Observations and Comments on EPA/DOE Mercury CEMs Demonstration at Holnam's Holly Hill, SC Facility

David L. Constans

Gossman Consulting, Inc.

Rex Jameson

Holnam, Inc.

Glenn Raynor

Holnam, Inc.

Presented at the AWMA International Specialty Conference on Waste Combustion in Boilers and Industrial Furnaces April, 1997


This paper covers observation and comments on a EPA/DOE Mercury CEM demonstration project performed on a large, long, wet cement kiln stack. Three CEMs were tested and compared versus EPA Method 29 and a modified Method 29 for total mercury emissions. This included challenging the CEMs with spiked mercury and mercuric chloride. This paper examines available data, recounts observations during testing and discusses CEM performance and operational difficulties.


In the proposed Hazardous Waste Combustor (HWC) regulation, (promulgated April 19, 1996) the EPA proposed that regulated units install mercury continuous emission monitors (CEMs). Subsequently, a series of demonstration projects were set up to investigate the ability of the various available CEMs to conform to the proposed performance specification. A performance test of three mercury CEM units was carried out at EPA's Incineration Research Facility in 1995. A report on that study was presented at the June, 1996 A&WMA conference in Nashville.

A continuation of this demonstration project resulted in the search for a hazardous waste combustor willing to participate in a mercury CEMs appraisal study. The selected site was Holnam's Holly Hill facility. The program began in April, 1996 with a somewhat unrelated PIC emissions study and was followed in July by the mercury CEMs installation and subsequent performance testing. This testing is now in its third and final phase, long term endurance of the CEMs. Data available to date demonstrates that the three mercury CEMs are unable to meet the performance specifications. The results do not tell us much about what happened along the way, however.


The proposed HWC regulation was intended as a melding of the RCRA hazardous waste regulations and the Clean Air Act. The legal and technical suitability of this marriage will undoubtedly be discussed at length elsewhere. It is not, however, a part of this paper. What is worthwhile knowing are the assumptions and representations made in the HWC regulation that led to the requirement for mercury CEMs. First, in the proposed HWC regulation(1), the EPA stated that "...the MACT floor level is based on hazardous waste fuel control." While acknowledging that "Raw materials and fossil fuels also contribute...." to mercury emissions the EPA basically ignored data that did not support this MACT floor basis. 80% to 90% of the mercury emitted from the Holnam Holly Hill Kiln #2 is from the raw feed and coal. (Based on the examination of the metals mass balance provided in the 1992 COC Test Results. Similar results have been seen for other kilns as well.) This leaves the facility with the ability to control only about 10% to 20% of the total mercury input by mercury restrictions on hazardous waste fuels. Second, the EPA assumed that carbon injection and subsequent removal in "...conjunction with hazardous feed rate control of mercury..."(2) would achieve the "below-the-floor" emission rate of 50 µg/dscm, even though carbon injection as a method of control of mercury is an unproved technology in cement kilns. Third, the mercury CEM monitoring requirement was justified by the EPA based on classing, "...all HWCs are subject to the regulation as major sources." There is considerable debate that such is indeed the case for cement kilns. Fourth, the EPA knew that mercury CEMs were being utilized in Europe and believed such monitoring should be required for HWCs. The proposed HWC regulation boldly states: "Several types of CEMs exist or are under development which measure mercury. Therefore, the rule proposes use of a mercury CEM to document compliance with the mercury standard."(3)

The performance test executed at EPA's Incinerator Research Facility in 1995 on the combustion products of a synthetic waste demonstrated poor performances for all three CEMs. In the report presented at the A&WMA conference, it was stated: "In summary, even though the RAs (editor: Relative Accuracies) in these tests were, at best, about 60 percent, had nine measurements at a given concentration been performed, RAs might have been at reduced levels on the order of 20%."(4) (That is in conformance to the performance specification.) Essentially then, statistics were blamed for the failure of the monitors to pass the performance specification.

