GB/T 36090-2018Gas analysis—Guide for quality assurance of on line automatic measuring system (English PDF)

3.1 Automatic measurement system (AMS)

Automatic monitoring system permanently installed on site, including analyzers, sampling equipment, sample pretreatment equipment, etc.

3.2 Standard reference method (SRM)

In order to verify the method used for standard measuring devices that are temporarily installed in the field.

3.3 CUSUM chart

The drift and accuracy variation accumulation and control maps are compared to the uncertainty components obtained in Procedure 1.

3.4 Drift

During unattended periods, the measurement function changes with a monotonic change in the calibration function over time.

3.5 Instability

Drift and measured value changes caused by changes in the calibration function during unattended periods.

3.6 Instrument reading

The measured value directly read from the AMS without using the correction function.

3.7 Unattended time

The maximum allowable time interval for which the operating characteristics remain within a predetermined range.

3.8 Precision

The degree of consistency between consecutive zero readings and continuous range readings over a specified time interval.

3.9 Response time

AMS response time when the characteristic value suddenly changes.

3.10 Range reading

The analog input is an AMS instrument reading that is approximately 80% of the measurement range.

3.11 Variability

The standard deviation of the difference between SRM and AMS parallel measurements.

3.12 Zero reading

Instrument readings when the AMS analog input content parameter is zero.

5.1 Overview

AMS should be evaluated in accordance with Procedure 1 prior to installation on the AMS site.

The evaluation laboratory shall comply with the provisions of GB/T 27025. Third-party laboratories can be hired when the factory does not have a qualified laboratory get on. The laboratory evaluation procedure (procedure 1) includes.

a) AMS functional test;

b) AMS uncertainty assessment and applicability assessment;

c) on-site verification of laboratory evaluation results;

d) Confirm the applicability of AMS.

5.2 AMS functional test

The functional test shall be consistent with the technical specifications given by the AMS supplier (manufacturer).

The functional test of AMS is shown in Appendix A. The main contents of the test include.

---AMS visual inspection;

---Zero and span check;

--- Drift test;

---Linear test;

---Interference test;

--- Response time test, etc.

See Appendix B for linear testing of AMS.

5.3 Measurement System Calibration Conditions

5.3.1 General

The calibration conditions will play an important role in determining the effect of the amount of influence on the measured value. Calibration of each input will involve The calibration condition at the time, and any change in calibration conditions in subsequent measurements will result in a deviation until recalibration is required. To this end, it should be as in 5.3.2 5.3.3 Determine the calibration conditions.

During the measurement, if the calibration is performed periodically, the change in the amount of influence between successive calibrations should be determined.

5.3.2 Chemical influence

Specify the maximum value xj,max of the chemical interference Xj in this field. In exhaust gas monitoring, if the value has no fixed information, then Use the maximum value given in Appendix C.

Specify the minimum value xj,min of the chemical interference Xj, which is usually zero.

Specify the given value of the calibration substance xj, cal. When calibrating by parallel measurement of AMS and SRM, xj,min and xj,max are usually used. The average value is taken as the value of xj, cal.

In order to be as close as possible to the matrix of the sample, it is sometimes possible to introduce the calibration material directly into the sample gas by adding a known amount of the measurement component.

In this case, the value of the chemical influence amount in the subsequent calibration is not constant. Direct estimation of the interaction between successive calibrations of dynamic processes Learn the maximum positive and negative deviation of the interference value.

5.3.3 Physical influence

If the value xj of the physical influence amount Xj (for example, temperature and pressure) is the same every time the calibration is performed, the value is taken as xj, cal, and Xj,max and xj,min should use the maximum and minimum values of the influence when measuring.

If the estimate is not the same when calibrating, the maximum positive and negative changes that occur during the measurement should be taken directly until the next calibration. (xj, max-xj, cal) and (xj, min-xj, cal) (note the sign contained in the value).

5.4 Uncertainty assessment and applicability assessment

5.4.1 Overview

The evaluation principle and general requirements for measurement uncertainty are specified in JJF1059.1. The measurement should be clearly defined.

It should be confirmed that the performance of the measurement system is valid.

The steps of the measurement procedure (eg, sampling, analysis, post-processing, and calibration) and materials (eg, reference materials) should be described. The required measurement quality should be clearly stated.

--- Required extended uncertainty Ureq (95% confidence);

--- Define the test value ctest of Ureq;

--- Define the average time of Ureq.

The evaluation process is shown in Figure 1.

5.4.2 Analytic functions, model functions, variance functions

The analytical function between the input quantity yi and the measured value c is given by equation (1).

If the influence quantity xj produces a correction term for the measured value c, the general model function of the measured value c is given by equation (2).

If the input is not related to the amount of influence, the variance of c is given by equation (3).

From equation (3), the square of the synthetic standard uncertainty uc is the weighted sum of the input and the square of the uncertainty of the influence, see equation (4).

If the uncertainty of the experimentally determined sensitivity coefficient (bj) is not negligible, then the term sum deltax2ju2(bj) should be included in equation (4).

