logo.gif (9477 bytes)
Back to the RMB Consulting & Research, Inc. Home Page
Index of all news articles posted to the RMB website
News specific to Compliance Assurance Monitoring
Information about RMB's Training Programs
Access to RMB's FTP Library
Read about RMB's current projects
Links and Resources
RMB's primary service areas
RMB Consulting & Research, Inc. corporate information
Send email to the RMB webmaster

Status of EPA's Continuous Particulate Mass (PM) Monitor Demonstrations

Ralph Roberson
RMB Consulting & Research, Inc.
5104 Bur Oak Circle
Raleigh, North Carolina 27612

Presented at the 1997 EPRI CEM Users Group Meeting
Denver, Colorado
May 14-16, 1997

Introduction and Overview

On April 19, 1996, EPA's Office of Solid Waste (OSW) proposed a rule in the Federal Register that would revise the national emission standards for hazardous waste combustors1. Among other things, the proposed rule would require the installation and operation of continuous particulate matter (PM) and continuous mercury (Hg) monitors. To support the continuous emission monitor (CEM) requirements of the proposed rule, EPA's proposal also includes several new CEM performance specifications: Performance Specification 11 (PS 11) -- specifications and test procedures for particulate matter continuous monitoring systems and Performance Specification 12 (PS 12) -- specifications and test procedures for total mercury continuous monitoring systems.

To develop technical support for the CEM requirements of the proposed rule, the Agency published, in February 1996, a notice in the Federal Register soliciting proposals from vendors that were interested in participating in the Agency's field demonstration of continuous PM and Hg monitors. The following six manufacturers responded to EPA's solicitation for continuous PM monitors: (1) Verewa, (2) Emissions SA, (3) Environmental Systems Corporation (ESC), (4) Sigrist, (5) Durag, and (6) Jonas, Inc. EPA's ongoing field demonstration of continuous PM monitors began in September 1996 and is being conducted on a hazardous waste incinerator at the Dupont Research Facility located in Wilmington, Delaware. The field demonstration is scheduled to conclude in May 1997. EPA recently published a Federal Register notice announcing the release of interim results from its continuous PM monitor demonstration and drawing some tentative conclusions from those results2. The purpose of this paper is to discuss those results and to provide comments on EPA's preliminary conclusions.

Regulatory Significance

The electric utility industry is not directly affected by the above-discussed EPA proposal, which would require the installation and operation of continuous PM monitors. However, we believe this rulemaking merits attention because of the general implications on CEM technology, and more importantly, because of EPA impending compliance assurance monitoring rule (CAM). EPA's CAM rule is well beyond the scope of this paper. However, suffice it to say we believe the CAM rule, when finalized, will require all utilities that have plants with enforceable particulate emission limits to periodically make a declaration that their plants were either in continuous or intermittent compliance with those particulate emission limits. We believe that a reliable and accurate continuous PM monitor could be one of several technological options available to the utility industry to make such determinations and declarations.

EPA's Initial (1995) Field Test Programs

Prior to the initiation of its current field demonstration program, EPA conducted two additional, but limited, field test programs of continuous PM monitors. In 1995, test programs were conducted at (1) a Rollins hazardous waste incinerator, located in Bridgeport, New Jersey and (2) a LaFarge cement kiln, which co-fires hazardous waste and is located in Fredonia, Kansas.

The Rollins Tests

According to EPA, the objectives of the Rollins tests were (1) to determine if technical information could be developed to support CEM vendor claims that continuous PM monitors could be used to assess compliance with emission standards and (2) to obtain field experience that would be useful in designing future field testing programs.

Three devices were tested at Rollins: a Sick RM2000 light-scattering instrument, a BHA CPM1000 time dependant optical transmission instrument, and an Emissions SA Beta 5M beta-gauge instrument. There were a number of problems with the field tests that make quantitative comparisons of the Rollins data to other field results difficult. For example, valid instrument calibrations could not be developed because reference method test data could be obtained for only two particulate concentrations. Also, the range of particulate emissions during testing was less than one-third of EPA's proposed PM standard for hazardous waste combustors, and only eight valid reference method tests could be performed. Fortunately, or unfortunately, the results from Rollins do confirm that light sensitive continuous PM monitors, such as light-scattering and time dependent optical transmission instruments, have a step function increase in their responses when entrained water droplets are present in the flue gas.

