Research On Self-Turning Control Algorithm Of Nonlinear System Based On C-R Fuzzy Model

Most of the controlled objects are complicated nonlinear systems and showthemselves more and more characteristics such as, first, the complexity: both thestructure and the parameters of the system are multidimensional, variational andnonlinear; second, the uncertainty: system itself and the outside environment includemany unknown and uncertain factors; third, the high performance: it needs to considerall kinds of complications when designing controller, because of the conflict betweenthe diversity of the control target and the target itself.Neglecting some of the nonlinearity or substituting them with linearity, linearmodel is only the approximate mathematical description of the actual system. Althoughthe conventional linearity theory is mature theoretically, it has been unable to satisfy usin industry area which becomes more and more developed day by day. So more andmore attentions have been paid to the study of nonlinear system both at home andabroad. In this paper, the author studies self-adaptive control of the nonlinear systembased on the C-R fuzzy model:Firstly, C-R fuzzy model structure of the complex system and identificationmethod are studied, including Fuzzy Input Space Clustering Algorithm (FISCA) andFuzzy Output Space Clustering Algorithm (FOSCA). Because of the blindness ofchoosing the initial factor in practical using, the author changes the initial membershipand makes them close to the work station. The number of judgment is reduced bychanging the condition of final judgment.Secondly, the C-R fuzzy model has linear system analytic form at every moment,but for a time domain it can represent the nonlinearity of a system. Basing on thispeculiarity mentioned above, a new Generalized Minimum Variance Self-turning Control algorithm, which is based on C-R fuzzy model, is proposed.Thirdly, on the basis of changing the C-R fuzzy model to a C-R fuzzy step model,whose structure is similar to step response model, self-turning PID algorithm isresearched based on the C-R fuzzy step model.And at last, the clustering method based on relational grades is studied. Thenumber of fuzzy subspaces of the C-R fuzzy model is identified by using the clusteringmethod, and the identification method of the C-R fuzzy model is improved. A dynamicidentification method based on the clustering method by relational grades is proposed.