Neuro-Fuzzy Modified Smith Predictor for IPDT and FOPDT Processes Control

Abstract

Abstract In this paper, intelligent control approaches are introduced to overcome the problems highlighted in the standard Smith predictor. First, in order to overcome the steady state error in the Integrator Plus Dead Time (IPDT) process control due to disturbance loading, a new fuzzy logic control based SP is developed by intentionally introducing a model mismatch to improve the system performance in terms of disturbance rejection and robustness to process modelling errors. In addition, for the First Order Plus Dead Time (FOPDT) process control, a SP based neural network control scheme is proposed to deal with the process modelling errors and proved to provide a significantly improved robustness. The neural network (NN) was designed to work with different types of modelling errors. Simulation results show that this NN approach provides excellent performance in terms of robustness to modelling errors and high adaptability to the control of both IPDT and FOPDT processes.