The LaFarge Tests

A second field test program was conducted at a LaFarge cement kiln during May 1995. EPA's stated purpose for the LaFarge tests was to conduct a full calibration of the instruments in accordance with the International Standards Organization specification (i.e., ISO 10155) to assess further whether continuous PM monitors could be used to determine compliance with particulate emission limitations and to determine if ISO 10155 could be used as a basis for the Agency's proposed continuous PM monitor performance specification.

Two instruments were tested at LaFarge: the Sick RM200, which was also tested at Rollins, and an ESC P5A. The measurement principle of both instruments is light scattering. EPA was able to calibrate both instruments by obtaining a total of nine reference method runs. The runs were conducted at three different PM concentrations. EPA could draw the following conclusions from the LaFarge tests.

•Response of the instruments to changing PM concentrations was better than for the Rollins tests.
•The ISO specifications could be used as a basis for developing performance specifications; however, the LaFarge tests do not meet the statistical criteria of ISO 10155.
•The current PM reference method (i.e., EPA Method 5) exhibits significant variability when measuring low PM concentrations, primarily because of probe washing requirements and the difficulty in filter recovery.

EPA's Ongoing Field Demonstration

As previously stated, EPA's ongoing field demonstration began in September 1996 and is being conducted on a hazardous waste incinerator at the Dupont Research Facility located in Wilmington, Delaware.

Description of the Dupont Test Site

Of the three types of facilities affected by the proposed requirement to install and operate continuous PM monitors (i.e., incinerators, cement kilns, and light-weight aggregate kilns), EPA selected an incinerator because incinerators tend to burn a wide variety of waste as their primary feedstock. EPA believes a wide variation in feedstock is most likely to result in PM emissions with highly variable characteristics, which, in turn, represents the most challenging tests for continuous PM monitors. Also, hazardous waste incinerators tend to utilize wet scrubbers, which result in flue gases with high moisture concentrations - yet an additional challenge for continuous PM monitors.

EPA decided to conduct its long-term field demonstration at the Dupont facility because:

•Dupont was not only willing to participate but was willing to make necessary modifications to the site to facilitate the demonstration.
•The incinerator is equipped with a sampling platform capable of supporting six continuous PM monitors as well as the equipment and personnel required to conduct the reference method testing.
•The incinerator receives many different types of waste from a number of Dupont facilities and has the capacity to burn many waste streams simultaneously.

The Dupont incinerator uses a ram feeder for solid waste, a cylindrical chute for batched waste material, and also has a Trane Thermal liquid/gas waste burner. The Dupont incinerator is equipped with a series of air pollution control technologies including an afterburner, spray dryer, venturi scrubber, neutralizing absorber, chevron-type mist eliminator and finally an electro-dynamic venturi (EDV). The EDV exhausts through a steel stack approximately 110 feet in height and 4 feet in diameter at the location of the continuous PM monitors.

Continuous PM Monitors Tested

The six continuous PM monitors being tested at the Dupont incinerator represent three different measurement technologies: light-scattering, beta gauge, and acoustic energy. Each technology and continuous PM monitor is described below.

Verewa. The Verewa continuous PM monitor extracts a sample from the stack under isokinetic conditions at a nominal stack flow rate. However, isokinetic sampling normally is not maintained as stack flow changes. The stack sample is diluted for high dust loadings (>200 mg/dscm) or if the stack gas is wet or saturated, otherwise, it is measured directly without dilution. The sample passes through a heated probe and sample line and is collected on a filter. A filter tape mechanism allows long duration operation and positions the filter in either a "measurement" or "sample" location. In the "measurement" location, the attenuation of beta particles from a carbon-14 source is measured. Each filter location used for sampling is measured before and after sampling: the difference between these two measurements is proportional to the mass of particulate sampled. According to Verewa, attenuation of the beta particles is minimally sensitive to the composition of the particulate, thus site-specific calibrations should not generally be necessary. The sampled gas is dried and the flow rate measured, thus allowing reporting on a dscm basis. The instrument uses a dual source/detector arrangement to allow measurement of the previous sample while acquiring the current sample. Zero and span calibration checks are carried out at programmable intervals. The zero check is performed by measuring the same location on the filter tape twice, in succession, without collecting a sample. The span calibration is checked using a radiation attenuator inserted into the measurement beam. A typical sample requires 5 minutes to collect, 2 minutes to perform the blank filter measurement, 2 minutes to measure the loaded filter, and about 1 minute of tape transport time. Using the dual source/detector configuration, measurements are typically reported every 7.5 minutes. These times are programmable, however, so different sample and reporting times can be obtained depending on the sample loading.

