FUZZY CONTROLLERS - LIFTING THE LINEAR-NONLINEAR FRONTIER

This paper brings back some ideas related with the linear analysis of fuzzy controllers. A fuzzy control algorithm does not go further than a non-linear function described by its inference map. Firstly, it is possible to obtain the inference map corresponding to a classical PID controller by choosing adequately the linguistic terms, the membership functions and the table of rules. Then the inverse problem is presented: to obtain the closer PID to a given fuzzy controller. From these ideas, the paper focuses its attention on the obtainment of a useful tool to design fuzzy controllers at least as good as the PID that allows the system to follow a specified behaviour. The next step is to improve the fuzzy controller parameters. Finally, an industrial application of these ideas is shown. The design of a fuzzy controller over a clinker cooler with grill, improving P and PI controllers already functioning, is discussed.