AT HAZARDOUS WASTE INCINERATORS
Prepared for:
Shiva N. Garg, P.E.
U.S. Environmental Protection Agency
Office of Solid Waste
Waste Treatment Branch (WH-565A)
Washington D.C. 20460
EPA Contract No. 68-02-3887
Assignment Nos. 45/64
Prepared by:
Jon N. Bolstad
iuSC! c Env1ronn«ental Services
11440 Isaac Newton Square, Suite 209
Reston, Virginia 22090
September 28, 1987
-------
CONTENTS
Tables . . . . . ii
Figures
1. Introduction and Summary
2. Theory and Operation of Carbon Monoxide Analyzers
Introduction
Measurement Principles
Flue Gas Sampling
Operating Characteristics
3. General Application to Hazardous Waste Incinerators
Introduction
Sampling/Analysis Location Considerations
Flue Gas Component Considerations
Consideration of Incinerator Design and Operating Conditions
Trial Burn Considerations
Monitoring Costs
4. Review of Monitoring Systems
Introduction .
System Design Review . .
System Performance Test
Trial Burn Plan
5. Review of Trial Burn Report
Introduction . . . . .
CO Monitor Performance Test Review
Evaluation of Trial Burn Results for Carbon Monoxide
Permit Conditions and Waste Feed Cut-off System
6. Operating and Maintenance Procedures
Introduction
Emission Monitoring System
WasteFeed Cut-off System . .
Data Reduction and Reporting
Eqi 1pment Replacement and Repair
Bibliography 60
Appendices
A. Manufacturers’ Reported Specifications for CO Analyzers
B. Instrument Manufacturers’ List
C. Example Monitoring Systems and Component Lists
0. Quality Assurance and PErformance Specifications for Con-
tinuous Monitoring of Carbon Monoxide at Hazardous Waste
Incinerators
7
7
15
18
21
21
25
29
31
31
36
36
45
47
48
48
51
53
54
54
57
57
58
• . 63
• 72
• 76
• . 79
I
-------
TABL ES
1. Summary of CO Monitoring System Performance Specifications . . . 3
2. Sununary of Measurement Characteristics 8
3. Summary of Reported Precision and Accuracy 19
4. Example Interference Effects 26
5. Typical Existing System Costs 33
6. Cross-stack - GFC with Backup Analyzer Costs 34
7. NDIR Dual Analyzer System with Backup Costs 35
8. Monitor Performance Test Completion Checklist 50
F IGURES
1. NDIR Luft-type Analyzer Schematic 10
2. GFC Analyzer Schematics . . . . . . . 12
3. Sampling System Schematic 16
4. Performance Review Checklist 44
1 1
-------
SECTION 1
INTRODUCTION AND SUMMARY
Because of increasing public concern over problems that had emerged
regarding the treatment, storage, and disposal of hazardous waste, the U.S.
Congress passed the Resource Conservation and Recovery Act (RCRA) of
1976. To implement this legislation and subsequent amendments to it, the
U.S. Environmental Protection Agency (EPA) has periodically issued regu-
lations and guidelines. One of the main areas in which EPA issued regu-
lations is for the permitting and operating of hazardous waste inciner-
ators. These regulations which are included in Title 40, Part 264 of the
Code of Federal Regulations (40 CFR 264) require that hazardous waste
incinerators continuously monitor emissions of carbon monoxide (CO) and
that permit limits for CO emissions be established to ensure the proper
operation of the incinerator. This document is designed to provide
guidance to applicants, and permit-writers and reviewers at the state and
federal levels to aid in obtaining and operating CO monitors in a manner
consistent with the Intent of the requirements in the regulations.
Specific regulatory requirements for CO levels and continuous emission
monitors (CEMs) include the following:
• A permit for a hazardous waste incinerator must specify the
operating limits for the CO level in stack exhaust gas and
allowable variations in those levels. The incinerator must be
equipped with a functioning system to automatically cut off the
waste feed if those limits are exceeded. (40 CFR 264.345)
• CO must be monitored continuously at a point downstream of the
cornbustion zone and prior to release to the atmosphere. (40 CFR
• 264. 347)
To assist 1nthe review of permit applications, the Office of Solid Waste
(OSW) I 1 sulng this guidance to enable the permit writers to ensure that
CO monitors on hazardous waste incinerators adequately represent the CO
emissions during the trial burn and continue to do so during subsequent
operation.
EPA is also considering certain maximum requirements for CO emissions
to augment the current regulatory standard regarding trial burn test
results. CO limits are currently based on levels achieved during the
trial burn. Maximum CO limits are considered as a simple, reliable means
of ensuring that Incinerators operate at high combustion efficiency over
1
-------
the life of the permit. In addition, limiting CO , and thus ensuring high
combustion efficiency Is an Important step to minimizing the emission of
“products of Incomplete combustion” (PICs). Detailed discussions of
the technical and policy aspects of these limits are contained in a sep-
arate report (Ref. 1).
Currently, the regulations do not specify any time-averaging period
for establishing the required CO permit limits and common practice has
been to set short-term limits (5 minutes or less) based on highest recorded
values during the trial burn. The required tests for destruction and
removal efficiency (DRE) are usually the result of several hours of testing
and thus short-term CO limits do not relate directly to the DRE tests.
This document provides technical guidance for two major purposes: (1)
to ensure that the required CO measurements are made within known limits
of measurement error, and; (2) that the systems are operated and maintained
so as to continue to operate properly. A separate document Is being
prepared describing oxygen monitoring. Appendix 0 describes the required
performance characteristics of the monitoring system needed to measure CO
levels in the range of the limits under consideration. Should these
limits change, appropriate changes will be required in the performance
specifications. Table I sumarizes the numerical standards for the moni-
toring systems; the text of Appendix D decribes how these parameters are
measured and calculated. The format of Appendix 0 is taken from the CO
CEM guideline prepared by the Office of Air Quality Planning and Standards
(OAQPS). The OAQPS CEM requirements for CO specific monitors were developed
for petroleum refinery catalytic crackers and although the requirements
are not all directly applicable to hazardous waste incinerators, the
general approach is.
Section 2 of this document discusses in detail the various types of
analyzers available, theory of their operation, and the operation of
ancillary equipment. CO monitors can be classified according to the
analytical methods currently available, as follows:
• Nondispersive infrared spectrometry - Correlation detection
(ND I R)
• Nondispersive infrared spectrometry - Gas-filter correlation
detection (GFC)
o Electro-analytical methods including catalytic oxidation (EAC)
and polarographic (EAP)
• Gas chromatography (GC)
Although there are other, more exotic analytical methods that can be used,
these are the most common in commercially available monitors. Appendices
A and B provide lists of the manufacturers’ specifications for specific
models and a list of instrument manufacturer’s addresses and phone numbers.
2
-------
TABLE 1. SUMMARY OF Co MONITORING SYSTEM
PERFORMANCE SPEC IF LCAT IONS
Parameter Sped fications
Low range High Range
Measurement Range, ppmv 0-200 max 0-1000 mm
System Response Time 1.5 minutes to 95%
Drift, Zero (as operating)* 10 ppmv, 20 ppmv,
(guide only) 24 hour 2 hour
Drift, Span* 20 ppinv, 50 ppmv,
24 hour 2 hour
Precision* the lesser of the lesser of
5.0% FS or 2.0% FS or
10 ppmv 20 ppmv
Linear lty* the lesser of the lesser of
10.0% FS or 5.0% FS or
20 ppinv 50 ppmv
* expressed as the sum of the mean absolute value plus the 95% con-
fidence interval of a series of measurements
3
-------
Other considerations discussed in Section 2 are the methods used to
obtaln4Iue gas samples for the CEMs including extractive sampling, cross-
stack analysis, and In-situ analysis. Extractive sampling which is the
method used by many, If not most, CEMs on hazardous waste incinerators
requires that a properly conditioned sample be extracted from the stack
or duct and transported to the measurement chamber of the Instrument.
Primary concerns include: the sampling location; the probe, sample lines,
and prime mover or pump used to extract the sample and transport it to
the measurement chamber; and sample conditioning. Finally, typical oper-
ating characteristics such as precision and accuracy, instrument sen-
sitivity, response time, and calibration are discussed in Section 2.
There are a number of special considerations that affect the design
and operation of CEMs on hazardous waste incinerators. These issues are
discussed In Section 3. The four major topics covered in Section 3 are
sampling/analysis location, interferences to the analytical method, design
and operating characteristics of the specific incinerator, and some con-
siderations related to the trial burn.
Although the RCRA regulations do not specify an exact location for
CEMs on hazardous waste incinerators, they are usually placed in one of
three locations, I.e., at or near the combustion zone, at or near the
point of discharge to the atmosphere, or somewhere in between those loca-
tions. Because the combustion zone is quite turbulent, hot, and dirty,
sampling in that location is quite difficult. Prior sampling to establish
the concentration profile during stable and unstable conditions is impor-
tant to determine whether a single or integrated multiple-point sample is
required. If the sample is to be extracted from the stack, EPA Performance
Specification 1 (40 CFR 60, Appendix B) provides some guidance for location
of sample sites. In addition, care must be taken to ensure that potential
in-leakage prior to the sampling port does not create stratification of
the sample gas stream, although the 02 correction provisions of the limits
being considered minimize the concerns about dilution of the stack gases.
Problems associated with intermediate points include ensuring the repre-
sentativeness of the sample, damage to the equipment from high temper-
atures, high dust concentrations, and acid gases, and physical constraints
related to access and operating environment.
For CO monitors that are based upon the absorbance of infrared energy,
any other gases present that absorb infrared energy at the same wavelengths
will interfere with the measurement. The most common interferents are
carbon dioxide and water vapor. Carbon dioxide interferes positively,
i.e., the measured CO concentration will be greater than the actual level
if carbon dioxide is present. Vapor-phase water also interferes posi-
tively. Techniques are available to correct for these specific-wavelength
interferences. Liquid-phase water and dust attenuate the entire spectrum
of the Infrared beam and these effects are usually handled by the signal
processing electronics. C,’oss-stack GFC analyzers are designed to pass a
reference and measurement beam alternately through the path to eliminate
the effects of droplet or particle interference. The sample conditioning
process used by extractive-sample NDIR instruments eliminates most of the
problems of interference caused by particles and liquid-phase water.
4
-------
Proper sample conditioning Is an important consideration in the design
of CEll systems. Sample conditioning systems are designed to remove or
mitigate-the Impact of temperature, moisture, and particles. In hazardous
waste Incinerators, acid gas removal may also be important. Although
high temperatures, I.e., above 250F, prevent the condensation of moisture
or acid gases, most Instruments require a cooler sample. The problem of
condensation is mitigated by including condensation and/or permeation
devices in the conditioning system. Particles cause the same attenuation
problems as water droplets and are filtered from the sample. Since most
of the particles created in hazardous waste incinerators burning process
sludges may be fine, i.e., <10 micrometers ( ) in diameter, more efficient
filters with a high pressure drop may be needed on CEM sample conditioning
systems.
Because the flue gas characteristics depend upon the design of the
incinerator Itself, consideration must be given to the unit’s characteris-
tics in the design of the CEM system. Hazardous waste incinerators can
be characterized as liquid injection, rotary kilns, and fluidized bed
combustors. In general, liquid injection incinerators have operating
conditions that produce fewer problems for CEMs than the other two types.
Because rotary kilns are used primarily to incinerate wastes with high
solids content, wastes that cannot be pumped or tanked, and wastes that
have solid residues, they usually have dusty flue gases that require
extensive dust removal. Fluidized bed incinerators also usually generate
flue gases with high dust loading. In general, hazardous waste inciner-
ators, and rotary kilns in particular, tend to operate at low CO concen-
trations, e.g., 20 ppm, with high spikes, e.g., >2000 ppm, multiple ana-
lyzers with different operating ranges or monitors with auto-range changing
are likely to be necessary.
Section 4 presents an approach by which the permit writer can review
a proposed monitoring system. This approach is presented in conjunction
with Appendix 0 which provides recommended quality assurance and perfor-
mance specifications for continuous monitoring of carbon monoxide at
hazardous waste incinerators. The review process Is divided into the
following elements:
• System design
• System performance test
• Trial burn plan
In the system design review, recommendations are provided for the fol-
1 owl ng:
• Sampling location
• Sample interface
• Analyzer performance review
• Data handling system review
This section also includes specific recommendations for reviewing the
procedures, test schedule, and data reduction and test report format for
the system performance test which Is required to ensure that the CEM system
5
-------
meets t specIflcatIons contained In Appendix D. Finally, monitoring
dur1n . . trIal burn and data reduction and reporting for the trial burn
are d *ed.
SeetTon 5 suggests procedures for reviewing the trial burn report
itself and translating these test results Into permit conditions. The CO
monitor performance test must be reviewed to ensure the completeness of
the data, to calculate or verify the quality assurance (QA) values, to
compare the QA values from the test with the specifications, and to iden-
tify remedial actions If needed. The CO limits under consideration make
interpretation of trial burn results less cumbersome, but the concen-
trations measured during the trial burn must be compared with 10 minute and
60 minute averages and QA data must be considered In the calculations of
the CO concentrations. The treatment of anomalous data must be consistent
with established practices. Oxygen correction procedures must be reviewed
to ensure that they are appropriate. Finally, the waste feed cut-off
system operation must be reviewed by evaluating the combustion control
logic and comparison of setpoints to standards.
The final section of this document, Section 6, focuses on operating
and maintenance procedures. Operation and maintenance of the CEM requires
calibration, audit, and preventive maintenance. The frequency of these
procedures are ‘as required’, daily, weekly, monthly, and annually or
semi-annually with increasing attention and effort as the frequency of
activity decreases. Procedures for the operation and maintenance of the
waste feed cut-off systems should include the control logic and testing
of the following:
• Analog-to-digital signal conversion precision and accuracy
• Digital signal Integration accuracy
• Mathematical processing and calculation algorithms (hardware or
software or both)
• Corresponding output (control) signals
Data reduction and reporting requirements are recommended. Finally,
equipment replacement and repair requirements are discussed. Specific
recome t1ons are impossible to make because of the diversity of possible
syste t general requirements are discussed.
S 4 appendices are included. Appendix D Is the recommended
standap determining the adequacy of the CEM; it includes performance
standards for the systems and describes the procedures required to demon-
strate compliance with the performance standards. Partial lists of instru-
ment vendors are given, along with the manufacturers reported instrument
performance in Appendices A and B. Appendix C shows some typical CEM
system schematics.
6
-------
TABLE 2. SUMMARY OF MEASUREMENT CHARACTERISTICS
Measurement Principle
Characteristic NDIR GFC EA(C,P) GC
cont.
cont.
cont.
discrete
Sample Type
extract
cross
stack &
extract.
In-situ
extract
Major Positive
Interferents
(unconditioned)
CO 2
H 2 0(v)
CO 2
H 2 0(v)
hydro-
carbons,
combust-
ibles
none
Analytical Range*
Low
High
0.5 ppm
50 S
10 ppm
100 %
20 ppm
20 V.
0.1 ppm
100 %
Other Analytes**
limited
yes(4)
no
many
Conditioning extensive none minimal moderate
Requirements dust, dust dust,
water, acid gas
acid gas
* Analytical range includes lowest detectable and highest measurable
concentatlon for the analyzer type; the range of a particular analyzer
will notbe as; great; cross stack instrument ranges are dependent on path
length, I.e., stack diameter.
** Other analytes Indicates the number of other parameters that can be
measured with a single instrument; a given type can often be built to
measure other analytes.
8
-------
Nondisgersive Infrared SDectrometrv (NDIR) - Conventional Detection
NDIR- Is the most common technique In use today for measuring Co in
gases, although GFC is gaining some acceptance for stack gas measurements.
The NDIR method takes advantage of the property of carbon monoxide to
absorb ER energy at distinctive wavelengths. A beam of IR energy Is
generated and passed through a gas sample. Any CO in the sample will
absorb ER at the characteristic wavelengths and, through appropriate
selection of filters and detectors, the degree of attenuation caused by
CO can be measured. Figure 1 is a simplified schematic of the major
components of an NOIR analyzer.
The source beam is chopped” with a motor-driven filter wheel which
passes the beam, optionally through narrow-bandpass filters, alternately
through the reference cell and the sample cell. The reference cell Is a
sealed unit filled with an Inert, CO-free, dry gas such as helium or
nitrogen. The sample cell is a flow-through unit and the sample gas is
constantly pushed or pulled through it at a known and fixed flow rate and
pressure. The sample cell is lined or coated with gold foil which is
highly reflective of ER radiation to minimize cell wall absorption. The
transmitted energy from each cell is directed to the detector section of
the analyzer where the differential energy between the beams is converted
to an electrical signal, amplified, and displayed on a meter of some type.
In the “Luft microphone (pictured In Fig. 1), a sealed chamber is
placed at the end of each cell and the two chambers are separated by a
flexible diaphragm. The energy difference heats one chamber more than the
other and the resultant pressure Increase deforms the diaphragm. This
diaphragm distortion Is translated into an electrical signal (condenser
microphone) and the resulting signal is amplified and processed (linear-
ized) to obtain an analog voltage corresponding to the amount of Co in
the sample cell. The voltage Is displayed on an analog voltmeter calibrated
to the concentration scale for which the instrument was designed. The
instruments are designed to optimize performance in the measurement range
of interest: cell geometry, chamber and microphone design, and detector
selection are combined to yield optimimum electrical output and inter-
ference rejection; signal processing electronics are built to provide a
linear output from the specific Instrument components.
The other detector types also employ differential measurement of the
transmitted energy. Photodetectors are photovoltaic cells In which the
incident radiation is directly converted to an electrical voltage. Dif-
ferential measurement Is obtained by alternately measuring the energy
transmitted through the reference and sample cells. Early versions were
not sensitive enough for environmental analysis, but advances in solid--
state electronics have permitted the use of photodetectors In source and
ambient monitors. Thermocouples and thermistors measure temperature dif-
ferences created by the transmitted energy differences, either directly
(thermocouples) or Indirectly (thermistors).
Like most devices with mechanical parts, NDIR analyzers are subject to
malfunction from vibration, mechanical misalignment, friction, etc. The
9
-------
Figure
Analyzer
NDIR Luft—type
LUFT—TYPE
DETECTOR
BANOPASS
F ILTERS
A
IR
SOURCE
a
BEAM COLLIMATORS
BEAM CHOPPER
1.
Schematic
-------
manufact*ers have solved, or at least minimized, problems related to
choppe btor failure, IR source spectral decay, detector chamber leaks,
responw jinearlty, and cell materials. The instruments are not main-
tenance WPQe, however, and a regular program of inspection and preventive
maintenance Is needed to maintain on-line reliability. Many of these
instruments were designed for use In coal-fired power plants; application
to HWIs has identified some additional difficulties. Acid gas (HC1)
corrosion of cell walls, presence of interferents, and instrument rugged-
ness are problems that users and suppliers must consider in a given appli-
cation.
