m20 - quality assurance of continuous emissions monitoring systems-geho1105bjxz-e-e

Technical Guidance Note(Monitoring) M20

Quality assurance of continuous emission monitoring systems - application of BS EN 14181 and BS EN 13284-2

Environment AgencyVersion 1

September 2005

Foreword

Technical Guidance Document M20 is issued by the Environment Agency as one of a series providingsupport to its regulatory officers, process operators, and those with interests in stack emissionsmonitoring. M20 has been written to support the application of a new CEN standard, BS EN 14181,Stationary source emissions quality assurance of automated measuring systems, and a supportingstandard, BS EN 13284-2, Stationary source emissions - Determination of low range massconcentration of dust - Part 2: Automated measuring systems.

This TGN provides guidance on:

The monitoring requirements of EU Directives for large combustion plant and wasteincineration.

Choosing CEMs that meet the performance requirements of these EU Directives. Demonstration of suitability through MCERTS product certification. Determining whether CEMs meet the uncertainty allowances specified by the above EU

Directives.

Assuring that CEMs are located in the optimum position. Functional tests to assure that CEMs have been installed and are operating correctly. Calibration using a Standard Reference Method (SRM). On-going surveillance to assure the correct operation of CEMs, by examining drift and

precision during continuous operation.

Annual surveillance tests of CEMs.

The TGN is in five sections, covering:

The regulatory background to BS EN 14181 and BS EN 13284-2. The suitability, selection and installation of CEMs. The functionality and calibration of CEMs. On-going quality assurance of CEMs. Annual surveillance tests for CEMs.For anybody carrying out work under BS EN 14181 and 13284-2, this TGN should be read inconjunction with the standards and their associated Method Implementation Documents (MIDs),which the Environment Agency has published.

Feedback

Any comments or suggested improvements to this TGN should be e-mailed to Richard Gould [email protected]

Record of Amendments

Version number Date Amendment

Version 1 publishedSeptember 2005

Contents

1. General guidance on quality assurance and calibration ............................................................. 11.1 Introduction............................................................................................................................................... 11.2 Regulatory framework and standards for monitoring ............................................................................... 11.3 Scope and structure of BS EN 14181 ....................................................................................................... 21.4 Roles, responsibilities and delegation of responsibilities.......................................................................... 4

2. Suitability of CEMs (QAL1) and MCERTS................................................................................. 52.1 Basic rules for selecting CEMs................................................................................................................. 52.2 Suitable ranges.......................................................................................................................................... 62.3 Selection procedures for CEMs and sampling systems ............................................................................ 7

3. Calibration and validation of the CEM (QAL2) ............................................................................ 83.1 QAL2 requirements .................................................................................................................................. 83.2 Location of CEMs..................................................................................................................................... 83.3 Management system requirements............................................................................................................ 93.4 The functional tests................................................................................................................................... 93.5 Calibration and validation....................................................................................................................... 113.6 Example QAL2 calculations and variability checks ............................................................................... 173.7 Frequency of QAL2 checks .................................................................................................................... 173.8 Performing an AST instead of a QAL2 .................................................................................................. 173.9 Significant changes to operating conditions and fuels ............................................................................ 173.10 Repeated exceedences of the calibration range ................................................................................... 18

4. Ongoing quality assurance during operations (QAL3)................................................................ 184.1 QAL3 - general ....................................................................................................................................... 184.2 Zero and span checks.............................................................................................................................. 194.3 Use of control charts ............................................................................................................................... 20

5. Annual surveillance test (AST).................................................................................................... 225.1 Purpose of the AST................................................................................................................................. 225.2 Functional tests ....................................................................................................................................... 225.3 Parallel measurements with a SRM ........................................................................................................ 225.4 Example AST determination................................................................................................................... 23

6. Status of this guidance.................................................................................................................. 23

References ............................................................................................................................................ 24

Glossary................................................................................................................................................ 25

Appendix 1: Example QAL2 calibration functions and variability tests ....................................... 26A1.1 CEM measuring SO2 ........................................................................................................................... 26A1.2 CEM measuring NO............................................................................................................................ 30

Appendix 2: Example Shewhart and CUSUM control charts......................................................... 35A2.1 Shewhart chart..................................................................................................................................... 35A2.2 CUSUM chart...................................................................................................................................... 35

Appendix 3: Example AST determination........................................................................................ 37

TGN M20 Version 1 September 2005 Page 1 of 39

Quality assurance of continuous emission monitoring systems applicationof BS EN 14181 and BS EN 13284-2

1. General guidance on quality assurance and calibration

1.1 IntroductionThe primary role of this Technical Guidance Note (TGN) is to provide guidance on the application ofEuropean Standard BS EN 14181, Stationary source emissions Quality assurance of automatedmeasuring systems1 and a supporting standard, BS EN 13284-2, Stationary source emissions -Determination of low range mass concentration of dust - Part 2: Automated measuring systems2 oninstallations falling under the European Directives for the incineration of waste (2000/76/EC3), andlarge combustion plants (2001/80/EC4). In this TGN these Directives will be abbreviated to the WIDand LCPD respectively.

Both Directives specify the use of international and national standards for monitoring, and defineperformance requirements for CEMs through specified uncertainty budgets or allowances for accuracyand precision. BS EN 14181 and BS EN 13284-2 were developed to support the quality assurance ofmonitoring requirements for these Directives, by including provisions to ensure that CEMs meet therequired performance specifications both before and after installation.

For simplicity, throughout this document reference to BS EN 14181 will also refer to BS EN 13284-2.

Key point

This TGN summarises the requirements of BS EN 14181 and BS EN 13284-2, andprovides guidance on how to perform each of the required tasks. For anybodycarrying out work under BS EN 14181 and 13284-2, this TGN should be read inconjunction with these standards and their associated Method ImplementationDocuments (MIDs).

1.2 Regulatory framework and standards for monitoring

1.2.1 Monitoring requirements in the WID and LCPD

The WID and LCPD specify performance standards for monitoring in two ways. Firstly, the Directivesprescribe the use of CEN standards for monitoring and calibration, or if CEN standards are notavailable, then the use of ISO, national or other equivalent international standards that will providedata of a suitable quality. Many of these standards include performance specifications for CEMs.

Secondly, these EU Directives specify overall performance requirements for both continuous anddiscontinuous monitoring through uncertainty allowances expressed as a percentage of the ELV.These uncertainty allowances are expressed as a 95% confidence interval (CI) and specify overallrequirements for accuracy and precision.

Key Points

Monitoring under the WID and LCPD must be performed according to therequirements of CEN standards, or ISO, national or other international standardsif CEN standards are not available. The standards are stated in TGN M2.

The WID and LCPD specify requirements for monitoring accuracy and precisionthrough 95% confidence intervals.


1.3 Scope and structure of BS EN 14181BS EN 14181 applies only to CEMs permanently installed at WID and LCPD installations. It does notapply to portable CEMs, such as those used in SRMs, or CEMs used in PPC installations outside thescope of the WID and LCPD. The requirements for such CEMs, or SRMs which used instrumentaltechniques, are described in other applicable standards for monitoring.

Also BS EN 14181 only applies to the CEMs themselves and not the data collection and recordingsystems used with CEMs.

BS EN 14181 specifies three quality assurance levels (QALs) and an annual surveillance test (AST).These are:

QAL1 A procedure to demonstrate that the CEM is suitable for the intended purposebefore installation, by meeting required performance standards and theuncertainty allowances specified in EU Directives.

QAL2 A procedure to calibrate the CEM once it has been installed, using SRMs andthen verify whether it still meets the required uncertainty allowances, onceinstalled.

QAL3 A procedure to maintain and demonstrate the required quality of the CEMduring its normal operation by checking the zero and span readings.

AST A procedure to evaluate the CEM to show that it continues to function correctlyand the calibration function is still valid.

These quality assurance levels follow a logical sequence to demonstrate the suitability of the CEM, itscorrect installation, commissioning, and calibration, followed by continuing and correct operation (seeFigure 1).

Figure 1 the sequence of quality assurance levels in BS EN 14181

QAL 1

The first level of quality assurance, QAL1, demonstrates the potential suitability of the CEM before itis installed on a stack. In England and Wales, MCERTS product certification at an appropriatecertification range (see sections 2.2 and 2.3) is taken as evidence of compliance with the QAL1requirements.

SuitableCEMs

CorrectInstallation &Calibration

ContinuingFunctionality

QAL1 QAL2

QAL3

AnnualSurveillance

Test


The WID and LCPD specify uncertainty allowances expressed as 95% Confidence Intervals. Whilstthe procedure for determining uncertainties is described in BS EN ISO 149565, making use ofperformance testing data, the procedure is both complex and prone to differing interpretations.Therefore the Environment Agencys approach is to make use of the known linear relationshipbetween uncertainty allowances and certification ranges, whereby the suitability of a CEM isdetermined by using certified ranges. This simplified yet proven approach is described is section 2.

QAL 2

The second level of quality assurance, QAL2, specifies procedures to verify that the CEM has beeninstalled correctly, calibrated using SRMs and that the CEM still meets the uncertainty requirementsof the EU Directives. The correct installation and functionality of the CEM is verified throughinspection and through a set of functional tests, examining performance characteristics such asresponse time and linearity.

SRMs are used to calculate a calibration function for the CEM. The uncertainty of the installed CEMis then determined by calculating the variability of the calibration function. The effectiveness of thistest requires at least fifteen repetitions of each applicable SRM spread out over at least three days.

Once installed, the uncertainty of the CEMs measurements may increase due to specific (andsometimes unique) factors at the installation, such as the position of the CEM or its sampling system,environmental conditions, stack gas conditions and uncertainties of calibration gases. Therefore,depending on the intended application, care must be taken in choosing a CEM to ensure it will meetthe QAL 2 requirements at that specific installation.

The QAL2 procedures are carried out when:

the CEM is installed; then at least every three years (under the WID) or five years (under the LCPD); whenever there is a significant change in plant operation which changes the emissions; after a failure of the CEM so that significant repair is required and affects calibration; after an significant upgrade or other significant change to the CEM affecting calibration.Any changes that do not affect the calibration of the CEM will not require a repeat of the QAL2procedure. Further guidance on significant changes is given in section 3.8.

QAL 3

The QAL3 procedure ensures that the CEM remains within the required specifications duringcontinued use. QAL3 achieves this by requiring the plant operator to regularly measure the drift andprecision of the CEM. This data is then plotted using control charts such as Shewhart or CUSUMcharts and then using the outputs of these charts to determine when the CEM needs maintenance.The frequency of the drift measurements depends on the maintenance interval determined during theMCERTS6 performance tests, and can be anything from one week to several months. The use ofCUSUM charts, however, requires drift tests to be carried out at least weekly. In many CEMS theQAL 3 tests are conducted automatically within an instrument and therefore occur more frequently.

AST

The annual surveillance test (AST) is a mini-QAL2 test. Its key objective is to check whether thecalibration function determined during the QAL2 tests is still valid. The AST consists of the samefunctional tests as those used in QAL2, but the calibration function is checked by using a smallernumber of repetitions of the SRMs (typically five repetitions). If the calibration function is still valid,


then no further action is required. If the AST shows that the calibration function is no longer valid,then a full QAL2 is required.

1.4 Roles, responsibilities and delegation of responsibilities

1.4.1 Roles and responsibilities

The responsibilities of CEMs manufacturers or suppliers, test laboratories, process operators and theregulator under BS EN 14181 are shown in Table 1.

Table 1 Actions and responsibilities within BS EN 14181

Organisation Roles and requirements

CEMs manufacturers and suppliers Achieving and maintainingcertification of CEMs to the applicableMCERTS performance standards

Supplying, correctly installing,commissioning and maintainingappropriate, MCERTS-certified CEMsto applicable installations

When appropriate, co-operating withprocess operators and test laboratoriesto calibrate CEMs

Test laboratories Achieving and maintaining MCERTSaccreditation for the SRMs

Performing the SRMs for the QAL2and AST procedures

Either performing or supervising thefunctional tests specified for the QAL2and AST procedures

Process operators Using CEMs certified to theappropriate MCERTS performancestandards

Performing the QAL3 procedures

Submission of QAL2, QAL3 and ASTreports as required by the regulator

Maintaining QAL3 records, otherrecords and information as specifiedwithin BS EN 14181, and retainingQAL2 and AST reports for periodsspecified by the regulator

Regulators Specifying BS EN 14181 requirementswithin permits or variations to permits

Checking operator compliance


Key Points

Process operators have overall responsibility for complying with BS EN14181.

The SRMs for the QAL2 and AST must be performed by test laboratoriesaccredited to ISO 17025 for the MCERTS performance standard for stackemission monitoring.

1.4.2 Delegation of roles

The requirements of BS EN 14181 are complex and we recognise the need for both co-ordination andco-operation between all organisations involved in the work. Our preference is for test laboratories toundertake all of the activities specified in QAL2 and the AST. However, as CEMs suppliers oroperators may be required to input the calibration function into the CEM, and as test laboratories maylack the familiarity with a variety of CEMs to perform some or all of the functional tests, we allow thefollowing flexibility:

Operators or CEMs suppliers may perform the functional tests for the QAL2 and AST, asspecified in Annex A of BS EN 14181, provided that the test laboratory oversees and auditsthese activities according to the requirements of Annex A in BS EN 14181.

The operator or an independent third-party organisation may project manage the activitiesspecified in BS EN 14181, subject to the following requirements:

The project manager must be able to demonstrate independence from the CEMsmanufacturer and supplier.

The project manager commissions the reference monitoring and the activities specified inAnnex A of BS EN 14181.

Either the third-party project manager or the test laboratory may perform the statisticaltests specified in BS EN 14181.

The project manager is responsible for producing the QAL2 or AST report.

2. Suitability of CEMs (QAL1) and MCERTS

2.1 Basic rules for selecting CEMsThe following guidelines apply when selecting CEMs:

Determinands The CEM is to be MCERTS certified for the determinandsspecified in the WID or LCPD where continuous monitoring isrequired.

Ranges The CEM is to be certified for a range that is suitable for theapplication (see sections 2.2 and 2.3).

Stack gasconditions

The operator should ensure that specific site conditions do notreduce the performance of the CEM to below requiredstandards.

Proven suitability The operator is recommended to ensure that the intended CEMis proven on comparable installations.


Particulatemonitors

Generally, particulate monitors may be sensitive to changes inflow rate, particle size distribution and changes in particleshape. Therefore the operator should determine whetherspecific stack conditions could potentially undermine theintegrity of the monitoring data.

The reference materials used in the automatic or manual zeroand span check procedures (as required for QAL3) should bedocumented by the manufacturer and assessed as part of theMCERTS certification process.

2.2 Suitable rangesWhen CEMs are tested and subsequently certified, the MCERTS certificate states the certified range.In some cases a CEM may have more than one range.

In general, the lower the certified range, the better the performance of the CEM is likely to be. This isbecause the majority of performance standards are expressed as a percentage of the range. Forexample, if the performance requirement for cross-sensitivity is 4% of the range and a CEM has acertified range of 0 to 75 mg.m-3, then the cross-sensitivity will not be more than 4% of 75 mg.m-3,which is 3 mg.m-3. A CEM with a certified range of 0 to 200 mg.m-3 may have a maximum cross-sensitivity up to 4% of 200 mg.m-3, or 8 mg.m-3.

The main performance characteristic that is not range-dependent is linearity (or lack of fit). Thereforeas an extra assurance, if a CEM is to be used for higher ranges than those certified, CEMsmanufacturers should ideally have had the linearity evaluated over the higher ranges during MCERTStesting. If this is not the case, then the linearity over the higher ranges should be evaluated eitherbefore installation or immediately afterwards.

If there is any doubt about a CEMs performance for a particular application, reference should bemade to the MCERTS test results.

Key Point

Generally, CEMs with lower certified baseline ranges will perform satisfactorily athigher ranges, since the lower the certified range, the better the performance.

To simplify things the Environment Agencys approach for selecting suitable CEMs is to apply rangemultipliers, whereby the lowest certified range is not more than 1.5x the daily average (DA) ELV forincineration processes and not more than 2.5x the DA-ELV for large combustion plant and other typesof process. As there is a linear relationship between certified ranges and uncertainties, thesemultipliers provide assurance that CEMs with appropriate ranges will meet the uncertaintyrequirements specified in the WID and LCPD. This approach is the same as that employed inGermany and will be the same as that within the future CEN standard for the performance of CEMs,which is expected to be published in 2007.

The CEM shall also be able to measure instantaneous values over the ranges which are to be expectedduring all operating conditions. If it is necessary to use more than one range setting of the CEM toachieve this requirement, the CEM shall be verified for monitoring supplementary, higher ranges.

