD QAO, %
50
50
"Compound recovery objective taken from SW-846 Method 8270B, Table 6, Column 5.
bD = Detected.
cRecovery QAO is the same as for naphthalene.
-------
N»ubenz»m-d5
QAO Ranges
— Terphenyt-dl4
— Nifrobenzene-dS
Octafluoro- —i
! bi phenyl,
j 2-Fluort>-
j bi phenyl
15
OctafluotobPanyi
cr
5
0
30
120
>137
<18
18
23
72
115
137
Surrogate Recovery, %
Figure 20. Histogram of SVOC surrogate recoveries from Method 0010 samples.
TABLE 62. SVOC SRM ANALYSIS RESULTS
Analyte
Concentration, mg/kg
Recovery, %
Analyzed
Certified
value
Acceptable
range8
Achieved
QAO
rangeh
Acenaphthene
<0.26
0.23
0.16-0.59
NCc
47-145
Acenaphthylene
0.29J
0.19
0.14-0.24
153
33-145
Anthracene
0.96J
1.10
0.70-1.50
87
27-133
Benzo(a)anthracene
1.47
1.80
1.50-2.10
82
33-143
Ben?o(b)fluoranthene
2.31
2.80
2.20-3.40
83
24-159
Bsn7o(k)fluoranthene
2.11
1.43
1.28-1.58
148
11-162
Benzo(ghi)perylene
1.48
1.78
1.06-2 50
83
D-2!9d
Ben7.o(a)pyrene
1.50
2.20
1.80-2.60
68
17-163
Chrysene
2.26
2.00
1.70-2.30
113
17-168
Diben7(a,h)anthracene
<0.332
049
0.33-065
NC
D-203
Fluoranthene
3.16
3.54
2.89-U9
89
26-137
Fluorene
0.32J
0.47
0.35-0.59
68
59-121
Indeno( 1,2,3-cd)pyrene
1 46
1.95
1.37-2.53
75
D-171
Naphthalene
2.62
4.10
3.00-5.20
64
21-133
Phenanthrene
3.02
3.00
2.40-3.60
101
54-120
Pyrene
2.45
3.00
2.40-3.60
82
52-115
"Established by the NRCC.
''Taken from Table 6, Column 5, of Method 8
270B.
cNC
dD =
= Not calculated.
Detected.
113
-------
TABLE 63. SVOC MEASUREMENT MDLs: OBJECTIVES
AND ACHIEVED
MDL, fig/dscm
Analyte Objective Achieved
Acenaphthene 1
0.91
Acenaphthylene 1
0.93
Anthracene 1
1.21
Benzo(a)anthracene 1
0.70
Benzo(b)fluoranthene 1
0.94
Benzo(k)fluoranthene 1
0.38
Benzo(ghi)perylene 1
1.25
Benzo(a)pyrene 1
0.99
Chrysene 1
0.66
Dibenz(a,h)anthracene 1
1.12
Fluoranthene 1
1.21
Fluorene 1
0.93
Indeno(l,2,3-cd)pyrene 1
1.08
2-Methylnaphthalene 1
0.78
Naphthalene I
0.91
Phenanthrene 1
0.67
Pyrene 1
0.63
multi-metal CEM tests, and the nine Method 29 sampling trains and two field blank trains,
collected for the mercury CEM tests, along with two sets of MS/MSD samples. All trace metal
analyses were completed within the required method hold times of 28 days, for mercury, and
180 days, for the other metals.
Table 64 summarizes the results of the three Method 29 train field blank sample analyses.
The data in the table are broken out by the two sampling train fractions analyzed for all metals
except mercury — the front half (probe wash plus filter) and the back half (impinger trains).
Aliquots of these samples (front half and back half) were separately analyzed for mercury. Also,
two additional impinger samples were analyzed for mercury. Results from the additional mercury
analyses are combined in the back-half data for mercury in the table.
Table 64 also notes the average blank concentration for the three field blanks analyzed,
as well as the standard deviation of this average. In cases where a given metal was detected in
respective fractions of all three field blank trains, the calculation of the average and standard
deviation is straightforward. In cases where a given metal was not detected in any of the
respective fractions of the three trains, a standard deviation was not calculated. In cases where
a given metal was found in the respective fractions of two blank trains but not detected in the
114
-------
TABLE 64. METHOD 29 FIELD BLANK ANALYSIS RESULTS: MULTI-METAL
CEM TESTS
Analysis result, total fig/sample
Low
Intermediate
High
Standard
concentration
concentration
concentration
deviation Average +
Metal
test
test
test
Average
(cr)
3 a
Front half
Antimony (Sb)
0.36
<0.30
<0.40
<0.40
0.40
Arsenic (As)
2.70
2.35
2.25
2.43
0.19
3.01
Barium (Ba)
7.37
5.73
7.23
6.78
0.74
9.00
Beryllium (Be)
<0.10
<0.10
<0.10
<0.10
<0.10
Cadmium (Cd)
0.38
<0.10
0.52
0.33
0.17
0.86
Chromium (Cr)
5.47
4.25
4.84
4.85
0.50
6.35
Cobalt (Co)
<0.10
<0.10
<0.10
<0.10
<0.10
Lead (Pb)
<0.20
<0.20
<0.20
<0.20
<0.20
Manganese (Mn)
3.59
2.59
3.22
3.13
0.41
4.37
Mercury (Hg)
<0.40
<0.40
<0.40
<0.40
<0.40
Nickel (Ni)
7.61
6.24
• 7.50
7.12
0.62
8.98
Selenium (Se)
6.44
6.20
7.11
6.58
0.39
7.74
Silver (Ag)
<0.20
<0.20
<0.20
<0.20
<0.20
Thallium (Tl)
<0.30
<0.20
<0.50
<0.30
<0.30
Back half
Antimony (Sb)
1.23
0.47
<0.60
0.77
0.33
1.77
Arsenic (As)
<0.60
0.53
0.74
0.62
0.08
0.88
Barium (Ba)
2.60
1.46
1.27
1.78
0.59
3.54
Beryllium (Be)
<0.15
<0.15
<0.15
<0.15
<0.15
Cadmium (Cd)
<0.15
<0.15
0.22
<0.20
<0.20
Chromium (Cr)
0.68
1.26
1.01
0.98
0.24
1.70
Cobalt (Co)
<0.15
<0.15
<0.15
<0.15
<0.15
Lead (Pb)
0.84
1.57
<0.30
0.90
0.52
2.47
Manganese (Mn)
53.90
56.90
65.40
58.73
4.87
73.34
Mercury (Hg)
<3.70
<3.70
<3.70
<3.70
<3.70
Nickel (Ni)
<0.45
1.49
1.00
0.98
0.42
2.25
Selenium (Se)
<0.75
1.78
1.62
1.38
0.45
2.74
Silver (Ag)
<0.30
<0.30
<0.30
<0.30
<0.30
Thallium (TI)
<0.45
<0.45
<0.75
<0.45
<0.45
115
-------
third, the detection limit for the third blank train sample was used in calculating the average and
standard deviation.
The data in Table 64 show that measurable quantities of 7 of the 14 metals were present
in the field blank front-half samples, and of 8 of the 14 metals in the back-half samples. Based
on this, it was decided to blank-correct the RM measurement data. This blank correction
proceeded as follows.
The requirement that an RM sample component have a metal quantity greater than 3
standard deviations above the average blank sample quantity to be considered significantly
different from the blank was imposed. This quantity, the average blank level plus 3 standard
deviations, is given in the far right column of Table 64.
In the test program data, this was not a constraint for the RM front-half measurement
results. The quantity of metal analyzed in all RM front-half samples was greater than the average
blank plus 3 standard deviations for all 14 metals. Therefore, all RM front-half measured
quantities were blank-corrected by subtracting the average front-half blank quantity. For the six
metals not detected in any field blank front-half sample, the blank correction was zero. Measured
levels in RM front-half samples were significantly greater than the field blank front-half detection
limit for these six.
The situation was less straightforward for the back-half samples, however. For these
samples, three cases existed. In the first case, an RM sample quantity was greater than the average
field blank quantity plus 3 standard deviations. For these, the blank correction was performed in
the same manner as for the front-half samples noted above. In the second case, the RM sample
quantity was less than the average field blank quantity, or was not detected at a detection limit
lower than the average field blank level. In this case, the quantity of metal in the RM sample was
said to be less than (not detected at) the measured RM sample value. In the third case, the
quantity of metal in the RM sample was between the average field blank quantity and this average
plus 3 standard deviations. In this case, by failing the requirement to be considered significantly
different from the blank, the RM sample quantity was said to be less than (not detected at) the
average field blank quantity.
After blank-correcting both the front-half and the back-half RM data in line with the
above, the result was a set of corrected RM front-half data having definitive metal quantities in
all RM front-half samples for all 14 metals. In other words, there were no non-detects or "less
thans" in the front-half sample data. In contrast, most metal concentrations (other than mercury)
were non-detects or "less than" some level in back-half samples. For these, one can define a flue
gas concentration range for a given metal in which the lower bound of the range corresponds to
setting the back-half amount equal to the "less than" amount (detection limit) and the upper bound
of the range corresponds to setting the back-half amount equal to zero.
For all but one metal, manganese, the total RM train sample quantity was a definite level
(i.e., back-half level definitive and not "less than"), or the difference between the lower bound and
the upper bound, defined above, was small. In these cases, the RM concentration was calculated
116
-------
based on the definite level or the upper bound. These are the RM concentrations discussed in
Section 4.3. For manganese, the difference between the lower and upper bounds as defined above
was significant. The difference was as large as 24 percent of the lower bound value, for the low
concentration test, and 63 percent, for the intermediate concentration test. Of course, the origin
of these differences was the high level of manganese found in both RM back-half samples and
field blank back-half samples. In no case was the level of manganese in an RM back-half sample
significantly different from the level in a blank. However, the high measured levels in both RM
and field blank samples caused the "less than" amount in RM back-half samples to be large. Thus,
for manganese, the flue gas RM concentrations discussed in Section 4.3 are based on the lower
bound, this bound resulting from setting the back-half amount equal to the "less than" amount as
defined above.
Table 65 summarizes the mercury analysis results for the two Method 29 field blank
sample sets analyzed. Results from each back-half sample fraction analyzed are shown separately
in Table 65 instead of being combined as was done in Table 64. The data in Table 65 show that
no mercury was detected in all but one field blank sample fraction — fraction 3B of the field
blank prepared on the intermediate concentration test day. The lowest RM train sample fraction
mercury level among all nine RM samples collected over the mercury CEM test week (all from
one of the RM sampling trains collected on the low mercury concentration test day) is also given
in Table 65.
