结合SVM和改进证据理论的多信息融合故障诊断  

Multi-information fusion fault diagnosis using SVM & improved evidence theory

作　　者：向阳辉[1] 张干清[1] 庞佑霞[1] 

机构地区：[1]长沙学院机电工程系,长沙410003

出　　处：《振动与冲击》2015年第13期71-77,共7页Journal of Vibration and Shock

基　　金：国家自然科学基金(51475049);校人才引进科研基金(12004);湖南省'十二五'重点建设学科资助项目;湖南省教育厅资助科研项目(14C0094)

摘　　要：为了综合合理利用设备多个方面特征信息来提高故障诊断的准确性,提出一种结合支持向量机(Support vector machine,SVM)和改进证据理论的多信息融合故障诊断方法。该方法通过混淆矩阵获取各SVM局部诊断证据对各故障模式的可靠度,赋予不同的权重系数,并对由各SVM局部诊断硬输出判决矩阵构造出的基本概率分配进行加权处理,从而实现SVM和改进证据理论在多信息融合故障诊断中的有效结合。实验结果表明,各SVM局部诊断证据的加权融合处理能够显著降低各局部诊断间的冲突,所提方法可以有效提高故障诊断的准确率。In order to comprehensively and reasonably utilize much feature information of equipments to improve the accuracy of fault diagnosis,a method of multi-information fusion fault diagnosis was proposed based on support vector machine（SVM）and improved evidence theory.The reliability of local diagnosis evidence of each SVM for every failure mode was acquired with a confusion matrix to give different weight coefficient.The basic probability assignments constructed with a hard output decision matrix from the local diagnosis of each SVM were processed weightedly to realize the effective combination of SVM and improved evidence theory in multi-information fusion fault diagnosis. The experimental results showed that the weighted fusion treatment of local diagnosis evidences of all SVMs can significantly reduce the conflicts between local diagnoses;and the proposed method can effectively improve the accuracy of fault diagnosis.

关 键 词：支持向量机 证据理论 故障诊断 多信息融合 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置] TP391[自动化与计算机技术—控制科学与工程]