Diagnosis Problem Solving Using Fuzzy Relations

This paper deals with the restoration and the identification of the causes (diagnoses) through the observed effects (symptoms) on the basis of fuzzy relations and Zadeh´s compositional rule of inference. We propose an approach for building fuzzy systems of diagnosis, which enables solving fuzzy relational equations together with design and tuning of fuzzy relations on the basis of expert and experimental information. The essence of tuning consists of the selection such membership functions for fuzzy causes and effects, and also fuzzy relations, which minimize the difference between model and experimental results of diagnosis. The genetic algorithm is used for solving the optimization problem. The proposed approach is illustrated by the computer experiment and the real example of diagnosis.