NONLINEAR PID CONTROL FOR COMPLEX INDUSTRIAL PROCESSES

Abstract

PID controllers have been widely recognized as an effective and practical too] for most linear processes. However, its performance may deteriorate when there exist severe nonlinearity and mode] uncertainty in the process. An effective way to cope with the problem is the introduction of nonlinear PID control in which the PID parameters are some nonlinear functions of the error and derivative of the error. With this flexibility, a nonlinear PID controller may improve the performance for processes with nonlinearity and model uncertainty. Due to the complexity of industrial processes, it is difficult to find a general structure of nonlinear PID controllers which can handle all kinds of nonlinearity and model uncertainty. The idea of designing different kinds of nonlinear PID for different kinds of processes will be more realistic. Design of an nonlinear PID controller usually consists of two steps. The first one is to determine a proper structure of nonlinear PID controller for a certain type of processes. The second is to develop a tuning formula for setting the controller parameters. In this thesis, four kinds of nonlinear PID controllers are proposed to control four different types of processes. Firstly, an nonlinear PID controller is designed for unsymmetrical processes. Two modified relay tuning methods are developed to find out the system dynamics and a gain scheduling method is used to design the nonlinear PID controller. Secondly, a nonlinear PI controller is developed for pH process as well as the linear process with parameters uncertainty. The controller has variable structure and can tune the control action to be stronger or weaker according to the current system performance. Thirdly, another auto-tuning nonlinear PID controller is developed to control a multiple-tank process. The nonlinear controller is designed based on extended linearization. Gain margin and phase margin method is used to give the practical tuning formulas. Final1y, identification and nonlinear PD control algorithm are also presented for continuous-time nonlinear mechanical systems. The methods presented in the thesis are supported by both theoretical analysis and simulation. The author believes that these methods have practical value and will contribute as useful modules within the commercial controllers, especially for complex industrial processes.