Key Point

When selecting a new CEM operators shall select a CEM with a certification rangewhich is not more than 1.5x the daily average ELV for WID installations and notmore than 2.5x the daily average ELV for LCPD installations.


2.3 Selection procedures for CEMs and sampling systems

2.3.1 CEMs already installed at a site

If CEMs already installed at an installation at the time the WID or LCPD permit is issued do not meetthe requirements for ranges in section 2.2, then the CEMs may still be used if they fulfil the QAL2 andQAL3 requirements of BS EN 14181. In simple terms, CEMs with ranges higher than those requiredstill may pass the QAL2 and QAL3 requirements, but the risk of failure increases as the certified rangeincreases.

If the CEMs do not meet the QAL2 and QAL3 requirements and cannot be adjusted or modified tofulfil the requirements, then the operator will be required to replace them within one year with CEMswhich do have suitable ranges based on the ELV multiplier rule.

2.3.2 New CEMs

New CEMs shall meet the requirements of the ELV multiplier rule.

Table 2 shows a selection of daily average ELVs for installations under the WID and LCPD, togetherwith the certification ranges and allowable uncertainties.

Table 2 Baseline ranges, ELVs and uncertainties

ELV,mg.m3

Certificationrange, mg.m-3

Allowableuncertainty,%

Allowableuncertainty,mg.m-3

NOx incineration 200 300 20% 40NOx large combustion plant, solid/liquid fuel 200 - 600 500 - 1250 20% 40 - 120NOx large combustion plant, gaseous fuels 200 - 300 300 - 750 20% 40 155NOx large combustion plant, gas turbines 50 - 120 125 - 300 20% 10 - 24SO2 large combustion plant, solid/liquid fuel 200 - 850 500 - 2125 20% 40 - 170SO2 large combustion plant, gaseous fuels 35-800 88 - 2000 20% 7 160SO2 incineration 50 75 20% 10CO incineration 50 75 20% 10HCl incineration 10 15 40% 4Particulate matter, large combustion plant 30 75 30% 9Particulate matter, incineration 10 15 30% 3Particulate matter, co-incineration 30 45 30% 9Total organic carbon, incineration 10 15 30% 3

NOTE 1: NOx is expressed as NO2. Therefore if a CEM measures NO alone, then the measurement must be convertedto a NO2 equivalent. For example, if the range for NO is 0 to 100 mg.m

-3, then the range for an NO2equivalent (or total NOx) will be 0 to 153 mg.m-3.

2.3.3 Sampling systems for extractive CEMs

Extractive CEMs comprise the analyser(s) and additional devices for obtaining a measurement result.As well as the analyser(s) this includes the sampling system. It is the complete system, including thesampling system, that has been tested and certified.

There are several types of sampling system, such as:

Simple heated lines coupled to heated analysers that measure gases in a hot, wet form. Heated lines and chiller-driers, delivering the sampled gases to the analyser in cooled, dry form. Heated lines and permeation-driers, delivering the sampled gases to the analyser in cooled, dry

form.


The stack-mounted probe is coupled directly to a permeation drier, which then passes the cooled,dry sample gas via an unheated line to an analyser.

There are also many variations of these basic forms and as analysers are typically designed for usewith specific types of sampling system, testing and subsequent approvals will certify a CEM with astated type of sampling system.

As industrial processes often differ in their requirements, some flexibility is allowed in the selection ofthe sampling system with the CEM. However, the installed CEM must not deviate from the type ofsampling system specified on the certificate to ensure the CEM is not degraded, such that it no longermeets the required performance specifications.

Such allowable variations could include:

A different length of sampling line to that which was tested. A different brand or model of sampling system, so long as there is evidence from third-party

testing that the alternative components meet the required performance specifications and havebeen tested on analogous systems.

Additional manifolds and heated valves used to allow more than one analyser to share a samplingsystem.

Key Point

MCERTS and BS EN 14181 have provisions for systems integration. As long assampling systems conform to the type originally tested and certified and there isevidence from third-party testing that the sampling system installed does notdegrade performance below the MCERTS requirements, then the alternativesampling system is permitted.

3. Calibration and validation of the CEM (QAL2)

3.1 QAL2 requirementsQAL2 requires operators to assure that CEMs are installed in the correct location, that there issufficient access to maintain, assess and control them, and to ensure that CEMs are both calibrated andoperating correctly. To this end, BS EN 14181 specifies two parts to QAL2, which are:

A set of functional tests and checks to ensure that the CEM has been installed correctly and isfunctioning at, or better than, the required performance levels.

A calibration and validation exercise, which consists of a set of parallel measurements usingSRMs, followed by a set of statistical operations and tests.

3.2 Location of CEMsOperators should follow the provisions for location and access described in TGN M17. Spatialvariations in temperature, pressure, flow rate and stack-gas concentration should not be greater than15% across the stack in any plane. Test laboratories should use ISO 103968 when characterising thestack gas conditions and assessing the intended location of the CEM.

Before installing the CEM, the stack gas must be characterised in order to determine whether there arevariations across the stack, such that the sampling position will have a significant bias on the readings.It is critical that CEMs are not only located at a point where there is access and other provisions for the


effective and continued operation of the CEM, but also in a location which complies with therequirements of applicable standards such as BS EN 13284-19 and ISO 10396. This is because thesample must be representative, or the bias in the readings will be so great that the CEM may not meetthe requirements of QAL2.

Additionally, the CEM must be located at a point where the sample is representative, and the SRM andthe CEM (or its sampling location) should be located so that they do not interfere with each other.

Key point

Spatial variations in temperature, pressure, flow rate and stack-gas concentrationshould not be greater than 15% across the stack in any plane. Test laboratoriesshould use ISO 10396 when characterising the stack gas conditions andassessing the intended location of the CEM.

3.3 Management system requirementsBS EN 14181 requires operators to have a systematic approach to managing and maintaining theCEMs, documented through procedures within an existing management system, such as those meetingthe requirements of BS ISO 900110 or BS EN ISO 1400111. Operators are also recommended to refer tothe requirements of BS ISO 1001212 for measurement management systems. Such procedures shouldinclude specific provisions for CEMs covering:

Selection. Maintenance and servicing. Responsibilities and training of personnel. Calibration, quality assurance checks and controls. Records and data management. Prevention of unauthorised adjustment of the CEM and its data recording devices. Maintaining availability spares, contingencies and back-up monitoring.

Key Point

Operators can use management systems based on BS EN ISO 9001, BS EN ISO14001 or BS ISO 10012 to provide for the management aspects of BS EN 14181.

3.4 The functional testsBS EN 14181 requires a set of functional tests to be carried out as part of the QAL2 (there are similarrequirements for the AST - see section 5). Some of the tests must be performed by the test laboratory.The remaining tests may be performed by either the test laboratory, the supplier of the CEM or theoperator, provided that:

The test laboratory has overall responsibility for the tests and carries out periodic audits of theorganisation performing the tests.

If the test laboratory or operator perform the tests, then its personnel must be able todemonstrate sufficient experience and training to do so. For example, the test laboratory shouldaudit the capability and training records of the operators staff who are responsible for thefunctional tests.


Table 3 shows the functional tests that have to be performed, indicating which checks the testlaboratory must perform and the checks that may be performed by the operator, supplier or testlaboratory. For completeness Table 3 also include the tests that have to be carried out in the AST.

Table 3 Designation of responsibilities for the functional tests in QAL2 and AST

QAL2 ASTActivityExtractive

CEMsIn-situCEMs

ExtractiveCEMs

In-situCEMs

Responsibility

Alignment &cleanliness

Any

Samplingsystem

Any

Documentationand records

Test laboratory

Serviceability Test laboratoryLeak test AnyZero and spancheck

Any

Linearity () () AnyInterferences AnyZero and spandrift (audit)

Test laboratory

Response time AnyReport Test laboratory

The linearity check is specified for the AST but not for the QAL2 test. However, we recommend thatthis check is performed during the QAL2 as well because BS EN 14181 permits the use of referencematerials to extend the calibration range, subject to certain conditions. The linearity check mayprovide the necessary data in order to extend the calibration range, or to provide a means of calibrationif the spread of data is insufficient to determine a valid calibration function.

The manual zero and span checks shall be performed using the same procedure as for the MCERTSperformance tests. Typically the zero and span checks require the use of reference materials. However,in the case of particulate monitors, these checks will require the use of surrogates for zero and span.

The inevitable downtime incurred in the CEMs through the functional tests need not be subtractedfrom the annual availability allowance specified in the MCERTS performance specifications.

