Extraction Method Of Fuzzy Control Rules Based On Case-based Reasoning

Fuzzy control is a kind of control method that applies the fuzzy system theory to the field of industrial production control.For systems with complex characteristics such as time delay and large inertia or difficult to describe precisely,the traditional control methods such as PID control effect is often poor,and fuzzy control method can achieve better results.The fuzzy controller is composed of fuzzy reasoning,knowledge base,fuzzifier and defuzzifier.Among them,fuzzy control rules in knowledge base,as the core problem of designing fuzzy logic controller,are one of the important research directions in fuzzy control field.The current rule extraction method has achieved some application results,however,there are still two key scientific problems,the quality of the extracted rules is not high and the extraction efficiency is low,have not been fully solved.In order to improve the quality and efficiency of the extracted rules,this paper uses a cognitive reasoning–Case-based reasoning(CBR)to extract the fuzzy control rules,which is derived from cognitive psychology,and studies the structure and function of the extraction rule model,the maintenance and revision strategies and experimental tests in the extraction process.the main work is as follows:(1)Design the structure and function of CBR extraction rules.According to the classical CBR reasoning model,a CBR-based rule extraction model is designed to realize the cyclic reasoning solution process of fuzzy control rules,includes generating initial rule base,rule clustering,generating fuzzy rule base,reducing rule base,similarity-based rule extraction,heuristic rule revision,rule storage and so on.All the rules used to design the fuzzy controller can be extracted from the sample data.(2)Design the methods of rules maintenance and revision methods.In order to reduce the adverse effects of redundant rules on fuzzy control performance,a similarity-based rule maintenance(SRM)method is used to calculate the similarity between a rule and other rules one by one in the rule base to remove the rule of which the similarity is 1 to other rules.In addition,in order to improve the rationality of the control rules,a heuristic rule revision(HRR)method is used to modify the reuse rules.According to the design of the revised principle,consider three case,that is,the rule of which the similarity is 1 to the target rule does not exist,the rule of which the similarity is 1 to the target rule exist and the number is unique,or numerous,to modify and obtain the final fuzzy control rule.(3)Experimental study.Experimental study includes three parts,that is,the development of experimental platform,performance testing and comparative experiments.Developed an experimental platform based on MATLAB.The platform uses the MATLAB-GUI to compile the operation interface,which is used to extract fuzzy rule.The fuzzy rule extraction program in the m file is called by the GUI object,which can easily and intuitively observe the extraction process and control performance of the control rules.The performance test mainly focuses on the performance of each component in the CBR extraction rules algorithm and the robustness of the fuzzy controller.The comparison experiment is the test of the control performance and anti-jamming ability between fuzzy controller designed by CBR-based extraction fuzzy rules and the traditional fuzzy controller and PID controller.The results show that this method has the advantages of application.