A new procedure for the determination of calibration curves for a gas chromatograph used in natural gas analysis

In the present paper, the authors propose an alternative methodology for the determination of the calibration curves of a gas chromatograph employed in natural gas analysis. Such methodology, coming from a modification to the regression technique known as Generalised Distance Regression (GDR), was applied to the data obtained during the calibration procedure of the gas chromatograph lodged in the bQuality ControlQ laboratory of the Snam Rete Gas (San Donato Milanese, Milan), accredited with the Calibration Service in Italy (SIT). Binary mixtures, gravimetrically prepared according to existing international standards in the abovementioned laboratory, were used as reference standards for the instrument calibration. In order to obtain the gas chromatograph calibration curves, the GDR technique was first applied, to consider the heteroscedastic uncertainty both in the dependent and independent variables. However, for some chemical components, a polynomial curve of order higher than one were obtained. In general, since the gas chromatograph is affected by time drifts, it would be favourable to have straight lines as calibration curves. In this way, the periodical check (usually made daily) would be extremely simplified. In fact, to determine the drift entity for a line, two checkpoints would be necessary, and if the line would pass through the origin, a single point would be necessary. In order to simplify the management of the time drift check procedure, the authors introduced a modification to the GDR technique, obtaining the Modified Generalised Distance Regression, which allows one to impose a straight line as calibration curve for each component, obtaining an unbiased estimation of the calibration coefficients and of their variance–covariance matrix.