The Extraction Method of Fuzzy Rules in the Electromagnetic Design of Large-Scale Hydraulic Generator

Abstract

The paper proposed a new fuzzy rule extraction method from the electromagnetic design of Large-scale hydraulic generator with the characteristics of small sample, high dimension and fuzzy. By the adjacency matrix, the related electromagnetic design parameters are extracted as the rules background. The unsupervised classification algorithm based on the Gauss mixture model is adopted to extract Gauss mixture composition from the distribution spaces of the parameters to build the fuzzy interval set of parameters. It solves the fuzzy interval division problem by the existing rule extraction methods and reduces the complexity of interval division in high dimension sample. According to the fuzzy division results, the electromagnetic design samples are transformed into the rule set. Considering the relative matching degree of rules combining the membership distribution and number advantages of each rule, the paper uses the relative matching degree as the explanation degree to evaluate rules. The rules with high explanation degree are finally adopted as the effective electromagnetic design rules.