Knowledge Representation Using Fuzzy Logic Based Characteristics
Nasredin Chaker, Rainer Hampel · 2000
Rule based systems for Fuzzy Control, Fuzzy Supervising and Fuzzy Diagnosis are non-linear multi-input, multi-output systems. The result of the signal processing within the system is a high dimensional characteristic field. For optimizing the behaviour of the system, we have a large number of degrees of freedom. The reconstruction of the knowledge base from the characteristic field as starting point is impossible. With this background, the paper describes the quality of a two dimensional Fuzzy Controller characteristic field in connection with the necessary deformation for the compensation of non-linear effects. We will demonstrate that one should define two restrictions for the characteristic field: continuity without local extrema and differentiability. Under such conditions we need only two free degrees for optimizing the Controller behaviour using the characteristic field deformation. With help of this experience, the high dimensional Controller will be cascaded. A subjective decision for it is to distinguish the type of the input variables in dominant, non-dominant, and optimization variables. The quality of the cascade is depending on the final characteristic field and the completeness of the rule base. An example will demonstrate the effects of cascading. The cascading concept is realized in the so called High Speed Matrix Controller (HSMC), whose structure is described in this paper, as well. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.