Empirical comparison of methods of fuzzy relational identification

A number of methods have been proposed for the identification and self learning of relational fuzzy models. The paper compares some of the methods and looks at their tolerance to noise and to the choice of initial fuzzy ranges. Data from runs of a simulated fed-batch fermenter are used as a test case. The results show that identified relational models can give as good results as rule-based models and can be made to be very tolerant of noise in the identification data