Emissions SA. The Emissions SA Beta 5M uses a heated sampling probe to obtain an isokinetic sample, and isokinetic sampling is maintained automatically. The sample is collected on a filter which, at the end of the sampling period is moved, by means of a continuous filter tape mechanism, to a measurement location between a carbon 14 beta particle source and a detector. The beta transmission through each blank filter is determined before sampling begins. The sampling duration is programmable and determines the mass concentration detection limit. At high PM concentrations, sampling time is kept short to prevent sampling excessive amounts of particulate, and is usually set at 2 minutes for typical applications. Analysis takes 6 minutes, and thus a measurement is made every 8 minutes. At the end of each sampling period, the probe nozzle is temporarily closed, opened, and closed again in order to re-entrain any particulate that may have deposited in the probe.

Environmental Systems Corporation. ESC's model P5A light scattering monitor monitors the back scattered light (180o) from an infrared light emitting diode (LED). The instrument is said to have a nearly constant response to particles in the 0.1 to 10 micron range and a measurement range of 0.0005 to 8 gr/dscf. The probe volume is located 4.5 inches from the end of a probe containing both the transmitting and receiving optics that is inserted into the flow through a standard flange. The probe is purged with air to keep the optics clean. Only one point of access to the stack is required. The instrument automatically carries out zero and span calibrations.

Sigrist. The Sigrist instrument is an extractive light scattering monitor. This device extracts a heated slipstream from the stack, a small portion of which is sampled in turn and passes through a scattered light photometer. The entire sample is then returned to the stack. The sampling rate is set to be isokinetic at a nominal stack flow rate, but the sampling rate is not constantly adjusted to maintain an isokinetic rate. Rather, a constant sampling rate is maintained. The photometer measures the light scattered at 15 from a incandescent bulb emitting over the range 360 to 2800 nm. A double beam compensation measuring method is used in which the light path is split, and the intensity of the reference path adjusted by an attenuator to equal the intensity of the measurement path. The amount of adjustment necessary is the output signal. This approach is intended to make the output independent of fluctuations or aging in the optical and electronic components, including the buildup of dirt on the optics. Sigrist claims that the instrument does not experience calibration and zero drift. Periodic cleaning and checks with optical filters supplied with the instrument should be performed every 6 to 12 months.

Durag. The Durag instrument is a light scattering monitor measures the light scattered by the PM at an angle of approximately 120 degrees. The light beam is generated by a halogen lamp (400-700 nm) modulated at 1.2 kHz., and the sample volume is located in a region 80 to 280 mm from the stack wall. The instrument is designed to perform automatic zero and span checks, and provides automatic compensation for dirt on the optics, although the optics are protected by an air purge system. Stray light from surface reflections of the transmitted beam is minimized through the use of a light trap mounted on the opposite side of the stack. The instrument is normally mounted directly on the stack wall, thus providing for an in-situ measurement. For applications where water droplets are expected, a hot bypass system can be purchased.

Jonas, Inc. The Jonas acoustic energy PM monitor uses shock waves caused by the impact of particles with a probe inserted into the flow to measure the particulate concentration. The device counts the number of impacts and also measures the energy of each impact. This information, coupled with knowledge of the flow velocity, allows calculation of the particulate mass. Since the probe distorts the flow, changes in flow velocity and particle size distribution will, in principle, change the instrument response. This instrument can also report real-time particle size distributions; however the robustness of these distributions has not been independently determined.

Discussion of EPA's Results to Date

The fundamental problem with EPA's proposed approach to continuous PM monitoring continues to be the fact that commercially available instruments, especially those that are based on the principal of light scattering, do not provide a direct measure of particulate mass emissions. Of course, by direct measure, we mean that the instrument must measure particulate mass and the volume of flue gas from which that mass of PM was sampled. As EPA observes in its Federal Register notice, the characteristics of the emitted particulate matter exhibit significant variability, and "this variability in the particulate properties causes a varied response from the PM CEMS2."

