Comparison of Continuous Opacity Monitor and …Opacity Monitor and Method 9 Observations to...

April 2002 AISE Steel Technology 53

Ipsco Steel Inc., Montpelier Works, wasinterested in replacing the continuous opac-ity monitor (COM) installed on its EAF/LMFbaghouse stack due to New SourcePerformance Standard (NSPS) requirements.The COM had been a high maintenancedevice due to high vibrations on the nega-tive pressure baghouse stack. Further, theCOM had a documented problem measuringthe 3% opacity standard applied to the EAFemissions. Bag penetration by sparks in thenegative pressure, pulse jet baghouse hadbeen a continuing maintenance problem,and the COM had shown itself to be ineffec-tive as a tool for identifying failed bag com-partment locations. When opacity was iden-tified in the stack, the COM provided theoperator with no useful information as toemission source. When more than one com-partment was involved, successive isolationof individual compartments provided nocompartment identification informationwhen using the COM as the emission indi-cator.

To provide operators with a more proac-tive and useful tool, Ipsco investigated thefeasibility of using a broken bag detectionsystem. After a review of broken bag detec-tion system technologies, a triboelectric basesystem was selected.

Due to the lower levels of particulate changedetected by the triboelectric technology andthe problems associated with using a COM inthis application, Ipsco applied to the U.S. EPAto use alternative monitoring technology onthe EAF/LMF baghouse. The use of a tribo-electric broken bag detection system was pro-posed for replacement of the COM originallyinstalled on the stack. The proposed technolo-gy change was conditionally approved by the

U.S. EPA, provided thatIpsco run a 45-day com-parison test to establishthe relative relationshipof the triboelectric sig-nals to opacity as meas-ured by the COM. Ascompliance with theNSPS 3% opacity stan-dard for the EAF bag-house was establishedby visible emissionobservations using U.S.EPA Method 9, suchobservations were alsoincluded in the compari-son by Ipsco.

InstrumentMeasurementMechanismsFundamental compli-ance with the particu-late standard for EAFemissions is based onparticulate mass emis-sions from the bag-house. With continuousmonitoring of massemissions remaining apractical impossibilityfor fume control sys-tems, the use of surro-gate measurements hasbeen adopted to char-acterize baghouse par-ticulate emissions. Thesurrogate measurementadopted by the U.S.

Comparison of ContinuousOpacity Monitor and Method9 Observations toTriboelectric Broken BagDetector Signals on an EAFBaghouse

A 45-day test program com-

paring output of a triboelec-

tric broken bag detection

system to a continuous opac-

ity monitor was conducted

on an EAF/LMF baghouse at

Ipsco Steel Inc., Montpelier

Works, from Oct. 12 through

Nov. 25, 2000. During the

test, elevated opacity was

artificially created to better

identify the comparison at

opacities higher than 5%.

U.S. EPA Method 9 visible

emission observations were

made concurrently during

the surrogate particulate

introduction periods.

Authors

Joseph C. Wesselman, Corporate Director—Environmental Affairs, Ipsco Enterprises Inc., Lisle, Ill., andCharles W. Askins, President, AG Environmental Services, Inc., Victor, N.Y. ([email protected])

54 AISE Steel Technology April 2002

EPA in the NSPS for the EAF baghouse wasrelative opacity. Opacity is the measurementof the amount of light lost when passingthrough an emission plume, and is charac-terized by percent loss of light. The opacitymeasurement range is 0 to 100%, with zerobeing no loss and 100% being a completeloss of light (fully opaque).

U.S. EPA established a standard method ofvisual evaluation for particulate emissionsthat uses a trained human observer to estab-lish opacity of exhaust plumes. This proce-dure is found in 40 CFR 60 as ReferenceMethod 9 (U.S. EPA Method 9). The require-ment for visual opacity measurements usingU.S. EPA Method 9 was established in theIpsco EAF NSPS to allow operators and reg-ulators to monitor EAF baghouse compliancewith the particulate mass emission standard.

The measurement of opacity (by eitherMethod 9 or a COM) and the measurementof triboelectric change are all surrogatemeasurements for the mass emissions fromthe baghouse, and measure the amount ofparticulate present in an exhaust gas stream.None of these systems measure true particu-late mass emissions.

COM SystemOpacity can be measured using an instru-ment such as a COM. The COM uses thesame basic principle as Method 9 observa-tions, measuring the loss of light associatedwith passage of a light though an exhaustplume. In the case of a COM, the source oflight is a beam that is directed across a ductor stack striking a mirror on the oppositeside of the duct. The mirror is aligned toreflect the light back to the light source. Theinstrument measures the reduction in lightintensity after this double path length, andcalculates the loss as relative opacity.

COM opacity measurement is based on theprinciple that particles present in the gasstream affect the amount of light returning,as particles have inherent and variable char-acteristics of light absorbency, refraction orreflection. These characteristics are associat-ed with the size, shape and structure of therespective particle, and quantity of particlespresent in the gas stream, but are not associ-ated with the respective particle density ormass. COM presumes that all light lost is dueto particles in the gas stream.

As there is neither a correlation to the gasflow volume nor the concentration of partic-ulate present, mass emission associated witha gas flow cannot be measured by COM.