Key points

The test laboratory shall have overall responsibility for the functional checksspecified in the QAL2 and AST, but the following checks may be carried outeither by the operator, CEMs supplier or test laboratory: alignment andcleanliness, sampling system integrity, leak test, manual zero and spancheck, linearity, interferences and response time.

Although QAL2 does not require a linearity check, this test is stronglyrecommended to provide supporting data for calibration.


3.5 Calibration and validation

3.5.1 The standard reference methods (SRMs)

BS EN 14181 requires SRMs to be used to calibrate and validate CEMs. BS EN 14181 bases thisrequirement on generic standard BS ISO 1109513, which describes how monitoring equipment iscalibrated using a SRM. It is based on three premises for its effectiveness and accuracy. These are:

There is a spread of data over the required range of the monitoring system. There is a linear relationship between the CEM data and the SRM data, when both sets of

measurements are valid.

The SRM is accurate and precise. The SRM is linear and has no scatter.Only test laboratories that are accredited to BS EN ISO/IEC 17025 for the MCERTS performancestandards for manual stack-emission monitoring for the applicable SRMs may perform the SRMmeasurements during QAL2 and AST. However, the test laboratory may be an external third partylaboratory or part of the operators organisation. The applicable SRMs are defined in TGN M214. Werecommended that the three days for the QAL2 reference tests should be spaced apart so that the datacan be analysed after each day. This is especially relevant for particulates.

Key Points

Only test laboratories accredited to ISO 17025 for the MCERTS performancestandards for manual stack emissions monitoring for the applicable SRMsmay perform the reference monitoring tests in QAL2 and the AST.

The SRMs are prescribed in TGN M2. The SRM data should have a wide spread over the measurement range, a low

scatter and show a linear response to an increase in the value of thedeterminand

The test laboratory may be an external third party laboratory or part of theoperators organisation.

The calibration function within QAL2 and the AST are based on the premisethat the SRM is sufficiently accurate and precise, as well as producing anadequate spread of data over the applicable range of the CEM.

We recommended that the three days for the QAL2 reference tests should bespaced apart so that data can be analysed after each day. This is especiallythe case for particulates.

3.5.2 Calibration using a SRM

Figure 2 illustrates the principle of linear calibration using a SRM in which the SRM data is comparedwith the CEM data and is used to derive a calibration function. The CEM itself may have a bias in onedirection or another, depending on gain of the CEM and its offset relative to zero. A calibrationfunction, in its simplest form, can then be described by equation (1):

yi = bxi + a + (1)where:


y = the SRM values for i = 1 to n

x = the CEM values for i = 1 to n

b = the slope of the calibration function

a = the intercept of the calibration function

= the deviation between the actual and the expected valueIt is unlikely that the SRM will produce a perfect straight line with no scatter when monitoring stackemissions. However, it is essential that the scatter - or imprecision - of the SRM is less than that ofthe CEM being calibrated and validated. Therefore the test-laboratory must characterise the SRMthoroughly and assure that the uncertainties are not only well known and understood, but also wellbelow the uncertainties specified in the WID and LCPD (ideally, no more than half the uncertaintiesspecified in the WID and LCPD.) If these conditions are not fulfilled, then BS EN 14181 will not bean effective tool for calibration of the CEM. In a worst case, it may either fail or mis-calibrate a CEMdue to deficiencies within the SRM application, rather than any faults or a lack of accuracy andprecision within the CEM itself.

Figure 2 the principle of linear calibration using a SRM

CEM data calibration

y = 1.0804x + 0.0513R2 = 0.9981

0

20

40

60

80

100

120

0 20 40 60 80 100 120

SRM data, mg.m -3

CEM data, mg.m-3

-3

3.5.3 Instrumental SRMs

Test laboratories may use SRMs that are based on either manual methods or instrumental methods.The following two conditions shall be met if a test laboratory wishes to use instrumental SRMs:

The test method using the instrumental technique shall be validated and accredited by UKAS oran equivalent body. To check equivalence of alternative methods to SRMs, refer to BS EN TS1479315.


The monitoring equipment used shall be MCERTS certified for all the applicable determinandsand the appropriate ranges, if available. The certification may be either under the scheme forCEMs, or the scheme for portable monitoring systems (for Type I systems).

MCERTS certification is critical, since any instruments used to calibrate and validate CEMs must bedemonstrably at least as good as the CEMs being calibrated. The ranges of the MCERTS certificationare especially important, so that the instrumental system used within the SRM has an uncertainty thatis at least as low as that of the CEM (see section 2.2). For example, a portable monitoring system witha minimum certified range of 0 to 200 mg.m-3 for both CO and SO2 would not normally be suitable tocalibrate and validate CEMs installed on incinerators, where the required ranges for CEMs are 0 to 75mg.m-3 for both these determinands. Monitoring systems used within SRMs shall undergo appropriatequality control and assurance checks at least annually and preferably more frequently.

Furthermore, the instrumental systems used within SRMs shall also undergo the same quality controlsand quality assurance as CEMs covered by BS EN 14181. Indeed, all applicable national andinternational standards for instrumental techniques, such as ISO 1084916 for NOx and ISO 793517 forSO2 monitoring systems specify similar quality assurance and control procedures to BS EN 14181.

Key Points

If instrumental techniques are used within SRMs, then the monitoringsystems used shall be certified to the MCERTS performance standards forthe applicable determinands and appropriate ranges.

Monitoring systems used within SRMs shall undergo appropriate qualitycontrol and assurance checks at least annually and preferably morefrequently.

3.5.4 The acquisition of data

When conducting the parallel measurements, the test laboratory must take the measured signalsdirectly from the CEM (eg expressed as an analogue or digital signal) during the QAL2 and AST tests,using the installations data collection system. This operation must be quality assured by theorganisation that has overall responsibility for the QAL2 and AST tests.

3.5.5 Spread of data

The test laboratory must select a set of representative operating conditions which cover as wide arange as possible, but deliberately modifying the process to artificially increase emissions is notpermitted. Ideally operators should select a time when the emissions are likely to be their highest andmost varied, but the process may not be deliberately varied in order to create higher than normalemissions. For example, when bag filters are replaced, emissions of particulate are temporarily higherand this produces an ideal time to measure a wider range of emissions.

3.5.6 Number of data points

QAL2 specifies at least fifteen sets of valid data when performing the SRMs and it is advisable toobtain at least 18 or 19 sets of data to ensure sufficient valid data sets. Three of the data points must beat zero, or near zero, where near zero is defined as a value that is no more than 5% of the ELV. Ideallyzero values should be measured when the installation is not producing emissions and if this is notpossible, then the test laboratory may use surrogate values.

Figures 3 and 4 illustrate the impact of an insufficient spread of data. In Figure 3, there are threevalues near zero and the R2 value for the linear regression line is an acceptable value of approximatelyR2 = 0.9. However, if all the values are clustered (Figure 4), then the linear regression line and hencethe calibration function is very different, and the R2 value is unacceptably low (about 0.1). A similar


case arises when the data is clustered too near to zero and when there is not a sufficient spread of dataacross the measurement range.

The QAL2 and AST tests must provide at least fifteen and five sets of data respectively. The datashould ideally be spread over at least 50% of the ELV with at least three data points at, or near, zero.

The emphasis is also on valid sets of data which cover the ELV therefore the test laboratory isadvised to carry out a greater number of tests in order to meet the minimum requirement. Ifpracticable, the data from the CEM and SRMs should be plotted on a chart as the QAL2 and AST testsprogress, as this will indicate whether the spread of data is sufficient, whether the data has enoughvalues near zero and whether there are any obvious outliers.

If the test house assesses that a data set is considered invalid then the reasons for this should be notedin the QAL 2 report (for example: changing process conditions, error in SRM, failure of instrument).

Figure 3 An example of an adequate spread of data for a set of QAL2 measurements

y = 1.0486x - 3.1R2 = 0.8904

-50

0

50

100

150

200

250

300

350

400

450

0 50 100 150 200 250 300 350

CEM data, mg.m -3

SR

M d

ata,

mg.

m-3

Figure 4 An example of a set of QAL2 measurements with no measurements near zero

y = 0.6867x + 85.09R2 = 0.1496

0

50

100

150

200

250

300

350

400

0 50 100 150 200 250 300 350

CEM data, mg.m -3

SR

M d

ata,

mg.

m-3


Key Points Spread of Data

The QAL2 and AST tests must provide at least fifteen and five sets of datarespectively.