This paper also states: "OSW (Office of Solid Waste) is planning a demonstration of mercury CEM on actual industrial waste combustors during 1996. No further pilot scale testing of mercury CEMs is planned." Indeed, by the time this paper had been presented, Holnam Inc. had agreed to allow their #2 kiln at Holly Hill, South Carolina to be used as a mercury CEMs demonstration test facility. Based on observations made during this mercury CEMs demonstration at Holnam's Holly Hill facility it would appear that EPA's OSW and the CEM vendors would have benefitted from additional pilot scale testing prior to subjecting the CEMs to the difficulties of analyzing cement kiln flue gas.

Description of The #2 Kiln at Holnam's Holly Hill Facility

Holnam operates a large, long wet kiln at Holly Hill, South Carolina. Hazardous waste fuel is utilized, providing about 40% of the energy consumed by the kiln. Table 1 provides the significant physical operating parameters.

During the proposed test sequences the kiln was to be operated at normal conditions. There were no provisions made to perform any testing at maximum rates as would be required by a Boiler and Industrial Furnace (BIF) regulation compliance test. The only requirement was that the comparison testing, that is where the CEMs were compared to the standard emission sampling and analytical method, was to be done while the kiln was utilizing hazardous waste fuel. Nor was any extraordinary record keeping required, the routine monitoring and analysis already required by BIF was deemed sufficient.

The CEMS Installed For Testing

Three vendors agreed to provide monitors for the testing. Each of the three are discussed in turn including a brief description of their method of operation.

Hg-Mat2 (Seefelder Messtechnik). The Hg-Mat2 was supplied and supported by EcoChem Technologies. This unit is designed to measure total mercury concentrations in flue gas, yielding a reading in µg/dscm. A sample of the flue gas is extracted, non-isokinetically, at a rate of about 1.5 l/min. This sample is transported through a heated PTFE sample line maintained at 200°C. The sample is passed through two reactors within the monitor that cools the gas and converts the mercury compounds into elemental mercury. The vendor is hesitant to divulge the reactant solution, however it is most probably tin chloride. A demister removes unwanted moisture as the gas with the elemental mercury is carried into the detector. The detector is a UV photometer operating at 253.7 m. The gas is then dried and the volumetric flow rate measured so that the instrument can express the result in dry standard cubic meters. Daily zero calibration can be programmed into the instrument at preselected intervals. This unit allows an internal calibration check utilizing mercury permeation tubes. The Hg-Mat2 is more sensitive to interference from SO2 in the flue gas than the other two units, because it does not compensate for SO2 which has an UV adsorption frequency very close to that of mercury.

Perkin Elmer MERCEM. The P&E unit was supplied by Wheelabrator and supported with technicians from Germany. A sample of the flue gas at a rate of 17 liters per minute is pulled through a stainless steel probe and transported through a heated PTFE sample line maintained at 185°C. A slip stream of this flow, about 0.5 liters per minute, is directed into a reactor. The flue gas is reacted with tin chloride in a reactor to convert all of the mercury compounds to elemental mercury. This vapor phase mercury is cooled and dried before entering a gold/platinum trap where the mercury amalgamates onto the gold. At the end of a predetermined time, the trap purges with dry air and a photometric baseline is established. The trap then is heated to 750°C driving the mercury off and the concentration is measured in a U.V. photometer. The entire cycle time is about six minutes. Technically, this unit does not conform to the proposed performance specification due to the long response time. Also, as noted in the text later in the paper, the recorded concentration values were consistently one-third to one-half of the values recorded by the other two monitors.

Verewa (HM-100). The Verewa unit was supplied by Monitor labs and supported by them. This unit measures total mercury as well. The system extracts about 2 liters per minute of flue gas out of the duct through a stainless steel probe and transports the sample to the instrument via a heated PTFE sample line. The sample line is maintained at 120°C. In the instrument the sample gas is first heated to 800°C. The gas is then mixed with hydrochloric acid to convert all mercury and mercury compounds to mercuric chloride. This is then reacted with sodium borohydride at 10°C to reduce the mercury chloride to elemental mercury. This mercury vapor is then carried into a UV photometer (253.7 m). The photometer is double beam. The mercury laden gas passes through the first beam, is then scrubbed of mercury by the use of activated carbon, and the mercury free gas then passes through the second beam. In this manner the effect of interfering gases can be electronically deducted from the sample gas if they too are not absorbed onto the activated carbon.

