Intellectual Approach to the Design of Fuzzy Systems Based on Multi-objective Evolutionary Modeling

Over the past decades, fuzzy systems (FSs) have become widespread in many areas of practical application due to their universal approximating properties, as well as their ability to process inaccuracies and uncertainties in the behavior of complex systems without requiring accurate mathematical models. However, the design of such systems is not a simple objective since there is a lack of sufficiently complete expert information about the object being modeled, and designing an FS involves automated design methods based on experimental data. In this regard, for the FS design the authors propose a new intellectual approach based on the development of the theory of multi-objective evolutionary modeling which is based on the Pareto-optimality principle involving evidence combining theory or Dempster-Shafer (DS) theory) and provides selecting from the Pareto-optimal sets the solutions that best suit the multi-objective design aims, and increase the objectivity of the selection among the best solutions in multi-objective algorithms. The results of experiments in solving the problem in the field of railway automation related to the prediction of the cut rolling speeds in splitting up of train on a hump yard are presented.