In metrology, it is essential to analyse the instrumental drift of measuring instruments and measurement standards. Each reference instrument is periodically calibrated according to a frequency determined by the laboratory. Calibration establishes the metrological state of the instrument on a certain date of calibration. However, it is necessary to know the state of the measuring instrument during or after the calibration.
Reliable accounting for drift plays an important role in maintaining measurement accuracy. Otherwise, it can lead to significant measurement errors. Accounting for time drift is mandatory when conducting international comparisons of national measurement standards. The drift uncertainty can be evaluated from its history of successive calibrations. In the absence of such a history, the magnitude order of the calibration uncertainty can be estimated.
The analysis of the long-term drift of travelling measurement standards is limited to examples of key and supplementary comparisons of measurement standards of electrical capacitance. Quite a lot of such comparisons were conducted both by the Consultative Committee for Electricity and Magnetism (CCEM) and by most of the Regional Metrology Organizations (RMOs). There are international standards and guides that describe various statistical methods of analysing the measurement results.
For capacitance measurement standards, time drift is predictable and nearly linear. For comparisons of measurement standards, a linear model is more than often applied, as a travelling measurement standard with excellent stability characteristics is used. The consistent results have been obtained. The linear model was applied to estimate the drift of travelling measurement standards during the key and supplementary comparisons (COOMET.EM-K4, COOMET.EM-S4, and COOMET.EM-S13) of measurement standards of electrical capacitance. The estimation of the long-term drift of measurement standards of electrical capacitance as travelling measurement standards for comparisons using a polynomial regression are presented.