The data in Table 65 show that the lowest RM back-half sample mercury level was at
least a factor of 2.8 higher than the MDL for the fraction, or the fraction amount in the one
fraction 3B field blank sample containing detectable mercury. The lowest RM front-half sample
mercury level was a factor of 1.7 higher than the front-half MDL for the field blank sample
collected on the low concentration test day. The MDL for the front-half field blank sample
collected on the intermediate concentration test day was unusually high. The usual MDL for this
sample fraction is the 0.4 jig level achieved for the low concentration test day sample. Based on
TABLE 65. METHOD 29 FIELD BLANK ANALYSIS
RESULTS: MERCURY CEM TESTS
Mercury analysis result, total pg/sample
Intermediate Low
Sample concentration concentration Minimum
fraction test test RM sample
Front half <5.0 <0.4 0.7
Back half
Sample 2B <6.0 <6.0 16.6
Sample 3A <0.4 <0.4 7.0
Sample 3B 4.1 <2.8 15.1
117
-------
the results in Table 65, no blank corrections were performed for mercury in the RM data for the
mercury CEM tests.
Trace metal analysis precision and accuracy were assessed by preparing and analyzing
MS/MSD sample sets and measuring the percent recovery. The MS/MSD sample analysis results
are presented in Table 66, for the metals other than mercury, and in Table 67, for mercury. The
data in Table 66 show that, excluding silver recovery from fraction 1A samples, 97 out of 100
individual spike metal recovery measurements, or 97 percent, were within the accuracy QAO range
of 60 to 140 percent recovery. As the completeness QAO for this measurement was 70 percent,
the accuracy QAO as measured by spike recovery from MS/MSD samples was met. However,
spike recoveries for silver in the Method 29 front half MS/MSD samples (fraction 1 A) were less
than 6 percent. For this reason, all test program RM results for silver are considered unreliable.
Trace metal precision was also measured by calculating the RPD of each MS/MSD sample set.
The data in Tabic 66 show that, again excluding silver recovery from fraction 1A samples, 46 out
of 50 RPDs, or 92 percent, met the precision QAO of 30 percent RPD. Thus, the 70 percent
completeness QAO for the precision assessment, as measured by the RPD of the MS/MSD
samples, was met.
For mercury, the data in Table 67 show that all 16 mercury recovery measurements from
MS/MSD samples met the recovery QAO, and that all 8 RPDs measured for MS/MSD samples
met the precision QAO. Thus, both the accuracy and precision QAOs for the mercury
measurement were achieved.
As an independent check on trace metal analysis accuracy, a certified SRM obtained from
the National Institute for Standards and Technology (NIST), NIST 1633b flyash, was submitted
for analysis. Analysis results, compared to NIST-certified and non-certified concentrations, are
given in Table 68. The data in the table show that seven of the nine analytical recoveries
measured, or 78 percent, were in the accuracy QAO range of 60 to 140 percent recovery. As the
measurement completeness objective was 70 percent, the accuracy QAO, as measured by the
analysis of an SRM, was met. The most significant outlier in the SRM analysis was cadmium,
for which the analysis result was nearly 4 times the certified SRM concentration. By itself, this
result would cast suspicion on the RM cadmium analyses. However, all matrix spike sample
analyses for cadmium were quite acceptable. For this reason, the cadmium result in the SRM
analysis is considered an outlier.
Table 69 shows the MDL objectives and achieved values for the trace metal analyses.
The achieved levels represent the sum of the MDLs reported by the laboratory for the individual
Method 29 sample fractions. MDLs reported by the laboratory correspond to analytical capabilities
for clean samples. However, as discussed above, Method 29 train field blank samples contained
significant amounts of several of the trace metal analytes. In fact, the blank correction discussion
above noted that an RM level for a given metal was not considered different from the field blank
unless the RM level was greater than the average field blank level plus 3 standard deviations.
These field blank average plus 3 standard deviations concentrations are also given in Table 69.
For all metals except manganese, these concentrations represent the sum of the average plus 3
standard deviations levels noted in Table 64 divided by a hypothetical 3-dscm flue gas sample
118
-------
TABLE 66. TRACE METAL RECOVERIES FROM MATRIX SPIKE SAMPLES
Recovery, %
Sample
Ag
As
Ba
Be
Cd
Co
Cr
Mn
Ni
Pb
Sb
Se
T1
QAO
Matrix Spike, set 1
Fraction 1A MS
3.3
60.6
91.2
72.4
83.6
89.8
92.6
77.0
73.2
71.8
70.0
84.4
57.6
60-140
Fraction 1A MSD
2.6
65.6
88.0
88.4
89.2
76.8
97.2
93.0
92.2
64.6
76.4
94.0
72.8
60-140
RPD, %
23.7
7.9
3.6
19.9
6.5
15.6
4.9
18.8
23.0
10.6
8.7
10.8
23.3
30
Fraction 2A MS
92.2
69.2
89.6
85.0
93.4
64.4
90.6
102
85.8
85.0
85.0
83.2
73.4
60-140
Fraction 2A MSD
93.2
67.0
62.0
92.4
101
90.8
94.8
96.2
90.4
85.8
84.6
84.0
78.4
60-140
RPD, %
1.1
3.2
36.4
8.3
7.8
34.0
4.5
5.9
4.8
0.9
0.5
1.0
6.6
30
Matrix Spike, set 2
Fraction 1A MS
5.3
9.7
102
59.3
91.6
73.6
95.6
96.5
88.1
80.3
79.8
91.5
65.6
60-140
Fraction 1A MSD
0.8
70.5
82.4
88.5
90.4
84.5
94.5
87.3
69.5
80.8
83.8
74.5
62.7
60-140
RPD, %
146
152
21.3
39.5
1.3
13.8
1.2
10.0
23.6
0.6
4.9
20.5
4.5
30
Fraction 2A MS
92.3
66.1
84.1
87.9
95.1
90.8
89.4
91.9
86.6
89.0
83.3
83.0
80.2
60-140
Fraction 2A MSD
82.3
63.9
84.9
81.9
95.3
81.2
80.9
80.8
79.0
79.2
79.3
75.3
72.5
60-140
RPD, %
11.5 3.4 1.0 7.1 0.2 11.2 10.0 12.9 9.2 11.7 4.9 9.7 10.1 30
-------
TABLE 67. MERCURY RECOVERIES FROM MATRIX SPIKE SAMPLES
MS set 1 MS set 2
Sample
Spike
recovery, %
RPD, %
Spike
recovery, %
RPD, %
Recovery
QAO, %
MS
MSD
MS
MSD
Fraction IB
81.0
80.8
0.2
87.5
68.2
24.8
60-140
Fraction 2B
92.6
95.2
2.8
83.6
94.4
12.1
60-140
Fraction 3A
92.8
86.2
7.4
87.2
86.9
0.3
60-140
Fraction 3B
87.8
73.2
18.1
73.8
62.9
15.9
60-140
RPD QAO, %
30
30
TABLE 68. TRACE METAL SRM ANALYSIS RESULTS
Concentration, mg/kg
Recovery,
Analyte Analyzed Certified value Non-certified value %
Antimony (Sb)
<4
6
NCa
Arsenic (As)
130.5
136.2
96
Barium (Ba)
610
709
86
Cadmium (Cd)
3.05
0.784
390
Chromium (Cr)
171.4
198.2
86
Cobalt (Co)
55
50
110
Lead (Pb)
52.1
68.2
76
Manganese (Mn)
117.2
131.8
89
Mercury (Hg)
<4
0.141
NC
Nickel (Ni)
120.3
120.6
100
Selenium (Se)
16.87
10.26
164
Thallium (T!)
<5
5.9
NC
Recovery objective
60-140
aNC = Not calculated.
120
-------
TABLE 69. TRACE METAL MEASUREMENT MDLs:
OBJECTIVES AND ACHIEVED
Analyte
MDL,
Objective
pg/dscm
Achieved
Field blank
average + 30,
Mg/dscm
Antimony (Sb)
5
0.35
0.72
Arsenic (As)
2
0.35
1.30
Barium (Ba)
20
0.18
4.18
Beryllium (Be)
0.2
0.09
<0.09
Cadmium (Cd)
2
0.09
0.35
Chromium (Cr)
10
0.26
2.68
Cobalt (Co)
5
0.09
<0.09
Lead (Pb)
20
0.18
0.89
Manganese (Mn)
2
0.18
1.44
Mercury (Hg)
10
1.42
<1.42
Nickel (Ni)
5
0.26
3.74
Selenium (Se)
30
0.44
3.49
Silver (Ag)
3
0.18
<0.18
Thallium (Tl)
2
0.26
<0.26
volume. For manganese, the average plus 3 standard deviations is taken from the front-half value
only, in keeping with the above blank correction discussion. Back-half manganese levels were so
high for both RM and field blank samples that no RM back-half sample could be distinguished
from the field blank; thus, all RM back-half manganese concentrations were assumed to be zero.
The data in Table 69 show that the MDLs achieved by the laboratory easily met the MDL
objectives for all metal analytes. Furthermore, the data in the table show that the average plus 3
standard deviations concentrations from the field blanks were also below the MDL objective for
all of the metals.
6.4 TECHNICAL SYSTEM REVIEWS
At the request of the EPA Work Assignment Manager, the sampling and analysis efforts
under this test program were subjected to technical system reviews (TSRs). The TSR of the field
and laboratory operations at the IRF was performed on September 13 and 14, 1995. The TSR of
the laboratory operations at Triangle Laboratories, the laboratory that performed the trace metal
and mercury analyses, was conducted on August 23. 1995.
121
-------
The TSR of the IRF uncovered only minor issues, which were immediately addressed and
which did not affect data quality. One concern was identified in the TSR of Triangle Laboratories,
but prompt corrective action prevented any effects on data quality. The TSR reports are included
in Appendix G-l.
122
-------
SECTION 7
DEVELOPER CLAIMS
As noted in Section 1.3, each developer was offered the opportunity to review the draft
of this test report, and based on this review, to submit a "vendor claims" discussion for inclusion,
without modification, in this final test report.
Four vendors representing five of the CEMs tested took advantage of this opportunity.
Each of their discussions is included, as received, in the following subsections as follows:
ORNL (VOC CEM)
EcoLogic (VOC CEM)
EcoChem (SVOC CEM)
SNL (Multi-metals CEM)
EcoChem (Hg CEM)
Section 7.1
Section 7.2
Section 7.3
Section 7.4
Section 7.5
123
-------
Intentionally Blank Page
-------
ORNL VOC CEM DEVELOPER COMMENTS
125
-------
Intentionally Blank Page
-------
ORNL DSITMS CEM for VOC's
Developer Claims and Self-Assessment / Test Comments Section
We are well pleased that the field-transportable Direct Sampling Ion Trap
Mass Spectrometry (DSITMS) instrumentation developed at ORNL was able to do
a credible job monitoring VOCs in flue gas during the EPA sponsored test at the
Incinerator Research Facility. The test results have revealed several areas where
improvements can be made to the field sampling interfaces that will allow more
routine monitoring of emission processes like incinerator stacks and bring the
relative accuracies of the CEM response to sanctioned levels. These areas are
discussed below.