NOIR - Gas Filter Correlation Detection (GFC )
Gas filter correlation analyzers also use the infrared absorption
properties of CO to quantify the amount present in a sample, but do so in
a manner different from conventional NOIR analyzers. In GFC analyzers, a
reference cell containing a sample of the pure (or a high concentration)
gas of interest is used to filter the source beam to remove the exact IR
absorption spectra of the analyte gas. A second cell containing an inert
gas, such as N 2 , Is used as the measurement beam. As both the measurement
beam and the reference beam pass through the sample gas, some of the dis-
advantages of separate sample and reference paths can be eliminated. The
effect of source intensity variations, spectral shifts, optical misalign-
ments, and certain Interferences can be minimized through the detailed
comparison of IR spectra afforded by the GFC instrument. A narrow-bandpass
filter Is placed In front of the detector so that the measurement is
taking place at the CO absorption peak (- 4.7 m). The usual detector is
a photocell and the photocell output is electronically synchronized with
the gas filter chopper to make the measurement comparisons. Simplified
schematics of GFC analyzers are shown in Figure 2.
One of the major advantages of GFC units is that the gas filter analyzer
permits automatic correction for source/detector aging. Cross-stack con-
tinuous emission monitors for CO are GFC analyzers, but there are also
extractive GFC units. GFC cross-stack units are either double-pass design
where both the source and detector are located on the same side of the
stack and the measurement beam is sent across the stack and returned by a
retro-reflector, or single-pass units which have the source and detector
located opposite sides of the stack. The various manufacturers have
used a i- P of approaches to dealing with such problem areas as alignment
shifts stack movement, maintaining clean optical windows, and compen-
sati— temperature changes and stack radiation effects, and
interr ‘ion mechanics and electronics.
Electro-analytical Methods
Catalytic Oxidation --
Catalytic oxidation sensors use the heat of combustion of carbon monox-
ide to quantify the amount of CO present In a sample. Several different
methods are used to convert the heat of combustion into a measurable
entity. The most common method Is measurement of the current change in a
Wheatstone bridge resulting from the temperature-driven resistance change
11
-------
COULE
SO1 cE,OETECToR
Figure
OO1LE -
RETROREFLECTOR
I- —
2
2. GFC Analyzer
Sch
enat ics
WJ A
IL1
Wi_ J7 :
GAS FILTER
!JHEEL
RETROREFLECT
AND HOUSING
12
-------
of a thermistor caused by oxidation of the combustibles. The Wheatstone
bridge consists of a combination of series and parallel electrical circuits
arranged so that electrical flow among the various paths can be balanced
and an Imbalance in one leg (as from a resistance change) can be measured
in another leg of the bridge circuit. Manually variable resistors are
used to set the zero and span of the instrument. Normally, two separate
catalyst beads are used in the system; one is used to compensate for
ambient temperature changes and is not exposed to sample gas, and the
other is the measuring bead.
As CO monitors, EA units are not sufficiently specific; any other
combustible compounds oxidized at the temperature/catalyst combination used
for the sensor yield a positive interference. As combustion optimization
monitors, this interference is actually a benefit in that unburned fuel is
also measured. However, the response factors for some of the materials
burned In HWIs Is low In comparison to fossil fuels so these units are less
effective as HWI combustion monitors than as boiler combustion monitors.
In recent years, the sensitivity of these units has been improved so that
the detection limit is as low as 100 ppm rather than 1000-2000 ppm found
In earlier instruments.
The units are usually placed on or near the stack, with an extractjye
probe having the sensors mounted at the end of the probe outside the stack.
The extractive probe systems are optionally combined with an 02 sensor
measuring the same gas as the combustibles/CO sensor. The in-stack sensors
do not have the capability for calibration with standard gases so instru-
ment zero and span are set electrically. Some of the extractive systems
have provisions for flowing calibration gases in the probe.
Polarographic --
The use of voltammetric (polarographic) analytical procedures has been
applied to CO measurement with some success. An electrochemjcal cell is
used with specifically chosen electrolytes to measure the product of
reaction of the pollutant of interest and the reagent. The reaction pro-
duces a current flow as the pollutant is oxidized or reduced in the cell
and this current flow is measured with a galvanometer. A semi-permeable
membrane Is used between the cell electrolyte and pollutant gas and the
system Is designed so that the rate of diffusion of the pollutant through
the membrane is solely dependent on the concentration of the pollutant.
Thus, the rate of reaction, and rate of reaction product reaching the
sensing electrode, are directly proportional to the analyte concentration
in the sample gas.
Each manufacturer claims sensitivity and interference-rejection cap-
abilities based on their particular cell design. Generally these units are
small, portable and designed as explosion or other warning monitors. They
are commonly used as toxic gas detectors for prevention of worker exposure
to unsafe concentrations of CO. More sophisticated versions have been
used for monitoring ambient concentrations of carbon monoxide where the
operating environment can be well controlled and the sampling system used
to minimize interferences. These ambient systems generally do not function
very well in an Industrial environment.
13
-------
Gas Chromatograohv
Gas chromatography provides precise and specific quantitative results
even though It Is not generally considered a continuous monitor. Chroma-
tography offers the potential for analyzing a sample for several different
analytes with little or no additional cost over a single analyte. The
process of chromatography Involves using a stationary medium which has
the ability to retard the passage of one or more chemical compounds rel-
ative to other compounds. In general, gas-liquid chromatography (often
referred to as gas chromatography) Is used for analysis of organic com-
pounds and gas-solid chromatography (GSC) is used for fixed-gas (SO 2 ,
Ca 2 , CO etc.) analysis. Both types have been employed for stack gas ana-
lysis. In gas-solid chromatography, (GSC) the column is filled with a solid
material, referred to as the support, that is coated with a viscous liquid,
known as the stationary phase. These columns are referred to as packed
columns and the combination of solid support and stationary phase is the
packing. In gas-solid chromatography, then, the solid is the packing. The
packing is chosen for the particular analysis to retard the passage of
the compound(s) of interest with respect to other compounds present in
the gas sample. Detection is accomplished by measuring some property on
the gas exiting the column and converting this property change to an
electrical signal.
A number of detector types are available and selection depends on the
required sensitivity, presence of interferences, nature of the analyte, and
concerns for the operating environment. Conron detectors for stack gas
analysis Include flame Ionization (FID), photoionization (PlO), thermal
conductivity (TCD), electron capture (ECD), flame photometric (FPD), and
Hall detectors. Each has some particular advantages and disadvantages and
selection involves weighing the advantages and disadvantages. CO is usually
measured with a TCO using GSC as the separation process. CO can’t be
directly quantified with an FID, but can be indirectly by reduction of
the CO to methane and measurement of the methane with an FID. This in-
direct method is most often used for measurement of total hydrocarbons
but could be adapted to CO-specific measurements. ECOs respond well to
metallo- and chlorinated organic compounds. FIDs, on the other hand,
respond to the C-H bond and thus very well to aliphatic and aromatic
hydrocarbons, but poorly to oxygenated (aldehyde and ketone) and chlor-
inated compounds. PIDs can be selected to exhibit a much greater sen-
sitivity to aromatic compounds than to aliphatic ones. Most detectors
used for GC analysis are non-specific; that is they respond to a variety
of compounds, although the response will vary depending the compound.
Chemical-specific detectors like FPDs are not generally required for GC
analysis as the separation and identification is performed by chromato-
graphic process. What Is required is a detector which responds to the
compound of interest and much of the choice centers around the detection
limits. A semi-specific detector, such as the P 10, can be used to simplify
the chromatographic separation process if there is an interferent which
Is difficult to separate from the analyte but which doesn’t generate a
significant detector response.
14
-------
GC analysis Is a semi-continuous process; GCs are riot continuous moni-
tors within the definition of the performance specification. Each analysis
is performed on a discrete parcel of sample gas and can be repeated as
soon as the analytical cycle time (usually several minutes) has elapsed.
The advantage of GC analysis is that several components can be measured
with a single instrument. Disadvantages are absence of continuous output
and expense. Typical laboratory GCs are not sufficiently rugged for an
industrial process monitor environment and industrial process GCs are
relatively expensive, especially compared to NDIR analyzers.
FLUE GAS SAMPLING
Regardless of the type of measurement system used, the sample being
measured must be representative of the flue gas composite. The end use of
the data Is a significant element of what constitutes a representative
sample. A sample taken at a given point may be representative for one
purpose but not for another. For example, a sample location downstream of
quench air addition would not be representative for combustion control
purposes but would be representative for correction of stack gas CO con-
centration to 7% 02.
Extract ive Samol ing
Sample Extraction and Transport - -
The sampling extraction and transport system consists of a probe, sample
lines, arid sample prime mover. The sample probe is the interface between
the stack gas and the sampling system and should be constructed of mater-
ials which will withstand exposure to the stack gas but not change or
react with the sample. This also applies to the rest of the sampling
system, but the probe is constantly exposed to the most severe service,
the rest of the system is exposed to less severe service as a result of
conditioning. Figure 3 displays a schematic of a complete system and
includes a number of components which may not be present in any particular
system. Some manufacturers, for example, combine particle filtration and
moisture removal In a coalescing filter.
Probe . The sample probe may either be a single or multiple inlet type
and may be equipped with a coarse particle filter. Combustion-zone or
hot-location probes often need to be cooled and air, steam, or water are
used as coolants. Stainless steel, Hastelloy’, Inconel’, and ceramic
materials are used for probe construction. For low temperatures and
small ducts, probes are sometimes constructed of Teflon’.
Samole Lines . The most common material for sample lines is Teflon’,
although stainless steel and, sometimes, polyolefin tubing are used suc-
cessfully. Depending on the instrument conditioning requirements and
environmental conditions, the sample lines may need to be heat-traced
with steam or electrical-resistance heaters. Sample lines are sized to
obtain the required flow rates with minimum pressure drop, while minimizing
residence time.
15
-------
STACK
PRO8E/
COARSE
FILTER
FiNE
FILTER
PERñEAT I ON
DRYER
Figure
3. Sanpl
Ing
Systen
Schenat
VES
SMPLE
PuMP
AIR-COOLED
CONDENSER / REHEATER
DRAiNS AEOIUl
FILTER
VENT
CALI8RATIO
GASES L
I
I
IC
-------
Prime Mover . Samples are extracted from the stack with pumps or with
ejectors or eductors, or both in combination. Each mode has advantages and
disadvantages for a particular application. Eductors and ejectors use air
or steam to Induce a vacuum by aspiration and are common industrial fluid
movers, simple (no moving parts), inexpensive to operate, and offer easy
flow regulation. Disadvantages include mixing of the sample and ejector
fluid at the unit discharge which limits the use to locations downstream
of any use of the undiluted sample gas. Pumps are relatively expensive
and prone to mechanical failure but maintain the integrity of the sample
gas and can easily be used in either pressure or vacuum service.
Sample Conditioning - -
The sample conditioning system is critical to proper operation of the
extractive analyzer monitoring system. Most analyzers require a particle-
and moisture-free gas for accurate and continued operation and the focus
for conditioning systems is on dust and water removal. In some cases,
other interfering components are removed, but general practice and sound
design is to use an analyzer which is not subject to or corrects for
other interferences. Sample gas temperature and pressure are also con-
trolled. The sample conditioning elements can be interspersed throughout
the sample train. Coarse particle (>20 jLm) filtration is usually done at
the inlet end of the sample probe, and often the filter is configured so
that it can be periodically back-flushed. It is desirable to have provi-
sions for Introducing calibration gases as near to the front of the sample
system as possible so that the entire system is evaluated, including the
conditioning system. Following the coarse filter are fine filters, acid
gas scrubbers, gas coolers and condensers, moisture droplet filters, and
gas dryers. Heaters and pressure regulators are usually the final stage
of gas conditioning before delivery to the analyzer.
Cross-Stack Sam l ing
The cross-stack analyzers, such as the GFC instruments described pre-
viously, avoid many of the problems of the extraction systems such as
sample line leaks, material corrosion, and conditioning system malfunctions
because the sample remains in the stack. However, calibration of the
entire system with standard gases is not practical (e.g., filling the
stack with calibration gases); interferences cannot be removed from the
sample gas so corrections or elimination must be done optically and/or
electronically; and Instrument repair and maintenance is much more dif-
ficult for an Instrument on the side of the stack than one in a control
room.
The sample Interface for cross-stack analyzers consists of optical
windows exposed to the stack gas environment and through which the IR
beam passes. The instruments are usually bolted to flanges located on
opposite sides of the stack wall. Double-pass instruments usually have
only a retroreflector on one side and the source/detector on the other;
single pass Instruments have the IR source module on one side and the
detector on the other with various control elements located in the source
and/or, detector module or a separate control module. Some form of keeping
the optical windows clean Is provided, usually by purging with clean air.
11
-------
In-situ Sampling
In-situ analyzers are those where the sensing/detection element is
placed directly In the gas passage and there are no true TM in-situ” ana-
lyzers for CO. Many of the EA units are referred to as in-situ because the
sensor is located on the stack and a small quantity of sample gas is
diverted into the probe, passed over the sensor, and discharged back into
the stack. Thus the sample has not been extracted in the conventional sense
although it has been conditioned to some extent. For the most part, these
analyzers for CO are improvements on the combustibles” analyzers used
for combustion control purposes, so they have been designed for use in
more severe environments than the emission monitors. At this time, there
are no EA analyzers meeting the recommended performance specification
although some units nearly comply. If a complying EA unit with an extrac-
tive probe were used, stratification in the sample gas can be tolerated
because the probe can be designed to sample gas from several points along
its length. These extractive probes are usually steam ejectors which
induce a sample gas flow over the sensor and discharge the steam/gas
mixture back into the duct downstream from-the probe.
OPERATING CHARACTERISTICS
Each of the analyzer types discussed in the previous section has its own
set of operating characteristics and some of these have already been
described. In this subsection, some quantitative comparisons among the
types are made. These comparisons are intended to provide the reader with
a basis to choose the best-suited analyzer for a particular application
rather than rate these instruments independent of their application. Many
of the parameters are inter-related in that an improvement in one area
may be to the detriment of performance in another area. For purposes of
comparing instruments, either generically (by type) or specifically (by
manufacturer and model), the data reported in the manufacturers literature
are the best available and are used in the following comparisons.
The major parameters of concern for monitoring CO from HWIs are sen-
sitivity, precision and accuracy, response time, and instrument drift.
Other factors of Interest include calibration requirements and procedures,
linearity, Interference rejection, and environmental requirements. The
source of the definitions and explanations in the succeeding subsection
is the Oualltv Assurance Handbook for Air Pollution Measurement Systems:
Volume I Principles , USEPA-EMSL, EPA-600/9-76-005, Dec. 1984, and will
hereafter be referred to as the QA Handbook.
Precision and Accuracy
Table 3 summarizes the general range of precision and accuracy for the
various analyzer types as reported by manufacturers. Detailed information
on specific analyzers can be found in Appendix A.
The data In the table would suggest that precision and accuracy are
not a problem for CO analyzers, especially when one considers that values
of ±10% are often used as a general rule for laboratory analyses. However,
18
-------
the values really represent the best obtainable performance in controlled
conditions. When a complete system Is operating in the real world, it is
much more difficult to achieve the kind of results shown in Table 3. The
performance criteria of Appendix D reflect very good, but achievable,
performance of a complete system. In general, Appendix D requires t5%.
TABLE 3. SUMMARY OF REPORTED PRECISION AND ACCURACY
(ANALYZERS ONLY)
Parameter
NDIR
GFC
EAC
EAP
Precision,
%FS
Lowest
.5%
1.0%
1.0%
1.0%
Highest
2.0%
1.0%
2.0%
.
nr
Typical
1.0%
1.0%
2.0%
Accuracy,
%FS
lowest
1.0%
2.0%
1.0%
nr
Highest
2.0%
3.0%
2.0%
nr
Typical
2.0%
2.5%
2.0%
nr
* nr - not reported
Sensitivity
In general, most of the analyzers reviewed offer minimum detection
limits (MDLs) of approximately 2 to 3% of the full scale measurement range.
Conversely, the full scale range is 30 to 50 times the MDL. These factors
become significant when both high range and low range values are of con-
cern, as they often are in the case of HWIs. If both high and low con-
centrations of CO (peaks and long-term averages of near-baseline levels)
then a single analyzer selected to measure one or the other accurately
will not likely measure the remaining one very accurately. The significance
of the operating range or quantification range will be discussed in a
later section.
19
-------
ResDonse Time
The NOIR analyzers have the slowest response time of the units described
here, while the GFC instruments have the quickest. The response time of a
gas chromatograph is nearly always much longer than the response of con-
ventional “continuous” analyzers. Most NDIRs (analyzer only) have response
times on the order of 30 seconds, while the GFC units are 1-5 seconds.
The EA units show response times in the range of 2-30 seconds, while GC
operating times are usually several minutes. In addition to the actual
instrument response time, the time for conveying the sample to the measure-
ment section must be included. For the cross-stack and in-situ units,
this time is effectively nil. For the GC and NDIR systems, which are
extractive and require a sample extraction/conditIoning system, this
time can be as much as several minutes, although the sampling systems are
generally designed to achieve residence times of less than 1-2 minutes.
Many of the faster-responding units have adjustable response times so
short-term spikes can be damped out. These systems are often designed for
combustion control purposes (setting fuel flow rates, modulating air
dampers, controlling fan speeds) so a very rapid response time can be
detrimental. The extractive systems have a built-in damping effect due to
the internal volume of the system. The use of the analytical results,
therefore, needs to be considered when evaluating the desirability of a
particular response time. Appendix D establishes a maximum permitted
response time for emission measurement purposes. If other uses of the
data are to be made (e.g. combustion trim control), then faster response
times may be necessary and cross-stack types are probably needed.
Cal ibrat ion
The extractive systems permit the use of standard calibration gases to
calibrate the entire monitoring system, while the cross-stack units must
be calibrated using surrogate calibration devices. Most extractive systems
(refer to Fig. 3) are, or can be, designed to flow calibration gases
through nearly all of the system, although the probe, and sometimes the
coarse filter, are not included in the calibration ioop. The recommended
procedures for system calibration are included as Appendix D to this
manual.
20
-------
SECTION 3
GENERAL APPLICATION TO HAZARDOUS WASTE INCINERATORS
I NTRODUCT ION
This section descrIbes the application of continuous CO analyzers to
hazardous waste incinerators and highlights possible problem areas and
their solutions. There are four major topics to be considered when design-
ing or evaluating a system for a HWI. These are sampling/analysis location,
presence of interferences to the analytical method, consideration of the
design and operating characteristics of the specific incinerator, and
some considerations related to the trial burn.
SAMPLING/ANALYSIS LOCATION CONSIDERATIONS
The RCRA regulations require that CO be monitored continuously, either
in the stack (264.345(b)(I)) or at a location downstream of the combustion
zone but prior to release to the atmosphere (264.347(a)(2)). The rules,
therefore, permit some latitude in selection of a monitoring location.
