Design of a data-oriented PID controller for nonlinear systems

Abstract

It is difficult to maintain the desired control performance by applying a fixed PID controller because most process systems have nonlinearity. The CMAC-PID controller which uses Cerebellar Model Articulation Controller (CMAC) has been proposed as the effective method to these systems. However, some problems of CMAC related to the learning time or the huge memory needed are pointed out due to its structure. The issue of the learning time has been solved by the authors' method with the Fictitious Reference Iterative Tuning (FRIT) approach in recent years. In this paper, the method to deal with the problem of the required computer memory is considered by using Group Method of Data Handling (GMDH) approach. According to this method, even though the proposed GMDH-PID tuner keeps the same performance as the CMAC-PID tuner, the required computer memory is drastically reduced. The effectiveness of the proposed method is examined by a simulation. © 2012 Springer-Verlag.