New types of knowledge about system dynamics for intelligent control system design.

Abstract

This is a thesis by publication. This thesis comprises ten published/submitted journal articles including eight research articles and two review articles. Five of these journal papers have been already published or finally accepted for publication. This thesis, based on research undertaken in the area of intelligent and non-model-based control, aims at broadening knowledge about system dynamics applicable for control system design. Currently, mathematical models of systems, experimental input-output data and experts’ knowledge in the form of fuzzy rules (linguistic expressions) are three types of knowledge about systems dynamics which are employed in control system design. These types of knowledge are used to define the number and the positions of controllers in the control system (architecture of control systems) and/or the mathematical form of controllers (controllers’ structure) and/or controllers’ parameters. Defining control systems at these three levels (architecture, controllers’ structure and parameters) forms the process of control design. In the area of non-model-based control, some cases of unexpected poor control performance were observed by the author. For instance, neuro-predictive method controls process plants very well. This method is based on input-output data. Thus, this control technique seems to promise a good control performance in general. However, attempts to use this technique in yaw angle control of a model helicopter were unsuccessful regardless of the effort spent on tuning the controller parameters (see Chapter 3); similarly, unsuccessful outcomes resulted for feedback fuzzy control of model helicopter pitch angle (see Chapter 4). This thesis has two main contributions: firstly, this thesis provides explanation for the aforementioned unexpected poor performances through introduction of two new concepts: Control Inertia (see Chapter 3) and Generalized-Type-Zero (GTZ) Systems (see Chapter 7). It is shown here that high control inertia systems witness a considerable repeating overshoot, and GTZ systems need consistent control input to maintain their desirable control output. Secondly, based on these two new concepts, this thesis offers new control methods with developing new control ideas: fuzzy brakes and steady state control laws which can improve the performance, energy consumption and suitability for implementation of control systems. The proposed methods are shown to be usable for a wide range of systems. The merits of the proposed control approaches were indicated theoretically and practically. As a result, being/not being high control inertia and being/not being GTZ were used as new types of knowledge about system dynamics applicable in control system design.