Opacity, whether measured by Method 9 orCOM, is a qualitative surrogate for the quanti-ty of particulate matter being emitted. TheNSPS for opacity from the EAF/LMF baghouse

is <3% opacity averaged over six minutes ofcontinuous readings. The human eye can dis-tinguish opacity to the level of 5% increments,and the minimum detectable change recordedby the human eye is 5%. Therefore, the EAFNSPS of <3% opacity is not a visible emissionstandard. COM is able to measure opacity infractions of a percent, however it has beenidentified by ASTM D6216-98 to be accurateonly to the level of 10% opacity.1

Triboelectric Broken BagDetector SystemThe measurement principle of a triboelectricbroken bag detection system is based onmeasuring small changes in the electricalcharge of an energized probe placed in theexhaust gas stream. The probe is a stainlesssteel or other metallic material rod that isenergized with a DC electrical voltage.Particulates present in the gas stream strikethe probe and change its electric field, simi-lar to the release of static electricity. Thesmall changes in the electric field associatedwith the particle impacts are measured inpicoamps and are readily quantifiable. Thetriboelectric signal is an analog output dis-played as a percent of scale. The impact ofno particles is measured as 0%, with the rel-ative increase of particle strikes measured upto 100% of the scale.

Similar to COM, particle characteristics ofsize, shape and structure, as well as thequantity of particles present in the gasstream affect the relative change in triboelec-tric signal. However, these factors are notrelated to density or mass of the respectiveparticle.

Equipment Specifications andDescriptions System specifications for instruments used inthe comparison are described here.

COM System — The COM system installed byIpsco for use on the EAF/LMF baghouse wasmanufactured by Sick, Inc. Measurements aredisplayed on a monitor as instantaneous opac-ity and most recent six-minute average opaci-ty. The COM instrument is installed on thebaghouse stack at the sample testing port levelon the stack, approximately 75 feet abovegrade. The total stack height is 150 feet abovegrade. The monitor is located at grade in thebaghouse control building immediately east ofthe baghouse structure.

The signal from the COM reads out as adigital display on the console face of thecontroller. For the purpose of tracking COMdata during this comparison, the COM 4-20mA signals (instantaneous and six-minuteaverage opacities) were connected to an


Auburn computer. A real-time readout wasavailable on the PC monitor, and a record ofthese signals was tracked and stored on thePC hard drive. COM system specifications arepresented in Table 1.

Broken Bag Detector System — The bro-ken bag detection system installed by Ipscofor use on the EAF/LMF baghouse was man-ufactured by Auburn Systems, Inc., and usesthe triboelectric principle of particle detec-tion. Triboelectric system specifications arepresented in Table 2.

Tribolink detectors are located in fourlocations in the exhaust plenum of the bag-house. Two Tribolink detectors (A-2, B-2)are located in the east compartment groupplenum, and two detectors (A-1, B-1) arelocated in the west compartment groupplenum. Each detector group includes twoprobes. The detector configuration designplaces the probes such that there are eightcompartments upstream of the B groupprobes, and 14 compartments upstream of

the A group of probes.Figure 1 illustrates thegeneral orientation of thebaghouse compartmentsand the respective loca-tions of the Tribolinkdetectors and the COM.

EAF/LMFBaghouseThe EAF/LMF baghouse is anegative pressure, pulse jetcleaning design that wasinstalled for fume control ofan EAF and LMF meltshop subject to the NSPSfor EAFs. The negative pressure baghouse has astack exhaust, and thus, must install and oper-ate a COM on the exhaust stack.

The EAF is a twin shell DC single electrodedesign, initially created to produce at the rateof 164 tons per hour. At this base design rat-ing, the EAF baghouse was sized and permit-ted to operate using 24 compartments, withone compartment off-line for service and/orcleaning during operations. The baghouse(under a phased PSD permit) was permittedand built to allow for the EAF to operate at aproduction rate of 200 tons per hour. At the

200 tons per hour rate, the baghousesystem was designed to operate with28 compartments, with one compart-ment off-line for service and/orcleaning during operations.

During the surrogate particulateintroduction tests and the 45 daycomparison period, the EAF wasoperating within the 164 tons perhour permitted production range.

Equipment QA Prior toComparison TestingPrior to beginning the comparison, amanufacturer's calibration was con-ducted on the respective instrumentand monitoring system. A manufactur-er's calibration conducted by Sick dur-ing the period of Sept. 25-28, 2000,insured that the COM conformed tothe manufacturer's specifications. OnSept. 26, 2000, the output signal fromthe COM was connected to the

Tribolink PC and verified as to its equivalencewith the COM console display. The signal inter-face to the computer continued to operate with-out problems throughout the comparison peri-od. Each of the Method 9 observers used duringthe comparison possessed current certifications.

Problems with COMPerformanceAs noted previously, the COM was initiallycalibrated by the manufacturer prior to

Triboelectric System Specifications

Manufacturer Auburn Systems, LLC

Model TribolinkTM

Number of probe 4locations

Number of 2sensors/location

Detector material 316 Stainless Steelof construction

Probe temperature -40 to 450°Frange

I/O interface PC

Operating system Windows 98platform

Table 2

COM Specifications

Manufacturer Sick, Inc.

Model OMD 41

Number of locations 1

Material of Carbon Steel construction

Ambient -4 to 131°Ftemperature range

Relative moisture 95%

Signal output 4-20 mA

Operating system Dedicated COM software

Table 1

EAF baghouse compartment config-uration and broken bag detectorlocations.

Figure 1


beginning the U.S. EPA requested compari-son on Oct. 12, 2000. The manufacturer'stechnician was on site for two days duringthe pretest calibration period. However, theCOM continued to perform erraticallythroughout the test period, and was recali-brated a total of three times before and dur-ing the comparison. Manufacturer's calibra-tions were conducted on the following dates:

• Sept. 25-28, 2000• Oct. 18, 2000• Nov. 6, 2000

Ipsco was concerned that the COM shouldoperate as closely as possible to the manu-facturer's performance specification and theU.S. EPA Performance Specification 1,Appendix B of 40 CFR 60 (PS-1). Therefore,the company elected to have all of the cali-bration work done by the manufacturer'strained technicians, presuming that a superi-or level of calibration would be obtained.After the completion of the comparisonstudy, the COM was evaluated by a qualifiedengineer for conformance with PS-1. Thisconformance certification was requested bythe U.S. EPA, and the instrument was certi-fied as conforming to PS-1.

