On-line Learning in Self Organising Fuzzy Controller with Application

A Self Organising Fuzzy Controller for linear or non-linear processes with a minimum knowledge about the process is developed with the only assumption that the output of the process is monotonic with respect to the input (as most industrial processes are). Beginning with an empty rule-base a fuzzy process model is progressively on-line built. In this way the controller gradually learns how to control the process. The inference and defuzzification mechanisms have their background on the Fuzzy Equality Relations Theory. A simple technique for on-line adaptation of a similarity factor is proposed. Feedforward effect is introduced in the definition of this similarity factor and in the defined error_in_advance value . This controller was successfully applied in simulation for both the regulation and the tracking problems. Practical essays were made on a real non-linear thermal process and the developed controller was able to provide satisfactory performance.