The choice of a monitoring location cannot be separated from the objectives
of the monitoring. CO measurements serve to indicate good combustion (CO
below proposed guidelines) and acceptable DRE (CO less than trial burn
values). The trial burn evaluation provides DRE results averaged over the
time period used for sample collection (usually 20 minutes or greater)
and the minimum time-averaging period In the proposed CO limits is 10
minutes. This suggests that the instantaneous CO concentration at a given
time is not particularly important except as it affects the time-averaged
value. Thus, It Is not necessary to measure the CO concentration either
spatially, at the end of the flame or combustion zone, or temporally,
within seconds of the actual occurrence. What is necessary is to measure
the conc ratlon at a time and place which provide reproducible measure-
ments. ! t1me factor for most HWIs Is not significant because the res-
idence of flue gases from time of generation in the combustion chamber
to dischrn’ge to the atmosphere rarely exceeds five seconds. Choice of a
physical location for a sampling probe or cross-stack path requires con-
sideration of dilution of the flue gas by non-combustion gases (e.g.
ambient air or baghouse cleaning air), flue gas stratification, presence
of potentially interfering components, and structural considerations.
General
The primary concern In selection of a sampling location is that there
be no stratification of gas-phase pollutants, i.e., that the concentration
21
-------
be uniform across the stack or duct at the point (or points) of sampling.
Generally, this means that the sample point should be as far downstream
from any potential in-leakage or reaction as possible. Fortunately, finding
a well-mixed location is not as difficult as finding a suitable location
to sample particulate matter; gaseous pollutants are not subject to dif-
ferentiation by inertial forces as are particles. The criteria contained
in EPA Reference Method 1 (40 CFR, Title 60, Appendix A) provide a basis
for selecting sampling locations. These specifications are intended for
selecting representative velocity locations. However, they are often used
for selecting monitor locations without consideration of the intent. The
minimum duct diameter criteria of Method 1 suggests that the sample point
be at least 8 equivalent duct diameters downstream of any disturbance.
The guidelines for siting continuous emission monitors (EPA-450/2-82-026)
are different from Method 1 and suggest that monitor locations be a minimum
of 2 equivalent duct diameters downstream from any disturbance. If the 8-
diameter criteria of Method 1 can be obtained without much difficulty,
additional assurance of homogeneity is obtained over a lesser value, but
it is generally not a significant improvement. The 2 diameters down-
stream/one-half diameter upstream criteria will generally produce accept-
able results.
More important to obtaining representative CO measurements is the
homogeneity of the sample obtained in the absence of in-leakage of ambient
air or other diluent gas and absence of chemical reactions (like combus-
tion) at the sample point. For these purposes, a sample location in the
discharge stack is usually acceptable. While the gas velocity profile
might not be acceptable for particle sampling because of flow irregular-
ities, it is usually acceptable for gaseous pollutant sampling. Potentially
stratified locations should be avoided but If this is not feasible, then
a stratification check should be performed. A stratification check can be
done measuring the CO? and 02 concentration at several points across the
duct cross section using an Orsat analyzer, or using a portable Continuous
analyzer for CO, C0 2 , or 02. In some cases, flow-straighteners are in-
stalled to eliminate problems with velocity profiles for manual sampling
(HC1, particulate matter) and these will actually create stratification by
preventing mixing of the flue gases. If flow directors are required to
obtain valid particle measurements, the location is not recommended as a
CEM location and alternatives should be sought. If no alternative is
available, stratification checks should be done with and without flow
straighteners and judgement exercised regarding whether the flow modifiers
have to remain or are removed after the ORE tests. The best solution to
this problem is likely to be a multiple-point sampling probe.
There may be a number of reasons to monitor at a location other than
the stack discharge as there are some disadvantages to locating a monitor
in the stack. Alternative locations may be desired becaus.e of multiple
unit discharges through a common stack, desire to monitor for combustion
control purposes, difficult access to sample probes and/or conditioning
systems, or to minimize sample line lengths. Whatever the reasons, other
locations are acceptable as long as the criteria regarding homogeneity
are met. If the location is under pressure, then any leakage at the sample
point would be out of the system and dilution would not be a factor. If,
22
-------
however, the duct were under vacuum (e.g. after the air pollution control
device and before the ID fan), then in-leakage could be problematical.
The CO limits under consideration require correction of the measured CO
concentrations to a 7% 02 dry basis so in-leakage is not important as
regards compliance with the limits but it is important as to whether or
not it creates a non-homogeneous gas stream.
Combustion Zone
The combustion zone is not a particularly good sampling location for a
continuous monitor for several reasons. The combustion zone is quite
turbulent and obtaining a representative sample of the flue gas stream is
very difficult. The high temperatures require special and expensive probe
systems to withstand the thermal and chemical stresses of the environment.
The high particle concentration places great demands on the particle
filtration system, and the probe system can alter the chemical composition
of the localized gas stream by chilling it or serving as a reaction site.
Perhaps the most salient disadvantage to combustion zone monitoring relates
to instrument availability.
Equipment exposed to conditions as severe as those in the combustion
zone have a tendency to fail often. The RCRA regulations mandate no incin-
eration of hazardous waste unless the monitors are operational. Thus, any
monitor downtime shuts off waste feed. This can, and often does, create
an upset to an otherwise steady-state and acceptable combustion process
because of instrument failure. The net result is worsened combustion
from stopping and starting waste feed and results in higher emissions of
unburned or partially burned material.
If it is necessary to monitor CO in the firebox, the location should
be selected carefully. The location should be as far from the flame front
as possible to minimize damage to the probe and reactions with the flue
gas. It is necessary to define the concentration profile at the sampling
point by measuring CO concentrations at several points in a plane normal
to gas flow. The equal-area method of Ref. Meth. I is recommended for
locating sample points. The disturbance criteria will almost certainly be
violated by the nature of gas passage design, so the number of sample
points should be as many as possible. Conventional profile sampling is
conducted with at least 12 sample points for ducts of the size commonly
encountered In HWIs. This number represents a balance between the physical
difficulties of sampling and obtaining a statistically sound sample. The
sampling should be repeated at least three times during each of three
conditions: (1) stable operation at high load; (2) unstable operation at
high load; and (3) unstable operation at low load. These data will permit
selection of an appropriate point for sampling. If the concentration
profile remains constant, a single point at the location of the average
concentration can be used. A constant profile means that the concentration
at a given point relative to a reference point remains constant even
though the actual concentration may vary. If the profile changes with
load or conditions, an integrated multiple-point sample is recommended.
Otherwise, the measured CO value will be representative during one con•
dition but not during another.
23
-------
Exhaust Stack
In most cases, a stack location is acceptable and desirable for ob-
taining the needed CO data. The flue gases have been partially conditioned
by the air pollution control devices (APCDs), the gas stream is reasonably
homogenous, and the physical situation is generally amenable to install-
ation of monitors. The time delay between occurrence of change in con-
centration and its appearance at a stack location is acceptable.
The sampling point location criteria contained in EPA Ref. Meth. I
provide the basis for selecting a sample site, with the following caveats.
Method 1 is intended for obtaining representative samples during multiple-
point isokinetic manual sampling for solid and liquid particles and ad-
dresses the velocity profile of the gas stream with regard to size dif-
ferentiation of the entrained particles. Gaseous pollutants are not subject
to this inertial separation so the criteria are not directly applicable.
Method 1 requires that a sampling location be at least 2 equivalent duct
diameters downstream of any flow disturbance. At any location closer than
2 diameters, the flow is presumed to be too turbulent to obtain accurate
velocity readings. This turbulence, however, works to the advantage of gas
monitoring by ensuring a well-mixed gas stream. The condition to be avoided
is one where there is the possibility of diluent gas in-leakage into a
non-turbulent gas stream. This condition would likely create stratifica-
tion. One need not be concerned about dilution of the stack gas; the
proposed CO limits require correction of measured values to a standard
oxygen basis (7% 02). However it is good practice to locate monitor sites
as far from any possible leaks as possible, especially for negative pres-
sure locations. If a candidate location is under vacuum (upstream of a fan
inlet), then care must be exercised to ensure that the gas stream at the
location is well mixed so that any leakage will be distributed over the
entire gas stream. In these cases, the sample site should be as far as
possible from the source of potential in-leakage; certainly not less than
the 2-diameter criteria of Method 1. If there is a question about the
degree of mixing or presence of stratification, then a sample traverse
should be performed. In positive-pressure systems (downstream of a fan),
any leakage will be out to the atmosphere and will not affect the accuracy
of the monitored concentration.
The procedure discussed above concerning combustion zone monitoring
can be applied for any stratification check. The load considerations will
probably be less significant than for a combustion zone location, but
should not be overlooked. For the most part, duct locations will exhibit
a constant degree of stratification at varying loads because the bulk
flow of gas is created by mechanical means, and phenomena such as on-going
reactions are not a factor. At least twelve sample points are usually
necessary to identify stratified gas streams. If stratification is found,
and the location is the only practical one, additional sampling will be
necessary to quantify the degree and nature of the problem. If at all
possible, however, stratified locations should not be used. For purposes
of CO monitoring, stratification Is considered present if a single point
shows a concentration different from the average value by more than 15%
24
-------
of the average value or if two values are different from each other by
more than 20% of the average value.
Locations between Stack and Combustion Zone
There may be valid reasons for monitoring CO at locations other than
those discussed in the previous sections. For example, the operator of a
facility with a caustic wet scrubber for acid gas removal might prefer to
measure CO 2 upstream of the scrubber to avoid the problems of CO 2 absorp-
tion. Carbon monoxide absorption by scrubber media is not a probTem because
CO is practically insoluble. It would probably be cost-effective to monitor
CO at the same location. Generally, the types of locations can be clas-
sified as pre-APCD, mid-APCD, and post-APCD but before the stack. Each
type has a particular set of circumstances to be considered, but the
factors discussed above remain valid. Basically, the closer to the combus-
tion zone the sampling location is, the more severe the sampling conditions
will be and the more conditioning that will be required. As the gas stream
receives treatment, the sample conditioning requirements are lessened.
The most common problem encountered in monitoring these intermediate
locations is finding a representative location. The incineration facilities
are designed to occupy the least amount of space, use the minimum amount
of materials of construction and to have the minimum amount of ductwork,
consistent with such other criteria as temperature stresses, and fabri-
cation and maintenance requirements. Other common problems are high
temperatures, high dust concentrations, acid gases (primarily SO 2 and
HC1), and physical constraints factors of access and operating environment.
FLUE GAS COMPONENT CONSIDERATIONS
Interferents
For those CO analyzers that use infrared absorption as the method of
detection, any flue gas component which absorbs JR in the same wavelength
region as CO is a potential interferent. The most common of these are CO 2
and water vapor. Carbon dioxide and water vapor coincidentally have molec-
ular bonds which absorb JR energy in the same region as CO. While the
absorption peaks are not identical to those for CO, they are similar and
close enough to interfere unless measures are taken to eliminate the
interference. Most manufacturers have gone to great lengths to minimize
the effect of these interferents and have been fairly successful through
judicious use of precision optical filters, sophisticated electronic
signal processing, and other techniques. These interferences, if present,
yield a positive result even if no CO is present. For the purposes of
RCRA monitoring, these interferences yield conservative results because
they indicate the presence of CO even if there is none. If the interference
level is too great, however, real changes in the CO concentration can go
undetected, or at least unnoticed.
Table 4 below demonstrates the effect of various CO and CO 2 concen-
trations with analyzers with different CO 2 rejection ratios. A 5,000:1
CO 2 rejection ratio means 5,000 ppm of CO 2 will be measured as 1 ppm.
25
-------
TABLE 4. EXAMPLE INTERFERENCE EFFECTS
Case I - CO 2 Rejection Ratio - 5,000:1
Condition 1 Condition 2
( a) (b) (a) (bL
True Values
CO 2 10% 10% 5% 5%
CO 150 ppm 0 150 ppm 0
Measured Value 170 ppm 20 ppm 160 ppm 10 ppm
Case 2 - CO 2 Rejection Ratio - 1,000:1
Condition 1 Condition 2
( a) (b) (a) (bL
True Values
CO 2 10% 10% 5% 5%
CO 150 ppm 0 150 ppm 0
Measured Value 250 ppm 100 ppm 200 ppm 50 ppm
A CO 2 rejection ratio of 5,000:1 is fairly typical; there are instruments
with CO 2 ratios as low as 1,000:1 and lower. When selecting an instrument,
the actual (or estimated) flue gas conditions and the acceptable measure-
ment error will determine what capability is required.
Water vapor also Interferes with NDIR analysis. Like CO 2 and CO, water
vapor absorbs IR In the same region by molecular absorption. Manufacturers
provide water vapor rejection ratios In their specifications. Liquid-
state water, or droplets, also interferes with the measurement but the
nature the Interference is different than vapor-phase interference.
This drqplet Interference is the same as that caused by dust particles,
i.e., scattering of the transmitted beam from reflection and refraction by
the particles, and generally affects the entire spectrum of the beam. The
problem of droplet or particle Interference is handled by both cross-
stack and extractive analyzers. The cross-stack units use the sample path
for both reference and measurement beams and eliminate the effects of
intensity attenuation by electronic comparisons of the spectra. The con-
ditioning system for extractive systems removes the droplets and particles
prior to analysis. The conditioning system also removes virtually all of
the vapor-phase water, as well.
26
-------
Measurement Basis
Another consideration related to the flue gas components is the basis
of the CO concentration measurements. The limits now being considered
identify the components of a standard basis as 7% 2 and dry. The pre-
ceeding discussion addressed water as an interferent and pointed out that
cross-stack and extractive systems handle the problem differently. By the
nature of the measurement systems, cross-stack analyses yield a wet-basis
measurement and extractive systems yield a dry-basis measurement. Most
emission standards expressed as concentrations (ppm, %, gr/dscf) are dry-
basis standards and the CO limits currently being considered will likely
be expressed as dry-basis. The procedures for correction to a standard
basis are discussed in Section 4 under data handling system review. A
separate manual discusses oxygen monitoring. The data generated by all
the analyzers needs to be corrected to 7% 2; the data generated by cross-
stack systems needs to be further corrected for stack gas moisture content.
Measurement of Other Pollutants
Carbon monoxide is the only atmospheric emission that the RCRA regula-
tions specifically identify for continuous monitoring. The rules do require
monitoring of several process parameters on a continuous basis, but these
requirements only peripherally affect monitoring for CO. There are other
pollutants monitored at HWIs for a variety of reasons. The limits under
consideration require oxygen monitoring to correct the measured CO values
to 7% 02. Some state agencies and some EPA regional offices require moni-
toring for one or more emitted pollutants, including sulfur dioxide,
nitrogen oxides, total hydrocarbons (THC or total unburned hydrocarbons,
TUHC), and opacity, to verify compliance with air quality regulations.
Some permittees are required to monitor some combination of C0 2 , 02, THC
or TUHC, opacity, or stack gas flow rate. The effects of these requirements
should be considered in the selection of the analyzer type and in the
sampling location.
If THC is to be measured as a combustion indicator, an extractive
system will probably be used because the only practical method of analysis
is with an FID analyzer which requires that the sample be pumped to the
analyzer. In this case, it is likely that CO will also be monitored at
the same location and from the same sample line. In this example, the CO
analyzer should be placed ahead of the THC analyzer or in a split stream
from the main sample line because the FID analyzer uses a destructive
detector which would tend to oxidize any CO in the sample gas to CO 2 and
thus give erroneous CO values. If several parameters are to be monitored,
the different conditioning requirements and analytical methods of different
analyzers may conflict and the system should be designed to avoid these
conflicts or minimize any bias that might result.
Samole Conditioning
Sample conditioning systems are designed to remove or minimize the
negative effects of three major elements of the sample gas stream: temper-
ature, moisture, and particles. Removal of interfering species is also
- 27
-------
done in the conditioning system if the analyzers internal interference-
handling capabilities are inadequate but this can usually be avoided.
Some stack gases will also contain significant quantities of acid gases
(HC1, SO 2 ) and these will be removed to some extent as a result of moisture
and temperature control.
Any sampling extraction location requires particle filtration. There
are two aspects of particle filtration; removing the particles from the
sampled gas stream and maintaining the filtration system in an effective
operating condition. Locations upstream of the particle removal devices
have a much higher particle loading than a stack location so particle
removal is more problematic. Filtering the particles from the gas stream
is not technically difficult but maintaining a reasonable pressure drop
can be difficult. Systems are designed with prefilters, coarse filters
and fine filters using porous media which can either be cleaned by back-
purging or removed and thrown away. Some sophisticated systems include
multiple filtration loops and are designed to automatically change fil-
tration loops when the pressure drop in the system exceeds a setpoint.
The particle size distribution is an important factor in selecting fil-
tration media. HWIs disposing of waste solvents and the like tend to
generate small particles (< 10pm) while HWIs disposing of process sludges
can generate very small or very large particles (> 50 pm in diameter)
depending on the process. Large particles are much easier to remove from
the gas stream than small ones but a preponderance of large particles is
generally associated with a high dust load so filters tend to plug more
readily. Fine particle removal requires a more efficient filter; usually
at the expense of high pressure drops.
High temperatures can be a benefit if the analyzer can accept a hot
gas in the measurement chamber because the sample gas can be kept above
the water and acid dew points. In the context of an analyzer, temperatures
above about 150°F are considered high. A sample gas temperature of >220°F
is required to prevent water vapor condensation; prevention of sulfuric
acid condensation can require temperatures as high as 320°F. If the ana-
lyzer and sampling system can withstand these temperatures, sample line
heaters can be used. More often, though, cooling of the gas and moisture
and acid gas removal are desired and are accomplished with controlled
condensation and/or permeation devices.
Moisture removal is required for several reasons. Very few extractive
analyzers operate very well for very long with condensation in the measure-
ment section. Some units exhibit an interference to water in the vapor
state. Further, the presence of acid gases (condensible or not) with free
moisture creates severe corrosion problems. Most extractive systems contain
water removal systems. The dryness of the sample gas after conditioning is
expressed by the dew point and most systems are designed to achieve a
dew point of 30-50F° below the lowest expected operating temperature of
the analyzer. Figure 3 (Section 2) shows a sampling system with all of
the major components of a conditioning system. A typical system will not
include all elements, nor are all elements needed; the schematic is iflus-
trative of the types and locations of devices involved.
28
-------
CONSIDERATION OF INCINERATOR DESIGN AND OPERATING CONDITIONS
Unit Tvoe
The type of combustion source has some bearing on the design and oper-
ation of the monitoring system. The flue gas characteristics are defined
by the combustion type. The three major types of combustion chambers for
hazardous waste incineration are the liquid injection (LI), the rotary
kiln (RKI), and the fluidized bed combustor (FBC). Many units are built
with multiple combustion chambers and include more than one type.