The circumstances associated with therepeated calibrations during the comparisonstudy by event date are discussed below.

Sept. 25-28, 2000 — As previously noted,this calibration was performed with theintent that the COM would be at its bestoperating performance before beginning thecomparison. When the calibration was com-plete, the COM appeared to operate in accor-dance with the Method 9 observations madeon the stack exhaust plume.

Oct. 12, 2000 — Prior to beginning the sur-rogate particulate comparison program onOct. 12, 2000, it was observed that the COMwas reading an average opacity of 4.9%,despite the fact that the meltshop was notoperating, and there were no emissions. Theweather was clear—the temperature was65°F with prevailing winds at 5 to 10 mphfrom the southeast. Surrogate particulate(iron oxide) was introduced into the bag-house exhaust plenum to verify the relativeperformance of the COM. The COM didrespond to increased levels of exhaust opac-ity. However, the unit never returned to 0%following dust introduction when visibleemission observations using Method 9showed an opacity of 0%.

Given the poor performance of the COMat low level opacities, a second calibration ofthe instrument was requested. Routine dailyand additional Method 9 observations were

made during the time period of Oct. 12-18,when the COM was recalibrated to confirm0% opacity in the exhaust plume, althoughthe COM indicated opacities ranging from 3to 10% during this time period.

Oct. 18, 2000 — The COM was inspected onOct. 18, 2000 by a technician from SpectrumSystems, the manufacturer's representative,and some reprogramming was done. Thecomplete instrument package was recalibrat-ed. When the technician left the site theCOM was reading 0% opacity—the sameopacity level recorded by stack exhaustplume Method 9 observations.

Nov. 1, 2000 — Surrogate particulate (ironoxide) comparison was scheduled for Nov.1, 2000, and the COM was observed at 0%opacity prior to beginning the introductionof particulate. The weather was clear; thetemperature was 70°F with prevailing windsat 10 to 15 mph from the southeast. TheCOM registered no change in relative opaci-ty when the surrogate particulate was intro-duced into the exhaust plenum of the bag-house. Yet, Method 9 observations indicatedopacities ranging from 5 to 25%. It appearedthat the COM was in a flat-line signal mode,even though no adjustments had been madeto the unit since the technician calibrated theunit on Oct. 18, 2000. The COM continued toread 0% opacity or nearly 0%, despite intro-duction of surrogate particulate.

A third COM recalibration was requestedfrom the manufacturer, with the requirementthat Sick provide the best technician avail-able to correct the problems experienced to-date with the COM unit. The opacity indicat-ed by the COM remained at 0% until Nov. 6,2000.

Nov. 6, 2000 — The COM was againinspected on this date by a Sick technicianfrom its German home office. The technicianmade programming changes to correct theflat-line signal induced by portions of theprogramming done by the Spectrum Systemstechnician. The complete instrument pack-age was again recalibrated. After recalibra-tion by the Sick technician, the COM erro-neously read false-positive opacitiesbetween 5.0 and 6.0% when there were nomeltshop operations. The Sick technicianinstructed Ipsco engineering personnel onthe procedure to zero the COM during non-operating periods, and this procedure wasperformed. After the zero procedure wasperformed, the COM read 0% opacity duringnonoperating periods, as well as duringoperating periods as verified by Method 9observations. It became evident after a few


days that the COM read a true 0% opacityonly on dry days (low relative humidity).When relative humidity was high, the COMread a baseline of 3.0 to 4.0% opacity, evenduring nonoperating periods. This elevated,false-positive baseline opacity reading wasevident during the two days of surrogate par-ticulate comparison done on Nov. 15-16,2000.

Broken Bag Detector SystemPresently, there are no triboelectric-basedsystem performance specifications publishedin 40 CFR 60, Appendix B. However, the U.S.EPA published a guidance document for suchsystems.2 A broken bag detector system cer-tified calibration was completed by Auburnon the broken bag detector system duringthe period of Sept. 20-21, 2000. During thiscalibration, all of the detector probes wereremoved, inspected and cleaned. The systemoperated throughout the comparison periodwithout hardware or software problems.

General Objectives of theInstrument ComparisonThe U.S. EPA did not establish any specificobjectives for the comparison beyond that itmust be conducted for a period of 45 days.To better qualify expectations for the com-parison, several objectives were identified byIpsco for the 45-day comparison. The fol-lowing objectives were accepted by the U.S.EPA prior to beginning the comparison:

• Identification of a baseline normalrange for triboelectric signals fromeach of the four detector locations.

• Comparison of the relative detectionranges for the COM and triboelectricsystem.

• Comparison of the real-time reliabilitybetween the COM and the triboelectricsystem.

• Introduction of iron oxide into each ofthe two clean-side plenums upstreamof both detector groups in the respec-tive plenum to compare system per-formance at opacities above 5%.

• Establishment of the relative relation-ship between the COM and triboelec-tric systems.

• Establishment of the relative relation-ship between the triboelectric systemand Method 9 observations.

Comparison Study DiscussionAs previously noted, the comparison studyprogram began on Oct. 12, 2000. Findingsfrom the 45-day comparison and a specificreview of the two surrogate particulate intro-duction tests performed to evaluate tracking

of higher opacity levels in the exhaust plumeare discussed.

Comparison Test Program — Ipsco gener-ally divided the comparison test programinto two programs: comparisons during sur-rogate particulate introduction and compar-isons during scheduled operations.