One of the most striking trends in the test data that was not highlighted in
the test report is that the ORNL results were systematically lower than the
reference method results. This trend is shown in the table below with a negative
sign indicating a result lower than the reference method and a plus sign
indicating a higher result than the reference method. The letters in the column
header correspond to the low (L), intermediate (I) and high (H) VOC
concentration and the numbers in the column header correspond to the reference
test number. The only compounds that did not demonstrate this trend on all three
test days were 1,2-dichloroethane and 1,1-dichloroethene.
| Compound
LI
L2
13
11
12
13
HI
H2
H3
| Benzene
Carbon Tetrachloride
Chlorobenzene
Chloroform
1,2 -Dichloroethane
+
-
+
+
+
+
-
-
-
1,1-Dichloroethene
-
-
-
+
+
+
-
-
-
Tetrachloroethene
Toluene
1,1,1-Trichloroe thane
J Trichloroethene
-
In order to understand what might prompt a systematically low reading
from the DSITMS CEM, it is important to remember that the Sorbent Trap /
Thermal Desorption (ST/TD) DSITMS method employed during this test requires
two measurements to report concentrations of VOCs in the flue gas. Specifically,
these measurements are the mass of VOC detected and the volume of flue gas that
passed through the sorbent trap. These two measurements are then combined to
give a VOC concentration. The ST/TD/DSITMS method has been used in many
field studies with a demonstrated relative standard deviation in the 10-20% range
127
-------
ORNL DSITMS CEM for VOC'a
for the measurement of sample mass collected on the sorbent media. Thus, we
are very confident that the mass measurement was not at fault for the low CEM
response.
The primary cause for the low CEM response was a poor volume
measurement. Since we were not specifically funded to create an incinerator
stack CEM, we made use of the materials at hand. The volume measurement
method we used was to multiply the sample flow rate, obtained using a Dwyer
flowmeter (i.e. a rotometer), by the sampling time, measured using a stop watch.
This method is clearly manual and not suitable for an automated CEM, however,
it was available. The problem with this arrangement was that the flowmeter was
calibrated for dry nitrogen while the sample gas still contained a substantial
fraction of water. Since the presence of water in the flowmeter will yield an
artificially high flow rate (unless the flowmeter is plugged by a condensed
droplet), the sample volumes calculated will be too high and the VOC
concentrations reported would, therefore, be too low. The calculation of VOC
concentrations in the flue gas was further complicated by the fact that the sorbent
traps adsorbed some amount of water prior to the volume measurement. A
correction for the presence of water was attempted by estimating that the sample
flow was saturated with water at a given temperature. This calculation led to a
water content of 6.6% by volume compared to the 34% by volume reported in the
IRF data. It is quite possible that the actual water content at the flow
measurement point was higher than the value used to calculate the DSITMS
results. Substitution of a larger value for the water content in the calculation of
VOC concentration in dry standard units would lead to a higher VOC
concentration. The solution to this problem is straightforward. That is, to use a
computer controlled, mass flowmeter downstream of the sorbent trap and a
condenser to perform a dry volume measurement. A DSITMS interface with
these components would provide dramatically better results and help meet the
automation criteria for a permanently installed CEM.
One potential problem involving the use of a filament-based, electron
ionization method was highlighted in the ORNL test results when a systematic
decrease in the ST/TD/DSITMS CEM response was observed during the latter part
of the high concentration test. This decrease was ascribed to a partial filament
failure using the response from the Continuous Air Monitor (CAM) DSITMS
CEM, which was running concurrently with the ST/TD/DSITMS at this point, as a
basis for comparison. Periodic calibration checks and re-calibration could be used
to identify and eliminate this problem in a future, automated version of the
DSITMS CEM. Also, alternative ionization methods could be employed that are
less susceptable to degredation or failure caused by overexposure to air and
water.
The potential of sample breakthrough is a consideration when using
sorbent-based gas sampling. It should be emphasized that the sorbent traps used
in this study were constructed with three sorbents of increasing retentivity. In
effect, a "backup trap" is built-in. Therefore, sample breakthrough is not
considered to be a significant contributor to the systematically low CEM response.
The effect of water on these traps has been examined and the results indicate that
if a drying step (i.e. flowing dry nitrogen through the trap before thermal
128
-------
ORNL DSITMS CEM for VOCs
desorption) is performed, recoveries are close to 100%. A drying step was
performed on all flue gas samples in this study.
The CEM test results obtained during this study also provided an estimate
for the frequency of calibration and zero drift checks. Average calibration drift
measured at the end of the test day (ca. 10-12 hours after initial calibration)
ranged 36-76% while a midday calibration check (ca. 5-6 hours after initial
calibration) was a much more acceptable 11%. Therefore, 4-5 calibration checks
every 24 hours should insure proper operation of the CEM and further improve
the relative accuracy of the CEM. It remains to be demonstrated whether the
calibration checks can be used to adjust the CEM response or whether a
calibration of the desired concentration range will be required. Since calibration
and zero measurements will be automated in any future versions of the DSITMS
CEM, these checks and calibrations should not pose a significant problem.
Further improvements in the CAM are ongoing. The goal is to achieve
lower detection limits for VOC and SVOC compounds using this "raw-gas"
interface. The CAM has the advantage of providing more a real-time response
even though both interfaces could be "continuous" monitors. The advantage of
real-time response is the ability to detect transient process events.
129
-------
Intentionally Blank Page
-------
ECOLOGIC VOC CEM DEVELOPER COMMENTS
131
-------
Intentionally Blank Page
-------
rov/c chsmical
SPFCML1STS
ECO LOGIC
0*0
Developer Claims and Self Assessment/Test Comments
Introduction
The Airsense 500 and CIMS 500 Mobile Chemical Ionization Mass Spectrometers are developed
and manufactured by V & F Analyse - und Messtechnik GES M B.H. in Absam, Austria. ECO
LOGIC is the North American supplier for both instruments V & F first developed the CIMS
500 in 19S6. Since then more than 150 instruments are in operation worldwide in a variety of
production, research, and industrial applications. V & F developed the Airsense 500 in early 1995
due to industry's demand for fast continuous monitoring with greater sensitivity. Both
instruments use a patented ion transfer technique and low energy chemical ionization to minimize
or eliminate fragmentation of sample molecules and produce single ion products for ease of
analysis. (See H-l-2 Appendices for Principles of Operation)
The low energy or chemical ionization technique is based on the interaction of neutral molecules
(sample gas) with a low energy ion beam (source gas). Generally sample gas molecules with an
ionization potential (eV) equal to or lower than the incoming ion beam will be ionized. Thus it is
important to select the appropriate source gas based on the compounds of interest. An option of
three source gases, each carrying a specific ionization electron voltage potential, is available to the
instrument operator. The three source gases are Krypton, Xenon, or Mercury.
Table 1 lists the EPA's target VOC list, their ionization potentials and molecular weights, the ion
products using Xenon or Mercury, and the selected masses for continuous monitoring. Xenon is
capable of ionizing all ten compounds but with a degree of fragmentation for some compounds
The eV of Mercury is high enough to ionize only five of the ten compounds. Although Mercury
would have provided higher sensitivity for the five compounds, the decision to demonstrate the
instrument's capabilities for ten compounds outweighed higher sensitivity Spectra I and 2 show
the uncalibrated response using Xenon to I ppm multi-mix calibration gas for the ten target
compounds The x-axis shows the mass range in atomic mass units (amu) and the y-a.xis
represents detector response in counts/second. For most of the chlorinated VOC's, the Xenon
will knock offa single chlorine atom. For continuous calibrated monitoring, up to ten individual
masses can be selected Table 1 lists the masses selected for the test program
To demonstrate instrument ruggedness and ease of set up, the Airsense was positioned directly at
the scrubber exit sampling location A ten foot heated sample line was connected from the sample
port directly into the instrument's sample inlet which consists of heated stainless steel
components. This technique allows for maintaining sample integrity while eliminating sample
handling and potential VOC losses
Discussion
!-!3 Dcnris S'... f\ik k: •: Ontario, On.ul.l, N",B 2K0
RofUsixxl i.jl'h '> I
r.K i" I')) MSl'i-'1.' >5
Preceding page bisnS*
-------
The following table shows the compounds of interest with ionization potentials and the masses
selected for monitoring.
COMPOUND
•V
MOLECULAR
WEICHT
•Nt
12.2 rV
Ut
10.4 cV
ATOMIC MASS
X«
Benzene
92
73
78
78
78
Chlorobcnzene
9.1
112
112.114
112.114
112
Chloroform
11.3
IIS
83,83
•
83
1J Dighloroeihane
III
98
62.6-1
-
63
Tcwchlorocthenc
11.6
166
131.133.166
-
166
Toluene
88
92
91
92
91
Trichlcroeihenc
9.5
130
95.97
130.132
95
Ctrbon
Tetrachloride
11.5
152
117.119
•
117
1,1 Dichlorethene
96
96
61.63
96.98
61
1.1.1 Trichlofoelhane
110
132
97.99
•
97
Btrgraph plotter Output - ECO tocic INT, 2nc
SPECTRUM 1 - CAL. XIXTJRS II
USEPA IRF - AIRSENSE 500
Bargraph ?lott«r Output - ZCO LOGIC INT. I!ic
5PECTRUM 2 - CAL. MIXTURE 12
USEPA IK? - AIJISENSC 500
-12*
C2CH
kj
>0 79 88 98 199 110 12B |3B J40 tSQ IW 17®
1 Ln-,
lit 129
134
-------
Trend Graphs 1,2 and 3 display a portion of the data collected over the entire three test days.
For the purpose of this report, the trend graphs represent every 10th data point collected for each
compound. The monitoring speed for each compound was approximately every 1 second for Test
Day 1, and every 0.6 seconds for Test Days 2 and 3. This equates to several thousand
measurements per compound in a 10 hour test program.
Trend Graph 1 displays the data collected on Test Day 1 (07/31/95). The expected target flue gas
concentrations were in the 1 to 2 ug/dscm range. The trend graph shows several hundred ppb for
many of the compounds through the RM sampling periods. These high concentrations were
determined to be partially a result of background interferences that were present in the gas stream
and not compensated for on the instrument.
