Abstract
A fuzzy logic and genetic algorithm (GA)-based approach to identifying nonlinear systems is proposed. In fuzzy modelling how to partition input domain is a critical issue. Some approaches by using clusters have proved successful, but they are not applicable when data are distributed uniformly. In this paper we propose an approach using a GA with tree-structured individuals, in which each individual defines a partition of the whole input domain, and a linear function is employed in the action part of each rule. This tree-structured-based GA approach can partition a domain in a flexible manner: some parts of the domain can be partitioned into high resolution and others have gross resolution, ensuring a high degree of approximation to a nonlinear system without increasing the computational load. For the search for the optimal solution, we use a GA to generate and evolve models by which a suboptimal solution can be found quickly