ALGORITHMS OF DATA PREPARATION FOR PARAMETRIC SYNTHESIS IN PRELIMINARY DESIGN OF THE TECHNICAL OBJECTS BASED ON EXPERT INFORMATION PROCESSING

Abstract

The article describes the issues of preparation of the expert information for parametric synthesis at the preliminary design of the technical objects under uncertain parameters. Within B. Liu’s uncertainty theory, the algorithms for generating critical variants of the uncertainty distribution functions reflecting the achievement of the “best” and “worst” for a decision-maker (DM) of the numerical characteristics of functions that depend on the uncertain parameters are developed. It is proposed to carry out the formation of uncertainty distribution functions providing a compromise between the expert views, in two stages including the grouping of uncertain values reflecting the views of some experts, and establishment of the compromise functions of uncertainty distribution  for each selected expert group. The algorithm of uncertain values clustering on the fixed number of classes based on the k-means method is developed. An interactive algorithm of uncertain values grouping is suggested. In terms of this algorithm, uncertain values act as graph nodes, whereas the threshold value determined by DM and the distance between uncertain values define the presence or absence of connection between these nodes. Using the proposed expert information processing algorithms, the uncertainty distribution functions are formed, which are the input to algorithms for calculating the numerical characteristics of functions that depend on non-deterministic parameters.