TREND GRAPH #1 - TEST DAY J5/07/J1
ECO LOGIC AIRSENSE 500 DATA
-OulM I
-eiHten
i———CJH'll
I ¦ i . — C?w»Clt
Spectra 3, 4, and 5 displays the uncalibrated response to the instruments background with
nitrogen, ambient air, and the process gas respectively. The mass range was set at 55-120 amu,
which is the range for the majority of the compounds of interest (see Table 1). The low level
interferences were not compensated for, this in addition with the low signal expected for the
target flue gas, the Airsense results in respect to relative accuracy (RA) were compounded The
results were not entered for RA calculations
135
-------
Bargcaph Plotter Output - SCO LOGIC INT. IKC
SPECTRUM 3 - KI7R0CIH SACKCROUKD
USEPA IRP - AIRSEKSE 5C3
S&r^rapb Plotter Output - ECO LOGIC Itrr. INC
SPECTRUM 4 • IRT ROOM AIR
USEPA ZRF - AIRSENSE 500
bS
Tfrrt
rt'if f""
rWi
95 9S 195 US 5S tS 7S BS
3Arsr«ph PXnttir Output - ECO LOGIC 1ST. IMC IPA02-1 /
SPECTRUM 5 - P-12-TiSr PRCCZSi CAS
L'SiPA 13-* - AIRSENSE SCO
,U.1»
41m
u
i.» t-.-l
US
hopper laAdad, not running ?:C8
The Airsense operators chose to leave the instalment off over the first night prior to Test Day I.
It was then switched back on two hours prior to the pre-test calibration (see Trend Graph 1). A
high zero and calibration drift resulted from this as the internal vacuum improved through the day
and the instrument's background decreased Coincidently, as the background decreases, the
instrument's sensitivity increases. The Airsense was left nmning for the remainder of the test
program.
It is important to note the continuous real-time data that was generated for this test. Fast
measurements allow the process operator to immediately observe changes in concentrations as
incinerator conditions change. The first and last 30 minutes of the trend graph represents the
calibration and zero responses to direct input of I ppm multi-mix calibration gas and nitrogen
respectively. The 90% of full response was typically a few seconds. This could be in milliseconds
depending on the application and speed of analysis that is desired. Response to zero was similar
The remaining data shows the immediate response to the process gas. The Airsense demonstrated
that the planned step changes in target contaminant feed rate only resulted in gradual decreases.
Trend Graphs 2 and 3 display the data collected for Test Days 2 (08/02/96) and 3 (08/04/96).
The instrument's calculated RA's for these days improved as the target flue gas concentrations
increased and as the operator's technique for stack gas background compensation improved.
136
-------
Pf
Ca Hfc lo«
TREND GRAPH #2 -TEST DAY 8S/08/02
ECO LOGIC AIRS E N S E 500 DATA
f
PaaUTaat
Calibration
Wa«#r Praaturt
cerr«ct*4 lor
Qu*nclt Ta«fc
| «):«!
MM
#3 | 1S42
Flnlahatf RM
• 1 ft 19 44
Startad RM
Jf.rf.d RM 02 | n:12
i1 ft 1.01
Flnlahad AM
#2 • 1 3:45
Start** KM
iJ S K:0S
St«p Cn«ng«
• 1 | 15 41
Organic Faa4
Off m 17:92
TIME
o r. P *> - ft
TREND GRAPH #3 -TEST DAY 95/08/04
ECO LOGIC AIRSENSE 500 DATA
Po«t-T»st
Calibration
Pra-T
C»lib
Fmlthtd RM
«1B £ 11:35
Flnishtd RM
•3 ft 15 10
Loatitd Hoop"'
A 15:10
Start** RM
Flniihcd RM
*16 fi 10:00
•2 0 13:39
Started RM
•3 £ 12:40
Organic F««4
Startad RM
Off £ 1C;28
•1 a t.17
srartad RM
•1 t 12:04
Wttta ft Organic
Ftad 4 t: 10
Jtap Chans*
#1 C 13:20
s s s s
S 8 8 8 8
137
-------
Real time emission monitoring capabilities were highlighted in Test Day 2 at 8:40 a.m. when the instrument began
sampling process gas. The signals began rising rapidly, especially benzene. At 9:08 a.m. the incinerator operators
corrected the water pressure of the Quench tank. The signal dropped immediately with this correction. A further
correlation can be made between the high and erratic benzene peaks throughout Test Day 2 with the inconsistent
ERF Combustion Gas data seen in Figure I.
Trend Graph 3 shows similar trends in regards to calibration periods, waste processing, and waste step changes.
The one exception is the high toluene peaks accompanied by more stable benzene readings. This possibly correlates
with the stable IRF Combustion Gas data collected for that day, see Figure 2.
Figure 1
0 300 (ppm)
0 21 (%)
Figure 2
0-21 J
AftmUfffltf NO*
s
C07
RM1
RM3
^X<"1.
F'tc« cftanga
5>grM' evfpof paditm
FJ:„c».nQ. fJIKWjngt
/
cfung*
9 11 13 15
Time ol Day (hf)
Kiln, Afterburner, and Stack CEMs
Test 2VOC 8/2/95
3cm
*200 Jppm)
0-21 (%)
0 240 (p^m)
0 2< (*)
AMI
RM2 RM3
jd
HI
02
C02
> n 13 15
Time of Day {hi)
Kiln, Afterburner, and Stack CEMs
Test 3VOC 8/4/95
138
-------
Conclusion
• The Airsense 500 calculated RA's for the three days improved as the target flue gas concentrations were
increased. Operator experience for this particular application improved through the three Test Days
• The instrument's ruggedness was demonstrated with setup directly at the scrubber exit sampling locations, in
over 100°F ambient temperature and continuous sampling of untreated gas.
• Rapid on-line measurement and data presentation capabilities was demonstrated with the incinerator
operator's ability to see immediately on screen the effects of incinerator process changes on the VOC
concentrations
• Ease of operation was demonstrated with minimal manual effort required to operate the instrument. During
the RM test runs, the Airsense operator was required only to manually input comments regarding process
conditions.
• Single point calibrations were performed using 1 ppm multi component calibration gases.
• These instruments are commercially available and have been for years. The rapid continuous measurement
capabilities for VOC's, sulfur and nitrogen compounds, PAHs, and many more have been actively utilized
worldwide for a variety of applications.
• Application development, not instrument design, is a necessary requirement for these types of studies if the
Airsense is to be considered as a regulatory tool We agree with the importance of further studies of this
magnitude and welcome the opportunity to be involved.
• For the present, industrial emission inventories and combustion process optimization are programs that
could benefit from cost effective and time effective rapid monitoring techniques as opposed to time and
labor extensive regulatory techniques.
139
-------
Intentionally Blank Page
-------
ECOCHEM SVOC CEM DEVELOPER COMMENTS
141
-------
Intentionally Blank Page
-------
Developer's Comments...
EcoChem Realtime PAH Monitor
Summary
EcoChem appreciates the opportunity to have contributed to this important project with the PAS tOOOe totalparticulate-bound
polycyclic aromatic hydrocarbon (PAH) monitor. EcoChem believes that the program results demonstrate the capability of the PAS
1COOe to provide real-time measurement of particulate bound PAH and process monitoring for combustion diagnostics. In addition to
the findings and conclusions in previous sections of this report. EcoChem would like to draw attention to the following information and
issues:
• monitor specification summary,
• additional analysis of the relative accuracy test results.
• detection charactenstics of the monitor, and
• use of the PAS 1000c as a combustion monitor.
This information provides further analysis of the monitor's performance, and offers additional monitor operational characteristics not
assessed in the scope of this evaluation.
Analyzer Specification
The photoelectric aerosol sensor (PAS) works on the principle of photoionization of aerosol bound polycyclic aromatic hy drocarbons
(PAH). Sample gas containing products of incomplete combustion is first passed through an clectrofilicr to remove any charged
aerosols. Using a UV light lamp, the electrically neutral flow of aerosols is then ionized. The wavelength of the light is chosen such
that only the PAH coated aerosols arc ionized, while gas molecules and non-carbon aerosols remain neutral. The aerosol panicles
which have PAH molecules adsorbed on the surface then emit electrons which arc subsequently removed by application of an electric
field. The remaining positively charged panicles arc collected on a filter inside an electrometer, where the charge is measured. The
resulting electric current establishes a signal which is proportional to the lota! concentration of photoionizable particles in the
sample gas stream, and hence proportional to the total particiilnle-bour.d PAH concentration. The analyzer does not differentiate
between various PAH species, and the analyzer signal is a measure of total PAH adsorbed on the carbon panicles
Table 1. PAS 1000e monitor specifications
MOOEl PAS 1000e
Display Two 3 V4 digit LED displays
POWER 115 volts AC / 60 Ha -- 150 Watts Moritor, 100 Watts Pump
Range 0 tc 2 picoarrp, 0 to 20 picoamp and 0 to 100 p.coarr.p (user selectable)
ENSi, MTi ~ 1 -3 |ig /m PAH pef pcoarr.p (actual ca'ibration is site-specific)
RESPONSE TIME Selectable (7 sec , 0.7 sec and 70 milliseconds)
Analog Output 0 - 5 volt, and 0 to 20 miiilamp
Digital Output RS - 232
Sample gas Total Flew 4 liter/min. Heated probe with dilution system. Flue gas dilution set points cf 0 01, 0.02,
Q 05. 0.1, Q 2 and 1 0 RingsM exchangeable sampling nozzle Two rr.3SS flow controllers
Operating temp Mor.ltor <0 to 120 c~. Stack sampler 750'F
DIMENSIONS Mooter 8 3in x 21 2 in * 17 8 in ( H x W x D )
Pump 6.9 n * 19.7 m x 13.8 ir, ( H x W x D )
WEIGHT Moo ter 66 lb., Pu.-rp 22 :b.
DIGITAL transmitter Voltage r5 vo'ts , Resclutsci: 0 01 % of FS. Accuracy t 0 C2% of FSmax
Input; Voltage Output: R232
Da-A ACCUISITION SOFTWARE Sofrw^'e collects data frcn PAS 1000e, displays real-time strip chart and calculates averages
User-aefined calibration factors and averaging time Flat ASCII file output for further analysis (use in
standard spreadsheets e g EXCEL)
143
-------
EcoChem Realtime PAH Monitor
Relative Accuracy
It should be emphasized th.it the Photoelectric Aerosol Sensor detects only particle-bound PAH As a result of combustion, a
wide spectrum of PAH are produced. These range from the simple double-ringed compound like naphthalene which is
predominantly in the vapor phase (due to it's low boiling point) to the other high molecular weight PAJI (e.g. bcnzo(a)pyrene) which
are predominantly in the particle-bound phase (due to their high boiling points) The relative partitioning in the vapor and particle-
bound phases is dependent upon the system temperature, particle loading and the chemical properties.
