Context:

In the design process of Fuzzy Inference Systems based on Boolean and Kleenean Relations (FIS-BKR) there is a dilemma choosing the regular kleenean extensions of a given boolean function. The set of possible kleenean extensions of a boolean function has a lattice structure under the usual partial order of functions. The fuzzy convex combination proposed by Zadeh guarantees some properties related to this order.

Method:

The addition of a convex combiner just before the defuzzifier offers a solution to the above situation. The ISE (Integral Squared Error) and ITSE (Integral Time-weighted Squared Error) performance indexes were used on an application for tuning a liquid level control system.

Results:

The tuning process carried out on the FIS-BKR controller with fuzzy convex combiner using constant coefficients, implied an improvement of the controlled system up to 1.427% for ISE index and up to 21.99% for ITSE with respect to the extreme extensions.

Conclusions:

New evidence of convenient characteristics of FIS-BKR controllers with fuzzy convex combiner was presented when the performance indexes ISE and ITSE were evaluated. On the other hand, although in this work parameter tuning for convex combination was done by grid search (brute force), it would be interesting to study more effective optimization methods for this purpose.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )