基于网格式支持向量机算法的转轴裂纹故障诊断  

Shaft crack fault diagnosis based on grid support vector machines

作　　者：袁胜发[1] 李秀琼[1] 

机构地区：[1]华中农业大学工程技术学院,武汉430070

出　　处：《振动与冲击》2009年第9期155-158,共4页Journal of Vibration and Shock

基　　金：教育部新教师基金项目(课题编号:200805041079)

摘　　要：支持向量机(SVM)是一种对小样本决策具有良好学习性能的机器学习方法。常规SVM算法是从二类分类问题推导得出的,针对于故障诊断这种典型的多类决策问题,研究了一种网格式支持向量机多类算法,每个类别和其他2至4个类别之间采用常规SVM二值分类器进行分类,所需二值分类器总数少,可扩展性强。把转轴上不同位置的裂纹当作不同的故障,运用网格式支持向量机进行转轴裂纹位置故障诊断,结果表明该算法具有计算量小、诊断速度快、故障识别率高、容易扩展等优点,适合于较大规模的多类别故障诊断应用。Support vector machines is a general machine-learning tool that exhibits good generalization when fault samples are few. Since basic support vector machines are originally designed for two-class classification,a new multi-class classification algorithm named grid support vector machines was presented to solve the pattern recognition problem in a fault diagnosis as a typical multi-class classification case. With this rid support vector machines,every class constructed two-class SVM classifiers with less than 4 other classes,and the total number of two-class SVM classifiers is less. The rid support vector machines is simpler and more extensible compared with other methods of multi-class support vector machines. The cracks in different positions of a shaft were regarded as different classes of fault,and were diagnosed by the grid support vector machines. The results showed the new method distinctly improved the fault recognition accuracy and the diagnosis speed,and it was more suitable for practical application of multi-class fault diagnosis.

关 键 词：故障诊断 支持向量机 多类算法 网格式 

分 类 号：TH17[机械工程—机械制造及自动化] TP18[自动化与计算机技术—控制理论与控制工程]