研究目的
To revise the current Guide to the expression of uncertainty in measurement (GUM-93) and propose an algorithm for processing measurement results that addresses its inconsistencies with supplements GUM-S1 and GUM-S2, particularly in the evaluation of type A uncertainty and the consideration of non-Gaussian distributions of input quantities and non-linearity of model equations.
研究成果
The study presents a revised approach to processing measurement results and evaluating uncertainty, addressing the limitations of GUM-93. It introduces methods for obtaining unbiased estimates of the measurand and its uncertainty, considering non-Gaussian distributions and non-linear models. The kurtosis method is shown to provide accurate estimates of expanded uncertainty for symmetric distributions, with deviations from MCM estimates not exceeding ±2.5% for more than 5 repeated measurements.
研究不足
The proposed methods are valid only for normally distributed results of multiple measurements when the number of repeated measurements is more than 4. The kurtosis method is suitable for symmetric distributions, and the Monte-Carlo method must be used for asymmetric distributions.