Liquid Injection --
In general, the liquid injection units require less special consider-
ation than the other types. Comparatively, i x units operate with a more
homogeneous feed, at more stable conditions, lower flue gas dust loadings,
and often with lower temperatures because of waste heat recuperation. ix
units are designed to operate with lower excess air and potentially higher
CO than the other types. In terms of the time rate of change of CO con-
centrations, LI units change more rapidly than fluidized bed units but
less rapidly than RKIs. However, LI incinerators respond to intentional
process changes more rapidly than do the others, which is significant if
the CO monitor is being used for either trim or gross combustion control.
If CO concentrations are used for fine adjustment of operating parameters
(trim control), a rapid response time (< 15 sec) is required; if CO levels
are used for gross control, rapid response is not necessary and may even
be detrimental.
Rotary Kiln --
RKIs are usually used to destroy wastes which cannot be pumped, which
contain high concentrations of solids (percent level rather than ppm
level), where the solid residue has economic value (like plating sludges),
or when the waste cannot be tanked for one reason or another. RKIs are
usually equipped with an afterburner section to burn the volatile or
destructively distilled waste components. RKIs are subject to rapidly
occurring changes in CO, primarily in the form of upscale spikes. The Co
monitoring system needs a rapid response time (on the order of 10-15
seconds) if these peaks are to be accurately measured. Most RKIs are
designed to operate with low CO concentrations (high excess air) so the
difference between the average and peak concentrations can be great (20
ppm vs. 2000+ ppm). Since most analyzers do not have this great operating
range, multiple analyzers or at least auto-range changing is necessary to
measure both peaks and averages accurately. RKIs also generate dusty flue
gases so sample locations downstream of particle removal equipment are
preferred.
Fluidjzed Bed Combustors --
For the most part, FBCs are the most stable of the combustion chamber
types. FBCs are used for process sludges which are too “sticky” for RKIs
or for solids with low fusion temperatures. The most significant concern
relative to CO monitoring systems is the extremely high dust loading at
sample locations ahead of the dust removal equipment. The fluidizing medium
is sand or other refractory granular solid and the off-gas from the fluid-
29
-------
izing chamber entralns a substantial portion of the bed material. An
integral cyclone or other particle collector is used to recapture entrained
bed material and return it to the fluidized section. Quench air is often
added to the combustion products to minimize the thermal stresses on the
downstream components of the system. Even with °2 correction, large enough
volumes of quench air are added such that the stack gas concentration of
CO after quench addition is at the low end of the analyzer measurement
range and random variations can often conceal combustion-related changes
in the CO level. This is one of the cases where the stack may not be the
best monitoring location; the disadvantages of high temperature and high
particle loading may be offset by the advantages of obtaining CO data
representative of combustion and useable for process control. If a combu-
stion zone sampling location is being considered, the effect of dilution
or quench air on peak and average CO concentrations should be calculated.
If the dilutions would mask the changes in CO concentration needed for
combustion control, the stack location would not be the best choice.
Waste Characteristics
Generally, the nature of the waste to be burned does not have much
direct effect on CO monitoring as the effects appear in the combustion
design (type of unit, excess air, etc.). The major consideration for the
sampling system is the presence of waste components giving rise to acid
gases, particularly sulfur and chlorine. The amount of HC1 and S02 in the
flue gas needs to be considered when specifying the sample conditioning
system and the materials of construction for sample interface components,
and the concerns are much greater for sampling locations before acid gas
removal components of the APCD system. Samples extracted from the stack
generally need minimal acid gas conditioning because most HWIs are built
with acid gas removal APCD systems.
Waste Feed Cut-off Requirements
The purpose of the CO monitor is to discontinue waste feed to the
incinerator when some preestablished condition is reached. The monitoring
system and waste feed cut-off system interact and these interactions need
to be considered. Sections 4 and 5 discuss in detail the specific operation
of the waste feed cut-off relative to the CO monitor and DRE requirements,
but the technical issues related to integration of the two systems are
presented here.
When matching the cut-off system hardware and operation to the CO
system, several factors should be evaluated. The time-averaging aspects of
the CO analyzer/data processing system should be considered relative to
issuance of the signal to the cut-off system. Cut-off system parameters
to be considered include: response time (or lag time) of the cut-off
controller; bandwidth (signal level, overshoot, etc.); and whether the cut-
off controller is a proportional type (like a metering valve or pump
speed controller for combustion trim control) or an on/off type (like a
pump switch or shut-off valve). Other considerations include the method
of handling redundant analyzers and switching CO analyzers without gener-
3Ô
-------
ating a cut-off signal, procedures for calibrating analyzers without
generating a cut-off signal, and starting up a system on non-hazardous
fuel and dealing with the high CO values associated with a startup.
TRIAL BURN CONSIDERATIONS
The monitoring system used for the trial burn should be the same system
used for subsequent operation and for determining compliance with permit
conditions, Insofar as possible. In certain cases (new units particularly),
the trial burn may reveal problems with the CEM system and necessitate
modifications to the system. Generally, the shakedown period allowed by
the regulations will provide ample opportunity to remedy these problems
before the trial burn. In the event that modifications are necessary, the
modified system should be subjected to the same type of QA tests as would
a permanent system, with the exception of the 24 hour zero and span drift
test. A temporary system should be calibrated before and after each day’s
testing and zero and span checks conducted between test runs. Response
time tests need only be conducted once unless the system is altered during
the trial burn testing.
If possible, the monitor performance tests recommended in Appendix D
should be conducted before the trial burn. If the system meets the speci-
fications, the CO analyzer data will be directly comparable to the ORE test
data and to subsequent operating data and can be used to narrow the scope
of trial burn testing and target the DRE evaluation to those conditions
of most concern. The trial burn conditions should be checked to ensure
that the monitor can adequately detect the expected or desired responses.
MONITORING COSTS
Costs of monitoring CO and 02 have been estimated for three different
types of system. These three systems are: 1) a typical existing system;
2) a GFC cross-stack system with backup analyzer and; 3) an extractive
system with dual on-line NDIR analyzers with a backup unit. The cost data
include purchase and installation of the capital equipment and for con-
ducting the performance tests required by Appendix D. Tables 5, 6, and 7
summarize the cost elements of these systems along with “ball park”
values for the typical performance characteristics of the system. System
1 represents a coninon configuration in use at HWIs today and generally
would n t meet the performance specification. Systems 2 and 3 are different
configurations of systems which would meet the performance specifications.
The tot i Installed cost of these two systems are not significantly dif-
ferent from each other; the more expensive cross-stack GFC analyzer does
not require the sampling system. If an operator elected to comply with
the not-to-be-exceeded standards, the sophisticated data processing system
could be replaced with a simpler system costing $1O-12,000 each. The
back-up analyzer could also be eliminated and some savings in as install-
ation realized by the simpler data systems, but the costs associated with
performance testing would remain. The cost for a minimum system which
meets the performance specifications would be approximately $42,000 for a
cross-stack GFC system and $63,100 for an extractive NDIR system. These
data are based on the assumption that 2 separate extractive NDIR analyzers
31
-------
are required to meet both the high- and 1ow-range specifications. If
manufacturers can supply multiple-range analyzers which meet the specifi-
cations some cost savings may be effected.
32
-------
TABLE 5. TYPICAL EXISTING SYSTEM COSTS
System Description:
Single range analyzer with full-scale of 1000 ppm; no
continuous oxygen monitor; simple single-valued limit-
type control output signal (yes/no above setpoint)
System Performance:
Range
Zero/Span Drift, 24 hour
Minimum Detection
Precision
Accuracy
Response Time
Typical System Components:
Air or steam-educted probe with integral filter, sample
conditioner with moisture and fine particle removal, NDIR
analyzer, manual zero and calibration, “off-the-shelf”
data logger/recorder with capability of averages and
excursion calculations, strip chart backup.
Cost Estimates:
Sampling System $3,400
Analyzer $ 6,600
Data Recording/Processing $ 2,700
System Installation S 3,800
Performance Testing $ 6,400
Total Installed Cost $22,900
0-1000 ppm
± 25 ppm
± 25 ppm
± 2% FS
± 5% FS
± 2 mm.
33
-------
TABLE 6. CROSS-STACK - GFC WITH BACKUP ANALYZER COSTS
System Description:
Dual range analyzer; continuous oxygen monitor; simple
single-valued limit-type control output signal (yes/no
above setpoint) but microprocessor-based (PC-type) data
system, continuous correction to 7% 02
Performance Specifications:
Range 0-200, 1000 ppm
Zero/Span Drift, 24 hour ± 2.5% FS
Minimum Detection ± 2.5% FS
Precision ± 2.0% FS
Accuracy ± 5.0% FS
Response Time ± 30 sec.
Typical System Components:
Dual channel GFC cross-stack analyzer with dual range
calibration cells & oxygen measurement, backup GFC analy-
zer, microprocessor analyzer control, PC type data acqui-
sition system for historical trend analysis and propor-
tional process control.
Cost Estimates:
Sampling System (not req’d)
Analyzer $39,700
Data Recording/Processing $15,000
System Installation $13,600
Performance Testing $11,500
Total Installed Cost $79,800
34
-------
TABLE 7. NDIR DUAL ANALYZER SYSTEM WITH BACKUP ANALYZER COSTS
System Description:
Dual analyzers; continuous oxygen monitor; control output
signal and data processing by micro-computer (PC type)
System Performance:
Range
Zero/Span Drift, 24 hour ±
Minimum Detection ±
Precision ±
Accuracy ±
Response Time ±
Typical System Components:
Extractive system with back-purged in-stack filter, heated
sampling line, mechanical pump, sample conditioner, auto-
matic cal gas system, continuous 02 monitor and two sepa-
rate NDIR analyzers and a multi-range backup unit, PC
type data acquisition for historical trend analysis and
proportional process control.
Cost Estimates:
Sampling System
Analyzer
Data Recording/Processing
System Installation
Performance Testing
Total Installed Cost $ 85,910
0-200, 1000 ppm
2.5% FS
25% FS
2.0% FS
5.0% FS
2 mm.
$ 7,800
$34, 760
$15,000
$15, 150
$13,200
35
-------
SECTION 4
REVIEW OF MONITORING SYSTEMS
INTRODUCTION
The RCRA regulations specify that the Part B application contain a
trial burn plan which includes “. .a detailed engineering description of
the incinerator, including: ... stack gas monitoring and pollution control
monitoring system” (270.62(b)(2)(ii)(G)) , and “ a detailed description of
sampling and monitoring procedures, including sampling and monitoring
locations in the system, the equipment to be used, sampling and monitoring
frequency “ (210.62(b)(2)(iii). This section provides some guidelines
for reviewing (and preparing) the CO monitoring aspects of a Part B permit
application.
The review of the CO CEMS is a three-step process: (1) evaluate the
system design and operating specifications; (2) conduct the performance
test to verify that the “as-built” system performs as intended, and; (3)
establish permit conditions from trial burn CO data. For interim-status
facilities with operational CO CEMS, there may be advantages to conducting
the performance tests without spending a great deal of effort reviewing
the system design if a cursory review suggests that the system will likely
meet the specifications. Conversely, thorough design review of a new
system can avoid or at least minimize future difficulties with the accept-
ability of a system.
SYSTEM DESIGN REVIEW
Review of a proposed CO monitoring system consists mainly of comparing
design specifications with the performance specifications of Appendix 0
to this manual.
Sampling Location
Section 3 described the general considerations for a suitable monitor
location. The primary objectives are that the sample location point(s) or
path represent the entirety of a flue gas stream and that the flue gas
stream represents the combustion status of the incinerator. The stack is
generally the best place to locate the sample interface, whether it is an
extraction probe or a cross-stack installation. Problems of conditioning,
stratification, and inleakage are minimized, access is usually acceptable,
and the results reflect what is discharged to the atmosphere. If the
36
-------
proposed location is on the discharge stack or duct leading to the stack,
the review needs only consist of verifying a few items:
• Is the location within the distance-from-disturbance criteria?
• Does the sample interface include the central 50% of the cross-
sectional area (path for cross-stack units, points for extractive)?
• Are all potential sources of inleakage more than 8 duct diameters
upstream of the location?
If the answers are yes, then the site is acceptable. If not then additional
review must be done.
There are valid reasons for locating a monitor at a position other
than the stack and the relative importance of conflicting objectives will
determine whether a particular location is acceptable. No one set of
standards can be universally applicable and some judgement is required.
When considering a representative location for gaseous pollutant moni-
toring, the major concern is potential stratification of the gas stream.
Stratification, meaning the presence of a definite change in concentration
with respect to position in the duct cross section, usually results from
insufficient turbulence or mixing of flue gases after inleakage of a gas
with different composition or from localized chemical phenomenon such as
reaction or sorption. One of the advantages of cross-stack monitors is
that the instruments provide an average value over the path length and
thus lessen the effect of stratification. They do not, however, measure
the average in two dimensions so the result does not necessarily reflect
the average of the entire gas stream. If the proposed location is not
near any source of inleakage, the disturbance criteria will suffice to
ensure a representative sample. Concentration changes from inleakage most
often results from ambient air being drawn into the ductwork in the up-
stream, or negative-pressure side of the fan. Typical sources of inleakage
are joints and seals between fixed and movable part, vibration isolation
joints and seals, and thermal expansion joints. Engineering drawings for
the facility layout should be reviewed to determine the nearness of these
potential leak sources. Fan seals seldom result in stratification because
of the turbulence in the fan housing, but leaks here add to dilution of
the flue gas. For a number of reasons, dilution air is intentionally or
unavoidably added which may create stratification. Many incinerator designs
add quench or tempering air to cool the flue gas and minimize insulation
requirements and temperature stress to the ductwork. The fabric filter
collectors (baghouses) used for particle removal often use ambient air to
clean the bags (seldom creates stratified gas stream, but adds dilution
air) and baghouse operating temperature limits often require the addition
of cooling air near the inlet (often creates stratified gas stream).
Stratified gas streams caused by reaction are less common than leak-
related causes, but the drawings should be reviewed to determine if strati-
fication might exist. Sampling locations near the flame will exhibit
stratification because of the ongoing combustion process and should gen-
37
-------
erally be avoided. The discharge of acid gas scrubbers can be stratified
due to the combination of low design velocity (no mixing) and localized
variation In removal efficiency (concentration difference). This stratifi-
cation usually occurs with the less soluble gases like SO 2 and in large
scrubbers as are used in power plants, but it can occur in HWIs. Wet
scrubbers used on HWIs seldom remove CO but other gases, like SO 2 and
HC1, can be affected.
If there is a potential for stratification, alternative locations
should be investigated. As discussed in Section 3, the flow disturbance
criteria are Intended to apply to velocity-related issues of representative
sampling and these issues are not as important for gaseous pollutant
sampling as for particle sampling. In fact, if the gas stream shows smooth
velocity profiles or low velocities (<-1200 feet/mm), additional distance
from any source of concentration change is needed because less mixing
occurs at these conditions. Use of the 2/0.5 distance criteria of the CEM
Guidelines is limited to cases where no stratification is expected.
If no other location is suitable and there is a probability of strati-
fication, a stratification test (described in Appendix D) of the proposed
sampling should be performed. Obviously, this test can only be done on an
operating facility. For units not yet operating, use of an alternative
location is peferable to changing the system after installation. Using a
stratified location should be a last resort due to the intrinsic difficulty
of obtaining a quantitatively accurate sample. If a cross-stack instrument
is used to analyze a stratified gas stream, the instrument should be
placed so the measurement path is oriented perpendicular to any concen-
tration gradient. The path should represent the overall gas stream so
that the linear average provided by the long-path measurement is equivalent
to the calculated average for the entire gas stream. For extractive sys-
tems, a single sample point should be located at a point of average CO
concentration and this point should be within the inner 50% area of the
duct. If a multiple point sampling probe is used, the sample points should
be located in the central 50% of the area so the sample point average
equals the duct average.
In addition to stratification, inleakage can dilute the flue gas stream
and mask combustion-related changes in the gas composition. This masking
is not a common problem, but should be considered if the facility is
designed to use quench or tempering air. The effects of dilution cause
more problems for 02 monitoring than for CO or CO 2 monitoring. For example,
quench air is added to cool the flue gas from 1500’F to 500F. The 02
concentration of the gas prior to the point of dilution is 7% and the 02
concentration after dilution would be -16.8%. If the combustion condition
changed such that the pre-dilution 02 level was 5% (a change of 2 volume
percent 02), the post-dilution concentration would only change by 0.6
volume % from -16.8% to 16.2%). In this case, there is a good reason to
measure oxygen prior to dilution and It may then be desireable to measure
CO at the same location, even though the location is less than desireable
for other reasons (dust and temperature). Changes in CO concentration are
less masked by the dilution because there Is no CO in the dilution air.
For the same dilution and operating scenario as above, the CO changes
38
-------
would appear as undiluted concentrations of 500 ppm/750 ppm and as diluted
concentrations of 150/220 ppm. Thus, the CO monitor would need a lower
operating range to detect the changes but the proportional difference
would remain the same. This type of evaluation should be performed to
ensure that the pre-dilution sampling location is warranted.
SamDle Interface
The general considerations for the sample interface, which includes
all elements of the system from the duct to the analyzer measurement
section, relate to materials of construction and selection of appropriate
conditioning equipment. Because of the variety of systems which can be
employed, it Is virtually impossible to define the specific hardware which
comprises an acceptable interface. This section presents some general
guidelines as to the factors to be considered.
Cross-Stack Analyzers - -
The cross-stack analyzers have minimal sample interface requirements
other than those already discussed regarding selection of a monitoring
site. Two units bolt to flanges on opposite sides of the stack and special
optical windows are the only components in contact with the gas stream.
The units, the source or source/detector module and the detector or reflec-
tor module need to be placed in the same horizontal and vertical planes
perpendicular to the duct walls. The instruments can be adjusted, within
limits, to achieve and maintain proper optical alignment. The specific
location on the stack or duct should be as free from vibration as possible.
Generally, this requires some form of isolation of the sample location
duct segment from fans and motors. Many HWIs have a free-standing stack
on a concrete base with the induced-draft (ID) fan on the same pad. The
fan motor can cause stack vibration if the components are not isolated
by separate pads or vibration dampers. Because most of the measurement
system is mounted on the stack, it is important to provide good access to
the instrument housings.
Extractive Systems-Conditioning - -
Nearly all analyzers which require the sample to be conveyed to the
measurement section need some form of conditioning to ensure that the
sample is relatively free from interfering components and physically
compatible (temperature, pressure, etc.) with the sensor and/or detector.
The integrity of the extracted sample needs to be maintained. The recom-
mendations for materials and for conditioning are based on a stack location
following a wet scrubber so the flue gas is expected to exhibit the fol-
lowing characteristics:
• low-to-moderate amounts of HC1, generally less than 50 ppmv,
• low particle concentration (less than 0.08 gr/dscf) but with a
size distribution with a predominance of fine particles,(i.e.
less than 5pm,
• saturated moisture content (12-32% by volume), with temperatures
in the range of 125F to 160F,
39
-------
• gas composition of 5-15% C0 2 , 15-5% 02, 0-2000 ppm CO, 0-200 ppm
so 2 .
