Romanuke V. Theoretic-game methods of identification of models for multistage technical control and run-in under multivariate uncertainties

The investigation object is a finite-staged process of identification of models for multistage technical control and run-in under multivariate uncertainties; the investigation goal is in increasing efficiency of identification of models for multistage technical control and run-in under multivariate uncertainties of their parameters and relationships with minimax optimality principle in theoretic-game models of removing multivariate uncertainties regarding finiteness of the plays' horizon and unique selection of the optimal strategy; methods of investigation are based on general regulations of system theory, game theory, mathematical modeling, probability theory, optimality principles in gaming modeling, principles of functional analysis, methods of approximate computation; theoretical results - there has been created a methodology of identification of models for multistage technical control and run-in under multivariate uncertainties for resource-saving tracking of states of functioning of technical objects and their state-by-stage run-in, where there are methods of identification by practicing the discrete and continuous minimax strategies, methods of identification by determining sets of rational and superoptimal strategies, method of identification by evaluating the aftereffects of the minimax strategy application, there have been improved methods of parametrical identification by finding the set of all continuous minimax strategies in classes of the convex (concave) gaming models for removing interval uncertainties, there has been found the solution of the generalized problem of minimizing the maximal unbalance of requirements and resources for removing groups of interval uncertainties, two computational methods have been developed for evaluating a generalized functional integral in determining the identification superoptimal strategy; practical results - usage of the developed methods of identification allows optimizing processes of training and determining parameters of neural nets, high-precision indirect measurements and predictions, processes of searching faults, to allocate projector's resources optimally under partial uncertainty of requirements, and also to estimate unknown probabilistic distributions. Rate of implementation - results of the dissertation were implemented at SIECC AMIAU in Khmelnytskyy region, at LLC "Vzuteks" (Khmelnytskyy), at National academy of the State frontier service of Ukraine of B. Khmelnytskyy (Khmelnytskyy), at OJSC "Temp" (Khmelnytskyy), at SME "Hydraulics Vinnytsia-Service" (Vinnytsia), at concern "ORTUS-LTD" (branch in Vinnytsia). Field (scope) of usage - high-performance identifiers in neuronet systems and static scenario models for tracking states of functioning of technical objects and their state-by-stage run-in.