5.4.3 Analysis of source of uncertainty

5.4.3.1 Principle requirements

The source of uncertainty should be analyzed according to the actual measurement situation, especially the main ones that have the greatest impact on the measurement uncertainty. The source of certainty should not be duplicated or missing.

Measurement errors or any unexpected events should not be considered as a source of uncertainty.

Not all sources of uncertainty need to be included in the calculation plan. Any uncertainty component does not exceed the maximum standard uncertainty If the amount is 20%, the source of uncertainty is negligible.

5.4.3.2 Uncertainty related to response time

Response time is the source of uncertainty for continuous measurement systems.

The measured value may be affected by the previous sample (hysteresis effect), or by the sampling process (eg, mixing, reversible adsorption, etc.), or the test process Affected by (eg, electronic time constants, etc.), the actual impact will depend on the response time being measured. Claim.

--- The response time should be less than 25% of the average time, at which point its impact is negligible;

--- Under high dynamic conditions, the fluctuation of the measured value within 5% of the average time is higher than the test value (ctest), then the response time should be Less than 10% of the average time. At this point, its impact is negligible.

5.4.3.3 Uncertainty related to calibration

Sources of uncertainty associated with calibration include.

---The deviation of the experimental calibration function (deviation, miscalculation);

--- Uncertainty of standard samples or reference methods;

--- (Instrumental) drift/instability and so on.

5.4.3.4 Influence of input quantity

The amount of influence on the input includes.

--- a component of the gas that may change the measured value;

--- can change the physical quantity of the measured value, such as temperature, pressure, radiation, power supply voltage and frequency;

---Chemical interference substances;

--- The influence of the operator, etc.

5.4.3.5 Others

Other various factors of uncertainty include.

---Sampling and transport effects;

--- adsorption, desorption efficiency;

---Noise, etc.

5.4.4 Estimation of uncertainty components

5.4.4.1 General

The value of the measurement system performance characteristic may be a claimed value (manufacturer's description) or an actual measured value. Any one of them can be used Determine whether the required measurement quality is met. The values used in this process should be representative in the measurement procedure.

The value of any performance characteristic may cause measurement uncertainty, and the quantification of its impact can only be performed at the test level C=ctest.

For example, the influence of the influence amount Xj includes the system deviation deltac(xj) and the random error scxj( )[ ]. Taking the square root of the mean square error as For the measurement of standard uncertainty, see equation (5).

The uncertainty of the amount of influence (xj) is calculated as shown in equation (6).

If the upper and lower limits of an influence quantity deviation are known, the standard uncertainty u(xj) can be obtained by the calculation of equation (7).

It is assumed here that the probability distribution of the quantities is a uniform (rectangular) distribution, for example, temperature; note the notation including deltaxj,p and deltaxj,n.

If the two extreme values are zero-symmetric, then equation (7) is transformed into equation (8).

Typically, the value of the performance characteristic is the result of an experimental test. Unless this contribution is rated less, the uncertainty associated with this calculation is here Cheng Cheng should be considered as an additional contribution.

5.4.4.2 Repeatability and reproducibility

The reproducibility of the input can include a variety of sources of uncertainty, such as noise, calibration uncertainty, instability/drift, operator shadow The amount of influence and environmental impact.

Reproducibility only contains random parts of the source of uncertainty. Systemic impacts from the same source should be treated separately.

Calculate the recurrence standard uncertainty uR[c(yi)] of C=ctest using the recurrence standard deviation sR(yi) of the input quantity according to equation (9).

All sources of uncertainty may be covered by reproducibility. In addition to noise, the source of uncertainty should be introduced separately in this process.

According to the formula (10), the repeatability standard uncertainty ur[c(yi)] of C=ctest is calculated by the repeatability standard deviation sr(yi) of the input amount i.

5.4.4.3 Linear

If the applied linear calibration function does not match the actual calibration function, such as nonlinearity, it is called linear miscalculation. Corresponding to C=ctest Yi=yi, the linear miscalculation of the input quantity Yi measured in test, which is the difference between the measured value of Yi of the calibration function and yi, test deltayi, fit.

The standard uncertainty corresponding to the measured value c is calculated using equation (11).

If the linear miscalculation is specified as a symmetrical upper and lower limit, for example, a percentage value ± P, the equivalent of equation (8) can be applied.

5.4.4.4 Uncertainty of the calibration function

The experimental calibration function is obtained from the measurements. Because the number of measurements is limited, there is always a residual uncertainty in the applied function.

Determine the Yi=yi corresponding to C=ctest, and the uncertainty of the calibration function of the input quantity Yi in test is the standard deviation s(^yi). According to formula (12) A standard uncertainty component corresponding to the measured value c is calculated.

5.4.4.5 Instability/Drift

Corresponding to Yi=yi of C=ctest, the change of the measured value of the input quantity Yi in test is represented by instability, which includes the system of drift D(yi) The general term and the random term sinst(yi). The standard uncertainty of the measured value c is given by equation (13).

If instability (drift) is specified as a symmetrical upper and lower limit, for example, every q days is specified as a percentage value ± P, then available The equivalent of equation (8) is used to calculate the standard uncertainty.

5.4.4.6 Selectivity

The selectivity Ij indicates that the change in the measured value c is attributed to the change in the interference value xj […]