Installation of The CEMs And Manual Sampling Facilities

The sampling lines from the flue gas source to the monitors were to be restricted to less than ten meters in length. This required the installation of the probes into the duct between the induced draft fan and the stack. Such a sampling location does not conform to the EPA requirements for CEMS or flue gas sampling. To ensure that the location was comparable to the proper stack sampling location, a number of runs were sampled simultaneously at both locations for subsequent comparison. A trailer was parked next to the duct and stack and used as a shelter for the three CEM instruments. A platform was installed on the duct to allow probe and sample line maintenance as well as allowing the collection of manual method samples. A second trailer was parked on the other side of the stack. This trailer was set up as a manual sampling preparation and recovery laboratory, and was equipped with a vented fume hood on one counter top.

The three CEM cabinets were installed in one end of the mercury-CEM shelter/trailer. A computer system, installed in another room of the shelter, received data from each instrument into a data acquisition software package. A modem connection allowed the EPA contractor to access the data from their North Carolina office.

In addition to the duct sampling location there was a platform on the stack with four sampling ports. This sample location fully conforms to the EPA requirements for flue gas sampling.

Test Plan

The EPA contractor, Energy and Environmental Research Corporation (EER), wrote a test plan and QAPP for the demonstration. The plan included three phases.

The first phase was a Method 301 validation of a revised flue gas sampling and analysis method specifically for mercury, Modified Method 101A. The revision to the method was to improve the collection of mercuric chloride.

The second phase consisted of a suite of tests specified by the performance specification. Specifically: 1) A seven day zero and span calibration drift test. 2) A relative accuracy test audit (RATA). 3) Low, mid and high range calibration error tests utilizing mercuric chloride and mercury spiking of the flue gas. 4) An interference response test.

Phase three is the endurance testing phase. During this phase monthly calibration error and RATA tests were to have been performed. Also, any problems with the physical performance and maintenance of the CEMS was to be documented.

Observations And Comments on The Execution of The Performance Testing

The execution of the mercury CEM demonstration testing at Holnam's Holly Hill facility began in July 1996 and is expected to continue through April 1997. The following are observations and comments on the execution of the performance testing. Gossman Consulting, Inc. had been retained to observe the EPA's contractor (EER) with the purpose of acquiring insight into the problems associated with continuous emissions monitoring of a cement kiln stack for mercury emissions, hopefully ascertaining which monitor performed better than the others and generally how accurate such a CEM would be versus the manual method. For this purpose it was not necessary for each set of tests or each test day in a sequence be observed. Instead a variety of tests were selected for observation, frequently associated with prearranged visits from representatives of EPA and Department of Energy (DOE). DOE is sharing in the cost of the demonstration project.

It should be noted that the majority of the difficulties and delays noted in these observations stemmed from the experimental nature of what was being attempted. This is amply illustrated by the changes enacted in the field to the test protocol. These changes were instituted without a formal review process. In effect, the program evolved into an experimental program rather than being a rigorous test of performance versus an EPA performance specification as set out in the test protocol.

The following observations were made during the testing. A short paragraph describing the testing precedes each set of observations.

The purpose of this testing was to validate a modification of Method 101A that would allow the improved collection of mercury chloride for inclusion in the total mercury content of the flue gas. It was believed that the monitors would have a problem performing this task since the detector will sense only elemental mercury not mercury compounds. This is the reason that the CEMs have reaction chambers to convert mercuric compounds into elemental mercury. This validation is to consist of a series of tests where flue gas samples are extracted from the duct, and are then spiked with a metered volume of nitrogen as a carrier gas to which has been added a measurable quantity of either mercury chloride or mercury released out of permeation tubes that have been heated to a specified temperature. All of the sampling is done in a "quad probe" sample system. No traverse of the stack is possible. Because there are four sampling trains (two to be spiked, two that are not spiked) plus the mercury spike system and the sample train to measure the spike input the entire setup is very complex and crowded at the port.