The data should ideally be spread over at least 50% of the ELV with threedata points at or near zero.

3.5.7 QAL 2 for particulate CEMs

The following provisions apply to the calibration of particulate CEMs:

The calibration function may be linearised according to BS EN 13284-2. The error associated with the SRM increases significantly when the dust concentration is below 5

mg.m-3. Great caution should be applied when making a QAL2 assessment on processes with dustlevels lower than 5mg.m-3.

In processes with dynamic dust levels (associated with the cleaning cycle of arrestment plant) theresponse time of the instrument should be reduced during the QAL2 procedure to permit theinstrument to accurately follow these dynamics.

It is beneficial to conduct the QAL2 on non-consecutive days so that the validity and spread ofdata from one days testing can be considered before further testing is conducted.

3.5.8 Peripheral CEM measurements

BS EN 14181 specifies requirements for peripheral measurements. These are determinands whichneed to be measured but do not have performance characteristics assigned to them within the WID andLCPD. In BS EN 14181, peripheral measurements are:

Oxygen Moisture Temperature Stack gas pressure

CEMs for oxygen and moisture (if used) must be certified to MCERTS performance standards. Thesame applies to SRMs that use instrumental techniques. Functional checks should be performed onCEMs for oxygen and moisture (if used) although ordinarily a full QAL2 should not be needed for theinstallations peripheral measurements. However, if the CEM fails the QAL2 tests using the operatorsperipheral measurements, then the SRM peripheral measurements may be used instead. If the CEMthen passes the QAL2 tests, then the operator must fix the peripheral monitoring equipment as soon aspossible and verify its performance through QAL2 exercises.

SRM monitoring for oxygen is required in any event for the QAL2 tests for other determinands, so the15+ sets of oxygen SRM measurements can then be used to perform a QAL2 for oxygen. Whenperforming the variability test for oxygen and moisture measurements, the following virtual ELVs anduncertainty allowances shall be applied:

Oxygen: ELV = 21%, 95% CI = 10% Moisture: ELV = 25%, 95% CI = 30%

If CEM readings for moisture are also found to be erroneous when compared to the referencemonitoring and following the variability tests, then the SRM results for moisture shall also be used toperform a full QAL2 exercise on the installations CEMs which measure moisture.


CEMs for temperature and pressure shall be cross-calibrated using reference instruments that aretraceable to national standards.

Key Points Peripheral Measurements

If the CEM fails the QAL2 tests using the operators peripheral measurements formoisture and oxygen, then the SRM peripheral measurements may be usedinstead. If the CEM then passes the QAL2 tests, then the operator must fix theperipheral monitoring equipment as soon as possible and verify its performancethrough QAL2 exercises.

CEMs for temperature and pressure shall be cross-calibrated using referenceinstruments that are traceable to national standards.

3.5.9 Sample lines and delays

If the test laboratory uses instrumental techniques within SRMs, then differences between thesampling systems of the CEMs and SRMs can result in a difference in integration time, meaning thatsets of measurements starting and ending at the same time may be uncoordinated. Therefore the testlaboratory should carry out the following:

Establish if there is a difference in integration time; for example, by injecting a test gas into theCEM and SRM sampling probes at the same time and determining if there is a significantdifference in responses, taking into account performance characteristics.

If this test shows that there is a difference in integration time, then any SRM systems may beconnected at the same point on the sampling system as the CEM, in order to align the lag timesbetween sampling and analysis. If this is not practicable, then the test laboratory must measurethe differences in lag times and correct the data accordingly.

3.5.10 Establishing the calibration function and the test of variability

In carrying out the analysis of data the following steps are required:

Tabulate the CEM and SRM data. Plot the CEM data and SRM data together. Calculate the calibration function. As guidance, an indicator for a valid calibration function is a

correlation coefficient of the regression line of R2 = 0.9 or more for gaseous CEMs and 0.5 forparticulate CEMs.

Establish the valid calibration range (which should cover the ELV). Convert the data to calibrated and standardised values. Carry out the variability test.

3.5.11 Extension of the calibration range

In general, if the data is sufficiently linear to derive a valid calibration function, then:

The calibration range of CEMs monitoring gases may be extended by 10%. The calibrationrange of particulate monitors may be extended by 100%.


The calibration range may be extended further using reference materials so long as the resultingdata points are within the 95% confidence intervals of the calibration function.

3.5.12 Scatter of data points

If the emissions are


Note: If there is a significant change of fuel, then the operator should first perform an AST (refer to section 8 in BS EN14181 and this document). If the results fit within the 95% CIs of the calibration function, then no further testing isrequired. If not, then a full QAL2 is required.

However, a new QAL2 for changes in the process or fuel will not be needed if:

The operator can demonstrate that the change in process does not affect the emissions profileand the original calibration factor remains valid.

The thermal input is less than 10% per year for the alternative fuel, and/or; The change in fuel use can be shown to have no significant effects on emissions, when

compared to the original fuel.

If there is a change of fuel, then the operator should first perform an AST (refer to section 8 in BS EN14181.) If the results fit within the 95% CIs of the calibration range, then no further testing is required.If not, then a full QAL2 is required.

3.10 Repeated exceedences of the calibration rangeIf the emissions exceed the calibration range derived from SRM data, then the provisions of section6.5 of BS EN 14181 shall apply, such that a new QAL2 is required if:

More than 5% of the number of measured values from the CEM, calculated weekly, are abovethe maximum SRM value (+10%) for more than five weeks in the period between two AST orQAL2 tests.

More than 40% of the number of measured values from the CEM, calculated weekly, are abovethe maximum SRM value (+10%) for more than one week in the period between two AST orQAL2 tests.

4. Ongoing quality assurance during operations (QAL3)

4.1 QAL3 - generalThe purpose of QAL3 is to detect drift and changes in precision in the CEM by performing regularchecks of the zero and span readings. These may be checked using reference materials or using asurrogate method traceable to national standards.

An initial QAL3 check is required before starting to report data. Any changes in drift are thencompared to the MCERTS performance specification for the applicable certified range of the CEM todecide if any intervention is required e.g. maintenance of the CEM.

In order to determine whether any changes in the zero and span values are due to actual drift, orwhether they are caused by random deviations and the effects of influence factors such as variations involtage and ambient temperature, any changes to zero and span values are plotted on control chartsand compared to allowable variations expressed as standard deviations, or sAMS in BS EN 14181.

The sAMS can either be calculated using MCERTS test data, or determined by using span test gases.Then multiples of sAMS are used to set warning levels and alarm levels. When calculating sAMS, use themethod described in the examples in BS EN 14181 for determining sAMS, but using the followinginfluence factors:

Effect of ambient temperature. Effect of stack gas pressure for in situ CEMs. Effect of voltage. Cross-sensitivity to other determinands.


Detection limit.

When using test gases, several readings are taken and the standard deviation is calculated from thesereadings. This approach is simple and practical, but if the operating conditions at the time result in ahigh precision, then this can result in artificially low warning and alarm limits.

Two limits are set on the control charts, which are (i) a warning limit to show that the CEM is startingto drift out of control and (ii) an action or alarm limit to show that the CEM has drifted beyondspecifications and corrective actions are needed.

Whilst auto-corrections before the CEM drifts out of the control range are not recommended, suchauto-corrections may take place so long as the CEMs still meet the MCERTS specification for zeroand span drift.

4.2 Zero and span checksZero and span checks shall be performed using reference materials, such as test gases. If this is notpracticable or possible, then the CEMs supplier may provide surrogates, which are traceable tonational standards.

4.2.1 Test gases and reference materials

If test gases are used, then such gases used should be traceable to Primary National Standards andshould have certificates which meet the requirements of BS 4559-4. Suppliers of test gases should alsobe accredited to BS EN ISO/IEC 17025 to applicable standards such as the BS 4559 series. Test gasesare required for all the gaseous determinands with ELVs.

Gas-mixing systems can be used, as these are particularly useful for multi-point span checks. Suchsystems should meet the performance standards specified in USEPA Method 205.

Surrogate reference materials are required for performing zero and span checks on particulatemonitoring CEMs and these should be assessed as part of the MCERTS testing for their validity inproviding an appropriate QAL3 check.

4.2.2 Requirements on the CEMs and data recording systems

To carry out zero and span checks, the CEMs and the data recording systems have to be able to: Record both positive and negative values. Record any changes in readings from the previous zero and span checks Record zero and span data results for greater than one year. This permits auditing of the data at

the AST.

