期刊名称：International Journal of Hybrid Information Technology
印刷版ISSN：1738-9968
出版年度：2016
卷号：9
期号：8
页码：233-242
出版社：SERSC
摘要：Binary-tree support vector machine (SVM) has such advantages as small repeated operation workload, fast classification speed and dead zone inexistence, but the structural design can influence the classification accuracy thereof. In order to rationally design the structure of the binary-tress SVM, a multi-classification algorithm (AHP-BSVM) combining analytic hierarchy process (AHP) and binary-tree SVM is proposed in this paper. Firstly, the analytic hierarchy process is adopted to establish the evaluation system model so as to comprehensively evaluate multiple influencing factors for determining the weight values of various faults; then, the faults are ordered by the weight values and the structure of the binary-tree SVM is determined according to the fault sequence; finally, the proposed algorithm is adopted for fault diagnosis and analysis. The simulation experiment shows: compared with other algorithms, the proposed algorithm has higher recognition accuracy and higher classification accuracy, and is applicable to multi-classification, thus having good promotion prospect.
关键词：Binary-tree; Support vector machine; Analytic hierarchy process; Bearing ;fault diagnosis
