Kropachek O. Theoretical bases of the analysis and synthesis of computerized information-measuring systems for diagnosing dynamically non-stationary objects

The aim of the work is to increase the efficiency of computerized information measuring systems for the control and diagnostics of objects with a priori uncertain dynamic non-stationary properties with the limitation on the volumes of measurement information. The object of the study is the process of formation of parametric uncertainty in the primary and secondary system transformations information in computerized control and functional diagnostics of objects with dynamic properties.The subject of the research is the methods of analysis and synthesis of information-measuring technologies of control and diagnostics of objects with dynamic properties. Research methods are based on the theory of probability, random processes, multivariate statistical analysis (regression, dispersion, covariance, spectral) with the primary system transformation of information; Discriminant and cumulative analyzes with secondary system transformation of information; consecutive mini-max (maximizing) analysis during the optimization procedure; methods of mathematical, physical and computer simulation in the creation of computerized control and diagnostic systems and their components. Scientific novelty of the obtained results: for the first time, differential transformations were applied for single-model statistics of accumulated sums, which allowed obtaining additional information on changes in the velocity of the mathematical expectation of the power of vibration signals. For the first time, the frequency-time moments of the mutual spectral correlation between the one-model and differentiated statistics of the accumulated sums were used, which allowed to reveal parametric changes in periodically non-stationary vibration signals. The first methods of mathematical analysis of covariance expansion of two-dimensional frequency-time moments of the inter-spectral correlation of vibration signals have been developed, which allowed to increase the number of primary statistically independent correlation-spectral parameters that carry additional information on the spectral non-stationary vibration signals. Firstly, methods of frequency-time analysis of components of the neuron informational parameter of autocorrelation, which implement a co-relational shift by a quadratically transformed wavelet spectrum and increase in the sensitivity to disturbances spectral nonstationarity thermodynamic processes. For the first time, a method of correlation-spectral optimization of the number of components of the vector of measuring dynamic non-stationary signals with the use of, as a target function, the average risk of making statistical decisions, which eliminates the displacement of the global extremum of the function, is developed.