A new recurrent neurofuzzy network for identification of dynamic systems

In this paper a new structure of a recurrent neurofuzzy network is proposed. The network is based on two interconnected Fuzzy Inference Systems (FISs), one recurrent and another static, that intend to model the behavior of an unknown dynamic system from input–output data. In the proposed structure each rule involves a linear system in a controllable canonical form in order to reduce the online computational load and facilitate the online checking of the stability of the resulted network. The training for the recurrent FIS is made by a gradient-based Real-Time Recurrent Learning Algorithm (RTRLA), while the training for the static FIS is based on a simple gradient method. The initial parameter conditions prior to training are obtained by extracting information from a static FIS trained with delayed input–output signals. To demonstrate the effectiveness of the proposed structure, two nonlinear systems are identified.