Sugestie kilku uściśleń w Przewodniku GUM-2008 i zaokrąglanie wyników pomiarów wieloparametrowych

W pracy wykazano, że kilka zaleceń Przewodnika GUM o wyrażaniu wyników pomiaru jest niewłaściwych dla poprawnego przetwarzania danych w pośrednich pomiarach wieloparametrowych, a w przykładzie H.2 nie oszacowano poprawnie korelatora. Podano progi ograniczające minimalne liczby cyfr przy jednolitym zaokrąglaniu danych liczbowych multimezurandu z przykładami oraz proponowane zasady dalszego zaokrąglania poniżej tych progów. Zaproponowano kilka zmian w GUM. Z pracy wynika, że pilnie potrzebne są międzynarodowe zalecenia do poprawnego wyznaczania wyników pomiarów wieloparametrowych (Suplement 2 do GUM) oraz normalizacja prezentacji numerycznej ich danych.

this paper it is shown that recommendations of evaluation estima-tors of the value, uncertainty and correlation coefficients given by the international metrological Guide GUM-2008 are not useful enough for the proper evaluation of results in multidimensional indirect measurements. Mistakes are shortly discussed. It is found that smallest eigenvalue of correlation matrix obtained after calculations for output multi-measurand in the GUM H.2 example is negative and data is not proper as cannot be presented by 3-order normal distribution. Output data of the H.2 example are evaluated by two cases. First one is reference - when input data are considered as given absolutely accurate and thresholds of safety rounding calculating from them output data are find. In the second case, proper to wide application, method rounding of output is joined with accuracy of uncertainty of input data. Two variants of this method are compared. Few corrections of H.2 example are proposed. Formulas for the proper rounding of the multi-measurand numerical data up to thresholds, and recommendations for rounding below them are given. The minimal data structure for describing the multi-measurand is defined. From this work it results that in multivariable measurements urgently is needed standardization of the method of measurement data evaluation (Supplement 2), and also for their presentation which allows full numerical review of the multidimensional data, their wide dissemination and unlimited in time storage.