# ISO 6974-2:2012 pdf download – Natural gas -Determination of composition and associated uncertainty by gas chromatography 一 Part 2: Uncertainty calculations

ISO 6974-2:2012 pdf download – Natural gas -Determination of composition and associated uncertainty by gas chromatography 一 Part 2: Uncertainty calculations.

5.2 Principles

Uncertainties associated with the component mole fractions shall be calculated in accordance with ISOFIEC Guide 98-3.

For Type 1 analyses in accordance with ISO 6974-1 tho uncertainty calculation includes random and systematic uncertainbes from three main sources: uncertainty of the certified reference mixtures, uncertainty of analysis and uncertainty of the data fitting procedure.

For Type 2 analyses in accordance with ISO 6974-1, the uncertainty calculation includes both random elements and systematic errors introduced by the assumption of a linear response through the origin, the systematic errors being calculated from the results of the initial performance evaluation.

Subclause 5.3 describes methods for estimating the uncertainties of processed mole fractions calculated from raw mole fractions using the conventional normalization method, Annex A provides a method for use when the methane-by-difference approach (see ISO 6974-1:2012. Annex C) is employed,

ISO 6974-1 recommends the use of the generalized least squares (GLS) approach for calculation of the processed mole fraction. However, in some circumstances, an alternative approach using ordinary least squares may be acceptable and calculation of uncertainty in processed mole fractions in this situation is described in Annex C.

5.3 Step 9— CalculatIon of uncertainty of mole fractions

5.3.1 Determining the equations to be used

5.3.1.1 General considerations

The equations to be used In this step for calculating the uncertainty of mole fractions are given In 5.3.2 and

5.3.3. The equations to be used should be determined by following the three-stage process described In

5.3.1.2 to 5.3.1.4.

The following pornts should be taken into considerabon when selechng the equations to be used.

a) When usmg the mean normaiizatioW method (see 5.3.2). the following are calculated in turn for each analyte:

1) mean peak analyser response from all runs;

2) raw mole fraction;

3) normalized mole fraction.

b) When using the ‘run-by-run normalizatIon’ method (see 5.3.3). the following are calculated in turn for each analyte:

1) raw mole fraction for each run;

2) normalized mole fraction for each run;

3) mean normalized mole fraction.

B.1 Uncertainties of relative response factors for flame ionization detectors (FIDs)

The relative standard uncertainties of the relative response factors, as calculated for an FID and given in ISO 6974-1:2012. Table Dl. shall be taken to be equal to 2 %121. Alternative figures may be used ii determined by thoroughly validated oxpenmental procedures.

NOTE The method for determining the relative response factors for an FID Is given m ISO 6974-1:2012. 0.1.

B.2 Uncertainties of relative response factors for thermal conductivity detectors (TCDs)

The relative standard uncertainties of the relative response factors, as calculated for a TCD and given in ISO 6974-1:2012, Tab4e D.2, shall be taken to be equal to 10 %121. Alternative figures may be used If determined by thoroughly validated expenmental procedures.

NOTE The method for determining the relative response factors for a TCD is given in ISO 6974-1:2012. 0.2.

Thás annex provides an alternative procedure3l to the generalized least squares approach (see ISO 6974-1:2012, 6.5.5). The approach described in this annex has the benefit of being a more straightforward procedure with which to carry Out the calculations. In order to maintain the simplicsty of this alternative approach, it can only be applied when the analysis and calibration functions can be approximated in a first-order form.

Consider a set of data of points (xi. y,) forming a first-order calibration curve where x is the mole fraction of each standard and v, the Instrumental response. The equation for a first-order calibration function Is given by Equation (C.1).