Hybrid conventional/fuzzy control methods in process control

In this thesis, we focus on the combination of conventional control methodology with fuzzy sets and logic theory. The philosophy behind this study is that a hybrid control system is designed to take advantage of these two methodologies. The research not only covers the investigation of the hybrid structure of the control system, but also includes the hybrid system design and analysis. Two classes of fuzzy control systems in the form of hybrid in structure are proposed. One is hybrid fuzzy-Kalman filter controller which is developed for an open-loop unstable system, a ball-beam balance system. In this control system, the conventional Kalman filter is used as the state estimator and the controller is a fuzzy logic control. The other is a hybrid fuzzy internal model control system. The control system structure consists of two layers: a primary one which is the fuzzy internal model configuration and a secondary one which is a pure integrator aiming at eliminating the static error caused by the primary one.; Along the hybrid system design and analysis line, a model-based fuzzy control system design method is introduced. The model used in this control system design is constructed in the continuous-time domain by combining the local explicit model in the differential equation form with the global implicit model in the fuzzy IF-THEN rules form. An on-line identification algorithm has been proposed to adjust the parameters of fuzzy dynamic model according to the input-output measurements. Based on this fuzzy dynamic model, two types of model based fuzzy adaptive controllers are proposed for the nonlinear time-varying open-loop stable plants. One is fuzzy adaptive internal model controller in which the feedforward fuzzy controller is designed based on the identified fuzzy dynamic model. The other is fuzzy direct adaptive control system in which the controller is formed by a series of local adaptive controller. The parameters of fuzzy controller are updated directly on-line to make the plant output to follow the reference model output asymptotically. Conventional nonlinear systems analysis i.e. Lyapunov stability analysis and the robust analysis methods are used to justify the stability of these hybrid adaptive closed-loop control systems. Extensive simulations and experiments have been carried out for demonstrating the improved performance of each proposed hybrid fuzzy control system compared with the non-fuzzy control system.