4.2.3 Frequency of checks

If operators are using CUSUM charts (see next section) then weekly zero and span checks will berequired. If operators are using Shewhart charts, then the frequency may be based on the maintenanceinterval determined during testing for MCERTS certification, although we recommend using shorterintervals until sufficient data is available to lengthen the time between checks.

Users have the option to use instruments with either automatic or manual QAL 3 checks. The majorityof instruments use automatic self-checks since these tests can be conducted without additionalmanpower.


4.3 Use of control chartsBS EN 14181 specifies the use of control charts for QAL3 and describes two types of control chart,Shewhart charts and CUSUM charts. The latter are more complex, but provide more information onthe performance of the CEM. Other types of control charts may also be used.

4.3.1 Purpose of control charts

Under QAL3 the operator regularly checks the response of the CEM to zero and span referencematerials. If these readings are repeated over a sufficiently short period of time and the CEM has nothad a chance to drift then the actual readings will be due to variations in precision and allowableeffects of influence quantities. Over a period of time, as the operator collects more data, there is only avery small chance that the readings will change by more than three standard deviations, unless theCEM has truly drifted. The purpose of control charts is to plot such trends and give an indication ofactual or forthcoming drift.

The user either calculates or determines the standard deviation, sAMS for the operation of theinstrument under anticipated stack conditions. and then uses multiples of this standard deviation to setwarning levels and alarm (or intervention) levels. Two control charts are needed one for zero driftand one for span drift.

There is a choice to use either the manufacturers specifications on uncertainty or alternatively themaximum allowable uncertainty as defined in the MCERTS performance standards.

4.3.2 Shewhart control charts

The results are presented as a function of time. The values shown by the CEM can be expressed as anabsolute value or as the difference between the reading and the expected value of the referencematerial. Two control charts are needed one for zero drift and one for span drift. Figure 5 shows anexample template for span drift.

The target values on the two charts are the average value of zero and span readings, zy and syestablished during the initial QAL3. This should be carried out immediately after the QAL2 has beencompleted.

The associated standard deviations sAMS and sAMS are used to calculate the levels which will trigger analarm and possible intervention.


Figure 5 Example template for Shewhart Control Charts

The upper and lower warning levels are given by n

sy AMSS

2+ and

n

sy AMSS

2+

The upper and lower alarm limits are given by n

sy AMSS

3 and

n

sy AMSS

3

where n is the number of consecutive repetitions of the test carried out (n should be at least 10 for theinitial QAL3 but can be equal to one in the case of a repeat QAL3)

Once the control chart is set up, the results of the zero and span tests (averages of the n readings on theCEM) are placed on the chart in order to detect drift and/or changes in precision that requireintervention by the operator e.g. maintenance of the CEM and possible rejection of the results sincethe previous tests.

EN 14181 requires the operator to intervene when:

one or more data points are beyond one of the upper alarm limits;

three consecutive data points are beyond one of the warning limits; four points among 5 consecutive ones are beyond

n

sy AMSs i.e. half the alarm levels for span;

eight consecutive points are on the same side of the mean;

six consecutive points are either increasing or decreasing.

Upper alarm limit

Upper warning limit

Target value sy

Lower warning limit

Lower alarm limit

Value indicated

by the AMS

ns2

y AMSs +

ns2

y AMSs

ns3

y AMSs +

ns3y AMSs


4.3.3 CUSUM charts

Any limitation that the Shewhart chart has in detecting progressive changes or staged changes can beovercome by associating several successive control points, ie through the use of moving mean chartssuch as CUSUM control charts. To achieve this, the values calculated and entered on the chart are notthe last value but the average of several previous values.

The CUSUM, or cumulative sum chart uses all of the data and is therefore a more sensitive way todetect slight changes in the mean. If a target value C is being considered, then the operating principleof the chart is to calculate the difference between each new value and value C, and to add this to acumulative sum. This cumulative sum is then reported on the chart relative to the values measured.

As long as the measurement results are close to the target value, the CUSUM charts curve remainsclose to zero. A positive curve indicates that the results are greater than the target value and a negativecurve shows the opposite. Stepped changes in a data series is shown by an abrupt change in curveshape. A gradual drift produces slight but continuous changes in the mean.

4.3.4 Examples of the use of control charts

Examples of the use of control charts are given in Appendix 2.

4.4 Reporting

QAL3 records should include the following:

CEM details monitoring approach and technique, operating range, make and model. CEM changes details of change in make, model and serial number through the year. Manufacturers service visit records routine maintenance. Manufacturers call out records corrective actions taken. Operators routine maintenance and corrective actions. QAL3 baseline re-sets - summary. Zero and span drift plots. Zero and span drift tabulation.

5. Annual surveillance test (AST)

5.1 Purpose of the ASTThe Annual Surveillance Test (AST) is a mini-QAL2 whose purpose is to verify the continuingvalidity of the calibration function. In general the key points in section 3 also apply to the AST.

5.2 Functional testsThe requirements and responsibilities for carrying out the AST tests is the same as for QAL2 (seesections 3.4 and 3.5). The testing laboratory has overall responsibility for the functional tests, althougheither the operator or the equipment supplier may perform these. In such cases, these tests shall beverified by audit by the accredited testing laboratory and included in their report.

5.3 Parallel measurements with a SRMOnly the testing laboratory may perform the SRMs. It must be noted that five tests are a minimum andthat the testing laboratory is advised to carry out a greater number in case any tests are deemed invalid.If the testing laboratory is using instrumental methods for SRMs, then the SRM monitoring systemshall be operated continuously over the entire day of the AST. Zero and span checks shall take place at


least at the start of each day, mid-way through each day and at the end of the day. At least five sets ofdata can then be extracted over any 8 to 10 hour period within a day.

Any continuous monitoring systems used within SRMs shall be certified where possible to theMCERTS performance standards either for CEMs or portable equipment (Type 1).

5.4 Example AST determinationAppendix 3 provides an example AST determination.

6. Status of this guidanceThis TGN may be subject to review and amendment following its publication. The latestversion of the TGN can be found on the Agencys web-site at: www.environment-agency.gov.uk bytyping M20 into the search facility.

We also welcome feedback on its use. Any comments or suggested improvements to the TGN shouldbe e-mailed to Richard Gould at [email protected]


References

1. BS EN 14181. Stationary source emissions Quality assurance of automated measuringsystems.

2. BS EN 13284-2. Stationary source emissions Determination of low range concentration ofdust Part2: Automated measuring systems.

3. 2000/76/EC, Directive on the incineration of waste.

4. 2001/80/EC, Directive on limiting emissions of certain pollutants into the air from largecombustion plants.

5. BS EN ISO 14956. Air quality Evaluation of the suitability of a measurement procedure bycomparison with a required measurement uncertainty.

6. Performance Standards for Continuous Emission Monitoring Systems The EnvironmentAgencys Monitoring Certification Scheme (MCERTS), www.mcerts.net .

7. TGN M1. Sampling requirements for monitoring stack emissions to air from industrialinstallations. Environment Agency.

8. ISO 10396. Stationary source emissions Sampling for the automated determination of gasconcentrations.

9. BS EN 13284-1. Stationary source emissions Determination of low range massconcentration of dust Part 1: Manual gravimetric method

10. BS EN ISO 9001. Quality management systems Specification with guidance for use.

11. BS EN ISO 14001. Environmental management systems - Specification with guidance for use.

12. BS ISO 10012. Measurement management systems Requirements for measurementprocesses and measuring equipment.

13. BS ISO 11095. Linear calibration using reference materials.

14. TGN M2. Monitoring of stack emissions to air. Environment Agency.

15. BS EN TS 14793. Stationary source emissions inter-laboratory validation procedure for analternative method compared to a reference method.

16. ISO 10849. Stationary source emissions -- Determination of the mass concentration ofnitrogen oxides -- Performance characteristics of automated measuring systems.

17. ISO 7934. Characterization of air quality Part 4: Stationary source emissions Section 4.1Method for the determination of the mass concentration of sulphur dioxide: Hydrogenperoxide/barium perchlorate/Thorin method.


Glossary

Glossary of terms

AST Annual surveillance test

CEM Continuous emission monitoring system

CI Confidence Interval

ELV Emission Limit Value

LCPD Large Combustion Plant Directive

QA Quality assurance

QAL Quality assurance level

SD Standard deviation

SRM Standard reference method

UKAS United Kingdom Accreditation Service

WID Waste Incineration Directive


Appendix 1: Example QAL2 calibration functions and variability testsThis appendix gives two examples of establishing the calibration function and testing the variability ofa CEM under the requirements of QAL2.

