摘要
针对轴承故障检测问题,提出一种基于相关向量机(RVM)的故障检测方法。RVM算法基于贝叶斯估计理论,它产生的决策函数具有少数的相关向量,利用RVM算法松散特性,解决了支持向量机算法(SVM)计算复杂度高的不足。为进一步降低检测时间,以重构相空间投影系数为轴承故障特征。试验最后同传统的SVM算法进行了比较,结果表明所建议的方法在保持较高检测率的同时,提高了故障检测的时效性。尤其检测时间从0.67 s降低了0.005 9 s(100倍)。因此,该方法非常适合于在线故障检测等实时性要求很高的领域。
A new bearing fault detection approach based on relevance vector machine(RVM) is presented.The RVM is based on Bayesian estimation theory,its distinctive feature is that it can yield a sparse decision function that is defined by only a very small number of so-called relevance vectors.By exploring this sparse property of the RVM,the disadvantage of the support vector machine(SVM),i.e.,its computational complexity is solved.To raise the computation speed further,the projected coefficients in a reconstructed phase space modeled by normal training samples are used as bearing fault detection features.The proposed method is compared with a well-tested SVM classifier.The results illustrate that the proposed RVM approach could greatly reduce the computational complexity of the SVM while maintain its best detection accuracy.In particular,the RVM approach could reduce the detection time from 0.67 s for SVM to 0.0059s(nearly 100-fold reduction).Thus,the proposed RVM classifier is more advantageous for real-time processing fault detection.
出处
《振动与冲击》
EI
CSCD
北大核心
2008年第10期6-9,187,共4页
Journal of Vibration and Shock
关键词
故障检测
相关向量机
相空间重构
支持向量机
fault detection
relevance vector machine(RVM)
phase space reconstruction
support vector machine(SVM)