As it was anticipated that opacity levelsabove 5% would not occur during normalbaghouse operation, Ipsco determined that aprogram to artificially generate elevatedopacity levels would be required during thecomparison program. The test program forsurrogate particulate introduction was initiat-ed late in the comparison period because ofthe difficulties encountered with the COM.The surrogate particulate introduction wasdone Nov. 15-16, 2000. The meltshop did notoperate on Nov. 15, but did operate on Nov.16, providing a particulate base load associ-ated with fume collection for the comparisonstudy.

During the surrogate particulate tests,Ipsco established the following objectives:

• Identification of a baseline normalrange for triboelectric signals for eachof the four detector locations.

• Introduction of iron oxide into each ofthe two clean-side plenums upstreamof both detector groups in the respec-tive plenum.

• Introduction of the iron oxide for asufficient time period to allow for atleast two successive six-minute obser-vations by a certified Method 9 observ-er.

• Conducting a total of two tests in eachof the plenums.

• Achievement of at least a 3% opacityfor the tests, as identified by the COMand the Method 9 observer.

• Generation, if possible, of opacities ofat least 20% during the comparison.

Particulate Introduction Method —Creation of opacity through bag penetra-tions to allow for EAF dust leakage wasconsidered. However, this method of opac-ity creation was not used for several rea-sons:

• Bag penetration destroyed goodequipment.

• The amount of dust could not be reg-ulated.

• It was not possible to generate data fora single probe group. Once the bagswere penetrated, they would continueto leak particulate until replaced.

• Use of penetrated bags does not pro-duce a range of data for opacities.

58 AISE Steel Technology Month 2002

Instead, surrogate particulate was introducedusing a compressor, sandblasting pot and a sin-gle introduction point in the respective exhaustplenum. The dust introduction point was locat-ed up stream of the "B" probe locations in therespective plenum as shown in Figure 1.

Compressed air for conveying particulatewas supplied from a compressed air line locat-ed at the penthouse level of the baghouse.The sandblasting pot and hoses were movedfrom the introduction location in the east com-partment line to the west compartment line forthe successive comparison runs. The introduc-tion points were under negative pressure, anddispersion of the dust within the plenum wasaccomplished by the natural draft and turbu-lence in the respective plenum.

Particulate Size Distribution — A com-mercial grade of iron oxide (Fe2O3) was pur-chased for use as the surrogate particulate,and samples of the Fe2O3 and EAF dust col-lected by the baghouse were submitted forparticle size analysis at an independent labo-ratory. The Fe2O3 mean particle size was1.857 µm, while the EAF dust mean particlesize was 1.818 µm. The iron oxide used as asurrogate particulate had particle size distri-butions similar to EAF dust (Table 3).

Comparison of Method 9 OpacityObservations and Triboelectric Signalsduring Particulate Introduction — A com-parison was made between the Method 9opacity observations and triboelectric signaloutput during the Nov. 16 surrogate particu-late introduction. Method 9 observation datacorrelated very well for the four particulateintroduction periods on Nov. 16. Figure 2

illustrates the linear regression curves of theMethod 9 opacities to the triboelectric signalsfor the four introduction periods.

The Method 9 opacity points are for one-minute averages of 15-second readingsaccording to U.S. EPA Method 9. Triboelectricsignals are also based on one-minute averagesignal measurements for the correspondingperiod. Linear regression of Method 9 opacityto the triboelectric signals produced the equa-tions, with associated R2 values, presented inTable 4. Regression analyses of the B-1 and B-2 probes with the Method 9 opacity observa-tions showed similar correlations.

During the surrogate particulate compari-son conducted on Nov. 16, the weatherchanged. The sky became overcast; the tem-perature was 38°F at the start of the compar-ison, and snow flurries began by the end ofthe fourth test.

Comparison of COM and TriboelectricSignals during Particulate Introduction —Throughout the test period, the COM indicateda base level opacity at 3 to 4% while accordingto simultaneous Method 9 readings there wasno observable opacity in the exhaust plume.The false-positive baseline opacity was report-ed by the COM the entire day.

For comparison purposes, the same tribo-electric signal data was plotted against theCOM data. The COM opacity to the tribo-electric signals (A-1, A-2 probes) for the fourtests on Nov. 16 were plotted as logarithmicand linear regression curves (Figure 3). TheCOM data points are one-minute averages ofCOM continuous opacity output signal, andthe triboelectric data points are one-minuteaverages from that system's signal output.

The respective regression curve formulaefor the triboelectric to COM signal compari-son, with the associated R2 values, are pre-sented in Tables 5 and 6.

VE to Triboelectric Signal Regression andCorrelation Data

CorrelationLocation Regression (R2)

Test No. 1, Probe A-1 y = 1.5634 x – 0.1031 0.9646

Test No. 2, Probe A-1 y = 1.155 x + 7.0787 0.8210

Test No. 3, Probe A-2 y = 1.1969 x + 0.4331 0.9734

Test No. 4, Probe A-2 y = 1.358 x + 2.2674 0.9791

Table 4

Particle Size Distribution Summary

10% 25% 50% 75% 90%

EAF dust <1.235 <1.581 <1.874 <2.107 <2.276(µm)

Fe2O3 <1.022 <1.328 <1.780 <2.346 <2.819(µm)

Table 3

Linear regressions of Method 9 observationsto triboelectric signals on Nov. 16: (a) A-1probe, Test No. 1; (b) A-1 probe, Test No. 2; (c)A-2 probe, Test No. 3; and (d) A-2 probe, TestNo. 4.

Figure 2

(a) (b)

(c) (d)


The data plots for the comparison of thesesignals do not appear as linear as the Method 9opacity comparison (Figure 3). Generally, theR2 values for the regressions are much higherfor the logarithmic regressions than for the lin-ear regressions. The variation of the compar-isons is somewhat related to the variability ofthe COM data as compared with Method 9observations taken on a real-time basis.