It is our considered opinion that the results of EPA's continuous PM monitoring demonstration clearly document that the technology and instrumentation are inadequate to achieve the Agency's stated purpose (i.e., to assess continuous compliance with particulate emission limitations). Moreover, as we discuss in more detail below, EPA cannot overcome this shortfall in technology by simply relaxing PS 11.

Overview of the Field Demonstration. On balance, we believe that EPA's experimental design for its continuous PM monitor field demonstration is quite good. First, the demonstration includes six individual instruments that utilize three distinctly different measurement principals: (1) light scattering, (2) beta gauge, and (3) acoustic energy. Second, by conducting a long-term (i.e., 8-9 months) field demonstration, the Agency is able to properly evaluate the actual operability and reliability of each continuous PM monitor. For example, EPA is able to document how much maintenance each instrument requires; how easy each instrument is to service; how responsive each vendor is to requests for service, spare part orders, etc. Also, one would expect that over the long-term demonstration period, the continuous PM monitors would be exposed to a wide variety of incinerator operating conditions and thus, challenged by a wide range in particulate emissions.

Specific Criticisms of the Field Demonstration. Our most serious criticisms of EPA's field demonstration of continuous PM monitors focuses on how the Agency has characterized and interpreted the results to date. For example, in the Federal Register notice, EPA characterizes the Dupont incinerator as "a worst-case facility for this demonstration test program2." Of course, EPA must recognize that the probability of selecting a single hazardous waste incinerator from a population approaching 100 units and having that incinerator be the true worst-case unit is very remote. Moreover, a number of other observations in the Federal Register notice contradict EPA's assertion of worst-case facility.

For example, EPA states that it has not been able to conduct any tests in the presence of entrained water droplets. This is a very important observation for several reasons. Of course, it provides strong evidence that the Dupont incinerator is not a worst-case facility. More importantly, the absence of water droplets makes it impossible to evaluate just how the various instruments would respond to such conditions. The statement that light scattering devices "see" water droplets as particulate matter should be irrefutable. We had hoped that this field demonstration would provide definitive quantification to the degree of interference a facility can expect from entrained moisture. We suspect the impact of entrained water droplets on the performance of beta gauge instruments is less well known than for instruments that utilize light scattering. Even if it can be postulated that entrained water droplets have little effect on a beta gauge instrument's ability to measure, clearly the presence of water droplets can be problematic for the extractive sampling probe as well as for the filter paper substrates. Given the number of hazardous waste incinerators that are equipped with wet air pollution control devices, we do not believe a field demonstration of continuous PM monitors can be complete without evaluating the impact of entrained water droplets on the performance of the instruments.

Another problem observed by EPA is the inability of the Dupont incinerator to emit particulate matter at or above its permitted limit of 0.08 gr/dscf. We realize this causes some difficulty given that two of the statistical criteria (confidence interval and tolerance interval) in proposed PS 11 are to be evaluated at the emission limit. However, the fact that the Dupont incinerator stack is dry and relatively clean makes it neither a "worst-case facility" nor a particularly challenging location to evaluate CEMS.

Comparison With Proposed Statistical Criteria. Table 1 summarizes the results presented in Table 2 of the Federal Register notice by presenting only the results for the initial calibration tests. Also, according to EPA's notice, the results in Table 2, "do not include data outliers which have been excluded from the analysis, such as paired data outliers2." The results presented in our Table 1 indicate that the continuous PM monitors are not performing well, even compared to the relaxed (relative to ISO 10155) statistical requirements of proposed PS 11. In fact, of the 15 statistical calculations (5 monitors times 3 tests each), only four meet the proposed criteria of PS 11. Hopefully, EPA will not interpret these results as a need to relax further the requirements contained in proposed PS 11.

Our final comment on the statistics of the continuous PM monitor demonstration deals with the y-intercept of the various linear regressions. Table 2 tabulates the regression equations for the five continuous PM monitors based on the equations presented on Figures 2-46 through 2-50 of Status Report IV3. All of these individual figures are titled, "... Cumulative Data Base without RSD Outliers." In the regression equations shown in our Table 2, y, the dependent variable represents the predicted Method 5 concentration (mg/dscm) based on x, the independent variable, which is the response of the continuous PM monitor. While it is encouraging that at least four of the five continuous PM monitors have very similar slopes (the coefficients of x), the magnitude of the y-intersect term is very troubling. The large y-intercept is troubling because it means that even as the response of each continuous PM monitor approaches zero, all the regression equations predict significant Method 5 concentrations. In fact, for a continuous PM response of 0, these regression equations predict Method 5 concentrations ranging from a low of 11.7 to a high of 18.9 mg/dscm.

Additional Comments on EPA's Field Demonstration

Spare Parts and Service. The discussion in the Federal Register notice titled, "Limitations of the Test Program" is puzzling2. EPA states that a CEMS purchased by a facility usually comes with a supply of spare parts. This simply is not true; most CEMS vendors include a recommended list of spare parts which the facility may or may not elect to purchase. Most importantly, EPA's stated inability to secure US-based trained service technicians speaks volumes about the maturity of the technology and its supporting infrastructure. The key question is, how many instruments of each type would have to be installed in the various geographic regions of the U.S. before the overseas vendors could justify a network of factory trained technicians?

Absolute Calibration Audits. In the Federal Register notice, EPA observes that NIST does not have traceable standards for continuous PM monitors. Following is a quote from a Technical Memorandum we prepared in August 19964:

" ... do not believe that EPA's proposed Appendix to Subpart EEE - Quality Assurance Procedures for Continuous Emissions Monitors Used for Hazardous Waste Combustors is a realistic approach to calibration stability. Proposed Section 5.2.2 to the Appendix defines a procedure for conducting an absolute calibration audit (ACA) for continuous PM monitors. The centerpiece of the quarterly ACA are "NIST traceable calibration standards." The problem is that there are no such standards for the technological principles (i.e., light scattering and beta transmissivity) most likely to be employed as continuous PM monitors."

Clearly, we agree with EPA's observation regarding the lack of NIST traceable standards. We are concerned about EPA's statement in the Federal Register notice that a German TuV version of the NIST standards (referred to as linearity test kits), "are sufficient substitutes for the yet to-be-developed NIST standards to conduct ACAs2." EPA solicits comments regarding the advisability of modifying its ACA approach, given the fact that these linearity test kits are generally not available to affected facilities.

Before suggesting the acceptance of a German clone of NIST-traceable standards, we believe that EPA should acquire some of the linearity test kits and evaluate them in the context of the Agency's ongoing continuous PM field demonstration project. The obvious question to address is, how does the performance of a continuous PM monitor determined from using a German-made linearity test kit compare with the performance determined from the application of EPA Method 5? Certainly EPA should provide an answer to this important question prior to deciding to revise its ACA approach to utilize standards that are not NIST traceable.

Statistical Analysis. In the Federal Register notice, EPA requests comments on its statistical approach to data analysis, especially with respect to the Agency's treatment of "outliers." We support EPA's approach for dealing with paired reference method measurements. That is, if two concurrent Method 5 runs fail to agree within some pre-determined tolerance, than at least one of the tests is inaccurate, and both runs should be rejected. On the other hand, we believe that EPA must be very careful in rejecting responses from any of the continuous PM monitors. We can easily argue that using any type of statistical outlier test merely assumes away the problem. That is, the basis of an outlier test is that all of the observations are drawn from the same distribution, and, because a particular observation does not fit that distribution, it can be discarded. In reality, EPA does not have sufficient data from any of its installed continuous PM monitors to know that all of the measurements are from the same distribution. Therefore, we do not believe that EPA can or should reject responses from continuous PM monitors on the basis of statistical outlier tests.

Multiple Calibration Curves. In discussing the relatively poor correlation coefficients obtained from the regression analysis, EPA observes that it may not be possible to develop a single calibration curve for the Dupont incinerator that will be valid for every operating condition2. This is the exact situation that was predicted in our August 1996 Memo, which stated4:

"... the characteristics of the particulate emissions are quite different at the three different mass loading loadings (i.e., low, medium, and high). It is easy to imagine that as the particulate mass loading increases, there is an accompanying change in the particle size distributions. Since most, if not all, of the commercially available continuous PM monitors are sensitive to particle size distribution, it is clear that each mass loading regime needs to be characterized by a different calibration curve. Unfortunately, PS 11 provides for only a single calibration curve - that may reflect significant inaccuracies for different mass loadings as particle size distributions change."

Upper Range Calibration. EPA states that the Dupont incinerator did not emit particulate matter at its permitted limit -- much less at twice its permitted limit, as specified in the calibration procedures of proposed PS 11. Since the permitted limit could not be reached, EPA proposes to evaluate the confidence and tolerance intervals at the highest measured PM concentration, not at the emission limit. We believe that this is a reasonable approach. As explained in our August 1996 Memo, requiring a source to operate at twice its permitted emission limitation, even to conduct an instrument calibration, is unacceptable. It is difficult to comprehend that Dupont could not shut down enough control equipment to exceed its particulate emission limit. Is EPA implying that the Dupont incinerator does not need any control equipment to comply with its particulate emission limitation? To the contrary, we believe that Dupont management exercised prudent judgment and elected not to disable enough control equipment so as to exceed its particulate emission limit -- even for the sake of an EPA field demonstration program. Again, referring to the our August 1996 Memo, we do not understand how EPA believes it has authority to not only allow, but to require a source to deliberately exceed its emission limitation, not by a small amount but by a factor of 2.

Summary and Conclusions

Perhaps the most obvious statement to make concerning EPA's Federal Register notice is that the Agency must have composed the notice based on how it "hoped" the field demonstration would turn out -- not on what the data actually show. Else, how does one account for statements such as, "[d]espite this heterogeneity of PM characteristics, most of the CEMS were able to meet the DPS 11 data acceptance criteria with certain outliers deleted 3." As discussed above, of the 15 statistical calculations (5 monitors times 3 statistical tests each) that can be conducted with the field data, only four meet the criteria set forth in proposed PS 11, and this is after various "outliers" are removed.

Reliable and accurate continuous PM monitors could be useful for both affected sources and the regulatory community; however, the technology is simply not adequate at this time to require for continuous compliance determinations. This conclusion is not only supported by the fact that EPA has proposed very broad confidence and tolerance intervals as part of its PS 11 requirements, but also by the results of EPA's field demonstration. EPA should acknowledge this technology shortfall and move forward with additional research and development - if the Agency wishes to use PM monitors for continuous compliance determinations. The absolute worst course of action that EPA could pursue would be to further weaken the already weak statistical criteria in proposed PS 11. That is, if in response to the results of the field demonstration, the Agency proposes to further enlarge the acceptable widths of the confidence and tolerance intervals, EPA will simply ensure that data from continuous PM monitors are virtually useless.


1. 61 FR 17358. April 19, 1996. Hazardous Waste Combustors; Revised Standards; Proposed Rule.

2. 62 FR 13775. March 21, 1997. Hazardous Waste Combustors; Continuous Emissions Monitoring Systems; Notice of Data Availability and Request for Comments; Proposed Rule.

3. Status Report IV: Particulate Matter CEMS Demonstration, Volumes 1, 2, and 3. prepared by Energy and Environmental Research Corporation for EPA's Office of Solid Waste, dated February 12, 1997.

4. Roberson, Ralph L., "Technical Comments on EPA's Proposed Rule to Require Continuous Particulate Matter Monitoring," Memorandum prepared for the Utility Air Regulatory Group, August 15, 1996.


Statistical Results for EPA's Initial Calibration Tests

Manufacturer Correlation Coefficient Confidence Interval, % Tolerance Interval, %
ESA 0.55 26 38
Verewa 0.69 27 32*
Durag 0.72 22* 36
ESC 0.71 22* 36
Sigrist 0.64 25* 40
*Denotes those values meeting the requirements of proposed PS 11.


EPA Regression Equations for PM Monitors

Manufacturer Predictive Relationship
ESA y = 0.79(x) + 18.7
Verewa y = 0.85(x) + 11.7
Durag y = 0.79(x) + 17.7
ESC y = 1.00(x) + 14.8
Sigrist y = 0.73(x) + 18.9

| Home | News | CAM | Training | FTP Library | Projects | Links | Contact | Services | Feedback |

RMB Consulting & Research, Inc.
Last Revised: May 22, 1998