A1.1 CEM measuring SO2The installation was an incinerator and the CEM was an infrared instrument with a heated line, chiller-dryer and an optical bench that measures the stack gas sample on a cold, dry basis. The SRM used wasthe wet-chemical method using BS 6069-4.1 (ISO 7934) and the ELV was 50 mg.m-3 at 11% O2.Fifteen parallel measurements were taken over three days and distributed evenly. The results are givenin the table 15.

a. Tabulating the data

The data is tabulated showing the measurements of the parallel SRM tests and simultaneous CEMdata.

Table A1.1 - Measurements for the QAL2 test

Sample number SRMmeasured value,

mg.m-3

y

CEMmeasured signal,

mg.m-3

x

1 25.9 18.6

2 5.6 2.6

3 15.7 10.5

4 3.4 0.2

5 0.8 0.0

6 2.3 0.0

7 3.1 0.9

8 11.0 7.4

9 1.6 0.0

10 3.9 0.0

11 21.6 16.9

12 21.7 9.5

13 21.2 11.5

14 10.8 4.9

15 37.7 28.8

186.3 111.8

b. Plotting the data

The data is then plotted on a graph, with the SRM values on the Y-axis and the CEM values on the X-axis as in Figure A1.1. This gives a reasonable idea of the calibration function that is calculated usinglinear regression analysis. The regression equation is shown on this example together with thecorrelation coefficient of the regression line, which has an R2 value of 0.95.


Figure A1.1 - Plot of SRM-results y versus CEM measured signals x

y = 1.25x + 3.1R2 = 0.9517

0

5

10

15

20

25

30

35

40

45

0 5 10 15 20 25 30 35

CEM data, mg.m -3

SR

M d

ata,

mg.

m-3

c. Calibration function

The next stage is to calculate the calibration function, which is described by:

ii xbay +=

where:

xi is the CEM datayi is the SRM dataa is the intercept of the calibration functionb is the slope of the calibration function

From Table A1.1 it can be seen that:

ymax = 37.7 mg.m-3

ymin= 0.8 mg.m-3

Therefore ymax ymin= 36.9 mg.m-3

Section 6.4.2 of BS EN 14181 provides two methods for calculating a and b, depending on the spreadof data. In simple terms, the user employs Method a if the spread is greater than 15% of the ELV, orMethod b is the spread of data is equal to or less than 15% of the ELV.

Therefore to determine which method to use, calculate if the difference of ymax ymin is smaller orgreater than 15% of the ELV :

3

3max

m.mg5.7

m.mg5015.0

15.0

=

=

= ELVy


ELVyy 15.0m.mg9.36m.mg8.0m.mg7.37 333minmax >==

Since the difference ymax ymin is greater than 15% of ELV, a and b are then calculated by Method b:

( )( )

( )

=

=

= N

ii

N

iii

xx

yyxxb

1

2

1

xbya =

where:

33

1

.5.7.8.1111511

=

=== mmgmmgxNxN

ii

33

1

.4,12.18631511

=

=== mmgmmgyNyN

ii

and

( )( )

( )

=

=

=

N

ii

N

iii

mmgx

mmgymmgx

1

23

1

33

.5.7

.4.12.5.7

b

bmmgmmga 33 .5.7.4.12 =

The result of the calculation is:

33 .5.725.1.4.12a

25.1b =

=

mmgmmg

The calibration function then becomes:

iii xmmgxy 25.1.1.3ba3 +=+=

d. Calculation of the calibrated values

Once the calibration function has been determined, the next stage is to convert the data to calibratedand standardised values. These are shown in Table A1.2.

The calibration function is valid in the range from miny to maxy plus an extension of 10% of thisrange. For this CEM the valid range is:

3-max

-3min

mg.m2.39

mg.m 1.3

=

=

y

y

Therefore an extension of 10% either of the range becomes:


( ) ( )-33

-3-3-3-3

mg.m8.42mg.m0.0

mg.m1.32.3910.0mg.m2.39mg.m1.32.3910.0mg.m1.3

+ y

y

Hence the valid calibration range is from 0.0 mg.m-3 to 42.8 mg.m-3.

Table A1.2 - Calibration of the CEM

Number CEMmeasured signal

SRMmeasured value

CEMmeasured value A

Difference Squared difference

i x, mg.m-3 y, mg.m-3 y , mg.m-3 yy , mg.m-3 2)( yy

1 18.6 25.9 26.4 0.5 0.2

2 2.6 5.6 6.3 0.7 0.5

3 10.5 15.7 16.2 0.5 0.2

4 0.2 3.4 3.3 0.1 0.0

5 0.0 0.8 3.1 2.3 5.1

6 0.0 2.3 3.1 0.8 0.7

7 0.9 3.1 4.2 1.1 1.2

8 7.4 11.0 12.4 1.4 1.9

9 0.0 1.6 3.1 1.5 2.4

10 0.0 3.9 3.1 0.8 0.6

11 16.9 21.6 24.2 2.6 6.8

12 9.5 21.7 15.0 6.7 45.2

13 11.5 21.2 17.5 3.7 13.8

14 4.9 10.8 9.2 1.6 2.5

15 28.8 37.7 39.2 1.5 2.3

Sum 83.4

A CEM measured value (calibrated value) is based on CEM measured signal and the calibration function.

e. Variability test

The variability is accepted if:

v0 kD

where:

D is equal to the standard deviation of the value Di , which is the difference between measured SRMvalue, y, and calibrated CEM value, .

0 is the uncertainty laid down by the authoritieskv is the test parameter

D is calculated by:

6262

1

2 .53.2.4.83115

11

1

=

=

=

= mmgmmgDNN

iiD

where


iii yyD =

The uncertainty laid down by the authorities is 20% of the emission limit value ELV as a 95% CI. 0 istherefore calculated as:

-3

-30

mg.m5

96.1mg.m5020.0

96.1/

=

=

= ELVp

For 15 measurements the kv value is 0.9761, (refer to Table 1, section 6.7 of BS EN 14181.) The testfor variability then yields:

-3

-3-3

mg.m88.4

1976.0mg.m5mg.m53.2

This is less than the allowable limit for o and therefore the CEM meets the requirements.

A1.2 CEM measuring NOThe installation was a power station with a NO limit (as NO2) of 450 mg.m-3. The CEM used a heatedline coupled to a chiller dryer and an infrared analyser whilst the SRM was an accredited methodbased on ISO 10849. Fifteen parallel measurements were been taken during over days and distributedevenly.


The data is tabulated showing the measurements of the parallel SRM tests and simultaneous CEMdata.

Table A1.3 - Measurements for the QAL2 test

Sample number SRMmeasured value

y, mg.m-3

CEMmeasured signal

x, mg.m-3

1 277 297

2 270 307

3 227 271

4 248 296

5 298 271

6 388 287

7 347 286

8 243 252

9 305 256

10 345 285

11 209 228

12 38 66

13 7 13

14 14 4

15 10 2

3226 3121



The data is then plotted on a graph, with the SRM values on the Y-axis and the CEM values on the X-axis as in Figure A1.2.

Figure A1.2 Plot of SRM-results y versus CEM measured signals x

y = 1.0486x - 3.1R2 = 0.8904

-50

0

50

100

150

200

250

300

350

400

450

0 50 100 150 200 250 300 350

CEM data, mg.m -3

SR

M d

ata,

mg.

m-3

c. Calibration function

The next stage is to calculate the calibration function. The calibration function is described by:

ii xbay +=

where:

xi is the CEM datayi is the SRM dataa is the intercept of the calibration functionb is the slope of the calibration function:

From Table A1.3 it can be seen, that:

ymax = 388 mg.m-3

ymin = 7 mg.m-3

ymax ymin = 381 mg.m-3

Section 6.4.2 of BS EN 14181 provides two methods for calculating a and b, depending on the spreadof data. In simple terms, the user employs Method a if the spread is greater than 15% of the ELV, orMethod b is the spread of data is equal to or less than 15% of the ELV. Therefore to determine whichmethod to use, calculate if the difference of ymax ymin is smaller or greater than 15% of the ELV :


3

3

max

mg.m5.67

m.mg45015.0

15.0

=

=

= ELVy

ELVyyy 15.0maxminmax =>

Since the difference ymax ymin is larger than 15% of ELV, a and b are calculated by Method b:

( )( )

( )

=

=

= N

ii

N

iii

xx

yyxxb

1

2

1

xbya =

where:

33

1

mg.m208mg.m31211511 ===

=

N

iixN

x

33

1

mg.m215mg.m32261511 ===

=

N

iiyN

y

and

( )( )

( )

=

=

= N

ii

N

iii

x

yxb

1

23

1

33

mg.m208

mg.m215mg.m208

mg.m208mg.m215 33 =a

The result of the calculation is:

-3-3-3 mg.m1.3mg.m2080486.1mg.m215

0486.1

==

=

a

b

The calibration function then becomes:

iii xbxay 0486.1mg.m1.3-3 +=+=

d. Calculation of the calibrated values

Using the calibration function on the CEM measured signals the results listed in Table A1.4 areobtained.

