Research On Sample Data Based Fuzzy Rules Extraction Method And Its Application

Fuzzy controller is widely used in industrial process control. The establishment of fuzzy rule base is the core problem of fuzzy controller design. However, the design of fuzzy rule base is based on the understanding of the controlled system which is mainly derived from expert experience and the observational data. For complex systems, extraction of fuzzy rules based on human expert knowledge may suffer from a loss of accuracy. Thus, the research on the method for fuzzy rules’extraction based on sample data is of great importance. A novel method for extracting fuzzy rules based on sample data is proposed in this dissertation. Firstly, the input space is divided by means of the possibilistic fuzzy C-means algorithm. The fixed size least squares support vector machine is then used to determine the rear functions of each fuzzy rule. Combined with the membership values of input variables, the output of the fuzzy controller can be obtained by taking the weighted summation of each rear function.The main contents and innovations of this dissertation are as follows:(1) The antecedent part of fuzzy rules are usually constructed by fuzzy clustering algorithms which are sensitive to noise. To solve this problem, the possibilistic fuzzy C-means clustering algorithm is used to divide the input space. By combining the membership value with the typical value, the possibilistic fuzzy C-means clustering algorithm can reduce the influence of the noise data and get the best cluster centers.(2) The fixed size least squares support vector machine is proposed to get the rear functions which can be obtained from the sample data. Compared with the support vector machine, the proposed method has better computational efficiency. The model obtained by our method is of good sparsity and generalization ability.(3) The experiment is constructed with a real dataset produced by a reheating furnace in a steel rolling mill which is taken as the research object. The experiment’s results show that the proposed method can extract the fuzzy rules of the temperature fuzzy controller effectively.