Mikhail Vasilevich Lyakhovets
◽
Georgiy Valentinovich Makarov
◽
Alexandr Sergeevich Salamatin
The article is devoted to questions of synthesis of full-scale - model realizations of data series on the basis of natural data for modeling of controllable and uncontrollable influences at research of operating and projected control systems, and also in training systems of computer training. The possibility of formation of model effects on the basis of joint use of multivariate dynamic databases and natural data simulator is shown. Dynamic databases store information that characterizes the typical representative situations of systems in the form of special functions - generating functions. Multiple variability of dynamic databases is determined by the type of the selected generating function, the methods of obtaining parameters (coefficients) of this function, as well as the selected accuracy of approximation. The situation models recovered by generating functions are used as basic components (trends) in the formation of the resulting full-scale - model implementations and are input into the natural data simulator. The data simulator allows for each variant of initial natural data to form an implementation of the perturbation signal with given statistical properties on a given simulation interval limited by the initial natural implementation. This is achieved with the help of a two-circuit structure, where the first circuit is responsible for evaluation and cor-rection of initial properties of the natural signal, and the second - for iterative correction of deviations of properties of the final implementation from the specified ones. The resulting realizations reflect the properties of their full-scale components, which are difficult to describe by analytical models, and are supplemented by model values, allowing in increments to correct the properties to the specified ones. The given approach allows to form set of variants of course of processes on the basis of one situation with different set degree of uncertainty and conditions of functioning.
Dynamic maintenance model for high average-utility pattern mining with deletion operation