The calibration function is valid in the range from miny to maxy plus an extension of 10%. For thisCEM the valid range is:

3max

3min

mg.m319

mg.m1

=

=

y

y


3

3

mg.m3510

mg.m31910,10

y

y

As miny is negative, the valid calibration range is from 0 to 351 mg.m-3.

Table A1.4 - Calibration of the CEM


SRMmeasured value

CEMmeasured value A


I x, mg.m-3 y, mg.m-3 y , mg.m-3 yy , mg.m-3 2)( yy

1 297 277 308 31 970

2 307 270 319 49 2445

3 271 227 281 54 2944

4 296 248 307 59 3445

5 271 298 281 16 269

6 287 388 298 90 8083

7 286 347 296 51 2586

8 252 243 261 19 353

9 256 305 266 39 1559

10 285 345 296 50 2472

11 228 209 236 26 702

12 66 38 66 28 757

13 13 7 11 3 11

14 4 14 1 13 163

15 2 10 1 11 112

Sum 26871

A CEM measured value (calibrated value) is based on CEM measured signal and the calibration function.

e. Test for variability

The result of the variability test is accepted if:

v0 kD

where:

D is equal to the standard deviation of Di0 is the uncertainty specified in EU Directiveskv is the test parameter

D is calculated by:

362

1

2 .5.4526871115

11

1

=

=

=

= mmgmmgDNN

iiD


where:

iii yyD =

The uncertainty laid down by the LCPD is 20% of the emission limit value ELV as a 95% CI. 0 istherefore calculated as:

3-

3-

0

mg.m4596.1

mg.m45020,0

96.1/

=

=

= ELVp

For 15 measurements the kv value is 0,9761, (refer to Table 1, section 6.7 of BS EN 14181.) The testfor variability then yields:

-3

-3-3

mg.m43.9

1976.0mg.m45mg.m5.45

This is more than the allowable limit for o and therefore the CEM does not meet the requirements.The cause of the failure must be investigated and fixed, after which the QAL2 test is repeated. If theCEM continues to fail the variability test and it is the performance of the CEM itself which is thecause of the failure, then it must be replaced.


Appendix 2: Example Shewhart and CUSUM control chartsFor simplicity the examples given below are for span tests only. The analysis of the zero tests wouldfollow exactly the same procedures.

An operator checks a CEM using a reference gas. The average value for the span gas established in theinitial QAL3 is 80 mg.m-3 with a standard deviation s of 2.5. Each week, the operator repeats thezero and span check, performing one check each so that n = 1. Table A2.1 shows the readings and thecalculated values for the CUSUM chart.

Table A2.1 Drift data for QAL3

Week no. Readings fromobservations,mg.m-3

Cumulativeaverage,mg.m-3

Gap between Cand reading,mg.m-3

CUSUM, mg.m-3

1 82 + 2 + 22 79 - 1 + 13 80 0 +14 78 79.75 - 2 - 15 82 79.75 +2 +16 79 79.75 -1 07 80 79.75 0 08 79 80 - 1 - 19 78 79 - 2 - 310 80 79.25 0 -311 76 78.25 - 4 - 712 77 77.75 -3 -1013 76 77.25 -4 -1414 76 76.25 -4 -1815 75 76 - 5 - 23

A2.1 Shewhart chart

Figure A2.1 shows the drift data plotted on the Shewhart chart. The chart shows that the last readingreached the lower alarm limit. According to this chart, no further action is necessary. However, if thetrend in drift is examined there is a forewarning of an abnormal situation starting as early as the 6thresult, as all the subsequent results have been below or equal to the target value. It is likely the trendwill continue and maintenance on the CEM will be required in the near future.

A2.2 CUSUM chartFigure A2.2 shows the CUSUM chart. When compared to the Shewhart chart, the trend and changesare much clearer from an earlier date. As with the Shewhart charts, the warning limits and alarm limitsare set at multiples of the value of the standard deviation, but the actual values specified in BS EN14181 differ from those used in Shewhart charts.


Figure A2.1 Example Shewhart chart

Shewart Chart

72.5

75

77.5

80

82.5

85

87.5

0 2 4 6 8 10 12 14 16

Week no.

Res

ults

upper alarm limit

lower alarm limit

upper warning limit

lower warning limit

target value

Figure A2.2 Example CUSUM chart

Moving median chart

75

76.25

77.5

78.75

80

81.25

0 2 4 6 8 10 12 14 16

Week number

Res

ults

lower warning limit

lower alarm limit

target value


Appendix 3: Example AST determination

This example is based on the sulphur dioxide CEM in Example 1 in Appendix 1. The ELV is 50mg.m-3, the reference method was ISO 7934 (wet chemistry) and the calibration function is:

yi = 1.25xi + 3.1

The valid calibration range is 0 to 42.8 mg.m-3 . The specified uncertainty is 20% of the ELV, which is10 mg.m-3.

The steps are similar to those for the QAL2 procedure, beginning with the tabulation of data.


Five parallel measurements using the SRM were taken over one day, with the monitoring spread overthat day. The results are shown in Table A3.1.

Table A3.1 - Measurements for the AST testSample number SRM

measured valuey, mg.m-3

CEMmeasured signal

x, mg.m-3

1 2.5 1.62 5.6 43 3.8 3.94 1,1 05 18 15.9 28.9 22.4


The results are then plotted on a graph with the SRM values on the Y axis and the CEM values on theX axis as in Figure A3.1. This will give an indication of the trend line and in this case, the regressionline has a correlation coefficient of R2 > 0.99.

Figure A3.1 Plot of AST data

02468

101214161820

0 5 10 15 20

CEM data

SR

M d

ata


c. Calculation of measured values of the calibrated CEM

The calibrated CEM values are then calculated by using the calibration function on the obtained CEMmeasured signals. The calibration function was determined in the previous QAL 2 test, and it isdescribed by:

yi = 1.25xi + 3.1

Using the calibration function on the CEM measured signals the results listed in Table A3.2 areobtained:

Table A3.2 - Measured values of CEM


SRMmeasured value

CEMmeasured value*


i xi. mg.m-3 yi. mg.m

-3iy . mg.m

-3 yyDi = .mg.m-3 2)( DDi

1 1.6 2.5 5.1 -2.6 0.4336462 4 5.6 8.1 -2.5 0.5664723 3.9 3.8 8.0 -4.2 0.850294 0 1.1 3.1 -2.0 1.593745 15.9 18 23.0 -5.0 3.067694

Sum -16.2 6.511842Average -2.6

* Measured value (calibrated value) of the CEM based on CEM measured signal and the calibration function

d. Calculation of the variability

Then the variability D is then calculated according to the following equation:

( )=

=N

iiD DDN 1

2

11

where

iii yyD =

=

=N

iiDN

D1

1

Using the values from table 3.2 yields:

3

2

3 mmg

28.1mmg

51.615

1 =

=D

According to section 8.5 of BS EN 14181, the variability is accepted if:v05.1 kD

where:

D is the standard deviation of the Di0 is the uncertainty laid down by the authoritieskv is the test parameter


For five measurements the kv value is 0.9161 (refer to Table 2 in section 8.5 of BS EN 14181.)The test for variability then yields:

-3-3

-3-3

mg.m87.6mg.m28.1

1916.0mg.m55.1mg.m28.1

The value is less than the required value so the variability of the CEM is accepted.

Finally the calibration of the CEM is accepted if:

095,0 )1( +N

sNtD D

In this example, the Student t-value for four degrees of freedom and a CI of 95 % (one-sided) is equalto 2.132. The inequality above then yields:

33

3- mg.m22.655

mg.m28.1132.2mg.m6.2

=+

As the average value of Di is less than 6.22 mg.m-3, then the calibration function is still valid.