Figure 4 shows a comparison of COM andMethod 9 data taken on Nov. 16 during thesurrogate particulate introduction periods.Regression analysis of the data collected onNov. 15 during the set-up work indicated thesame type of relationship. The correlationshows that the COM readings are generallyhigher than the Method 9 observations.

Evaluation of the Regulatory Relationshipof COM and Method 9 Opacity to theTriboelectric Signals — Compliance with the3% opacity standard is based on an average ofopacities measured or observed during a six-minute interval. Comparison of COM andMethod 9 data to the triboelectric signals usinga six-minute average provides a different statis-tical population of points for comparison. The

data averaged over six min-utes for the respectiveinstrument or methoddiminishes some of theswings measured on aninstantaneous basis. TheCOM and triboelectricinstruments take measure-ments on a frequency oftwo times per second, andMethod 9 measurementsare made every 15 seconds.Using measurements for allof the probe groups, aver-aged over six minutes, log-arithmic regression analysisof triboelectric to COM sig-nals produced the curveand formula presented inFigure 5, with an R2 valueexceeding 84%. All datapoints are based on real-time measurements.

Analysis of Method 9opacity to triboelectricsignals produced a linearregression (Figure 6) sim-ilar to analyses of individ-ual probe groups com-pared to their respectiveMethod 9 observations.The distribution of the data does not providea basis for logarithmic regression, and theresulting linear regression function has an R2

value of 63%.

Logarithmic and linear regressions of COMmeasurements to triboelectric signals on Nov.16: (a) Test No. 1, A-1 probe; (b) Test No. 2, A-1 probe; (c) Test No. 3, A-2 probe; and (d) TestNo. 4, A-2 probe.

Figure 3

Comparison of COM to Method 9opacity during surrogate particulateintroduction on Nov. 16.

Figure 4

COM to Triboelectric Signal Linear Regressionand Correlation Data

Linear Correlation Location regression (R2)

Test No. 1, Probe A-1 y = 1.878 x + 22.485 0.4054

Test No. 2, Probe A-1 y = 0.7932 x + 31.156 0.3311

Test No. 3, Probe A-2 y = 1.4231 x + 8.8582 0.8854

Test No. 4, Probe A-2 y = 1.8411 x + 13.226 0.7561

Table 5

COM to Triboelectric Signal LogarithmicRegression and Correlation Data

Logarithmic Correlation Location regression (R2)

Test No. 1, Probe A-1 y = 24.205 Ln(x) – 7.1213 0.5782




Table 6

Regression analysis of six-minuteaveraged Method 9 opacity and tri-boelectric measurements.

Figure 6

Regression analysis of six-minuteaveraged COM and triboelectricmeasurements.

Figure 5

(a) (b)

(c) (d)


Given the above comparison relationships,COM readings appeared to be logarithmic inrelationship to triboelectric signals, whileMethod 9 opacity observations were linear inrelationship to triboelectric signals. Theserelationships can be better understood whenthe following observations are applied to thedata:

• The relationship of opacity (particu-late) in air is a logarithmic relationshipas established by the Beer-LambertLaw (light penetration in a fluid). Asemissions increase, opacity can neverexceed 100%.

• The triboelectric system is more of aparticle counter than a COM, andthough both are measuring surrogateparameters for mass emissions, theparticle counter is closer to a measure-ment of particles. Thereby, a logarith-mic relationship could be expectedbased on the Beer-Lambert Law.

• The relationship of opacity to tribo-electric signals is more linear at opaci-ties below 40%.

• The triboelectric system programmingcan be set to focus on lower versushigher levels of particulates passing theprobe. In the case of an EAF where thestandard is basically no visual emissions(<3% opacity), the triboelectric systemfocus is directed toward low particulateconcentrations.

• Method 9 opacity readings at 15-sec-ond increments eliminates much of thevariability incorporated in COM meas-urements made every 0.5 seconds.

• The 5% band for visual opacity meas-urements also dampens the logarith-mic nature of COM readings.

A review of the regression curves indicat-ed several items with respect to the perform-ance of triboelectric signal measurementswhen compared to opacity as measured byCOM or Method 9.

• The slope of the linear regressioncurves for the respective plenumprobes (A-1 to B-1 and A-2 to B-2) arevery similar. This indicated that thedust introduced into the respectiveplenums was generating a similarresponse from the respective probes,even though there was dilution associ-ated with the increased air volume asthe dust traveled down the plenumtoward the stack.

• The logarithmic and linear regressioncurves for the respective plenumprobes, by location (A-1 to A-2 and B-1 to B-2), are similar. The probe rela-tionships reflect the relative location of

the dust introduction point, which isupstream of the B probe groups. Thissimilarity in signal response from oneplenum to another supports the repro-ducibility of the data and helps identi-fy the location of the particulate intro-duction point.

• The logarithmic regression relation-ship of COM data and the linearregression relationship of Method 9data are indicative of the system's lin-ear performance in lower opacityranges. This is expected for the EAFmonitoring installation, because thesystem manufacturer configured it tomonitor more effectively in the relativeopacity range below 40%.

• The logarithmic correlation betweenhigher COM opacities and triboelectricsignals indicated that triboelectric sig-nals, as a percent of scale, are rela-tively insensitive to opacities above40%. Stated another way, triboelectricsignals will vary proportionately toopacities below 40% with the scale set-ting used by the manufacturer for thiscomparison test. The triboelectric scalepercent of 70 to 100% will indicateincreases and decreases in opacity.However, the indication of opacitiesbetween 40 and 100% will be con-tained within the 70 to 100% triboelec-tric signal range.

