Classification of Motor Faults with a Fuzzy Support Vector Machine Based on Environmental Memberships
With the widely use of motor in industry product,it is significant to improve the motor fault classification technique.Since support vector machine (SVM) is very sensitive to outliers and noise in the training set,the fuzzy support vector machine (FSVM) algorithm has been accepted as a useful fault classification method.However,the traditional FSVM methods lack of the general criterion in defining the fuzzy membership.In this paper,a criterion of fuzzy membership based on the environment around samples is proposed.By this criterion,the fuzzy membership is defined by two factors: average distance and ratio of nearby similar samples to nearby heterogeneous samples.The ratio and difference between two average distances (similar and heterogeneous) are taken as criterion to point out weather the sample is outline or noise.Then with the result of judge,the fuzzy membership is figured out by different formula based on the average distance between simple and similar samples.In order to validate the proposed method,seven experimental cases including one normal motor and six fault motors were conducted on the machinery fault test device.Experimental results showed the classification accuracy of the current FSVM was up to 92%.
motor fuzzy support vector machine fuzzy membership fault classification
Yilong Liang Bing Li Wei Chen Zhengjia He
State Key Laboratory for Manufacturing Systems Engineering,Xian Jiaotong University,Xian 710049,China;Shening Qiao;Xian Shaangu Power Co.,LTD,Xian 710075,China
国际会议
南京
英文
1-4
2013-08-20（万方平台首次上网日期，不代表论文的发表时间）