PARAMETRIC SYNTHESIS OF MODELS FOR MULTICRITERIAL ESTIMATION OF TECHNOLOGICAL SYSTEMS

Abstract

The subject matter of the article is the problem of multicriteria estimation of the properties of technological systems (TS) in the process of their structural-parametric optimization. The goal of the study is to increase the efficiency of procedures for multicriteria estimation of TS properties at the stages of their design and reengineering using the technology of comparative parametric identification of the preferences of a decision maker. The objectives are: to increase the adequacy of the additive-multiplicative model of multifactor estimation of variants of building a TS based on the Kolmogorov-Gabor polynomial; to develop an efficient method of parametric synthesis of additive-multiplicative models of multifactor estimating and selecting variants of building a TS based on a decision maker’s preferences; to carry out the analysis and give recommendations on the practical use of the suggested method of parametric synthesis of models of multicriteria TS estimation. The methods used are: system analysis, decision theory, identification theory, multicriteria optimization methods. The following results are obtained: to increase the adequacy of the models of TS multifactor estimation, it is suggested to use the utility function of partial criteria that makes it possible to realize not only linear, convex or concave, but also S (Z)-like dependencies on their values. To solve the problem of parametric synthesis of models of multicriteria TS estimation, the method of comparative identification of a decision maker’s preferences is improved on the basis of the procedures for calculating the Chebyshev point and the residual vector. The experimental study of the efficiency of the suggested variant of the method is carried out. Conclusions. The application of the suggested function in additive-multiplicative models of TS multi-factor estimation does not change the methods for selecting their parameters. The suggested improvement of the method of comparative identification of a decision maker’s preferences on the basis of the procedures for calculating the Chebyshev point and the residual vector for the parametric synthesis of models of TS multicriteria estimation enables covering all practically important situations of selection described by binary relations of equivalence, strict, and nonstrict preferences. The experimental study of the method confirms the increase in the efficiency of the procedures of parametric synthesis of models built on its basis in comparison with the method of group accounting of arguments on the basis of genetic algorithms. Practical application of the results obtained in the support systems for making multicriteria design and management decisions will improve their accuracy and, on this basis, increase the functional and cost efficiency of modern TS.