Recognizing that low-level COM readingswere not absolute (due to moisture interfer-ence) during the test comparison period onNov. 16, but were related to increases ordecreases in opacity, comparisons betweenCOM and triboelectric signals showed verygood correlation and reproducibility duringthis test period.

Comparison of COM and TriboelectricSignals during Normal Operations — Aspreviously noted, problems were repeatedlyexperienced with the COM during the 45-daycomparison period. Despite these problems,Ipsco determined to proceed with the com-parison study using the originally scheduledperiod, recognizing that these problemswere representative of actual circumstancesencountered by EAF operators who attemptto operate a COM on the stack of an EAFbaghouse.

Given the erratic behavior of the COMfrom Oct. 13 through Nov. 9, Method 9observations were made during all daylightperiods when the COM was indicating a six-minute average opacity of 3% or greater. Thisobservation data indicated that there wasonly one actual period of six minutes with an


average opacity >3% (4.8% opacity).However, the COM showed extensive peri-ods of opacity >3% (9,300 minutes) duringthis period.

A review of the data indicated that theCOM indicated false high opacities of >3%for a total of six days and portions of a dayon three other days when no Method 9observations supported these readings. Themajority of these false high readingsoccurred during the period between Oct. 12through 18, and could be associated with theinitial calibration on Sept. 25. The inflatedreadings ceased after Oct. 18, when a recali-bration and adjustment was done by a sec-ond technician. When surrogate particulatewas introduced on Nov. 1, 2000, it was foundthat the COM was completely insensitive tothat particulate introduction. It could beassumed that the 0% opacity readings, asso-ciated with insensitivity, began at or sometime after the reprogramming performed bythe manufacturer's representative on Oct. 18.The fact that Method 9 observations contin-ued to indicate opacities <3% throughout thisperiod (Oct. 18 through Nov. 1) provided afalse sense of assurance that the COM infla-tion of opacities had been corrected. In fact,the COM was most likely blind for most ofthis period. This fact remained undetected,because the COM should have recorded 0%opacities, as did the Method 9 observers dur-ing this same time period.

The COM was again recalibrated on Nov.6, 2000, and the unit continued to operate ina manner that better reflected actual emissioncharacteristics throughout the period of Nov.6-25. Some inflation of opacity was still aproblem due to moisture interference.However, these differences were less fre-quent and corresponding Method 9 observa-tions identified those nonrepresentativeopacities indicated by COM measurements>3% opacity.

By comparison, the triboelectric signalswere also tracked on a continuous basis. Asa result of the surrogate particulate introduc-tion tests, it was verified that the triboelectricsystem tracked both visible and invisible par-ticulate. Therefore, to evaluate normal oper-ation data some signal levels with anassigned significance had to be established.

In Ipsco's initial presentation of an operat-ing and response system based on use of thetriboelectric broken bag detection system tothe U.S. EPA, several signal ranges were pro-posed. Arbitrary levels were picked for illus-tration in the presentation. However, for pur-poses of this comparison, it was necessary tobetter quantify those ranges. The signalranges in Table 7 were used for summarizingdata tracked by the triboelectric system.

These ranges are for com-parison purposes and donot necessarily reflect thefinal ranges that will beimplemented by Ipsco forbaghouse and broken bagdetection system opera-tion. During the time peri-od of Oct. 12 throughNov. 7, the scale factor forthe triboelectric signalswas set at 200 picoamps. The percentagesindicated in Table 7 are percent of scale asdisplayed and recorded by the triboelectricsystem computer and would typically applyto the scale factor of 200 picoamps.

For the A-1, B-1 probe groups during thetime period of Oct. 12 through Nov. 7, therewere several signal periods that exceededthe caution range (50 to 69%), lasting a totalof 43 minutes. The alarm range for the A-1,B-1 probe groups was exceeded for a totalperiod of six minutes during this time peri-od. By comparison, the signals for the A-2,B-2 probe groups during the same periodhad a total time of 1187 minutes above thecaution range, but had no signals in thealarm range.

There were no exceedances of the 3%opacity standard (as measured by Method 9)during this period, with the exception of onesix-minute period on Oct. 13. The projectedtriboelectric signal ranges at the 200 picoampscale factor were indicative of the actualemission performance of the baghouse.

The historic record of triboelectric signalswas averaged for the plotting purposes. On areal-time basis, the system takes two meas-urements per second. However, the operatorcan select the averaging period that thesemeasurements are plotted by the system.During the time period of Oct. 12 throughNov. 7, data were tracked and plotted usinga six-minute average of measurements at thescale factor of 200 picoamps. At noon onNov. 8, the averaging period was reduced toa one-minute average to allow for bettercomparison to Method 9 observations andCOM tracking done during the particulateintroduction tests conducted on Nov. 15-16.The scale factor was adjusted to 1500picoamps on Nov. 8, to allow for better visu-al evaluation of the signals. The 1500picoamp scale factor reduced normal signalactivity to the bottom third of the operator'svisual screen display. This visual change inreal-time curve representation of triboelectricsignals increased the operators ability toidentify abnormal spikes quickly. The one-minute average and 1500 picoamp scale fac-tor were used for plotting the data for theperiod from Nov. 8-25. For comparison of

Proposed Signal Rangesfor the TriboelectricBroken Bag DetectionSystem

Normal operating 0 to 49%range

Caution range 50 to 69%

Alarm range >70%

Table 7


normal operating data, a different set of nor-mal, caution and alarm ranges needed to beestablished.

Given the linear regression analysis of thesurrogate particulate comparison done onNov. 16, those functions provided a basis forcalculating an approximate triboelectric signalthat is equivalent to 3% opacity. Using linearregression for the opacities below 40% pro-vided a conservative margin for correlating tri-boelectric signals to opacity, whether meas-ured by COM or Method 9 observations. Thelinear regression curves for the respectiveprobe points are summarized in Table 8.