The reason this needs to be emphasized is because the standard chemical analysis methods (e g Method 8270B) typically
determine overall total PAH (vapor and particle-bound) while the PAS instrument reports only total panicle-bound PAH Thus
caution should be exercised when an attempt is made to directly compare the results of the real-time PAH monitor with the results of
a chemical analysis method.
Table 4-16 of the main document provides a summary of (lie relative accuracies for the two tests which were conducted
After looking at the detailed values provided in the Appendi\ reports, we would like to comment on the results in Table 4-16
Low Concentration Test: The analytical results for the low-level SVOC Reference Methods indicate that the analvtes
are at or below the lower detection limits of the method (Table 2). For Low Level SVOC tests, all but or.c (naphthalene for
Reference Method 3) are below the louest calibrated lev el and the highest detected ana!\1e (naphthalene for Reference
Method 3 at 5.13 nig) is within a factor of two limes the mean of the detection limit for all the undetected analvtes (2.7 mg
±0.7). In addition, the mean of all detected analyics for all Low Level Reference Methods (3.7 mg ± I) is within a factor of
I 5 times the mean of the detection limits for all the undetected analvtes.
Table 2: Excerpted Data from Flue Gas Report, Gas Chronnatography/Mass Spectroscopy, Method 8270B
Test
Low Level, RM 1
Low - Level, RM 2
Low - Level, RM 3
Tarqet Ar.alytes
mq/train
mq/train
ma/train
Naphthalene
4.90 J
4.57 J
¦ ¦ 5.13
2-MethylnaphthaIene
<2.21
<2.21
<2.21
Acenaphthylene
<2.63
<2 63
<263
Acer.sphthene
<2 56
<2 So
<2 56
Fluorene
<2 64
<2 64
<2.64
Phenanthrene
3 85 J
3.10 J
3.86 J
Anthracene
<3 41
<3.41
<3.41
Fiuoranthene
<3.43
<3 43
<343
Pyrer.e
3 00 J
2.14 J
2.66 J
Benzcjajanthracere
<1.99
<1.99
<1.99
Chrysene
<1 87
<1 87
<1.87
Benzo(b)fl-Joranthene
<2.67
<267
<2.67
Benzcfkjfluoranthene
<1.07
<1.07
<1.07
Beruc(a)pyrene
<2.80
<2.80
<2 80
lndeno(1.2,3-cd)pyrene
<3.05
<3 05
<3 05
Diberz( a, h janthracene
<3.17
<3 17
<3 17
Benzc(g,h,i)perylene
<3.54
<3 54
<3.54
J - Be^cw lowest calibrated level. RM - Reference Method
The compounds which the PAH monitor is supposed lo detect fparticle-bound higher ringed
compounds) are predominantly in the non-detectable range. Hypothetical^ the sum of the undetectable
analvtes could result in a quantity of PAH on the order of what v^as measured. In this case, the
differential between the total Reference Method PAH concentration and the PAS total PAH concentration
would be reduced, resulting in a significantly lower relative accuracy.
144
-------
EcoChem Realtime PAH Monitor
Medium Concentration Test: For (his icst (Tabic 3), ihcrc is a reported anomaly with the pyrene analysis for Run 2.
For all other tests, the standard deviation of the results are less than 20 percent. However, for Run 2, pyrene(60 2 mg) is
over 2 times the mean of Runs 1 and 3 (28 2 mg). If the assumption is made that for Run 2. the value for pvrene
is equal to the mean of Runs 1 and 3. then the estimated value of total PAH for Run 2 is 30 mq/m1. This
estimated concentration correlates well with the PAS concentration of 33.2 mq/m'.
Table 3: Excerpted Data from Flue Gas Report, Gas Chromatography/Mass Spectroscopy, Method 8270B
Test
Medium level, RM 1
Medium - Level, RM 2
Medium - Level, RM 3
Tarqet Analytes
mq/train
mc/train
ma/t.'airt
Naphthalene
53.0
33 7
47.1
2-Methylnaph{haler,a
<2 21
<2 21
<2.21
Acenaphthylene
<2 63
<263
<2.63
Acenaphthene
<2 55
<2 56
<2.56
Fluorene
<2 64
<2.54
<2.64
Phenanthrene
47.4
31.2
45.8
Anthracene
<3.41
<3.41
<3.41
Flucranthene
<3 43
<3.43
<3.43
Pyrene
27.6
60.2"
28 8
Benzc(a)ar:hraeene
<1 59
<1 99
<1 99
Chrysene
<1 87
<1 37
<1.87
Benzc('o)(luorar,ther,e
<2 67
<2 67
<2 67
Benzo(k;fIjorar,ther.e
<1 07
<1 07
<1 07
8enzo(a)pyrene
<2 80
<2.30
<2.80
IndencK 1,2,3-cd)pyrsne
<3X5
<3.05
<3.05
Dibenzja.hjanthracene
<3.17
<3 17
<3.17
BenzoCg,h.i)perylene
<3.54
<3.54
<3.54
J • Below lowest calibrated level * Anomaly covered in case narrative.
Site-Specific Calibration Curve
Using data wheie the output of the PAH monitor has been compared with analytically determined PAH concentrations, a general
calibration curve has been generated (Figure 1). For greater detail, a site-specific calibration curve may be generated. In order to
compare the PAS signal from one analyzer to another, ozone measurements are made in the ionization unit of each analyzer.
At this time, there is fairly good evidence to indicate that there exists a universnl calibration of the PAS That is. there is an
approximate relationship between the chargc/m' and the total poniculntc-fcound PAH which is independent of the type of aercsol,
w ithin a factor of two. This calibration curv e provides relative values Tor screening and rca'-t:rac trending applications.
lo.coo
t.ooo
Total Partide • Sound
PAH (n3'm3)
100
R-Squared » 0 95
Paring str-ctj'e
CI bjrne.'
Ut::an Participate Matter
Fgelo*!
- Reg'essisn
IDG 1.000 10.000
C jr/eM ders.fy{picoarrp-sec'm3)
100.000
Figure 1. Response of the photoelectric aerosol sensor for different combustion sources.
145
-------
EcoChem Realtime PAH Monitor
In order to obtain accurate absolute numbers, a site-specific calibration curve can be generated where the monitor output is
compared to an analytically determined PAH concentration (e.g. obtained from sampling by Method 0010 and analysis with Method
8270B). A site-specific calibration curve would then accurately represent tlic particle size, charge and specific PAH distribution of
the source. A site-specific calibration curve based on reference method results (pg/m3) and raw PAS output (pA) is shown in Figure
2. Considering the limitations which we discussed in the Relative Accuracy section, we feel that the PAH
monitor produces reasonable values which match those obtained from reference method testing.
40
I
35
Medium Tesl, Run 2, Corrected * + + j
I
30
^ !
I
1
£
25-
20.
^ j
£
15
10 -
*
* Pyrene for Medium Test 2 ;
5 .
corrected to mean of Medijrr. f
0 ^
Test Runs 1 and 3 !
~
s
10 15 20 25 30 35 40 45
Rafmer.se Method (ogfm*3}
Figure 2. Site-specific Calibration Curve Developed from Reference Method Test Results.
Use of PAS 1000e as a Combustion Monitor
The signal from the Photoelectric Aerosol Sensor is .in excellent real-time indicator of the operational characteristics of a
combustion process. Previous studies (Ref: Pilot - Scale Evaluation of the Potential for Emissions of Hazardous Air Pollutants
from Combustion of Tire - derived Fuel, EPA - 600/R-94-070, 1994) have reported that the PAH monitor tracked disruptions in
the combustion process very well in spite of the fact that measurements of CO did no! sigiiificantK change. As an illustraticn of this
capability, changes in the PAH monitor output were correlated to incinerator process changes « hich occurred during the test
program (Figures 3, 4 and 5).
Figure 3 is an example of the step change obscned following tlte introduction of the SVOC feed to the process. In the
figure, an increase from the nominal baseline signal is evident, stabilising at an upscale level after the system reaches a steady state
condition.
The sensitivity of the PAS lOOOe to small changes in process operation is shown in Figure 4. The figure illustrates the
effect of opening sample ports on the PAH signal. For this and every other instance, an upscale spike followed by a down scale spike
was observed when a sample port was opened during rcfcr:r.cc method testing. Although the mechanism for this efTec! is not kaoun
(i.e. a change in combustion characteristics due to changes m incinerator pressure, or a increase in entrained particulate due to a
change in fan operation due to the pressure change), i; is clear that a momentary change in duct pressure results in an observable and
reproducible change in signal output.
In Figure 5, it is clearly observed that the PAH monitor signal decreased when the SVOC/YOC feed was shut-off, increased
when the water was sprayed into the system to Hush out the organic feed and increased again uhen the quenching system was shut
off. The increase in concentration during the water spray was obscn ed to same degree during the end of all test days This effect
146
-------
EcoChem Realtime PAH Monitor
may have been due to an increase in PAH concentration caused by a large charge of SVOC feed introduced into the process while the
organic feed system was cleared with water. During this period, the remaining feed may have been introduced into the process at a
higher flow rate than that fed during the test period, resulting in more PAH in the system. The increase in signal output aAer the
quench system was turned of may have been due to an increase in particulate loading in the duct, however, this change was not seen
in all tests.
In these examples, the use of the PAS lOOOe signal as a diagnostic indicator of both small and large changes in process
operation is shown. These observations lead one to conclude that the real-time PAH analyzer may be used as a Process Monitor for
combustion systems.
Feed introdyced
6:4 S 7:10
Time
Figure 3. PAH Monitor response during startup of system (9/73/95)
Opening of sample ports
for Reference Method Testing
13:15 13:55
Time
Figure 4 PAH Monitor response during opening cf sample ports (or Reference Method Testing (8/28/95)
Q uench
system
Probe
rem o v e d
VOC/SVOC
a i.