Tuesday, July 9th, 1996. During a discussion with the EER project manager, while displaying and explaining how the quad-probe was utilized, a mercury thermometer was dropped to the floor of the laboratory trailer. This was inside the trailer/lab where the impingers are emptied into sample transport containers and the rinsing and subsequent recovery of the rinsate occurs. An effort was made to immediately collect and bag the glass and mercury. It is reasonable to believe that the lab was now "contaminated" with mercury vapors.

Earlier that day the oven needed to heat the permeation tubes (a converted GC oven) had failed to operate and a special purpose oven had been ordered for delivery to the facility. Subsequent to that the converted GC oven began to operate properly.

EER proceeded to prepare for doing the sampling that was to begin the next day. Generally blank samples are collected during the sampling period not the day preceeding the sample period.

Wednesday, July 10th, 1996. 9:00 AM EER's field supervisor has stated that because of the small size of the impingers the run times will be cut in half from that proposed in the sampling protocol, one hour instead of two hours. This is a major change in the protocol of the test. The reason given for this is that the volume of moisture that would be condensed during the sampling would otherwise overwhelm the impinger volume. Consequently the spike times will also be one half of that proposed in the plan, that is 15 minutes instead of 30 minutes. The supervisor claims that the volume of gas sampled will be the same, about 100 cubic feet. If moisture levels are the problem as stated, the sample volume would also need to be cut in half.

10:40 AM to 2:30PM. EER was performing a leak check of the trains prior to starting testing.

2:30 PM Due to concerns regarding the hazards of flying debris, EER was instructed by Holnam to remove their equipment from the stack by noon the next day due to the possibility of Hurricane Bertha coming ashore near Charleston. EER proceeded to perform one run before removing the equipment. This run was not used in the report.

A RATA test had been scheduled for the week of July 29th and representatives of EPA's Office of Solid Waste were to visit the facility and observe a portion of this testing.

Monday, July 29, 1996. 9:30 AM, EER personnel were not on site. The CEM probes had been installed and a catwalk was in place to service the probes and allow manual sampling of the duct. Technicians from Perkin-Elmer Bodenseewerk and Hg-Mat2 were on site. The Hg-Mat2 was operating, however the PE and the Verewa were not, although both were installed.

10:30 AM. With no EER representative on site and only one of the monitors operating it was obvious that a RATA test could not be conducted.

This sequence of testing was to be the start of the 7-day drift test, subsequently to be followed by the RATA.

Wednesday, Aug. 21, 1996. 8:00 AM, met with EER's field supervisor and discussed the status of the demonstration project. This supervisor related that the Verewa instrument had been down the previous week. No specific reason could be given for the problem. (This was typical as the vendor technicians would not discuss the problems that the instruments had in dealing with the sample matrix.) Two technicians had rebuilt the Verewa on Monday and were performing a startup and calibration. All of the vendors had been having problems with their machines, due to particulate loading, high moisture and/or high Hg levels (the devices were designed for low level detection of about five micrograms, this despite the advertised 0-150 microgram range). As an example, the Hg-Mat2 sampling system had plugged off in less than two months of intermittent service. The field supervisor admitted that there will be a problem in supplying enough Hg/HgCl challenge spike to all of the monitors simultaneously as the P&E instrument demands so much gas at the probe. There was some discussion of pressurizing this supply system.

10:20 AM. The EPA project director arrived on site to observe the 7-day drift test discovered that the 7-day test would not be completed before the RATA was scheduled to start. And that since an evaluation of the 7-day drift test requires analysis of the M101A samples to confirm the spiked Hg/HgCl concentrations, the data to confirm the 7-day drift would be at least a week behind that. Consequently, the RATA may be for nothing if one or more of the monitors fails the drift test.

A RATA test was conducted the week of August 26, 1996 as scheduled the previous week.