The reviewer will need a schematic (like one presented in Appendix C) for
the system, which identifies the major components of the sampling train,
and a description of the operating characteristics and materials of con-
struction. An engineering drawing of the stack and duct system showing
monitor location Is also necessary.
Probe . The sample probe is exposed to more extreme conditions than any
other part of the sample interface so materials of construction are very
important. The probe material needs to withstand the stack gas environment
without significantly reacting with the extracted sample. Teflons, Has-
telloy, and Incoloy• are good choices for nearly any application; the
metal alloys have higher temperature limits. Teflon can be used up to
about 450’F and Is virtually inert; Hastelloy’ is an expensive nickel/-
chromium/molybdenum alloy useful for high temperatures and corrosive
environments; Incoloy is another trade name for a high nickel/chromium
steel alloy useful for high temperatures and corrosive environments.
Teflon• probes need to be relatively short, especially for higher temper-
atures, because it is a thermoplastic material and will soften and deform.
Some operators have achieved some success with PVC, although this would
not be advisable for units with high-sulfur feeds. Special ceramic mater-
ials are often required for combustion-zone sampling and cooled-jacket
probes are often necessary for temperatures above about 1800’F. The dis-
charge end of the probe should be equipped with valves and fittings so
that calibration gases can be flowed through the entire system. The piping
and valving for calibration gases can be very simple or very complex;
simple systems generally require much more manual attention in the form
of connecting gas cylinders and switching valves than the more expensive
multiple-line solenoid valve systems. The more complex systems offer the
ability to automate calibrations and eliminate some of the need for oper-
ator attention. The probe is sometimes built with integral coarse particle
filters (at the probe tip in the stack) and moisture receivers and should
be provided with some means of back-flushing to minimize downtime for
cleaning.
Particle filters . Particle filtration Is usually divided into coarse
and fine removal. Coarse filters are more important for sampling locations
upstreani of air pollution control devices and are usually exposed to more
severe operating environments. Most of the particulate matter has been
removed by the time the gases reach the stack so stack locations require
only fine particle filters; coarse filters may be used as added protection
for the rest of the system.
Coarse filtration for particles larger than about 10 microns is provided
by sintered stainless steel filters or sometimes porous ceramic elements.
Most coarse filters are In-line and replaceable only with some difficulty
so a blow-back system is used to clean the filter medium. They are usually
heated (by the stack gas if Inside the stack or by a heating element if
external to the stack) to avoid condensation and corrosion from condensing
40
-------
acid gases. Some systems, especially those using educted samples, use a
flow-through/bypass (or Slipstream) filter for coarse filtration. The
educted sample is drawn through the center of a tubular filter at a high
velocity and vented either back into the stack or to the atmosphere. A
second sample line is attached to the annulus between the outside of the
tubular filter and the filter holder tube and a portion of the bypass
sample is caused to flow through the tubular filter medium by placing the
annulus under vacuum. The source of the vacuum is often the same eductor
used for the bypass sample flow, but it can be a sample pump. The high
velocity in the tube center prevents clogging of the filter face and
permits effective filtration with little pressure drop.
For extractive analyzers, virtually all particles greater than 1 j.tm
diameter must be removed. Fine filtration IS accomplished with depth
filters using porous, disposable media which remove particles effectively
without creating excessive pressure drops. Most fine filter media are
disposable because they are not easily cleaned. Materials successfully
used for fine filters include spun glass fibers and polypropylene fibers.
A number of commercially available conditioning systems use a coalescing
filter as a fine filter/condensate trap with a final particle filter at
the analyzer inlet. The coalescing filter agglomerates liquid particles,
including water, often acting as condenser and also removes fine particles
at a low pressure drop.
Moisture Removal . Some analyzers, particularly flame ionization units
for hydrocarbon analysis, benefit from a high temperature sample gas, and
others, like NDUVs, can tolerate relatively high temperatures (>250F). A
heated sample line is used to keep the sample gas above the boiling point
of water and particle removal is all that is required. If the CO unit is
to use the same sample line, the moisture removal section will need to be
placed between the THC and the CO analyzers. NDIR analyzers require drying
of the sample to avoid interference problems from water vapor and nearly
all extractive system CO analyzers require a liquid-free gas sample gas.
Most current-generation NDIR analyzers, whether GFC or conventional, are
equipped with optical filters and signal processing to avoid significant
interference problems (<1% of instrument full scale) with water vapor
concentrations up to 3 vol. % for a measurement range of 250 ppm. This
requires a sample gas with a dew point lower than 115F. Most extractive
systems will use an air- or water-cooled condenser to reduce the stack
gas temperature, and condense and remove most of the flue gas water (to
less than 2-3%). Additional moisture removal is obtained with desiccant
dryers, permeation dryers, or refrigeration condenser/dryers. The per-
meation dryer is probably the easiest to maintain and does an effective
job at reasonable operating costs. The sample gas is simply passed through
a bundle of tubes made of inert semi-permeable membrane material which
allows water vapor (not droplets) to pass from the sample gas to the
purge gas flowing over the tube bundle. The operating limit for most
permeation dryers Is about 125F; if the sample gas exceeds the temperature
limit, the dryer will be permanently damaged. Refrigeration dryers are
very effective, but they are expensive to install and maintain and gener-
ally used only when a large volume (like plant instrument air) of sample
gas is being processed; most analyzers do not need sample gas as dry as can
41
-------
be obtained with a refrigeration unit. Desiccant units also yield a very
dry gas by passing the sample gas through a packed bed of calcium sulfate
or other solid desiccant. Desiccant beds are seldom used for CEMS because
the sorbent capacity is limited and some desiccarits will also react with
flue gas components.
Interferent Removal . As a general practice, one should not attempt to
remove interfererits in a sample conditioning system. There are too many
unknowns involving the operation of the interferent removal device to
operate a conditioning system with any reasonable assurance that the
interfering component, and only the interfering component, is being re-
moved. Further, most analyzers are designed to eliminate the effects of
most of the common interferences. If it is necessary, the most common way
is to use a solid sorbent for insoluble gases or a scrubber for soluble
ones. Chemical oxidizing scrubbers are also used to convert one compound
into a non-interfering one. No specific recomendations are given for
interferent removal.
Temperature. Pressure, Flow Rate Conditioning . Generally, a particular
analyzer has a limited operating range for these parameters and most
systems will be designed to operate within these limits. The flow rate
through the sampling system should be more than required by the instruments
and the excess sample vented. The pump should be as close to the probe as
possible so most of the sample system is pressurized and any leaks will
be out rather than in. Temperature control is obtained by the moisture
removal system and it is a good idea to reheat the sample after final
drying as an added precaution against condensation in the analyzer. Pres-
sure regulators may be required if the analyzer does not contain internal
sample gas pressure control or if the overall sample line is operated at
a high pressure (>20 psi). If an aspirator or eductor pump is used, it
has to be placed at the end of the sample line and will place the entire
system under vacuum with the highest vacuum occurring at the analyzer; many
analyzers are built to withstand pressure but not vacuum. The manufac-
turers’ specifications should be checked against the system design. Most
NDIR analyzers are not very susceptible to small changes in flow rate, so
precise regulation of sample rate is not required.
Pumo and SamDle line . The pump can be either mechanical or fluidic.
Mechanical pumps can be used in either pressure or vacuum service; fluidic
pumps like aspirators and ejectors can only be used as vacuum sources.
Many systems will use a combination of both. Fluidic pumps require a
large volume of steam, compressed air, or water; but are very simple to
operate. For most applications, the mechanical pump will be used. If the
pump Is located at the end of the system (vacuum service), consideration
of the materials of construction is focused on pump reliability because
the sample has been analyzed when it reaches the pump. Pump surfaces in
contact with the unanalyzed sample gas should be coated with or made form
an inert material. Teflon has proven to be the best all-purpose material
for most applications, but stainless steels can be used after the con-
ditioning system. The discharge should be vented away from the workplace.
42
-------
The sample line should be Teflon’ or stainless steel, although poly-
propylene can be used if only CO and 02 are being analyzed. Sample line
size depends on the number of analyzers and required flow rates, but
lines smaller than 3/16 in. LD. create high flow resistance for the
length of line likely to be required.
Analyzer Performance Review
This task is relatively straightforward and consists of comparing the
analyzer operating specifications as reported by the manufacturer with
the limitations in Appendix D. The instrument must be able to measure CO
continuously as defined in Appendix D and must meet the standards for the
range, response time, precision, linearity and span drift limits. The
zero drift should also be compared against the operating guide. If the
reported values are within the limits, the system should be acceptable
and will probably pass the performance tests. The comparisons should
reflect the actual Instrument being proposed including the specific ranges
to be used. The one area where there may be some difficulty is where a
single analyzer with range switching is being considered for both sets of
specifications. In this case, the review should address the capability of
the unit to meet both the high and low range specifications without inter-
mediate calibration, (i.e. that performance is maintained in each range
regardless of which range was used for calibration).
Data HandlinQ System Review
At the minimum, the data system needs to have a continuous output
meter, so the concentration at any time can be visually observed, and a
chart recorder which records the meter output over time. The performance
specifications permit the output reading to be updated at a maximum of 30
second intervals. This minimum system could consist of a display meter and
a recorder which recorded a value every 30 seconds. At the other extreme,
a sophisticated and completely automated data system would:
- record the analyzer output over time;
- calculate and report the data in the desired format;
- calibrate the system and record the results;
- generate malfunction and scheduled maintenance reports;
- generate the desired combustion controller input signals; and
- summarize emission and activity data formatted for submittal to
the regulatory agency.
A real system will fall somewhere between these two extremes.
There are two types of data to be recorded; emission data and system
operating data. A sophisticated system can handle most of this data. Most
systems will process at least the emission data and maybe the calibration
43
-------
ANALYZER MAKE All) MOOEL
OWNER/OPERATOR
SYSTEM DESCRPflON
ANALYZER PARAMTER
ALLOWABLE LMTS
Low Range H Range
THS I$TRLLeIT
Low Range II % Rang.
I 1O00
.1.On*..k
6%F$
e o’IP8
2.OIIPS
1o
20 o
50%
U .imy.l* iivOI
100%
2G pn
60f I
Sp i M L P r ’n
20 24-llfw
6GI Pt 2 t mu
SU GESTED LMTS
1 14$ P lSTRI.LU1T
io 20mm
a%ce
PU FORMANCE TEST REYLTS
TEST ARMETER
MJIWEER OF TESTS
ANALYZER PERFO 4ANCE
cM• Th*
3
II P
1I Do
3
.1
ML 2.law us i A
0
20150
I of
I . 24. r
BpsiML 24-Em
I fiof
00-1
00% P8
8p s Ow **A
10-100% of l d
v uisas
•
G0IIF 2 0 cr2 0 %F50
LOisuty.
•
10.0% rwao
5.osFe’oo
Figure 4. Performance Review Checklist
I VEW R 1ERtA
-------
data; the operating data will be left to plant operating personnel. The
data processing system, regardless of complexity, should provide records
of compliance with the CO limits and cut-off requirements. Typical reports
would not summarize emission levels over time; they would tend to be
incident reports, describing the frequency and magnitude of CO emissions
which necessitated waste cut-off, duration of cut-off periods, and tabu-
lating highest averages if no cut-of fs occurred.
Perhaps the most significant aspect of the data handling system is the
procedure for correcting the CO concentration to 7% oxygen on dry basis.
If correction is done “on the fly”, some form of automation is a virtual
necessity. A programmed microprocessor would read the output from both CO
and 02 analyzers and calculate a corrected value continuously. The recorded
data should include all three parameters for each data point, i.e. stack
gas CO, stack gas 02, and corrected CO. The stack gas CO values are not
used for any controT action, but should be recorded as verification of
proper operation of the system. The permit wifl specify limits based Dfl
oxygen-corrected values, so the 02 corrected values are used for the cut-
off system operation.
SYSTEM PERFORMANCE TEST
After the system has been installed, a performance test is required to
verify that the installed CEMS meets specifications of Appendix D. This
performance test needs to be conducted before the trial burn to ensure
that the CO data generated during the trial burn is accurate. A written
test protocol should be prepared, as a portion of the Part B or the trial
burn plan, describing conduct of the system performance test. The protocol
should contain descriptions of all elements of the test, including:
• test procedures, description of the type of calibration
standards to be used, method of introducing the standards
into the measurement system, number of repetitions of the
drift, precision, and linearity tests,
• the test schedule, including the recommended conditioning
period, a detailed schedule for the specification tests,
and the schedule for data reduction and evaluation,
• a test report format, including description of the pro-
cedures for calculations, data handling and data reduction,
and the forms used for recording results and operations.
Test Procedures
The basic procedures to be used for a performance test are described
in Appendix 0. Each individual system will require some specific methods
and procedures for performing the various tests; these site-specific
procedures should be included in the performance test protocol. For
example, one system may be built so an operator pushes a button, while
another requires an operator to manually connect a new gas cylinder to
the line and open and close the proper valves. The quality and source of
45
-------
calibration standards for the test and for the routine operation should
be described.
The performance specifications also describe the number and type of
particular tests required to demonstrate compliance with the specifica-
tions. There may be occasions where more tests are required, or where
modifications to the sequence are necessary and these types of deviations
and the reasons therefore should be discussed. Manufacturers of calibration
standards can provide documentation of the precision and accuracy of
their standards. In general, calibration standards should be manufacturer-
certified to ± 2% analytical accuracy, which means that the reported or
tag value (nominal value) is within 2% of the actual value. These standards
are suitable for day-to-day operation and calibrations, but use of refer-
ence standards such as “NBS-traceable” or “certified reference materials”
(CRMs) is recommended for the performance test. The reference standard is
used as a transfer standard to check the manufacturer-certified value of
a secondary standard. The system would be calibrated using CRMs and then
used to analyze the secondary standard. If analysis of this secondary
standard agrees with the nominal value the nominal value is used. If the
difference between the reference standard and the secondard standard is
greater than manufacturer’s allowance for the secondary standard, the
value obtained from the calibrated CEM is used.
Test Schedule
The test schedule description should be compared against the performance
specification to ensure that the proper time intervals and number of
repetitions are correct. While it is not always practical, the schedule
should allow enough time between conditioning and the performance test and
between the performance test and the trial burn to fix any problems that
are identified.
A schedule in a daily appointment-like format is suggested to depict
the sequence of individual test measurements. This aids in test conduct
and permits identification of conflicts and overlaps. While not part of
the monitor performance evaluation, testing of the waste feed cut-off
system can be scheduled to occur during the CEMS tests. Most operators
will test and debug the process controllers independently, but setting up
a test program for the combined CEllS and cut-off system during the CEMS
performance tests can eliminate problems commonly encountered in the
trial burn. The cut-off system is often deactivated for the trial burn.
Data Reduction and Test ReDort Format
Reduction of data from the performance tests is relatively straight-
forward. If a programmable or computerized data processing system is to
be used, it needs to be programmed to provide the appropriate results for
use in the comparisons. Some operators will prefer to use a temporary
system (e.g. strip-chart recorder) for recording the performance test data
because it is often simpler to record and analyze the data manually than
to program or reprogram a microprocessor system. In any event, the test
46
-------
protocol should provide a report format sufficiently detailed so the
final report amounts to filling in the blanks.
TRIAL BURN PLAN
Although additional effort is required to review the monitoring system,
the proposed CO limits for HWIs have simplified the task of establishing
permit conditions based on trial burn test results. In the past, trial
burn conditions were set to yield a “worst-case” operation so acceptable
performance of the incinerator could be reasonably assured under the most
difficult conditions. This remains the case for most operational variables
but not for CO concentrations. Waste cut-offs during a DRE test run are
highly undesirable. A cut-off during a DRE test run complicates completing
a valid run or invalidates the run. Thus, the trial burn test conditions
should be established to avoid waste cut-off during a test run. The trial
burn plan should identify the CO concentrations expected during the trial
burn and data and cut-off systems should be configured for the expected
CO concentrations.
Monitoring during Trial Burn
The CO CEMS in use during the trial burn must be in compliance with
the performance specifications, and the routine calibrations should show
continued compliance during the trial burn. Both the CEMS and waste feed
cut-off system need to be operating during the trial burn and the activ-
ation levels and logic for cutting off waste feed should be established in
such a manner that the CO limits are not exceeded. The impact of oxygen
correction of stack gas CO also needs to be considered. It will not be
sufficient to measure the average 02 level and assume that it will remain
constant over the course of the trial burn.
If a cross-stack instrument is used, the moisture content of the flue
gases at the point of CO monitoring needs to be measured during the trial
burn and the CO concentration corrected to a dry basis. The moisture
content of the flue gas will be relatively constant and a single correction
factor, developed during the trial burn, can be used for subsequent cor-
rections.
Data Reduction and Reporting
The trial burn plan should include detailed descriptions of the data
recording system. If the system is largely manual, the description can
consist mostly of data forms and equations. If an automated data system
will be used for most of the data evaluations, the trial burn plan needs
to contaln.a fairly detailed explanation of the hardware, software, and
interfaces for the system. In many situations, the data system isn’t
finished until the the monitors are operational.
47
-------
SECTION 5
REVIEW OF TRIAL BURN REPORT
INTRODUCTION
The trial burn is the focal point for many technical issues that have
not been resolved in earlier stages of the permitting process. Thus, the
results of the trial burn are usually expected to provide answers to many
questions of technical feasibility and performance of equipment and these
answers become the basis for making deferred decisions. However, the
primary purpose of the trial burn is to demonstrate the incinerator’s
capability to destroy hazardous waste and to document certain operating
conditions as representative of incinerator performance. The values for
these operating parameters are incorporated in the final permit, if issued,
to assure compliance with the performance standards during continued
operation. The concentration of carbon monoxide in the combustion gas
measured during the trial burn is an important parameter as is evidenced
by the requirement that waste feed be discontinued if the CO level(s) is
exceeded. This section discusses review of the trial burn report with
respect to CO emissions and establishing permit conditions.
The permit writer’s task is to ensure that the hazardous waste is
being destroyed in compliance with the regulations. In order to provide
this assurance, the data used to set conditions must be accurate. The CO
data generated during the actual DRE testing portion of the trial burn
forms the basis for the permit conditions and it is incumbent on all
concerned to see that the data were collected and analyzed properly. The
last part of this section suggests methods to review and analyze this
data.
CO MONITOR PERFORMANCE TEST REVIEW
Review of a report on monitor performance testing is fairly
straightforward and requires verification that the system was installed
as expected, that all the required tests were conducted, that the data
were reduced and calculated correctly, and finally that the data show
compliance with the specifications.
Monitor Location
In most cases, the monitor location will have been agreed to in advance
and should not be an issue. In some cases, however, a precise location
48
-------
may not have been established. In any event, the report should include a
drawing of the “as-built” CEllS location, for the probe if an extractive
system, or the instrument if an in-situ. The “as-built” structure is
usually somewhat different than the design arrangement and most review
and approvals are based on design drawings. Often, design drawings are
not updated when the facility is finished. These construction/design
variances are not particularly significant if the only construction is
the monitoring system (as for an interim status facility without a CO
monitor) but they can be significant for a new facility.
