Methodology for Assessing the Adequacy of Statistical Simulation Models

Abstract

Statistical simulation models of complex technical systems characterized by several indicators of operational efficiency were studied in this paper. The efficiency of obtaining knowledge on the examined systems depends on the quality of the models used. One of the basic properties describing the quality of a model is its adequacy – the complex property characterizing the degree of conformity of the values of the output parameters of the model with the object with the required accuracy and reliability. Current approaches to evaluating the adequacy of models are based on various subjective convolutions of confidence factors from research results to a generalized indicator the essence of which, as a rule, is not interpreted. The presented method of assessing the adequacy of statistical simulation models of complex technical systems with several performance indicators differs from existing methods by using a generalized indicator of adequacy, which is the probability of achieving the required confidence of all the accuracy requirements to determine each of the considered performance indicators. This indicator is a natural, unambiguously interpreted (the probability of satisfying the requirements for model adequacy) objective and generalized indicator of adequacy of the simulation model. For preliminary calculations we use the Parzen–Rosenblatt method and obtain the probability density function of distances between real and model indicators of effectiveness of the examined system. The required result is then obtained by the suggested algorithm of multiple integration of the density function using the Monte-Carlo method. Recommendations on the realization of the computational procedures foreseen by the method are given. The application of the method is illustrated by a description of a computational experiment.