Wednesday, Aug. 28, 1996. 8:00 AM, a discussion was held with EER's field supervisor about the status of the project. It was related that the RATA testing was to begin about 09:00 - 09:30 AM. The sample collection will consist of: one set of samples taken from the same location (the duct to the stack) as the fixed probes for the Hg CEMS, which is also a fixed probe; two sampling trains on the stack one of which is a standard M101A train, the other is a modified M101A (the newly designated M101B) train which had been modified to collect Hg chloride more effectively. These two trains traverse and move port to port as a normal M29 sample train would be utilized. Three runs are planned for Wednesday, four for Thursday and two for Friday, for a total of nine for each of three locations. Each run is to be about an hour long.

4:00 PM, the Verewa instrument recorded a large increase in Hg emissions about 3:15 PM. The technician believed he had a dirty lens in the device, and proceeded to correct it. Normally this device and the Hg-Mat2 device are very close together. This "fogging" of the lens was a common occurrence and was specifically referred to later in EER's draft report.

Thursday, August 29th, 1996. 9:30 AM, EER has started the first manual sampling run of the day. All three monitors are on line. The P&E device is still indicating about one-half to one-third of the Hg emissions values displayed by the other two monitors.

4:00 PM, the fourth manual sampling run had been completed. No problems have been reported with the instruments or the sampling.

Friday, August 30, 1996. 8:00 AM, the Verewa was indicating an extremely high value. The technician was not willing to discuss why this was the case. Earlier in the week the technician had indicated that this was due to a dirty lens. Clearly the Verewa Hg monitor was having problems with the sample matrix.

1:30 PM, the second run has been completed and the last of the recoveries begun of the nine RATAs in the series.

During September and October EER executed a second RATA, an interference response test and a calibration audit. However, since these tests were scheduled on last minute notice and/or conflicted with other commitments, GCI was unable to supply an observer. In early December EER scheduled a third RATA.

Wednesday, December 11, 1996. 11:00 AM, EER is currently executing manual sampling run number three of what are to be four runs on this day. The sampling was being done at the duct, no Hg or HgCl spiking was being done, no sampling on the stack was being done. The sample collection train is the M101B train with the runs one hour in length. The reagent blanks had been recovered in the morning, the field blanks are to be collected and recovered on Thursday. A 7-day drift test had not preceded this RATA, nor is one planned succeeding this test. It can only be assumed that it is believed by EER that the previously conducted seven day drift test is adequate. This is highly questionable since it is known that the monitors have been forced out of service by plugging of the lines and have been cleaned and re-calibrated since then.

Only two of the monitor vendors have technicians on site, the Hg-Mat2 and Verewa. The HgMat2 and the Verewa monitors are reading in the 15 - 20 mg/m3 range at the beginning of the RATA runs. The P&E is indicating a value of about 5+ mg/m3 , a typical value in relation to the other two machines. All of these Hg values are much lower than has been the case historically. There appears to be no reason for these values that can be attributed to operating conditions alone. The kiln was burning about 8.5 T/Hr of HWF and about ten tons per hour of coal/petcoke (HWF feedrate is about 43% thermal replacement) and the kiln raw feed rate was very close to the maximum BIF limit. Kiln operation was very stable.

1:00 PM, a Holnam instrument technician states that the SOx and NOx values for kiln 2, monitored by the CEM system, are as follows: The SOx values were in the 150 to 250 ppm range (uncorrected) and the NOx values in the 900 to 1100 ppm range. It was stated that this was very common for this kiln. This is important in that the SOx may interfere with the detector (a quartz glass optical cell where the selective absorption of radiation emitted by an UV light source, 253.7 m Hg line, is determined). This is particularly the case for the Hg-Mat2 However the interference is minimal to moderate at this concentration of SOx. The Hg values indicated by the monitors rose slowly over the day to the more normal 20 to 40 mg/m3 values.

As a matter of routine Safety Kleen performs an analysis of the kiln raw feed, the coal/coke and the HWF every Wednesday for physical properties and metals and chlorine. The mercury concentration in all of the feeds were below the detection limit of the analysis of 1.1 ppm. Even at such low input concentrations there is sufficient Hg in the raw feed and coal to produce 20 - 30 ug/dscm of Hg emissions.

The last run of the day finished about 3:30 PM. The monitors had indicated the following average values over the four runs: Verewa - 27 mg/m3 , Hg-Mat2 - 24 mg/m3 and the PE monitor - 5+ mg/m3 . The Hg-Mat2 was zeroed between the last three runs.

Thursday, December 12, 1996. The Hg-Mat2 unit required some maintenance to the reaction train consequently the first run of the day did not begin until 11:00 AM. Also the Verewa is cycling between 30 and 50 mg/m3 while the Hg-Mat2 does not indicate a similar cycle. The feed rates and operating conditions of the kiln are the same as Wednesday.

EER collected the field blank samples, first performing a leak check on the train then drawing ambient gas through the train. The last run of the day finished about 14:00. It would appear that the Verewa unit had drifted about four to five mg higher than the Hg-Mat2 unit. The P&E continued to indicate values about one-third to one-half of the other two units.

This next set of tests were a series of calibration error tests. That is, the monitors were challenged at three concentrations of Hg and Hg chloride.

Wednesday, February 5, 1997. 8:00 AM, the system for adding the Hg and HgCl spikes to the monitors was just being brought on line. The kiln was consuming 8.5 T/Hr of HWF, very stable in its operation at 2.7% kiln exit O2, and ~160 ppm CO in the stack. The kiln stack CEM indicated the following: 400 to 500 ppm of SO2 and ~1500 ppm NOX both had been steady over the last several hours. All of the mercury monitors were ranging wildly from zero to 150 µg but each was out of sync of the others. Obviously something was wrong with the test system as the monitors had been operating well prior to the installation of the Hg/HgCl spiking equipment, indicating stable readings in the 15 to 20 µg/m3 range.

9:50 AM, EER's field supervisor had been feeding zero gas, dry nitrogen, through the spiking system manifold and into the monitors. This had been going on for about 30 minutes and the monitors have been wildly out of sync with readings ranging from zero to 150 µg. This situation continued all morning with the field supervisor and EER's newly trained Hg CEM instrument technician struggling to determine what to do to correct it.

11:30 AM, the Hg/HgCl spiking system was brought down into the trailer to work on it more easily.

1:30 PM, zero gas or Hg spike gas continues to be injected through the manifold system. Two monitors are reading zero, one is reading 30 - 60 µg. Dry nitrogen purges appear to send the monitors crazy with the readings ranging wildly the full range of the monitors span value. This continues all afternoon.

4:30 PM, it was decided to return the equipment to its proper place on the platform next to the probe location, even though the system still does not appear to work. This effectively puts an end to the day.

Thursday, February 6th, 1997. 8:00 AM Today while feeding Hg spike to the monitors the Hg-Mat2 executed an auto zero and zeroed out the spike, resulting in a negative value when the spike was discontinued. Later the same thing happened with the Verewa. After an adjustment to the auto zero sequence timers this was corrected.

9:15 AM to 3:15 PM, the struggle continued alternating zero gas, Hg spike and stack gas trying to get the monitors and the system lined out. Part of the problem involves the P&E unit which requires a gas flow of about ten times the other units. The P&E demands so much the other units receive less of the spike gas than they should; or at least that was the speculation.

3:15 PM, while EER's field supervisor was performing a leak check on the Hg/HgCl spiking system manifold he decided that one valve appeared to be performing poorly and subsequently blew it out with a high pressure nitrogen purge. Thereafter the system and the monitors began to perform properly. One run on one monitor was then performed.

Discussion of Testing

In early February, 1997 EER issued a draft report of the mercury CEM demonstration project at Holnam's Holly Hill facility.

It would be improper to present the data contained in that report in this paper. The report is still in draft and may be changed substantially in its final issue. In addition the data have not been sufficiently reviewed to determine if the data meets the required quality assurance / quality control objectives, the data supplied in the draft report is insufficient for this purpose. However, it is possible to discuss the draft report in general terms, especially as it relates to the observations made above.

The first thing that is noted in the draft report is that only the Method 301 validation of M101B, the first RATA and the first calibration error test have been reported. All of this testing was conducted in July and August 1996. The calibration error test conducted in October, the # 2 and # 3 RATA tests, (conducted in September and December respectively), an interference response test from late September and an absolute calibration audit test from early October have not been reported in this draft.

