A relational approach for complex system identification

In this paper the issue of complex system identification with the aid of fuzzy logic techniques is addressed. Models based on fuzzy relational equations, i.e. fuzzy relational models, are presented. Two numerical methods to estimate the parameters of such models are proposed. The first one is an optimization based methodology using sequential procedures of quadratic programming to refine models previously estimated by other methods. From this methodology, a recursive algorithm suitable for on-line identification is derived. The performance of the methods proposed is evaluated by modeling a real dynamic process