Model-based fuzzy controller

Most fuzzy controllers developed to date have been of the rule-based type, where the rules in the controller attempt to model the operators response to particular process situations. These controllers require considerable 'knowledge engineering' in that someone has to gather a collection of rules from knowledgeable operators and then condense them into a consistent rule-base for the controller. An alternative approach to using fuzzy logic in a controller is described in this paper. Instead of attempting to model the operator's decision making process, this controller design uses a fuzzy model of the process itself and imbeds this in a relatively conventional model-based controller. The paper also describes two tests of the controller design. The first is a simple level control simulation, and the second is the temperature control of a laboratory heat exchanger. The results of this work indicate that the fuzzy-model-based controller described here can equal or even exceed the performance of more traditional control techniques, even on quite simple processes.