Using the four regression curves, the tribo-electric signal ranged from 20 to 35% for all ofthe probe groups—as a percent of scale that isequivalent to 3% opacity. It is important tounderstand that the signal from the respectiveplenums groups (A-1, B-1 and A-2, B-2) aresimilar, but are different for the comparisonperiod on Nov. 16. This is expected, becauseeach plenum has a baseline triboelectric char-acteristic that is somewhat different. Eachrespective baghouse compartment provided aslightly different baseline particulate load whenthe EAF is operating, as was demonstrated onNov. 16. This behavior is evidenced by the lin-ear regression formulae. The slope for the A-1,B-1 probes is approximately 1.0, while that forthe A-2, B-2 is approximately 1.5. The interceptfor the A-1, B-1 probes is about 31 while thatfor the A-2, B-2 probes is about 16.

Given the slight difference in probe groupreadings with respect to COM opacity, data

collected after Nov. 8 can be evaluated usingrelationships established during the surro-gate particulate comparison conducted onNov. 16. Recognizing that the triboelectricsignals included both visible and invisibleparticles, an alarm range equivalent to 3%opacity was selected as indicated in Table 9for the respective probe groups, and is appli-cable when using the scale factor of 1500picoamps.

For the A-1, B-1 probe group during thetime period of Nov. 8-25, there were severalsignal periods that exceeded the cautionrange (30 to 37%), lasting a total of 52 min-utes. The alarm range for the A-1, B-1 probegroups was exceeded for a total period oftwo minutes during this time period. Bycomparison, signals for the A-2, B-2 probegroup during the same period had a totaltime of 15 minutes above the caution rangeand had no signals in the alarm range.

When reviewing the data for the triboelec-tric signals it is evident that there was verylittle time when the triboelectric signals werein the alarm range. A total of two minuteswas tracked on Nov. 22 when the triboelec-tric signals for the A-1, B-1 probes wereabove the alarm level of 38%. Method 9observations and COM data indicated thatopacity of the stack emission was below 3%for that operating day.

From this comparison study, it was deter-mined that triboelectric signals are indicativeof baghouse performance and compliancewith a visual emission standard of 3% opaci-ty. Furthermore, triboelectric signals allowoperators to identify the baghouse compart-ment zone that contributes to increases intriboelectric signals while the baghouse con-tinues to operate in compliance with its 3%opacity standard. Interface signal informationfrom the baghouse PLC that controls the rowsequence of bag cleaning allows the opera-tor to identify the contributing compartmentas well as the row of bags contributing to theincrease of particulate being discharged tothe exhaust plenum. Given this information,the operator can take corrective actionsbefore the particulate contribution can con-tribute to a violation six minutes in length.

No such advance information is availableto baghouse operators from COM or Method9 stack exhaust plume observations. On thecontrary, with the difficulty in maintaining acalibrated COM the operator must alwaysconsider the condition of greater than 3%opacity to be a false positive and investigatethe COM before undertaking any meaningfulcorrective action on the baghouse.

Triboelectric signals remained primarilywithin the normal range selected for thisevaluation, with some periods of signals

Regression Curves under NormalOperating Conditions x = COM opacity (%) and y = triboelectric signal (% of scale)

Probe B-2 y = 1.479 x + 18.723

Probe A-2 y = 1.562 x + 14.806

Probe B-1 y = 1.0083 x + 31.33

Probe A-1 y = 0.71393 x + 31.735

Table 8

Signal Ranges for the TriboelectricBroken Bag Detection System for EachProbe Group

Probes group A-1, B-1 A-2, B-2

Normal operating 0 to 29% 0 to 24%range

Caution range 30 to 37% 25 to 29%

Alarm range >38% >30%

Table 9


increasing into the caution range and oneshort period when the signal exceeded thealarm level. Given that the alarm range andthe associated caution range were basedupon the respective regression curves for theplenum probe group, the regression curveappeared to predict the level of 3% opacityin a conservatively low manner. This provid-ed additional assurance that the 3% opacitywould not be exceeded if the study tribo-electric comparison ranges were actuallyused for baghouse operation.

The time duration of the triboelectric sig-nals may also be a factor in relating tribo-electric signals to relative opacity. As boththe COM and the Tribolink comparison relyon the size, quantity and structure of parti-cles to make their measurements, the quan-tification of a time variable with respect totheir relationship may not be significant. Thetriboelectric measurement includes particlesthat are invisible to a COM and therefore willalways have a signal duration factor relatedto this invisible portion of particulate. Nostudy work was done to specifically quantifyany time-related variable between triboelec-tric signal and COM opacity. However, usinga relatively short averaging period for the tri-boelectric signal will mitigate the differencein the duration relationship associated withthe invisible particles. The one-minute aver-aging period used for the data tracked afterNov. 8 provided a real-time historical curvethat was indicative of opacity measured byCOM and Method 9.

EAF Steel Production During theComparison Test — During the comparisonperiod, EAF steel production continued in anormal fashion. Beginning with Oct. 12through Nov. 25 (45-day period), steel wasmelted every day. As is typical of steel melt-ing operations, the number of heatsprocessed on a given day varied. The rea-sons for increased or decreased productionvary due to a variety of factors. However, itis important to assure that productionoccurred in a representative fashion duringthis comparison period. Daily EAF and LMFmelting records were collected and reviewedfor the entire period of the comparison test.From this review, the meltshop's productionduring the comparison test period was con-cluded to be representative.