1 9:40
20:30
Time
Figure 5. PAH Monitor response during shutdown cf system (3/30/95)
147
-------
Intentionally Blank Page
-------
SNL MULTI-METALS CEM DEVELOPER COMMENTS
149
-------
Intentionally Blank Page
-------
Developer Self-Assessment for the Sandia Laser-Spark Metals CEM
In Test Scries 2 of the EPA/IRF CEM tests, the Sandia Laser-Spark metals CEM recorded
measurements for three of the fourteen RCRA metals during the "intermediate" metal-
concentration tests and for four of the RCRA metals during the "high" metal-concentration
tests. None of the RCRA metals were measured for the tests at the "low" concentration
levels. Comparison of these results with the estimated detection limits of Table 4-21
indicates an obvious discrepancy between the apparent field performance of the Laser-
Spark monitor and the performance measured in the laboratory. Since the EPA/IRF tests,
we have thoroughly gone over our equipment, we have corrected some minor deficiencies
in the optical alignment, and we have made some system improvements. However, the
field performance of the Laser-Spark monitor in the EPA/IRF tests is unexpectedly low
even in view of the most recent laboratory measurements. We currently have two
conjectures as to why the EPA/IRF results were disappointing, which are discussed below:
1. The first explanation arises from uncertainties in the particle number density. During
earlier tests at the SAIC STAR Center, a pilot-scale waste-treatment facility, the Laser-
Spark monitor detected intermittent "bursts" of lead emission, the low occurrence rate
of which indicates the presence of very low particle number densities. Post-test
analysis of physical samples extracted during the STAR Center tests indicated that the
concentrations of all metals appearing in aerosol form were significantly below the
lower detectability limits that had been measured previously in the laboratory. At the
time, we believed that the low observed particle number-density was due to unusually
low metal concentrations. However, particle number densities are in general not well
known for facilities such as the STAR Center and the IRF, and we believe that it is
reasonably possible that particle number densities during the IRF tests were also
sufficiently low so that the Laser-Spark technique (as currently configured) did not
adequately sample the particles during the averaging periods that were selected.
2. The second explanation arises from uncertainties in the effect that large concentrations
of condensed water would have on the laser-spark measurements. The EPA/IRF
system experienced an equipment failure before our tests, and, as a consequence, the
exhaust blower pulled such a large draft on the system that a visible dense mist of
condensed water streamed through the measurement volume. We do not currently
know to what extent these extremely irregular conditions affected the optical properties
of the flow or the plasma-creation process, cither of which could in turn strongly
influence the indicated metals measurements.
Currently, efforts are being made to better characterize the particle size distribution in the
emissions from facilities similar to the IRF. At the same time, we are working on
improvements to the monitor that will improve the response for low particle number
densities. We believe that a combination of these efforts will help us to understand the test
151
-------
2
environment and will allow us to better respond to the range of particle conditions that are
present in facilities such as the IRF. We are not currently addressing the second
explanation above, because we believe that this is a very off-normal condition that would
ordinarily cause a facility to cease operations regardless of the measurements made by a
metals CEM.
152
-------
ECOCHEM Hg CEM DEVELOPER COMMENTS
153
-------
Intentionally Blank Page
-------
Developer's Comments...
EcoChem Mercury CEM
Summary
EcoChem appreciates the opportunity to have contributed to this important project with the EcoChem Mercury CEM (also referred to
as the Hg-Mat 2). EcoChem believes that the program results demonstrate the capability cf the Hg-Mat 2 to provide real-time
measurement of total mercury continuous compliance monitoring In addition to the findings and conclusions in previous sections of
this report, EcoChem wants to draw attention to the following information and issues:
• analyzer specification summary
• site location and en vironmental considerations.
• additional analysis of the relative accuracy test results,
• correction in the Zero and Calibration drift Table 4-23 provided m the main document,
• additional analyzer performance data from tasting conducted by TUV RHeinland in Germany
The TUV information provides further analysis cf the analyzer's performance, and offers additions' analyzer operational characteristics
not assessed in the scope of this evaluation These include, quantification of interference (rem other Hue gas components,
comparison of response from two independent analyzers, and additional comparison agairst independent reference methods.
Analyzer Specifications
The EcoChem CEM provides continuous emission monitoring of lot.nl mercury in combustion and pyrolysis system flue gas.
Specifications for the analyzer arc summarized in Table 1 The Hg-Mat 2 uses a fixed wave length atomic absorption spectrometer
and a sample conditioning sy stem designed to provide a continuous signal output of mercury concentration All components arc
housed in an air conditioned 19-inch rack. Microprocessor control provides fully automated operation and a menu driven operator
interface. The Hg-Mat 2 is a complete emission measuring system for mercury consisting of three functional units: the sample
conditioning section, detector section, and microprocessor unit
Table 1: Hg-Mat 2 Analyzer Specifications.
Analyzer Data
Hg detector:
Atomic absorption spectrometer with fixed wave length, UV photometer with reference beam
for lamp control, electrodeless mercury lew pressure lamp
low 0-0.15 mg/Nm3, High: 0 - 1200 mg/Nm5
0.001 mg/Nm3
approximately 2 mi.n. (response time T90)
approximately 60 min.
5 to 2C0°C
< 25C hPa(mbar) ever pressure
60 - SO l/h
4 mm Teflon, heated to 200°C
Measurement ranges:
Resolution:
Response time:
Warm up time:
Sample gas lemp:
Sample gas press:
Sample gas flow-
Sample line:
Signal Outputs
Measurement signal:
4 - 20 mA, 0 -1 VDC
3 relay outputs: Operative, Malfunction, Maintenance
Analyzer status:
Physical Data
Weight:
Dimensions:
Power requirements:
Operating temperature:
150 kg
163 X 60 x 60 cm (H x W x O)
220 VAC. +10%, -15%, 5C/60 Hz
5 to 35°C
155
-------
EcoChem Mercury CEM
Site Location and Environmental Considerations
Due lo the distance from the sampling location to the environmentally controlled trailer. EcoChem elected to position (he CEM in an
non-air conditioned bay in the Incinerator Research Facility building closer to the sample port This decision was made contingent
on the temperature not exceeding the 35 ®C upper operating limit of analyzer, thus exceeding the capacity of the installed air
conditioner. However, during the test program the ambient temperature did exceed 35 °C. and reached as high as 40 °C. Although
the Hg-Mat 2 was fully operational for the test program, the location turned out to provide a challenge at the upper end of the
analyzer's operating range (S - 35 °C). In typical operation, the unit would be located in a more environmentally controlled room or
housing, or a sufficiently sized air conditioner would be installed in the instrument rack.
Relative Accuracy
EcoChem agrees with the report's conclusion that relative accuracy values on the order of 30 percent enn be expected based on the
test results from this program. Evaluation of relative accuracy based on only three sets of reference method tests does not reflect the
results that would be obtained from a nine or tuelve test evaluation. This is a common requirement in the existing CEM performance
specifications, or required in the new proposed Mercury performance specification document "Draft Performance Specification 12
- Specifications and test procedures for totsi mercury continuous monitoring systems in stationary sources."
Tabic 2 provides a summary of relative accuracy estimates calculated from the three tests conducted at each test condition
(as reported in Table 4-22), and two estimations for a nine sample calculation based oh the three tests:
1. ESTIMATE i is lite relative accuracy winch is reported in main document Table 4-22. Here three sets of reference
method tests were used to create the relative accuracy numbers. In addition, the calculation assumes a "II orst*
case" limit (Equation 4-7 -- RA5 / RA, = 1) for extrapolating from 3 lo 9 data points.
2. ESTIMATE 2 is the relative accuracy calculated if we were to use three replicate sets of the three Reference
Method tests to create 9 data points. No extrapolation algorithm is applied for this calculation,
3. ESTIMATE 3 is similar to ESTIMATE I, Here »ve have three data points but use the "!3cst-case" limit (Equation
4-8 -- RA« / RA; = 0.31) for extrapolating from 3 to 9 data points.
Table 2 indicates that the relative accuracy could be significantly lower than those reported in the main
document (ESTIMATE 1i. Other than one value, ESTIMATE 2 and 3 result in relative accuracies close to or below
30 percent with several estimates at or below 20 percent.
Table 2. Relative Accuracy (RA) Estimations for Nine Tests based on the Three Test Data Set,
Hg Test Series
ESTIMATE 1
RA, %
based on 3 RM tests
"Worst-case*
extrapolation
ESTIMATE 2
RA. %
for 9 RM tests
3 replicate data
sets
ESTIMATE 3
RA. %
based on 3 RM tests
*Bes!-ease*
extrapolation
Low concentration
60
32
19
Intermediate concentration
92
54
29
High concentration
61
21
19
The good agreement between the reference method and the Hg-Mat 2 is further illustrated b> the following chart (Figure l).
The figure shows consistent agreement wi'.h variations in the process over the period of each test series. For tests at low and high Hg
156
-------
EcoChem Mercury CEM
test conditions, total measured mercury concentrations decreased during the day, as demonstrated by results from both methods.
During the medium Hg test condition, the decrease and subsequent increase of Hg concentration during the test day is reflected in
the results as well.
140
m 120
E
§i 100
c
o
80
t 60
ci
X
40
20
~ RM
¦ EcoChem
in ¦ i
cc
5
o
CM
5
a.
i
o
2
IE
•a
(N
(M
n
5
5
2
5
5
Of
tr
Of
Of
ir
TJ
XJ
XI
.c
-C
0)
a)
,g>
o
O)
5
S
"T~
X
I
Reference Method Test
Figure 1. Comparison of Reference Method (RM) and EcoChem Mercury CEM
Correction to Zero and Calibration Drift Table 4-23 of Main Document
In the main document the values for the Zero and Calibrniion drifts have been interchanged. The correct values for the EcoChem
CEM should read as :
Table 3: Corrected values for Zero and Calibration Drifts
Low Mercury Concentration
2D
0%
CD
0%
Intermediate Mercury Concentration
ZD
0%
CD
-10 %
High Mercury Concentration
ZD
0%
CD
-23 %
Additional Hg-Mat 2 Performance Data
The Hg-Mat 2 has been approved by TliV Rhcin!a:id in Germany fRef: Report on the Suitability Test Cf the Hg-Mat 2 by TUV
Rheintand, Cologne, Report No: 936/802 009/Hg August 1994 ) for monitoring total mercury emissions from incinerators. As
157
-------
EcoChem Mercury CEM
part of these long term evaluations, the analyzer is subject to interference and accuracy evaluations Results of these evaluations arc
summarized in the following sections to supplement the performance evaluation conducted during the 1RF tests. Table 4
summarizes the results of interference tests conducted to determine the response of the analyzer to other Hue gas constituents. Figure
2 provides additional Reference Method test data and Figure 3 illustrates the reproducibility of results between two instruments.
Table 4. TUV Rheinland Interference Test Data for Hg-Mat 2
Gas
Concentration
Carrier Gas
Hg Concentration
mg/m3
Interference
% Full Scale
o
o
15 %
N j
-
<1
15%
n2
93
<1
CO
484 rr.g/m3
n2
-
<1
434 mg/m3
N,
75
<1
Oj
20.5 %
N;
-
<1
20.5 %
Nj
81
<1
NO
496 mg/m3
n2
-
<1
496 mg/m3
n2
76
-2.8
NOj
133 mg/m3
n2
-
<1
133 mg/m3
n2
75
-12.6
13 mg/m3
n2
-
<1
13 mg/m3
n2
77
-1.8
HC1
180 mg/m3
Synth. Air
-
<1
180 mg/m3
Synth. Air
81
<1
S02
500 mg/m3
n2
-
11.6
500 mg/m3
n2
81
10.7
50 mg.'m3
Nj
-
1.4
50 mg/m3
Nj
76
1.3
n-Butane
43.6 mg/m3
Synth. Air
-
<1
43.6 mg/m3
Synth. Air
81
<1
nh3
12.5 mg/m3
Nj
-
<1
12.5 mg/m3
Nj
83
+ 1.3
N;0
97 ppm
Nj
-
<1
97 ppm
Nj
82
<1
Benzene
1.08 mg/m3
Nj
-
<1
1.08 mg/m3
Nj
82
<1
158
-------
EeoChem Mercury OEM
2
o
o
X
I"
UJ
S
UJ
(J
2
UJ
cr
UJ
u.
