A fuzzy neural-network approach for nonlinear process control

Abstract This paper proposes an internal model control (IMC) scheme using a fuzzy neural network for process modeling. A fuzzy neural network is most useful in an environment where first-principles-based descriptions are difficult to obtain, but partial knowledge about the process is known and input-output data is available. However, previously proposed fuzzy neural-network approaches are inadequate for modeling complex chemical process systems, as when the input dimension increases, the number of hidden nodes (rules) increases exponentially. A novel fuzzy neural-network structure using hyper ellipsoids is proposed to avoid this problem. A fuzzy neural network is trained using steady-state as well as transient data by back-propagation. The inverse of the process is obtained by a simple interval halving method. The proposed approach is applied to modeling and control of a continuous stirred tank reactor and a pH neutralization process. The results show significantly better performances in comparison with a PID controller.