Pogonets I. Entropy Methods and Computational Techniques for States Identification of Controlled Objects

The thesis proposes the explanation of methods to increase the informativeness of state identification and diagnosis of controlled objects based on the integrated correlation assessment of entropy. It has been investigated entropy characteristics of analog-to-digital encoders and analog-to-digital converter implemented in a variety of theoretical and numerical bases. The advantages of submission of information in data bases Krestenson's and Galois, characterized by maximum entropy condition. It has been analyzed and formalized the entropy assessment of linear and nonlinear sources of information with regard to their spectral characteristics. An entropy correlation model of single-channel, multi-channel, stationary and quasi-stationary sources of information has been developed. The analytical expressions of entropy correlation models for analytically defined signals and the whole set of signaling, statistical and spectral characteristics of sources of information were received. The expediency of using autocorrelation function of equivalence to calculate the entropy of a wide class of sources has been substantiated. The structure of special processors of determination of the correlation measure of entropy and the structure of a computer system equipped with special processors of this class as an information and integrating components was proposed. Modeling and implementing of special processors to programmable logic array and evaluation of structural and temporal complexity was performed. The verification of functions of developed special processor was done by example of the entropy assessment of sugar production processes.