Test Procedures
The report should contain a description of the procedures used to
conduct the test and these procedures should match those required in
Appendix 0. There are four separate test parameters to be evaluated:
range, span drift, precision, and linearity. The zero drift may also be
included. Appendix 0 describes specific numbers of repetitions and the
sequence of the tests. There are also some requirements for Intervening
measurements, i.e. switching to another gas between precision measurements.
The order of the tests may vary, but all must be present. The test pro-
cedure description should discuss the quality of the calibration standards
and the method of introducing the standards into the measurement path. This
method used should include the entire sample train from stack to analyzer.
The report should include a description of any automatic internal analyzer
calibrations used during the tests in sufficient detail that subsequent
inspections or tests can duplicate the conditions. These conditions should
be included in the permit to ensure that the system operation is consistent
with the performance tests.
Test Completion
All the required tests should be completed and Table 8 presents an
example checklist to verify completion of the required tests. This type
of checklist also serves to identify that the procedures were conducted
as intended. If the raw test data are presented on strip-charts, completion
of the checklist is easy; if the raw data are printouts of a data proces-
sing system, completion of the checklist is likely to be more difficult but
also more necessary for ensuring that all the data are present. This
table can easily be modified to include provisions for summarizing the
test results and calculating the required values, in a form similar to
the calculation forms of Appendix 0.
Calculations
Appendix 0 describes the equations and procedures to be used in deter-
mining whether a system meets the performance requirements. The data are
treated statistically by applying a confidence interval to the measurement
sets so that some consideration is given to inherent variances in the
measurement process and to the possibility of infrequent occurrences of
measurements exceeding some established limit. The calculations presented
in the report should at least be audited by independent calculation of
results for one test parameter and not much additional effort is required
49
-------
to check all of them. The basic statistical equations can easily be done
on most calculators and many calculators are pre-programmed with mean and
standard deviation routines. Table 8 can easily be adapted to include
provisions for data entry and calculation.
TABLE 8. MONITOR PERFORMANCE TEST COMPLETION CHECKLIST
Test Parameter
Initial
1
Meas
2
ureme
3
nt
4
Number
5
6
Final
Span Drift Tests
Required
Y
Y
Y
Y
V
Y
V
N
Compi eted
Zero Drift Tests
Required
V
N
N
N
N
N
N
Y
Completed
Precision Tests
Required
V
V
V
V
V
V
V
N
Completed
Linearity Tests
Required
V
V
V
Y
V
V
•Y
N
Completed
* Y - Required
N - Not required
Test Acceotance
The final step In performance test review is comparison of the test
results with the specifications. If the system meets the test specification
for all parameters, the system is acceptable and the trial burn data can
be used for combustion evaluation and permit conditions. If the zero
drift is greater than the guideline, the system is conditionally acceptable
50
-------
and the trial burn data should be examined closely to ensure that cali-
bration was maintained during the ORE tests. If the system fails to meet
any of the other specifications, the data for the trial burn are not
valid and should not be used for permit purposes.
EVALUATION OF TRIAL BURN RESULTS FOR CARBON MONOXIDE
The proposed CO limits should simplify the process of setting permit
conditions for CO from the trial burn. If the CO emissions during the
trial burn exceed the proposed limits and ORE is maintained, the limits
will stand. If the CO levels are less than the proposed limits, then the
permit conditions should reflect the lower levels. Further, the monitor
performance specifications provide a yes/no test for the acceptability of
the CO data thus relieving the reviewer of some of the burden of evaluating
questionable data. The reviewer must, nonetheless, evaluate the CO data
generated during the trial burn carefully and needs to consider some
other aspects of CO emissions related to combustion stability.
Test Procedures
Having established the performance characteristics of the CEMS, the
quality of the CO data is presumed acceptable and the actual monitoring
procedures need minimal review. The reviewer needs to verify that the
CEMS was operated tn the same manner as during the performance testing.
Section 6 lists daily inspection/observation parameters which can serve
as evidence of continued proper operation. Recording these parameters
during the trial burn and the performance testing Is recommended.
It is suggested that the CEVIS operating data be recorded more frequently
than daily during the test burn. In the event that the monitoring system
experiences some malfunction, a data record of the operation may avoid
the undesirable situation of having to reject the ORE test results for an
entire day or more. Repeating one test run is preferable to repeating a
complete trial burn test series. It is also important to verify that
internal analyzer calibrations and/or other system checks performed without
operator intervention are maintained at the same frequency and intensity
as they were during the monitor performance testing.
The last item regarding the trial burn testing is whether or not CO
data exists for all the ORE tests. If the data are absent from one or
more test runs or for a complete burn condition, that ORE data will pro-
bably not be acceptable for permit condition use. There are some cases
when exceptions should be considered (discussed in the following sections)
and some cases where absence of the CO data is immaterial or not par-
ticularly significant such as a test burn conducted with high excess
oxygen (and correspondingly low CO) to demonstrate scrubber particle
removal efficiency.
Data Review
There are a number of aspects of the trial burn CO results that need
to be reviewed for determining appropriate permit conditions. These aspects
- 51
-------
encompass the time bases for averaging, correction of stack gas CO concen-
trations to 7% 02 and dry basis, dealing with the rolling average require-
ment, and dealing with anomalous or missing data.
The evaluation of CO levels should be related to the ORE of the unit.
The only ORE data comes from the emission tests. A VOST test run may be
as short as 20 minutes and a modified Method 5 run may last as long as
seviral hours. It is desirable to match the CO data to the ORE test
data. The rolling average requirement of the proposed CO limits dictates
that the average concentration for the test run not be used; rather the
CO data must consist of several 10- and 60-minute averages. The number of
values to be used will depend on the length of the test runs and the
integration/update interval of the data processing system. By definition,
continuous CO monitoring requires an analytical update at least once
every 30 seconds. If we assume a 60-minute test run and 30-second updates,
there will be one 60 minute average and iii 10-minute average values. The
10-minute values are based on no average value until 10 minutes have
elapsed and one updated 10-minute rolling average every 30 seconds there-
after. If the test run lasted 70 minutes, there would be 21 values for
60-minute averages and 131 values for 10-minute averages.
In the above example, the 30-second update by the data processing
system has already integrated the measured concentration for that time
interval. While the CO is being measured continuously, by the definition
in Appendix D, the concentration is being reported at time intervals
deemed often enough to represent the combustion status of the incinerator.
The displayed value may be updated as often as the detector output changes,
but the recorded value or the value processed by the data system may be
different. The data system could be designed to update the average values
as often as the display is updated, even though the guidelines do not
require it.
Correction of the stack gas concentration to 7% 02 is subject to the
same data processing concerns as described above for CO and there are
some other considerations as well. The performance specification for oxygen
monitors requires the same maximum response time as the CO analyzer.
However, the response time for the analyzers can be different. The recom-
mended approach is to correct the CO values to 7% 02 as they are included
in the 10- and 60-minute average calculations. This does not address
whether the value used for 02 correction represents the same time base as
the CO value. If the oxygen monitor has a slower response than the CO
analyzer, the CO values need to be “held up” in the data system or vice
versa if the oxygen analyzer responds more rapidly. Another alternative
would be independent calculation of oxygen concentrations and correction
the rolllng.raverage CO concentrations with rolling average oxygen concen-
trations.
A relatively common occurrence is the appearance of one or two very
high values, compared to the rest of the data. To determine whether these
data points should be included in the data set used for establishing
permit conditions, a statistical test should be applied to the data.
52
-------
There are a number of techniques for evaluating whether a seemingly anom-
alous value is beyond the range of reasonable probability and these tech-
niques are described in most statistics references. One technique commonly
used Is the Grubbs’ °t” test for outliers. In general, disregarding these
outliers Is the more conservative approach because it tends to eliminate
extraordinarily high values which result in higher permit levels than
normal operation would indicate.
Another common concern is whether a test is valid if some of the data
are missing. With short-term testing as is done during trial burns, 90%
valid data collection Is considered acceptable. For CO continuous moni-
toring, this means that valid data should exist for 90% of the time during
which DRE testing was being done. Data completeness as low as 75% is often
considered acceptable where the data do not exhibit much variation (stan-
dard deviation less than 10% of the mean). The trial burn plan should
include objectives for data completeness.
The proposed limits define stable combustion, after an upset, as a 10-
minute average value less than the 60-minute standard. If permit limits
lower than the proposed limits are being considered, the trial burn 10-
minute averages can be used to establish a corresponding definition of
stable combustion.
PERMIT CONDITIONS AND WASTE FEED CUT-OFF SYSTEM
The currently proposed CO limits state that waste is not burned if the
CO concentrations are above certain limits (with limited exceptions);
this would not prevent CO levels from exceeding these limits. If the CO
concentrations during the trial burn exceed the proposed values, the
permitted waste feed cut-off trigger levels will be 100 and 500 ppm for
the 60- and 10-minute averages, respectively. If, however, the trial burn
data show that the unit has the capability to operate at lower levels and
meet the ORE and other standards, then the permitted waste cut-off levels
should be lower than the proposed levels. The trial burn report should
contain documentation of the cut-off actions and operation during the
testing so proper operation can be verified.
The suggested procedure for determining appropriate levels for the
waste feed cut-off is to determine the means and arithmetic standard
deviations for each of the runs and for each test run series for both the
10 and 60 minute averages. The waste cut-off levels should then be set as
the sum of the mean and 2 times the standard deviation for the highest
test run which met the DRE standards unless this value (2s + mean) is
greater than the applicable limit. If this sum is greater than the limiting
values, then whatever limits exist are to be used.
The trial burn data should be reviewed to determine if the CO values
for restarting waste feed after a cut-off were consistent with the proposed
emission limits, or whether some other value(s) is appropriate. Following
this review, the permit should include either the regulatory limit or a
lower one, if the data justify a lower value.
53
-------
SECTION 6
OPERATING AND MAINTENANCE PROCEDURES
INTRODUCT ION
In order for the data generated by the system to be useful to the
agencies and the operators, the CEMS must operate successfully on a con-
tinuing basis. Successful operation requires periodic checks of system
performance and maintenance of records to document both the data generated
and the activities performed to maintain the quality of that data. Certain
inspection/verification procedures, primarily for the waste feed cut-off
system, are required by the regulations to prevent malfunctions which
might result in avoidable releases of hazardous materials to the atmos-
phere. This section provides some guidelines for routine CEMS operation to
ensure continued compliance with the established standards.
The basic purpose of these QA procedures is to prevent a system malfunc-
tion from going undetected for an extended period of time and to minimize
the magnitude of such a situation. Most units will occasionally go out-of-
specification and these procedures will detect this occurrence. If a
system starts going “off spec” frequently or regularly, there is cause
for concern. Keeping records permits the operator and the agency to eval-
uate the significance of a particular case.
Most industrial facilities are accustomed to using written standard
operating and maintenance (0 & N) procedures to maintain process operations
within established limits. For this manual, the types of activities in-
volved In an 0 & N program are divided into three categories: 1) observa-
tion and inspection; 2) testing and verification; and 3) corrective action.
All activities require recordkeeping for historical and analytical pur-
poses. The following sections discuss, in general terms, the application
of these 0 & N procedures to the monitoring system and, to a lesser extent,
the waste feed cut-off system.
EMISSION MONITORING SYSTEM
The basic 0 & M program for a CEMS consists of calibration, audit, and
preventive maintenance activities. The performance frequency of these
activities is divided among “as required”, daily, weekly, monthly, and
annually or semi-annually. The level of attention or detail increases with
decreasing frequency of performance.
54.
-------
Daily
Two system checks should be performed daily: a system inspection to
confirm operating status of all elements, and; a calibration check of the
instrument zero and span.
System Inspection - -
The daily system inspection consists of visual observations of system
operating conditions and evaluation of the status against “pass/fail”
criteria. Each system will be unique and a site-specific checklist is
suggested to ensure that all pertinent items are checked. Typical para-
meters to be checked in an extractive system include:
• system pressure drops, (e.g. overall, across filters, or
across the analyzer);
• sample flow rates;
• verification of flow through permeation dryers and air-
cooled condensers;
• sample gas temperatures at appropriate points in the
sample line, (e.g. stack exit, condenser outlet, reheater
outlet and measurement chamber);
• operation of sample mover, (e.g pump noise or vacuum or
ejector operation);
• operation of strip chart recorders (verifying times and dates,
chart advance, pen response);
Where appropriate, a check value or range of normal values should be
established which define acceptable or normal operations. The performance
test can provide the baseline values for proper operation.
Calibration Checks --
The daily calibration checks do not need to duplicate those conducted
during the performance tests as the performance tests demonstrated the
ability for the system to hold its calibration for 6 days. Any internal
electronic calibration check routine used during the performance tests
should be maintained. Some analyzers (particularly the GFC-type NDIRs)
are equipped with an adjustable automatic calibration cycle which places
a CO calibration standard in the measurement path and recalibrates the
analyzer automatically. Other units have similarly named routines (auto-
matic calibration) which check internal electrical signal levels but do not
truly check the measurement section of the instrument. In either of these
cases, the “calibration” frequency should be maintained as during the
performance tests. The basic distinction between the two types is whether
the entire system is checked or only the analyzer or a portion of the
analyzer. If the system does not have provisions for automatic calibration
checks, the daily checks should place zero and calibration standards in
the measurement path for evaluation of instrument response. In the case
55
-------
of the electrical signal check, a zero/span calibration should be performed
weekly. If the span value deviates from the calibration by more than the
specified drift limits, the system requires a full cal bration and the
precision and linearity tests should be conducted. Otherwise, readjustment
of the span is adequate. The incinerator can remain on-line (with waste
feed) during these calibration checks as long as the calibration cycle is
shorter than 10 minutes. A daily full-system calibration check is only
required for those systems which do not use some form of automatic cali-
bration.
Weekly
The RCRA regulations require weekly (unless specifically permitted
less frequently) testing and verification of the waste feed cut-off system.
Many analyzers have a provision for preventing a false control signal
during calibration periods and most units used for HWI monitoring will
probably be equipped with this feature. During this weekly test (or other
frequency if permitted), the CEMS should be calibrated with a standard
value near the permit limits so that the operation of the hardware and
software responsible for the cut-off signal can verified. The specific
procedure for testing the system will depend on the how the controllers
and the CEMS are related. Normally, an instrument span would be higher
than.the allowable Co limits and would cause a cut-off signal if continued
long enough to exceed the time-averaged setpoints. The data system would
normally prevent this false signal. If this signal block can be deacti-
vated, then a calibration cycle can be used to test the hardware and
software for the cut-off. If not, an alternative method should be devised.
If automatic electronic calibration checks are used for the daily
checks, then a weekly test of the measurement system is needed. This
weekly test is only required for those systems where the automatic internal
calibration checks only the electrical circuitry and does not include the
measurement path. If the auto-cal places calibration standards (gases or
cells or filters) in the analytical chamber or path, then the weekly
check is not required. The net result of the daily and weekly calibration
checks should be that the complete system is checked at least once per
week.
Quarterl v
The precision and linearity tests of Sections 4.5 and 4.6 of Appendix
D shoiz’Tdbe repeated quarterly to ensure continued compliance with the
- performance specifications. This is, in effect, a multiple-point recall-
bration compared to the single-point checks done daily or weekly, and is
a more thorough evaluation of system performance.
Annually
Regardless of the operating history, the performance specification
tests should be repeated every two years. The operating history should
be reviewed annually and the performance test repeated if the history
suggests a decline in performance.
56
-------
WASTE FEED CUT-OFF SYSTEM
The proposed limits require that waste feed be discontinued if CO
emissions exceed either rolling-average limits or not-to-be-exceeded
limits. The rolling-average limits require some rudimentary data processing
system; the time/concentration limits only require a timer for cut-off
purposes, but some signal processing is required for the 02 (and moisture,
if required) corrections.
For the not-to-be-exceeded standards, some form of programmable process
controller is required to read the CO and 02 values, correct the CO con-
centration, and generate an output signal to the timer. If the corrected
CO value exceeds the setpoint(s) the timer needs tQ be the accumulating
type with an automatic reset every hour so it will accumulate the total
time above the standard, cut-off waste if the time limit is exceeded, and
set the counter to zero every hour. A second process controller (or chan-
nel) is required to cut-off waste if the instantaneous limit is exceeded.
If rolling-average limits are used, each facility operator will be
required to develop a control logic scheme that will discontinue waste if
one or both of the limits are exceeded and prevent feed until stable
conditions are reestablished. Description of appropriate testing of the
number of potential process control systems which perform the task would
occupy a volume at least the size of this manual. The basic elements of
testing such a system are:
• analog-to-digital signal conversion precision and accuracy
• digital signal integration accuracy
• verification of the mathematical processing and calculation
algorithms (hardware or software or both)
• testing the corresponding output (control) signals
Much of this testing is only done during the debugging phase. After
testing and debugging, the hardware and software tend to perform consist-
ently unless a catastrophic failure occurs.
DATA REDUCTION AND REPORTING
There are two basic types of data records that should be kept for the
CO monitoring system: operating logs and emission records. The operating
logs serve as documentation that the CO emission data are valid. No speci-
fic reporting requirements have been established, although some perniittees
are required to submit periodic reports to state air quality agencies and
sometimes to regional EPA offices. The records are required to be kept on
file and be made available for inspection. No specific report submittal
is suggested but all CEMS records should be retained.
The regulations specifically address recording whether the waste feed
was discontinued when the emission levels exceeded permit limits. Thus,
57
-------
the data records need to show, at a minimum, each occurrence of CO emis-
sions above the standards and the corresponding action regarding waste
feed. Some facility operators may elect to Include other data regarding
trends, process feeds, long or short-term averages or other statistical
summaries of the CO levels in the system records.
The data records can be used in a number of ways, some of which are
identified below:
• Using the regular inspection program to identify problem
areas and repair or replacement before catastrophic failure
occurs.
• Maintaining a record of inspections, repairs, calibrations,
and CEMS operating logs encourages attention to system
operations and provides a basis for diagnosing problems
and prevention of reoccurrence.
EQUIPMENT REPLACEMENT AND REPAIR
The many possible configurations for CEMS makes recommendations for
specific maintenance practices nearly impossible. In general, the manufac-
turers provide recommended preventive maintenance procedures tailored to
their instruments and these procedures are a sound starting point for
development of site-specific 0 & M manuals. The quality of the vendors 0
& N manuals varies and experience with a particular system is often the
best way to generate a useful 0 & N manual. Most manuals also provide a
recommended spare parts inventory, which often amount to a spare analyzer
in pieces.
