Application decomposition in space with a generative elements for solving problems of probabilistic diagnostics

Abstract

In spite of plenty developed methods for probabilistic diagnostics, obtaining the effective solutions for Non-Gaussian models of statistical data is an actual problem. In this article the application possibility analysis of decomposition in space apparatus with a generative element for the decision of detection and identification (recognition) tasks of random processes’ disorder, described by higher-order statistics, is conducted. The offered approach is positioned as a semi-parametric variant of statistical analysis, constituted on a compromise between simple non-parametric and optimal realization-difficult parametric methods. The article tells about the systems’ structure that is based on polynomial matched filter, designed for sequential detection and identification of disorder. The features of adaptation property implementation of such systems are analysed. The task of disorder’s sequential detection on average and variance of Non-Gaussian random sequences is researched by statistical modelling as an example. Obtained results confirm effectiveness of the proposed approach for solving of the probabilistic diagnosis’ tasks, which can be used to construct automated systems for monitoring and diagnosis of Non-Gaussian random processes in various application areas.