UJ
-------
Intentionally Blank Page
-------
REFERENCES
1. "Continuous Performance Assurance for Metals Emissions from Hazardous Waste Combustion
Systems: An ASME/EPA Joint Workshop," Cincinnati, Ohio, September 1993.
2. 40 CFR 266, Appendix IX.
3. "Test Methods for Evaluating Solid Waste: Physical/Chemical Methods," EPA SW-846, 3rd
edition, Revision 2, September 1994.
4. Fournier, D. J., Jr., W. E. Whitworth, J. W. Lee, and L. R. Waterland, "The Fate of Trace
Metals in a Rotary Kiln Incinerator with a Venturi Scrubber/Packed-Column Scrubber,"
EPA/600/R2-90/043, February 1991.
5. Fournier, D. J., Jr., and L. R. Waterland, "The Fate of Trace Metals in a Rotary Kiln
Incinerator with a Single-Stage Ionizing Wet Scrubber," EPA/600/R2-91/032, September 1991.
6. Whitworth, W. E., Jr., and L. R. Waterland, "Evaluation of the Impact of Incinerator Waste
Feed Cutoffs," Acurex Environmental report under EPA Contract 68-C9-0038, October 1995.
7. Fournier, D. J., Jr., and L. R. Waterland, "The Fate of Trace Metals in a Rotary Kiln
Incinerator with a Calvert Flux-Force/Condensation Scrubber System," Acurex Environmental
draft report under EPA Contract 68-C9-0038, January 1993.
8. "Quality Assurance Project Plan for Testing the Performance of Real-Time Incinerator
Emission Monitors," prepared by Acurex Environmental under EPA Contract 68-C4-0044,
Work Assignment 0-4.
161
-------
Intentionally Blank Page
-------
APPENDIX A
DRAFT PS FOR MULTI-METALS CEMs
163
-------
Intentionally Blank Page
-------
DRAFT PERFORMANCE SPECIFICATION - SPECIFICATION AND TEST
PROCEDURES FOR MULTI-METALS CONTINUOUS EMISSION MONITORING
SYSTEMS IN STATIONARY SOURCES
March 22,1995
1. Applicability and Principle
1.1 Applicability. This specification is to be used for evaluating the acceptability of multi-
metals continuous emission monitoring systems (CEMS) at the time of or soon after installation
and whenever specified in the regulations. The CEMS may include, for certain stationary sources,
a) a diluent (02) monitor (which must meet its own performance specifications: 40 CFR part 60,
Appendix B, Performance Specification 3), b) flow monitoring equipment to allow measurement
of the dry volume of stack effluent sampled, and c) an automatic sampling system.
«
A multi-metals CEMS must be capable of measuring the total concentrations (regardless
of speciation) of two or more of the following metals in both their vapor and solid forms:
Antimony (Sb), Arsenic (As), Barium (Ba), Beryllium (Be), Cadmium (Cd), Chromium (Cr), Lead
(Pb), Mercury (Hg), Silver (Ag), Thallium (Tl), Manganese (Mn), Cobalt (Co), Nickel (Ni), and
Selenium (Se). Additional metals may be added to this list at a later date by addition of
appendices to this performance specification. If a CEMS does not measure a particular metal or
fails to meet the performance specifications for a particular metal, then the CEMS may not be
used to determine emission compliance with the applicable regulation for that metal.
This specification is not designed to evaluate the installed CEMS' performance over an
extended period of lime nor does il identify specific calibration techniques and auxiliary
procedures to assess the CEMS' performance. The source owner or operator, however, is
responsible to properly calibrate, maintain, and operate the CEMS. To evaluate the CEMS'
performance, the Administrator may require, under Section 114 of the Act, the operator to
conduct CEMS performance evaluations at other times besides the initial test. See Sec. 60.13 (c)
and "Quality Assurance Requirements For Multi-Metals Continuous Emission Monitoring
Systems Used For Compliance Determination."
1
165
-------
1.2 Principle. Installation and measurement location specifications, performance
specifications, test procedures, and data reduction procedures are included in this specification.
Reference method tests and calibration drift tests are conducted to determine conformance of the
CEMS with the specification.
2. Definitions
2.1 Continuous Emission Monitoring System (CEMS). The total equipment required for
the determination of a metal concentration. The system consists of the following major
subsystems:
2.1.1 Sample Interface. That portion of the CEMS used for one or more of the
following: sample acquisition, sample transport, and sample conditioning, or protection of the
monitor from the effects of the stack effluent
*
2.1.2 Pollutant Analyzer. That portion of the CEMS that senses the metals
concentrations and generates a proportional output
2.1.3 Diluent Analyzer (if applicable). That portion of the CEMS that senses the diluent
gas (02) and generates an output proportional to the gas concentration.
2.1.4 Data Recorder. That portion of the CEMS that provides a permanent record of the
analyzer output. The data recorder may provide automatic data reduction and CEMS control
capabilities.
2.2 Point CEMS. A CEMS that measures the metals concentrations either at a single
point or along a path equal to or less than 10 percent of the equivalent diameter of the stack or
duct cross section.
2.3 Path CEMS. A CEMS that measures the metals concentrations along a path greater
than 10 percent of the equivalent diameter of the stack or duct cross section.
2
166
-------
2.4 Span Value. The upper limit of a metals concentration measurement range defined as
twenty times the applicable emission limit for each metal.
2.5 Relative Accuracy (RA). The absolute mean difference between the metals
concentrations determined by the CEMS and the value determined by the reference method (RM)
plus the 2.5 percent error confidence coefficient of a series of tests divided by the mean of the RM
tests or the applicable emission limit
2.6 Calibration Drift (CD). The difference in the CEMS output readings from the
established reference value after a stated period of operation during which no unscheduled
maintenance, repair, or adjustment took place.
*
2.7 Zero Drift (ZD). The difference in the CEMS output readings for zero input after a
stated period of operation during which no unscheduled maintenance, repair, or adjustment took
place.
2.8 Representative Results. Defined by the RA test procedure defined in this
specification.
2.9 Response Time, The time interval between the start of a step change in the system
input and the lime when the pollutant analyzer output reaches 95 percent of the final value.
2.10 Centroidal Area. A concentric area that is geometrically similar to the stack or duct
cross section and is no greater than 1 percent of the stack or duct cross sectional area.
3
167
-------
2.11 Batch Sampling. Batch sampling refers to the technique of sampling the stack
effluent continuously and concentrating the pollutant in some capture medium. Analysis is
performed periodically after sufficient time has elapsed to concentrate the pollutant to levels
detectable by the analyzer.
2.12 Calibration Standard. Calibration standards consist of a known amount of metal(s)
that are presented to the pollutant analyzer portion of the CEMS in order to calibrate the drift or
response of the analyzer. The calibration standard may be, for example, a solution containing a
known metal concentration, or a filter with a known mass loading or composition.
3. Installation and Measurement Location Specifications
3.1 The CEMS Installation and measurement location. Install the CEMS at an accessible
location downstream of all pollution control equipment where the metals concentrations
measurements are directly representative or can be corrected so as to be representative of the
total emissions from the affected facility. Then select representative measurement points or paths
for monitoring in locations that the CEMS will pass the RA test (see Section 7). If the cause of
failure to meet the RA test is determined to be the measurement location and a satisfactory
correction technique cannot be established, the Administrator may require the CEMS to be
relocated.
Measurement locations and points or paths that are most likely to provide data that will
meet the RA requirements are listed below.
3.1.1 Measurement Location. The measurement location should be (1) at least eight
equivalent diameters downstream of the nearest control device, point of pollutant generation,
bend, or other point at which a change of pollutant concentration or flow disturbance may occur
and (2) at least two equi%'alent diameters upstream from the effluent exhaust. The equivalent duct
diameter is calculated as per 40 CFR part 60, Appendix A, Method 1, Section 2.1.
3.1.2 Point CEMS. The measurement point should be (1) no less.than 1.0 meter from the
stack or duct wall or (2) within or centrally located over the centroidal area of the stack or duct
4
168
-------
cross section. Selection of traverse points to determine the representativeness of the
measurement location should be made according to 40 CFR part 60, Appendix A, Method 1,
Section 2.2 and 2.3.
3.1.3 PathCEMS. The effective measurement path should be (1) totally within the inner
area bounded by a line 1.0 meter from the stack or duct wall, or (2) have at least 70 percent of the
path within the inner 50 percent of the stack or duct cross sectional area, or (3) be centrally
located over any part of the centroidal area.
3.2 Reference Method (RM) Measurement Location and Traverse Points. The RM
measurement location should be (1) at least eight equivalent diameters downstream of the nearest
control device, point of pollutant generation, bend, or other point at which a change of pollutant
concentration or flow disturbance may occur and (2) at least two equivalent diameters upstream
from the effluent exhaust- The RM and CEMS locations need not be the same, however the
difference may contribute to failure of the CEMS to pass the RA test, thus they should be as close
as possible without causing interference with one another. The equivalent duct diameter is
calculated as per 40 CFR part 60, Appendix A, Method 1, Section 2.1. Selection of traverse
measurement point locations should be made according to 40 CFR part 60, Appendix A, Method
1, Sections 2.2 and 2.3. If the RM traverse line interferes with or is interfered by the CEMS
measurements, the line may be displaced up to 30 cm (or 5 percent of the equivalent diameter of
the cross section, whichever is less) from the centroidal area.
4. Performance and Equipment Specifications
4.1 Data Recorder Scale. The CEMS data recorder response range must include zero
and a high level value. The high level value must be equal to the span value. If a lower high level
value is used, the CEMS must have the capability of providing multiple outputs with different high
level values (one of which is equal to the span value) or be capable of automatically changing the
high level value as required (up to the span value) such that the measured value does not exceed
95 percent of the high level value.
5
169
-------
4.2 Relative Accuracy (RA). The RA of the CEMS must be no greater than 20 percent of
the mean value of the RM test data in terms of units of the emission standard for each metal, or
10 percent of the applicable standard, whichever is greater.
