Robustness Of Fuzzy Reasoning Base On A New Measuring Method Of Perturbation Of The Rules

Intelligent modeling of uncertainty problems and reasoning is the major cutting-edge topics of artificial intelligence. Fuzziness is one of the main forms of uncertainty problems. When establishing a lot of fuzzy reasoning systems, the priori knowledge (such as rules of the fuzzy reasoning systems) is uncertain, and probably the small perturbation will occur negative effects to variety of performances of reasoning systems, which can be controlled if the fuzzy inference system has better robustness. So evaluating and controlling the performances of rules perturbation has been an important study for fuzzy reasoning systems, the problems have been researched as follow:(1) Firstly a new perturbation measuring method of fuzzy set is proposed, the new method is not the norm form compared with the original method, and the former can measure the perturbation of some special fuzzy set bitterly. Secondly we testify the properties of the fuzzy sets perturbation measuring and analyse the characters of the new measuring method.(2) When the rules have made the biggest perturbation, System robustness of multifold fuzzy reasoning, multidimensional fuzzy reasoning and chain of fuzzy reasoning are studied by the new perturbation measuring method of fuzzy set as a measure, further present corresponding necessary and sufficient conditions. Perturbation results of the multi-dimensional fuzzy reasoning forms are simulated by water tank control system.