Findings and Conclusions It is important to note that the data, correla-tions and formulae established for the IpscoSteel, Montpelier Works baghouse are some-what unique to the facility. The site-specificnature of particulate generated by an EAF is

unique to that particular furnace and its fumecollection and control system. Many vari-ables effect the size, structure and quantityof particulate generated by a meltshop.These variables include: scrap type andcharge practices, type of steels produced,length of heat cycle, slag practices, use ofcarbon and limestone, length of fume controlduct, use of spark boxes, type of scrap pre-heating and type of offgas cooling.

General principles for the correlation andcomparison of COM and Method 9 opacity totriboelectric signals are generally applicableto all EAF baghouses. Without evaluating thisrelationship in other EAF operations, anduntil such data are generated, it should bepresumed that the specific correlation formu-lae between opacity and triboelectric signalswill vary from shop to shop.

The findings and conclusions of this studyare as follows:

• Both the COM and Tribolink systemstracked increases and decreases inbaghouse particulate emissions.

• Both the COM and Tribolink systemsare indicative of particulate present inthe gas stream, although neither sys-tem is able to generate mass emissionrates or concentrations in this applica-tion.

• The Tribolink system is able to detectchanges in particulate levels that areinvisible to COM, and as such can pro-vide the baghouse operator informa-tion on the deterioration of bags longbefore the emissions become a viola-tion of the opacity standard.

• The Tribolink system operated at theIpsco Steel, Montpelier Works was con-figured by the manufacturer to focus onthe lower opacity range of emissions,and this bias was observed during thesurrogate particulate introduction peri-ods. The Tribolink signals displayed amore linear relationship to COM opaci-ty at opacity levels below 40% and werelinear to Method 9 opacities observedduring the comparison.

• During the surrogate particulate com-parison, the Tribolink system responseto opacities above 30 to 40% tended tobe logarithmic in nature, and opacitiesas high as 80 or 100% would be trackedby the system as a percent of scale, withindicated levels still below 100% ofscale. This does not appear to be aproblem, because the NSPS for the EAFis a six-minute average below 3% andthe system had a linear response char-acteristic with both COM and Method 9opacities in this low range.


• Using the linear regression curves gen-erated during the surrogate particulatecomparison on Nov. 16, the triboelec-tric signal level associated with 3%opacity was predicted to be 25 to 30%of the scale when the scale factor wasset at 1500 picoamps. Using the spe-cific linear regression curves generatedfor the respective plenum groups, thedata was evaluated against a range ofsignals that had an alarm range set atthe predicted 3% opacity for therespective plenum group. Method 9and COM data collected during normaloperations supported these levels asindicative of the actual opacity ofemissions.

• The comparison ranges of triboelectricsignals used for evaluating the datafrom Nov. 8-25 showed that theTribolink instrument was able to char-acterize the baghouse performancewhile still being indicative of compli-ance with the 3% opacity standard.The triboelectric signal was foundabove the alarm level for only twominutes on Nov. 22 during this com-parison test. The COM and Method 9observations on Nov. 22 did not indi-cate any exceedances of the 3% opac-ity standard. Therefore, it can be con-cluded that the regression curve pre-dicted a 3% opacity, which is conserv-atively low when compared to actualopacity as measured by either COM orMethod 9.

• As the Tribolink system probes wereinstalled within specific zones of thedischarge plenums of the baghouse,they served as a better indicator of theactual area performance for baghousecompartments where the particulateincreases were originating. COM wasunable to provide this type of area per-formance indication because of itsstack location. The Auburn system abil-ity to interface with the baghouse PLCfurther enhanced area identificationcapability of the Tribolink system. TheAuburn computer/PLC interface cantrack the row and compartment of bagsbeing pulse cleaned at any given time,and this information can be displayedwith the alarm log data that indicatedincreased particulate above the pro-grammed caution or alarm levels. This

identification feature significantlyimproved the operator's ability to rec-ognize and react to problems longbefore they become visible opacity vio-lations.

• The Tribolink system generallyallowed the baghouse operator to beproactive in response to changes inbag filtering performance as the sys-tem was able to identify increased par-ticulates that are invisible to COM andMethod 9 observers.

• Even with repeated calibrations by themanufacturer, it was difficult to getgood performance from the COM foropacities below 5% during this com-parison study. By comparison, theTribolink system operated throughoutthe comparison period without addi-tional manufacturer recalibration.

• The Tribolink system has the abilityto modem interface with the manu-facturer's home office. This featurewas installed on the Auburn system atMontpelier Works. It allowed for trou-ble-shooting assistance to beaccessed by operators within a matterof minutes during normal businesshours. A modem interface for theCOM is not a feature available for thatinstrument.

Triboelectric signals can be correlated toopacity, and particles of less than 1µm canbe tracked by the system. Generally, the lin-ear comparison function for triboelectric sig-nals to opacity data is most directly applica-ble for opacities below 40%. This correlationrange presents no problem for EAF baghouseoperations that must comply with a 3% opac-ity standard. The Tribolink system provideda more precise and proactive tool for bag-house operators to maintain compliance withthe zero visible emission standard imposedby the 3% opacity level.

References1. ASTM D 6216-98, Standard Practice for

Opacity Monitor Manufacturers to CertifyConformance with Design and PerformanceSpecifications, ASTM International, WestConshohocken, Pa., 1998.

2. Office of Air Quality Planning and Standards(OAQPS) Fabric Filter Bag Leak DetectionGuidance, U.S. Environmental Protection Agency,Office of Air Quality Planning and Standards,Research Triangle Park, N.C., Sept. 1997, EPA-454/R-98-015. AISE

Comparison of Continuous Opacity Monitor and …Opacity Monitor and Method 9 Observations to...

Documents

Transcript of Comparison of Continuous Opacity Monitor and …Opacity Monitor and Method 9 Observations to...