In addition, recall that during the August 1996 testing period it was remarked that the seven-day drift test would not be completed prior to the first RATA test. The draft report does not include any data specifically identified as being from that seven-day drift test. There is data on various calibration error tests during that seven-day period, but no definitive statement on data illustrating that the seven-day drift test was successfully completed by any of the monitoring systems. Having observed the difficulty which all of the instruments illustrated in dealing with the sample matrix, it is highly doubtful that any of them successfully completed the seven-day drift test.

There are other items noted in the observations that failed to be mentioned in the draft report. As an example, the fact that a mercury thermometer was dropped and broken in the sample recovery lab is not mentioned. The reduction in the length of the sample runs from two hours to one hour, probably resulted in some data quality problems. This was a significant change in the protocol in that fluctuations in the spiked Hg and HgCl flow rates would have a more pronounced effect on the total Hg or HgCl delivered due to the shorter sample period. This was subsequently seen in the data in the draft report where, for example, Hg spiking rates for the three "high level" runs during the calibration error tests were: 87.1, 18.5 and 46.1 ug/dscm clearly more varible than what was expected. The draft report goes on to discuss mechanical problems as the probable cause of this variability, however the variability would have been mitigated somewhat by the longer run times as planned in the original protocol.

Earlier in this paper it was mentioned that the mercury CEM units tested at EPA's Incineration Research Center had failed to meet the performance specification requirement. In a paper presented at the A&WMA Conference in Nashville, the authors of that paper believed that if nine runs had been conducted at a given test concentration the relative accuracies would have been within the parameters of the performance specification. At Holnam's Holly Hill facility nine runs were performed at a consistent mercury emission rate, these runs utilized an improved reference method, M101B. Despite this, based on the data reported in the draft report, none of the mercury CEMS passed any of the performance tests; not the RATAs, nor the calibration error tests, nor apparently the seven-day drift test. Additioanally there still appear to be problems with the Hg/HgCl spiking and measurement systems. Without a reliable method of challanging the Hg monitors with metered Hg/HgCl concentrations the calibration of the instrument can not be confirmed or evaluated.

Clearly the current generation of mercury CEMs are "not ready for prime time" at least at long wet cement kilns with ESPs. The CEM's reported and implied problems in handling the flue gas matrix; the high level of moisture, the particulate loading, the oxides of sulfer and nitrogen, perhaps even the organic content, all combined to put the three CEMs at a severe disadvantage in meeting the performance specification let alone indicating an ability to operate for extended periods of time without failure.

Table 1. Significant Kiln Operating Parameters


Typical Value

Kiln Raw Materials Feed Rate (wet basis) ~250 Tons/Hour
Kiln Clinker Production Rate ~100 Tons/Hour
Primary Fuel Feed Rate 9 - 11 Tons/Hour
Hazardous Waste Fuel Feed Rate 8 - 8.5 Tons/Hour
Kiln Gas Exit Temperature ~400 F
Flue Gas Temperature in Stack 350 - 380 F
Flue Gas Oxygen Content 5 - 6 %
Flue Gas Carbon-monoxide Content 100 - 200 ppmv at 7% oxygen
Flue Gas THC Content 10 -17 ppmv at 7% oxygen
Flue Gas Particulate Content 0.0287 gr/dscf at 7% oxygen (1)
Flue Gas Oxides of Sulfur Content 150 - 400 ppmv (2)
Flue Gas Oxides of Nitrogen Content 900-1500 ppmv (2)
Flue Gas Moisture Content ~ 34%

(1) Taken from the Re-Certification of Compliance Test in 1995, Maximum flue gas velocity condition

(2) Not corrected to 7% oxygen.


1. Federal Register Volume 61, No. 77, page 17393

2. Ibid, page 17394

3. Ibid, page 17427

4. "Performance Tests of Mercury Continuous Emissions Monitors at the USEPA Incineration Research Facility", presented at the June, 1996 A&WMA conference