4.3 Calibration Drift The CEMS design must allow the determination of calibration drift
at concentration levels commensurate with the applicable emission standard for each metal
monitored. The CEMS calibration may not drift or deviate from the reference value (RV) of the
calibration standard used for each metal by more than 5 percent of the reference value. The
calibration shall be performed at a point equal to 80 to 120 percent of the applicable emission
standard for each metal.
4.4 Zero Drift. The CEMS design must allow the determination of calibration drift at the
zero level (zero drift) for each metal. If this is not possible or practicable, the design must allow
the zero drift determination to be made at a low level value (zero to 20 percent of the emission
limit value). The CEMS zero point for each metal shall not drift by more than 5 percent of the
emission standard for that metal.
4.5 Sampling and Response Time. The CEMS shall sample the stack effluent
continuously. Averaging time, the number of measurements in an average, and the averaging
procedure for reporting and determining compliance shall conform with that specified in the
applicable emission regulation.
4.5.1 Response Time. The response time of the CEMS shall be less than one-tenth the
averaging time. The response time shall be documented by the CEMS manufacturer.
4.5.2 Response Time for Batch CEMS. The response time requirement of Section 4.5.1
does not apply to batch CEMS. Instead it is required that the sampling time be no longer than the
averaging lime. In addition, the delay between the end of the sampling time and reporting of the
sample analysis shall be no greater than one hour. Sampling is also required to be continuous
except in that the pause in sampling when the sample collection media are changed should be no
greater than five percent of the averaging time.
6
170
-------
5. Performance Specification Test Procedure
5.1 Pretest Preparation. Install the CEMS and prepare the RM test site according to the
specifications in Section 3, and prepare the CEMS for operation according to the manufacturer's
written instructions.
5.2 Calibration and Zero Drift Test Period. While the affected facility is operating at
more than 50 percent of normal load, or as specified in an applicable subpart, determine the
magnitude of the calibration drift (CD) and zero drift (ZD) once each day (at 24-hour intervals)
for 7 consecutive days according to the procedure given in Section 6. To meet the requirements
of Sections 4.3 and 4.4 none of the CD's or ZD's may exceed the specification. All CD
determinations must be made following a 24-hour period during which no unscheduled
maintenance, repair, or manual adjustment of the CEMS took place.
5.3 RA Test Period. Conduct a RA test following the CD test period. Conduct the RA
test according to the procedure given in Section 7 while the affected facility is operating at more
than 50 percent of normal load, or as specified in the applicable subpart
6.0 The CEMS Calibration and Zero Drift Test Procedure
This performance specification is designed to allow calibration of the CEMS by use of
standard solutions, filters, etc that challenge the pollutant analyzer part of the CEMS (and as
much of the whole system as possible), but which do not challenge the entire CEMS, including the
sampling interface. Satisfactory response of the entire system is covered by the RA requirements.
Hie CD measurement is to verify the ability of the CEMS to conform to the established
CEMS calibration used for determining the emission concentration. Therefore, if periodic
automatic or manual adjustments are made lo the CEMS zero and calibration settings, conduct
the CD test immediately before the adjustments, or conduct it in such a way that the CD and ZD
can be determined.
7
171
-------
Conduct the CD and ZD tests at the points specified in Sections 4.3 and 4.4. Record the
CEMS response and calculate the CD according to:
CD = ^ Rcem ~ Ry ) x 100 . (i)
Ry
where CD denotes the calibration drift of the CEMS in percent, R^, is the CEMS response, and
Rv is the reference value of the high level calibration standard. Calculate the ZD according to:
ZD = ^ X 100 . (2)
rem
where ZD denotes the zero drift of the CEMS in percent, Rc^ is the CEMS response, Rv is the
reference value of the low level calibration standard, and R^ is the emission limit value.
7. Relative Accuracy Test Procedure
7.1 Sampling Strategy for RA Tests. The RA tests are to verify the initial performance of
the entire CEMS system, including the sampling interface, by comparison to RM measurements.
Conduct the RM measurements in such a way that they will yield results representative of the
emissions from the source and can be correlated to the CEMS data. Although it is preferable to
conduct the diluent (if applicable), moisture (if needed), and pollutant measurements
simultaneously, the diluent and moisture measurements that arc taken within a 30- to 60-minute
period, which includes the pollutant measurements, may be used to calculate dry pollutant
concentration.
A measure of relative accuracy at a single level is required for each metal measured for
compliance purposes by the CEMS. Thus the concentration of each metal must be detectable by
both the CEMS and the RM. In addition, the RA must be determined at three levels (10 to 50
percent of the emission limit, 100 to 200 percent of the emission limit, and 5 to 20 times the
emission limit) for one of the metals which will be monitored, or for iron. If iron is chosen, the
8
172
-------
three levels should be chosen to correspond to those for one of the metals that will be monitored
using known sensitivities (documented by the manufacturer) of the CEMS to both metals.
In order to correlate the CEMS and RM data properly, note the beginning and end of each
RM test period of each run (including the exact time of day) in the CEMS data log. Use the
following strategy for the RM measurements:
7.2 Correlation of RM and CEMS Data. Correlate the CEMS and RM test data as to the
time and duration by first determining from the CEMS final output (the one used for reporting)
the integrated average pollutant concentration for each RM test period. Consider system
response time, if important, and confirm that the pair of results are on a consistent moisture,
temperature, and diluent concentration basis. Then compare each integrated CEMS value against
the corresponding average RM value.
ft
7.3 Number of tests. Obtain a minimum of three pairs of CEMS and RM measurements
for each metal required and at each level required (see Section 7.1). If more than nine pairs of
measurements are obtained, then up to three pairs of measurements may be rejected so long as the
total number of measurement pairs used to determine the RA is greater than or equal to nine.
However, all data, including the rejected data, must be reported.
7.4 Reference Methods. Unless otherwise specified in an applicable subpart of the
regulations, Method 3B, or its approved alternative, is the reference method for diluent (02)
concentration. Unless otherwise specified in an applicable subpart of the regulations, the manual
method for multi-metals in 40 CFR part 266, Appendix IX, Section 3.1 (until superseded by SW-
846), or its approved alternative, is the reference method for multi-metals.
As of 3/22/95 there is no approved alternative RM (for example, a second metals CEMS,
calibrated absolutely according to the alternate procedure to be specified in an appendix to this
performance specification to be added when an absolute system calibration procedure becomes
available and is approved) to Method 29.
9
173
-------
7.5 Calculations. Summarize the results on a data sheet An example is shown is shown
in Figure 2-2 of 40 CFR part 60, Appendix B, Performance Specification 2. Calculate the mean
of the RM values. Calculate the arithmetic differences between the RM and CEMS output sets,
and then calculate the mean of the differences. Calculate the standard deviation of each data set
and CEMS RA using the equations in Section 8.
7.6 Undetectable Emission Levels. In the event of metals emissions concentrations from
the source being so low as to be undetectable by the CEMS operating in its normal mode (ie,
measurement times and frequencies within the bounds of the performance specifications), then
spiking of the appropriate metals in the feed or other operation of the facility in such a way as to
raise the metal concentration to a level detectable by both the CEMS and the RM is required in
order to perform the RA tesL
»
8. Equations
8.1 Arithmetic Mean. Calculate the arithmetic mean of a data set as follows:
* = - £ x. • (3)
n i
where n is equal to the number of data points.
8.1.1 Calculate the arithmetic mean of the difference, d, of a data set, using Equation 3
and substituting d for x. Then
4i = xt - y, . (4)
where x and y are paired data points from the CEMS and RM, respectively.
8.2 Standard Deviation. Calculate the standard deviation (SD) of a data set as follows:
10
174
-------
8.3 Relative Accuracy (RA). Calculate Ihe RA as follows:
4 * ^{SD)
RA = Jl ,
rbm
where d is equal to the arithmetic mean of the difference, d, of the paired CEMS and RM data
set, calculated according to Equations 3 and 4, SD is the standard deviation calculated according
to Equation 5, R ^ is equal to either the average of the RM data set, calculated according to
Equation 3, or the value of the emission standard, as applicable (see Section 4.2), and is the
t-value at 2.5 percent error confidence, see Table 1.
11
175
-------
TABLE 1
t-Values
n*
to
n*
n*
2
12.706
7
2.447
12
2.201
3
4.303
8
2.365
13
2.179
4
3.182
9
2.306
14
2.160
5
2.776
10
2.262
15
2.145
.6
2.571
11
2.228
16
2.131
»
' The values in this table are already corrected for n-1 degrees of freedom. Use n equal to the
number of individual values.
176
12
-------
9. Reporting
At a minimum (check with the appropriate regional office, or State, or local agency for
additional requirements, if any) summarize in tabular form the results of the CD tests and the RA
tests or alternate RA procedure as appropriate. Include all data sheets, calculations, and records
of CEMS response necessary to substantiate that the performance of the CEMS met the
performance specifications.
The CEMS measurements shall be reported to the agency in units of pg/m1 on a dry basis,
corrected to 68°F and 7 percent 02.
10. Alternative Procedures
A procedure for a total system calibration, when developed, will be acceptable as a
procedure for determining RA. Such a procedure will involve challenging the entire CEMS,
including the sampling interface, with a known metals concentration. This procedure will be
added as an appendix to this performance specification when it has been developed and approved.
The RA requirement of Section 4.2 will remain unchanged.
13. Bibliography
1. 40 CFR part 60, Appendix B, "Performance Specification 2 - Specifications and Test
Procedures for S02 and NO, Continuous Emission Monitoring Systems in Stationary Sources."
2. 40 CFR part 60, Appendix B, "Performance Specification 1 - Specification and Test
Procedures for Opacity Continuous Emission Monitoring Systems in Stationary Sources."
3. 40 CFR part 60, Appendix A, "Method 1 - Sample and Velocity Traverses for
Stationary Sources."
4. 40 CFR part 266, Appendix IX, Section 2, "Performance Specifications for Continuous
Emission Monitoring Systems."
5. Federal Register Vol. 58, No. 203 (10/22/93), proposed 40 CFR part 64, "Enhanced
13
177
-------
Monitoring Program."
6. Draft Method 29, "Determination of Metals Emissions from Stationary Sources,"
Docket A-90-45, Item II-B-12. and EMTIC CTM-012.WPF.
7. "Draft Enhanced Monitoring Reference Document," prepared for the Emission
Measurement Branch, Technical Support Division and the Stationary Source Compliance
Division, Office of Air Quality Planning and Standards, U.S. EPA, Contract No. 68-D2-0068
(9/30/93).
8. "Continuous Emission Monitoring Technology Survey for Incinerators, Boilers, and
Industrial Furnaces: Final Report for Metals CEM's," prepared for the Office of Solid Waste,
U.S. EPA, Contract No. 68-D2-0164 (4/25/94).
178
14
------- |