Multicriteria selection of the optimal variant of a complex system based on the interval analysis of fuzzy input data

摘要

At the early stages of design of complex technical systems the engineers have to solve the problem of choosing the optimal structure from the multitude of possible variants. Complex technical systems are characterized by a set of multiple quality criteria and the required parameters, determined both in a formalized and non-formalized way. Part of this original data is given, in the course of design, in the form of permissible range of system parameter changes. These factors inhibit the objective comparison of system parameters when choosing the structure of the designed system, and the heuristic and expert methods of decision making, that employ the coefficients of criteria value, do not always give a true description of the mathematical model. The aim of the research is to develop the technique for normalizing the heterogeneous quality criteria, given as variation intervals and reducing them to the form that is good for further comparison and the choice of the preferential variant of the system structure. The suggested approach is based on the complex application of methods of interval analysis, fuzzy sets theory, decision making theory. The method of normalizing heterogeneous quality criteria is based on establishing preference relations in the interval form. Drawing the complex vector of quality criteria allows to obtain the technique of multicriteria selection of the complex system variant on the basis of the set of possible alternatives. The method of creating interval preferences with the set of alternative system variants using heterogeneous vector quality criteria is a fresh approach and, as different from existing methods with hard-and-fast ranking, it allows to use instead of quality criteria the membership functions that show the degree of proximity of the system variant to the efficient or Pareto optimal variants. The article describes the case history of optimal choice of the radio-electronic system at the early stages of design.