A nonlinear system identification method based on fuzzy dynamical model and state-space neural network

A novel method of fuzzy modelling using multiple local state space neural networks is propesed to handle complex nonlinear dynamics. It combines fuzzy logic and neural networks by a sound framework. The overall nonlinear system is represented by a set of state-space neural networks, connected by fuzzy variables. The resulting neural networks can be directly represented as state-space format so that control and fault diagnosis based on state space equation becomes more straight and easier. The efficiency of this method is tested by applying to a typical nonlinear system: three water tank system.