Computer-Integrated Technology for Controlling Experimental Measurements with Unknown Statistical Regularities

Abstract

Modern production is impossible without performing calculations in the design and quality control of finished parts and their components according to technical requirements. All the initial numbers for the calculations, with rare exceptions, are products of measurements. The properties of experimental samples contain information not only about the object but also about the measurement procedures. In the process of quality control of products, correct and adaptive measurement processing technology plays a key role.

The paper considers the problem of reference defectoscopy, which uses only normative standards, and there are no defect standards for detecting defects or deviations in technical objects. It is assumed that as a result of the control, samples of measurements characterizing the normal state of the object are obtained. Therefore, all other states that differ from normal will be considered defective or abnormal. To solve such problems, statistical homogeneity criteria can be applied. Among the criteria for checking the homogeneity of samples of random variables, parametric and non-parametric criteria are distinguished. Parametric criteria can only be applied to data that have certain assumptions about a certain probability distribution law. That is, they are applied only to certain laws, and the unknown is preserved at the level of parameters. Non-parametric criteria are criteria that do not require knowledge of the parameters of the sample distribution and do not depend on the specific type of probability distribution law. Most non-parametric criteria are rank-based, namely, they are based on ranking data, i.e., transforming input data into a sequence by their magnitude, and then comparing this sequence with other sequences of data. These criteria can be used to compare means, medians, variances, and other characteristics of two or more samples.