There are some general guidelines to be considered for maintaining
parts and supplies. However, the decisions about the level of back-up
equipment and supplies Is best left to the operator. A minimal spare-
parts list is presented below:
• Consumable materials (60 day supply is adequate for most pur-
poses)
- recorder pens and chart paper
- sample line filters
- calibration gases
• Spare parts inventory
- fuses
- lamps for analyzers
- pump repair parts, like diaphragms, vanes, motors, or
complete replacement pumps
- valve repair parts like packing, valve stems, sole-
noids, or complete replacement valves
- sample lines and fittings
- condenser coils
58
-------
- pressure and vacuum gauges -
- flowmeters
- most any Item subject to destruction or corrosion by
contact with sample gas.
59
-------
BIBL IOGRAPHY
1. Guidance on Carbon Monoxide Limits for Incinerator RCRA
Permits , (Draft Report), Midwest Research Institute Con-
tract No. 68-01-7287, July 1987.
2. Mitre Corp., Guidance Manual for Hazardous Waste Inciner-
ation_Permits, U.S. Environmental Protection Agency, OSWER,
Washington, D.C., 20460, SW-966, July, 1983.
3. Mclnnes, Robert et. al., Feasibility Study for AdaDating
Present Combustion Source Continuous Emission Monitors to
Hazardous Waste Incinerators. Volumes 1 & 2 , U.S. Environ-
mental Protection Agency, OSW, Washington, D.C., 20460,
EPA-600/8-84-Olla,b, March 1984.
4. Bonner, T. et. al., Engineering Handbook for Hazardous
Waste Incineration , U.S. Environmental Protection Agency
ORD, Cincinnati, Ohio, 45268, SW-889, June 1981.
5. Liptak, Bela G., Editor, Environmental Engineers’ Handbook:
Volume II Air Pollution , Chilton Book Company, Radnor,
Pa., 1974.
6. Perry, Robert H. and Don Green, Editors, Perry’s Chemical
Engineers’ Handbook. 6th Edition , McGraw-Hill Inc., New
York, 1984
7. McGraw-Hill EncycloDedja of Science and Technology , McGraw-
Hill Inc., New York, 1982
8. Guidance on Trial Burn Reporting and Setting Permit Con-
ditions (Draft Report), U.S. Environmental Protection
Agency HWERL, Cincinnati, Ohio, 45268, Acurex Contract
No. 68-03-3241, November 1986.
9. de Lorenzl, Otto, M.E., Combustion Enaineering , Combustion
Engineering Inc. New York, 1953.
10. Baumelster, Theodore, Editor, Marks’ Standard Handbook
for Mechanical Engineers. 7th EdItion, McGraw-Hill Inc. ,
New York, 1967.
11. Stern, Arthur C., Editor, Air Pollution: Volume II . Ana-
lysis. Monitor ing. and Surveying , Academic Press, New
York, 1968.
60
-------
12. Jahnke, James A. and G.J. Aldina, Handbook : Continuous
Air Pollution Source Monitoring Systems , U.S. Environmental
Protection Agency Technology Transfer, Cincinnati Ohio
45268, EPA- 625/6-79-005, June 1979.
13. Gaseous Continuous Emission Monitoring Systems - Perfor-
mance SDeclfication Guidelines for S02. NOx C02. 02. and
I , U.S. Environmental Protection Agency OAQPS/ESED ,
Research Triangle Park, North Carolina, 27711, EPA-450/3-
82-026, October 1982.
14. QualIty Assurance Handbook for Air Pollution Measurement
Systems : Volume I. Princioles , U.S. Environmental Protec-
tion Agency ORD/EMSI, Research Triangle Park, North Caro-
lina, 27711, EPA-600/9-76-006, December 1984.
15. Michie, Raymond M. Jr. et al, Performance Test Results
and Comoarative Data for Designated Reference Methods for
Carbon Monoxide , U.S. Environmental Protection Agency
ORD/EMSL, Research Triangle Park, North Carolina, 27711,
EPA-600/S4-83-013, September 1982.
16. Ferguson, B.B., R.E. Lester, and W.J. Mitchell, Field
Evaluation of Carbon Monoxide and Hvdroaen Sulfide Con-
tinuous Emission Monitors at an Oil Refinery , U.S. Environ-
mental Protection Agency, Research Triangle Park, North
Carolina, 27711, EPA-600/4-82-054, August 1982.
61
-------
APPEND ICES
62
-------
APPENDIX A
MANUFACTURER’S REPORTED SPECIFICATIONS
FOR CO ANALYZERS
63
J3 W1P—5OOSZ4
-------
Manufacturer: ANARAD, Inc.
Model: AR-xxx
Type: Extractive NDIR
Ranges Available: 500 ppm to 100%
Precision: ± 1% FS*
Accuracy: ± 1% FS
Drift: ± 1% FS, 24 hours
Response Time: 5 seconds to 90% FS
Linearity: ± 2% FS
Comments: Manufactured in several models depending on
sophistication and use
Manufacturer: AMETEK, Inc.
Model: El-b
Type: Cross stack NDIR-GFC
Ranges Available: 1000 ppm
Precision: Not specified
Accuracy: ± 3% FS
Drift: Not specified
Response Time: 3 seconds to 90% FS
Linearity: Not specified
* FS - Full Scale
64
-------
Manufacturer: ANETEK, Inc.
Model: WDG-I11C
Type: Extractive catalytic oxidation
Ranges Available: 2000 ppm, 1%
Precision: Not specified
Accuracy: ± FS
Drift: Not specified
Response Time: 11 seconds to 90% FS
Linearity: Not specified
Interferences: H2, other combustibles
Comments: Combustion control analyzer
Manufacturer: Bailey
Model: OL 230 with 01 231 Auto Cal
Type: Extractive catalytic oxidation
Ranges Available: 1000 ppm, 5000 ppm, 1%, 2%
Precision: ± 0.5% FS
Accuracy: ± 1% FS
Drift: ± 2% FS, 30 day
Response Time: 2 seconds to 63% FS
Linearity: ± 1% FS
Interferences: Other combustibles
Comments: Not true CO analyzer, analyzer for combustion
control
65
-------
Manufacturer: CEA Instruments
Make: RIKEN
Modeh RI-200 Series
Type: Extractive NDIR
Ranges Available: 1000 ppm to 50%
Precision: ± 2% FS
Accuracy: Not specified
Drift: ± 2%, 30 day
Response Time: 25 seconds to 90% FS
Linearity: Not specified, (Linearizer optional)
Manufacturer: Dynatron
Model: 3100M
Type: Cross-stack GFC
Ranges Available: Depends on path lengths
Precision: ± 1% FS
Accuracy: ± 10 ppm or ± 2.5% FS Whichever greater
Drift: Not specified
Response Time: AdS. 1-250 seconds
Linearity: ± 1% FS
Comments: Single-pass
66
-------
Manufacturer: Horjba
Make: PIR
Model: 2000
Type: Extractive NDIR
Ranges Available: 500 ppm - 100%
Precision: ± 0.5% FS
Accuracy: ± 1.0% FS
Drift: ± 1.0% FS, 24 hours
Response Time: Adj. 0.5, 1, 2, 3, 5 seconds to 90%
Linearity: ± 1% FS with optional linearizer
Comments: Multiple range options
Manufacturer: Horiba
Make: VIA
Model: 500
Type: Extractive NDIR
Ranges Available: 50 ppm to 1000 ppm
Precision: ± 1% FS
Accuracy: Not specified
Drift: ± 2%, 24 hours
Response Time: Adj. 0.5-16 seconds
Linearity: Not specified
Interferences: Interference rejection ratios sped fied
for many gases
Comments: Dual range with range ratios of 1:2 or 1:2 5
67
-------
Manufacturer: Infared Industries
Model: IR-760
Type: Extractive NDIR
Ranges Available: 100 to 5000 ppm
Precision: ± 0.5% FS
Accuracy: Not specified
Drift: ± 2% FS, 7 days
Response Time: 20 seconds to 90%
Linearity: ± 2% FS
Comments: Other models (higher ranges) available. Dual
range models available.
Manufacturer: Land
Model: Land CO
Type: Cross-stack GFC
Ranges Available: Depends on path length
Precision: ± 1% FS
Accuracy: ± 4% FS
Drift: Not specified
Response Time: Not specified
Linearity: Not specified
Comments: Single pass
68
-------
Manufacturer: MSA
Model: Mini-CO
Type: Portable extractive electrochemical
Ranges Available: 2000 ppm
Precision: Not specified
Accuracy: Not specified
Drift: Span ± 2% FS, zero ± 1% FS, 24 hours
Response Time: 30 seconds to 90% FS
Linearity: ± 2% FS
Manufacturer: Syconex
Model: 6000
Type: Cross-stack GFC
Ranges Available: Depends on path length
Precision: ± 1% FS
Accuracy: ± 2% FS
Drift: ± 1% FS, 30 days
Response Time: Adj. 5 seconds to 95% typical
Linearity: ± 1% FS
Convents: Double pass
69
-------
Manufacturer: Teledyne Analytical Instruments
Model: 9300
Type: Extractive NDIR,
Ranges Available: 500 ppm up to 7.
Precision: ± 1% FS
Accuracy: ± 2% FS at constant temp, ± 5% over operating
temperature limits
Drift: Not specified
Response Time: 5 seconds to 90% FS
Linearity: ± 1% FS
Manufacturer: Teledyne Analytical Instruments
Model: MAX 5, Portable
Type: Extractive electro-chemical
Ranges Available: 1000 ppm
Precision: ± 1% FS
Accuracy: ± 2% FS
Drift: Not specified
Response Time: <30 seconds to 90% FS
Linearity: Not specified
70
-------
Manufacturer: Whittaker
Model: P 310
Type: Extractive electro-chemical
Ranges Available: up to 150,000 ppm (15%)
Precision: ± 1% FS
Accuracy: ± 2% FS
Drift: ± 2% FS, 24 hours
Response Time: 2 mInutes to 90% FS
Linearity: ± 1% FS
11
-------
I3WIP- 50052
APPENDIX B
INSTRUMENT MANUFACTURERS LIST
72
-------
ANARAD, Inc. Extractive NDIR
534 E. Ortega Street Other-02, C02, NOx
Santa Barbara, CA 93103 THC, H2S
(805) 963-6583
AMETEK, Inc. Thermox Instruments Div. Extractive NDIR, Cross-stack
150 Freeport Road GFC (NDIR
Pittsburgh, PA 15238 Other-02
(412) 828-9040
Bailey Controls Co. Extractive catalytic
29801 Euclid Avenue oxidation
Wickliffe, OH 44092 Other-02, combustibles
(216) 585-8500
Beckman Industrial Corp. Extractive NOIR
600 S. Harbor Blvd.
La Habra, CA 90631
(213) 690-7600
CEA Instruments Extractive NDIR
16 Chesnut Street Other-C02, CH4
P. 0. Box 303
Emerson, NJ 07630
(201) 967-5660
Dynatron, Inc. Cross-stack GFC (NDIR)
P. 0. Box 745 Other-Opacity, SO2
Wallingford, CT 06492
(203) 265-7121
Esterline-Angus Instrument Corp. Extractive NDIR
P.O. Box 24000
Indianapolis, IN 46224
(317) 244-7611
Horiba Instruments, Inc. Extractive NDIR
1021 Duryea Avenue Other-C02, S02, CH4,
Irvine, CA 92714 C3H8, C6H14, NH3
(714) 250-4811
Infrared Industries, Inc. Extractive NOIR
P. 0. Box 989 Other-02, NOx, S02, CH4
Santa Barbara, CA 93102
(805) 684-4181
73
-------
Interscan Corp.
21700 Nordhoff
P. 0. Box 2496
Chatsworth, CA 91311
(818) 882-2331
Land Combustion, Inc.
3392 Progress Drive
Suite E
Bensalem, PA 19020
(215) 244-1100
MSA
600 Penn Center Blvd.
Pittsburgh, PA 15235
(412) 273-5000
Process Analytics
Combustion Engineering, Inc.
p. 0. Drawer 831
Lewisburg, WE 24901
(304) 647-4358
(Formerly Bendix Process Instruments)
Syconex Corp.
1504 Highland Avenue
Duarte, CA 91010-2831
(818) 359-6648
Teledyne Analytical
16830 Chesnut Street
p. 0. Box 1580
City of Industry, CA 91749-1580
(213) 283-7181
(818) 961-9221
Thermo Electron Instruments
108 South Street
Hopkinton, MA 01748
(617) 435-5321
Western Precipitation (Joy Division)
4565 Colorado Blvd.
Los Angeles, CA 90039
(818) 240-2300
Cross-stack GFC (NDIR)
Other-C02, HC, S02,
H20, HCJ
Extractive NDIR
Extractive electro-chemical
Other-02, C02, H2S, THC
Extractive GFC (NOIR)
Cross-stack GFC
Other-S02, NO, C02,
H20, Opacity
Polargraphic (extractive)
Polarographic (extractive)
Cross-stack GFC (NDIR)
Extractive NDIR
Extractive NDIR
74
-------
Western Research and Development Extractive NDIR
#3 1313 44th Avenue, N.E.
Calgary, Alberta, Canada T2E 6L5
(403) 276-8806
Westinghouse Electric Cross-stack GFC (NOIR)
Combustion Control Division Other-02, combustibles
P. 0. Box 901
Orrville, OH 44667-0901
(216) 682-9010
Whittaker Corp. Extractive electro-chemical
12626 Raymer Street Other-S02, NOx, 02
North Hollywood, CA 91605-4307
(818) 765-6622
75
-------
APPENDIX C
EXAMPLE MONITORING SYSTEMS AND COMPONENT LISTS
76
-------
HE TD S NPLE I D E
RDiC T
• . • • . . •. .
I CCWTROL SIG1.
LDES
I. ’
SYSTEM COMPONENT LIST
ITDI NO.
INCOICL PROBE WIN DISPCSABL ILT $
PRESSURC-RECUL*TIWC
341S ID TUI.CK 1UBI P WI N TEMP. CCW7ROL WERT 1 MC)CT
c3
w,L.DI* w o i -ut . us U
1
C L SCING FILT • CCPSIPmTIGN PUST/PRRTICLE
c
111 UOL 14E*’D CCP1P RTPDIT WITh ST IWLESS ST L SAMPLE L3)C
TULON F! L1 HOLD FOR PD0RAIC F!
c r P1OELECTRI C COOL
3 CUSTOM-BLHLT TEMP CONT OLLXD IM7t R TD S rIPLINC C PCT
Ic iiri P LYSIS CPLIBRATION SE3 IN UJMIMJM CVLIP S
: :
P MJOL 3-I Y VALUES
C RSC
El LT ./PRO3E
SAMPLE JMP
. —1 LCWRAN
I CO
E 1 A PRQ E$SIP G
I oxv I c 1
P LY ] I
77
-------
D
D
D
D
D
SOJRCEI’XTECTCR
MO JLE AI’O HOUSING
a)
js. ills... S SSSIIII1S1SSIISI SSSISUSIISIIISSI 1111151111111 IllS su
CLECTRICRL LIP (SIO’ L *10 CONTROL)
I :
... :
S SI:
i—i CO *P LY
U CONTROL MODULE : 1.151 1 1 ?
1—i
U CONT Qi.I1QWLZ cD
TA PROCESSING
IP4TSTRU1V T CA3IPCT
SYSTEM COMPONENT LIST
I7 N .
)DG I7TI I
D
RETRCR LECTCR IN SOLT-ON IDTh PROOF HOUSING
Co *PiPLY SOURCE DETECTOR MODULE IN POLT-ON WThRPROO HOUSING
02 *ILV IN-SFTU PROBE/SOSOR I$IIT
UP ILJC L L11C WITh PCILR. CONTROL *10 OUTPUT CCPCUCTCRS
R*CK-TYPC 1W tGRA TD 1PS’RUPDIT CPB:pET
FUEL, *1R, *10 Ir S1t4 CUT-OFT PROCESS CCNTROLL S
POA1iPLANcES
RErROR LECTOR
MODULE *10 HOUSING
PROCESS
CONTROL L S
a)
78
-------
(3wiP—s OO .?I)
APPENDIX 0
QUALITY ASSURANCE AND PERFORMANCE SPECIFICATIONS
FOR CONTINUOUS MONITORING OF CARBON MONOXIDE
AT HAZARDOUS WASTE INCINERATORS
79
-------
1. Introduction
According to the RCRA regulations (40 CFR Part 2 64 .346(a)(2)), carbon
monoxide “must be monitored on a continuous basis at a point in the in-
cinerator downstream of the combustion zone and prior to release to the
atmosphere”. Another part of the rule (264.345(b)(J)) states that the
permit will contain a limit for the CO level in the stack exhaust gas.
These two requirements were adopted to provide the agency with some means
of evaluating the combustion efficiency of an operating hazardous waste
incinerator (HWI) at any point in time. The regulation do not define what
constitutes continuous monitoring, nor do they define what instrumentation
is acceptable for determining compliance with the rules. Owners and oper-
ators of HWIs and the regulatory personnel (EPA regional office and state)
have not had any specific guidelines for assessing the adequacy of moni-
toring systems.
In order to ensure that the regulatory intent is observed, there must
be some consistency to the application of continuous monitoring require-
ments. This set of specifications and procedures attempts to provide
this consistency. The technical objectives are to ensure that CO monitoring
data are of known quality and that the monitoring system is representing
the operation of the incinerator.
These guidelines provide a set of performance standards for the moni-
toring systems, describe procedures for comparing the systems to the
standards, and suggest procedures to maintain the performance level during
continued operation. A glossary and a set of definitions are provided to
convey the technical foundation for the performance specifications so the
80
-------
reader can apply the principles if a specific case warrants deviation
from the prescribed procedures. For example, several of the terms used in
the performance specifications are also used by the manufacturers to report
equipment performance. Sometimes test conditions are specified but the
reported or test conditions differ from company to company.
There are three basic sections to this guideline which correspond to
three phases of evaluating a monitoring system. The first of these is a
set of equipment specifications and criteria which can be used for a
“paper evaluation of analyzers and other components of the proposed or
considered monitoring system; the second is the performance verification
phase where actual tests are conducted to verify that the operating moni-
toring system does what it is supposed to do; the third section deals
with maintaining the system and regularly checking one or more “indicator”
parameters to ensure that it continues to perform as intended.
2. Definitions and Glossary
2.1. Continuous Emission Monitoring System (CEllS) . The CEllS is con-
sidered to be all the equipment used to generate data and includes the
sample extraction and transport hardware, the analyzer(s), and the data
recording/processing hardware (and software).
2.2. Continuous . A continuous monitor is one in which the sample to be
analyzed passes the measurement section of the analyzer without interrup-
tion, and, which evaluates the detector response to the sample at least
once each 5 seconds and provides an output signal to the data recorder or
meter at least once each 30 seconds.
81
-------
2.3. Monitoring system tvDes . There are three basic types of monitoring
systems-extractive, cross-stack and in-situ. Cross-stack and in-situ
systems perform the analysis without removing a sample from the stack and
an extractive system withdraws a sample from the stack for remote analysis.
2.3.1. Extractive . Extractive systems use a pump or other mech-
anical, pneumatic, or hydraulic means to draw a small portion of the
stack or flue gas and convey it to the remotely-located analyzer.
2.3.2. Cross-stack . Cross-stack analyzers measure the parameter
of interest by placing a source beam on one side of the stack and either
the detector (in single-pass instruments) or a retro-reflector (in double-
pass instruments) on the other side and measuring the parameter of interest
(S02, NOx, opacity, etc.) by the attenuation of the beam by the gas in its
path.
2.3.3. In-situ . In-situ analyzers place the sensing or detecting
element directly in the flue gas stream.
2.4. Span . Span or span value is the upscale or positive instrument
reading at or near but not exceeding the full-scale reading of an instru-
ment.
2.5. Calibration Standards . Calibration standards are quantities of
materials with known and relatively unchanging values for the measurement
parameter of interest. The values for the parameter are known with a
specified degree of accuracy and precision.
2.5.1. Gases . A calibration gas contains a specific concentration
of the analyte gas (e.g. CO or 02) in an appropriate diluent gas (e.g. N2
or air), usually supplied in high-pressure cylinders. The diluent is chosen
so as not to interfere with the analytical technique and the gases are
82
-------
available in a wide range of concentrations and different quality grades.
2.5.2. Gas Cells and ODtical Filters . These are calibration devices
for optical measurement paths which when placed between the source and
the detector produce a predictable response which is or can be correlated
to a specific analyte concentration. Gas cells are usually sealed chambers
filled with a calibration gas, although there are flow-through cells.
Optical filters are solid materials designed to transmit (or absorb)
electromagnetic energy of a specific wavelength or spectral composition.
2.5.3. Standard Reference Materials (SRMsL This term is used to
describe a particular class of calibration standards which has been cer-
tified by the National Bureau of Standards as to the properties of inter-
est.
2.5.4. Other . There are several ancillary standard materials that
may be used in system evaluations. These include audit samples used to
test analyst proficiency or analyzer performance, electronic signal gen-
erators to test certain aspects of instrument response, pressure and vacuum
gauges, and temperature measuring standards.
2.6. Instrument Range . The maximum and minimum concentration that can
be measured by a specific instrument. The minimum is often stated or
assumed to be 0 and the range expressed only as the maximum. If a single
analyzer Is used, either manually or automatically, on multiple ranges,
the performance standards expressed as a percentage of full scale applies
to all ranges.
83
-------
2.7. Drift (Zero or SDan1 . Drift is change in the response or output
of an Instrument from a set or calibrated value over time. Drift is meas-
ured by comparing the response to a calibration standard over time with
no adjustment of instrument settings.
2.8. ResDonse Time . The response time of a system or part of a system
is the amount of time between a step change in the system input (e.g.
change of calibration gas) until the data recorder displays 95% of the
final value.
2.9. Precision . Precision is the agreement among individual measurements
of the same property, usually under prescribed similar conditions and is
usually expressed as the standard deviation of a group of measurements.
The units, for this specification, are % of full scale. Repeatability,
reproducibility, and replicability are terms used to describe precision,
depending on the descriptive statistic used.
2.10. Accuracy . Accuracy is a measure of agreement between a measured
value and an accepted, or true, value and is usually expressed as the
percentage difference between the true and measured values relative to the
true value. For this specification, accuracy is measures by the linearity
test, i.e. accuracy at 1 particular measurement value.
2.11. Linearity . Linearity is a measure of the deviation of a measured
value from a value predicted by a straight line drawn between the zero
calibration and full-scale or span calibration points when calibration
value and analyzer response are plotted on rectangular coordinates. For
this specification, linearity is measured at mid-scale and is a measure
of accuracy at this level.
84
-------
2.12. Interferent . An interferent is a component of a gas stream which,
when present, yields an instrument response when the desired effect is no
response.
3. Monitoring System Performance and Equipment Soecifications
This section describes the characteristics of monitoring system perfo-
rmance that are consistent with the regulations and with sound engineering
judgment. The specifications are attainable with systems and components
available from a number of commercial vendors of emission monitors and
permit the use of different analyzer types. Table 3.1 summarizes the
standards and each of items are discussed in the following paragraphs. Two
sets of standards are given-one for low range measurement and one for
high range measurement. The high range standards relate to measurement
and quantification of short-duration high-concentration peaks and the low
range standards relate to the overall average operating condition of the
incinerator. The dual-range specification may require the use of dual
analyzers or at least dual range units as most commercial units do not
appear to have the capability of meeting both standards with a single
unit.
3.1. Measurement Range. In order to measure both the high and low
concentrations specified in the draft CO limits with the same or similar
degree of accuracy, it is necessary to establish specifications appropriate
to the concentration values and the time averaging periods. The ranges
are the full-scale measurement range of the instrument. With this approach,
time-weighted averages can be calculated and status relative to the limits
determined with assurance.
85
-------
Table 3.1
Summary of CO Monitoring Systems
Performance Specifications
Parameter
Specifications
Low range High range
Measurement Range, ppmv
Monitor Location
System Response Time
Drift, Zero (as operating)*
(guide only)
Drift, Span*
Precision*
Lineari ty*
0-200 max 0-1000 mm
No numerical standard
1.5 minutes to 95% FS
10 ppmv, 20 ppmv,
24 hour 2 hour
20 ppmv, 50 ppmv,
24 hour 2 hour
the lesser of the lesser of
5.0% FS or 2.0% FS or
10 ppmv 20 ppmv
the lesser of the lesser of
10.0% FS or 5.0% FS or
20 ppmv 50 ppmv
* expressed as the sum of the mean absolute value plus the 95% con-
fidence interval of a series of measurements
86
-------
3.2. Monitor Location. The recommended location for the CO analyzer,
if in-situ or cross-stack, or for the sample extraction probe for extrac-
tive systems is in the exhaust stack at least 2 equivalent duct diameters
downstream from any source of dilution or in-leakage.
3.3. Response Time. The specification for response time is based on
providing at least 6 distinct periods (or values) to comprise the ten-
minute average. Six ten-minute values can then be used to determine the
60-minute average.
3.4. Drift. The drift specifications also reflect consideration of the
averaging times. The long-term (24 hour) calibration and zero drifts are
not particularly important to peak measurements; thus the 24-hour drift
limits are proportionally less restrictive than the 2-hour limits. Each
set of drift tests, as described in the next section, requires several
repetitions of the test and permits one failure to meet the specification.
The zero drift values are presented as a guide to operational performance;
a monitoring system which does not meet the guideline during the perfor-
mance test should not be rejected for this reason only. The regulations
are not concerned with precise measurements near zero; the concern is for
elevated levels of CO. If the instrument exhibits excessive zero drift
but maintains performance at the upper levels, the operation is acceptable.
This is more likely to occur with systems with non-linear response but
inmost cases, zero drift causes a corresponding span drift.
3.5. Precision. All of the performance specifications involving com-
parison of one measurement value with another are interrelated. An analyzer
with a drifting span or calibration will be neither very precise nor very
87
-------
accurate. The precision specification presented here is repeatability and
is defined as the variability among measurements of the same sample with
the same analyzer at different times and is calculated as described in
5.3. The precision limits are different for the different ranges, both in
the relative and the absolute senses, reflecting the intent of the CO
limits and averaging times, and they are consistent with the drift specifi-
cation. Thus, the analyzer system Is permitted to exhibit a certain amount
of variation about some central value whether this variation is a result
of drift or random changes in the instrument response.
3.6. Linearity. The linearity specification evaluates the accuracy at
the mid-point of the measurement range. The common procedure for analyzer
calibration is to set a zero response with a zero gas and a full-scale
response with a span gas(or cell or filter), adjusting the display or
meter reading to the known concentration of the span value. The linearity
tests in Section 4 evaluate the accuracy of the system at the mid-point
of the measurement range by determining the difference between the measured
value and the expected value at this point.
4. Ouality Assurance-Monitoring System Performance Tests
4.1. General. This section describes the procedures to be used in
evaluation of CO monitoring systems at hazardous waste incinerators. The
procedures are adapted, for the most part, from Gaseous Continuous Emission
Monitorjno Systems - Performance Specification Guidelines for S02. NOx.
C02 02. and TRS , EPA Publication No. 450/3-82-026, October 1982, with
modification deemed necessary to apply the general procedures to HWIs.
88
-------
4.2. Location. The best or optimum location of the sample interface
for the monitoring system is determined by a number of factors, including
ease of-access for calibration and maintenance, the degree to which it
represents total emissions, and the degree to which it represents the
combustion situation in the firebox. As previously mentioned, the location
should be as free from In-leakage influences as possible and reasonably
free from severe flow disturbances. The sample location should be at
least 2 equivalent duct diameter downstream from the nearest control
device, point of pollutant generation, or other point at which a change
in the pollutant concentration or emission rate occurs and at least 0.5
diameters upstream from the exhaust or control device. The equivalent
duct diameter is calculated by Equation 5.1.
The sample path (for cross-stack monitors) or sample point(s) (for
extractive or in-situ monitors) should include the concentric inner 50%
of the stack or duct cross section. For circular ducts, this is .707 x
diameter and a single-point probe, therefore, should be located between
.141*diameter and .839*diameter from the stack wall and a multiple-point
probe should have sample inlets in this region. A location which meets
both the diameter and the cross-section criteria will be acceptable.
If these criteria are not achievable or if the location is otherwise
less then optimum, the possibility of stratification should be investi-
gated. To check for stratification, the concentration of CO should be
measured at several points in the duct at the potential location. The
concentration of 02 or C02 should also be measured as verification of
diluent leakage. For rectangular ducts, at least nine sample points located
at the centroids of similarly-shaped, equal area divisions of the cross
89
-------
section should be used. For circular ducts, twelve sample points (i.e.,
six points on each of the two perpendicular diameters) should be used,
locating the points as described in 40 CFR 60, Appendix A, Method 1.
Calculate the mean value for all sample points and select the point(s) or
path that provides a value equivalent to the mean. For these purposes, if
no single value is more than 15% different from the mean and if no two
single values are different from each other by more than 20% of the mean,
then the gas can be assumed homogeneous and can be sampled anywhere. The
point(s) or path should be within the inner 50% of the area.
4.3. Response Time. The response time tests as here described apply to
all types of monitors, but will generally have significance only for
extractive systems. The entire system is checked with this procedure
including sample extraction and transport (if applicable), sample con-
ditioning (if applicable), gas analyzer, and the data recorder.
Introduce zero gas (or zero cell or filter) into the system. For extrac-
tive systems, the calibration gases should be introduced at the probe as
near to the stack as possible. If any form of multiplexing” is being
used, the response time tests should be performed for each path in the
system. For in-situ systems, introduce the zero cell at the sample inter-
face so all components active In the analysis are tested. When the system
output has stabilized ( no change greater than 1% of full-scale for 30
seconds), switch to monitor stack effluent and wait for a stable value.
Record the time (upscale response time) required to reach 95% of the
final stable value on a data form like Figure 4. Next, introduce a high
level calibration gas (or cell or filter) and repeat the above procedure
(stable, switch to stack, stable, record). Repeat the entire procedure
90
-------
three times and determine the mean upscale and downscale response times.
The slower or longer of the two is the system response time.
4.4. Zero and Span Drift. The purpose of the drift checks is to deter-
mine the ability of the CEMS to maintain its calibration over a specified
period of time. The performance specifications establish a drift standard
related to span drift for the two analyzer ranges and suggest operating
values for zero drift. The long-term averaging (1 hour) corresponds to
the 24-hour drift and the 10-minute averages correspond to the 2 hour
drift. Each drift test is conducted several times and the system(s) are
allowed to exceed the limit once during the test. The monitoring system
should be operated for some time before attempting drift checks because
most systems need a period of equilibration and adjustment before the
performance is reasonably stable. At least one week (168 hours) of con-
tinuous operation is recommended before attempting drift tests.
Drift checks and precision and linearity checks should all be conducted
using the entire monitoring system. For extractive and in-situ systems
using calibration gases, this would include all sample conditioning equip-
ment, including external filters and moisture traps. For cross-stack
systems using calibration cells or optical filters, the measurement path
should be as nearly the same as the stack path as possible and include
lenses, optical filters, mirrors, beam splitters, etc., and the electrical
and electronic signal processing and control circuitry, including pressure
and temperature and other compensation circuits. During the drift tests,
no adjustment of the system is permitted except those automatic internal
adjustments which are part of the automatic compensation circuits integral
91
-------
to the analyzer. Subsequent CEll operation must include the same configur-
ation as used during the performance testing.
4.4.1. Two-hour drift. Select a span gas with a CO concentration
between 80 and 100% of the full-scale range of the analyzer. However,if a
range higher than 200 ppm is used, the span gas should be 160-200 ppmv
CO. The zero gas should have less than 1 ppm CO. At the beginning of the
test, zero and span the analyzer using the selected gases (or cells or
filters). After two hours and at two-hour intervals thereafter, alternately
introduce both zero and span gases, wait until a stable reading (Sec.
4.3) is obtained and record the values reported by the system. Repeat
this procedure for 12 hours, obtaining six values of zero and span gas
measurements. The procedure may use more than 12 hours and the two-hour
periods need not be consecutive but may not overlap. The difference between
the established or reference value for the span and the measured value
may not exceed the specifications in Table 3.1 more than once and the
mean of the absolute values of the differences plus the 95% C.I.(confidence
interval) must be less than the limit of Table 3.1. Calculate the results
according to Eqs. 5.2.n. The zero drift should, but is not required to, be
within the limits of Table 3.1
4.4.2. Twenty four-hour drift. The procedure for determining
compliance with the 24-hour drift limits is essentially the same as the
2-hour test, except that the time intervals are 24 hours instead of 2
hours. The system is calibrated initially and at least 24 hours later,
and at subsequent 24-hour intervals, zero and span gases (or cells or
filters) are consecutively introduced into the measurement path. The 24-
92
-------
hour periods may not overlap but may extend longer than 24 hours. Deter-
mination of compliance follows the same procedures and one of the six
values may exceed the performance specification but the average must not.
4.5. Precision. The precision test (one aspect of a calibration error
test) is much like the span drift tests, except that the instrument is
presumed to be at a steady-state condition. The test consists of repeated
analysis of the span gas (or cell or filter) with intervening changes in
the measurement input. The precision and linearity tests (Section 4.6)
can be conducted simultaneously by using a mid-scale calibration standard
as the intervening input change. The analyzer is calibrated with zero and
span standards. After the introduction of the span standard, a lower
concentration standard is introduced to the measurement section, and then
the span standard is reintroduced. This process (mid, span) is repeated 6
times, recording the instrument response to the span value. The average
deviation (Eq. 5.3) must be less than the precision limits of Table 3.1.
If the linearity test (Section 4.6) is performed at the same time, there
are additional requirements.
4.6. Linearity. The performance specifications are not intended to
eliminate instruments which do not have a linear response, but rather to
set limits on the allowed deviation between the actual value and the
predicted value. The basic procedure for testing linearity (or calibration)
error is to set the instrument zero and span with the appropriate standards
and then repeatedly measure a standard in the middle of the range. In
order to minimize bias from previous analyses, the sequence of standard
introduction should alternate between high and low standards prior to the
mid-level standard (e.g high, mid, low, mid, high, mid, low, mid, etc.)
93
-------
until 6 analyses of the mid-level standard are obtained, with three values
obtained from upscale approach and three values obtained from downscale
approach. If the precision and linearity tests are conducted at the same
time, additional high standard measurements need to be made and they can
be preceded by either low or mid-scale standards. Calculation of deviation
from linear response is done with Eq 5.4. If the system response is in-
tended to be non-linear, a multiple-point calibration is usually performed
to generate a calibration curve. The manufacturer most often provides
either an instrument-specific calibration curve or a recommended procedure
to generate one. Depending on the nature of the response of the instrument,
a 50% standard may yield a low response (e.g. 30 % of full-scale) or a
high response (i.e. 70 % of full-scale) and the calibration points are
concentrated in the region of the greatest rate of change. In these cases
the calibration error test standard should be greater than 45 % of the
instrument range and between the manufacturer’s recommended span and
intermediate calibration standards.
5. Equations
5.1. Equivalent Duct Diameter. To determine the round duct equivalent
diameter of a rectangular duct
2 * L *
D = _____
C (Liw)
where DC = equivalent diameter, I = length and W = width
5.2. Drift Calculation. Use the following equations to calculate system
drift for comparison to the performance specifications
di = IMVI-RVI
94
-------
where di = absolute value of individual drift measurements, MV1 =
Individual measured values of standard, and RV = reference value of
standard
1 n
d.
1
n i —i
where arithmetic mean of the absolute values of the drifts and
n — number of measurements (usually 6)
21 fl 1
X d. _A( d.) 2 “2
1=1 1 1=1 1
Sd — ________________________
n-i
where Sd standard deviation of the d 1 values
Sd
95 % C.!. — t * ______
In—
where 95% C.!. — the confidence interval for the data at a prob-
ability of 2.5% (two-tailed) and t values are the critical t values
for the Student’s “t” test. Some t values are presented below.
— t
3 4.303
6 2.571
12 2.201
* n values are number of samples; t values are corrected
for number of degrees of freedom
Drift = d ± 95 % C.!. (ppm)
95
-------
5.3. Precision Deviation. Use the same approach for calculating preci-
sion as for drift, using the individual differences between the measured
values and the reference value as the data points.
5.4. Linearity Deviation. The same calculations are also used for the
linearity checks, using the differences between the measured instrument
output and expected output for the reference standard as the data points.
6. References
6.1. Jahnke, James A. and G.J. Aldina, Handbook Continuous Air Pol-
lution Source Monitoring Systems , U.S. Environmental Protection Agency
Technology Transfer, Cincinnati Ohio 45268, EPA 625/6-79-005, June 1979.
6.2. Gaseous Continuous Emission Monitoring Systems - Performance
SDecification Guidelines for S02 NOx. C02 . 02. and TRS , U.S. Environmental
Protection Agency OAQPS/ESED, Research Triangle Park, North Carolina,
27711, EPA-450/3-82-026, October 1982.
6.3. Quality Assurance Handbook for Air Pollution Measurement Systems
Volume I. Princiøles , U.S. Environmental Protection Agency ORD/EMSL,
Research Triangle Park, North Carolina, 27711, EPA-600/9-76-006, December
1984.
6.4. Michie, Raymond M. Jr. et al, Performance Test Results and Comoa-
rative Data for Designated Reference Methods for Carbon Monoxide , U.S.
Environmental Protection Agency ORD/EMSL, Research Triangle Park, North
Carolina, 27711, EPA-600/S4-83-013, September 1982.
6.5. Ferguson, B.B., R.E. Lester, and W.J. Mitchell, Field Evaluation
of carbon Monoxide and Hvdro en Sulfide Continuous Emission Monitors at
an Oil Refinery , U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina, 27711, EPA-600/4-82-054, August 1982.
96
------- |