基于优化支持向量机的轴承故障诊断方法研究

以滚动轴承在正常、内圈故障、外圈故障和滚动体故障四种工况下的振动信号为研究对象,采用小波包变换的方法提取信号的能量熵,构成振动信号的特征向量。在此基础上采用支持向量机进行故障模式识别,建立支持向量机模型需要选择适当的核函数及相关参数,使用径向基核函数,需要设置的参数为核函数的宽度和误差惩罚系数,分别结合传统的网格搜索,遗传算法,粒子群算法优化支持向量机参数以提升分类性能。试验结果表明,采用优化后的支持向量机进行故障诊断可以大大提高诊断精度。

Research on the fault diagnosis of rolling bearing based on optimized SVM

With vibration signal of rolling bearing under the four conditions of normal,inner ring fault,outer fault and roller fault as research object,by using wavelet packet transform method energy entropy of the signal is extracted to constitute the vibration signal feature vector.On the basis fault pattern recognition is made by support vector machine,which model is built by selecting the appropriate kernel function and related parameters.With RBF kernel function and width of the kernel function and error penalty coefficient as parameters to be set,combined with traditional grid search,genetic algorithm and particle swarm algorithm respectively to optimize the parameters of SVM to improve classification performance.Test results show that,the optimized SVM for fault diagnosis can greatly improve the accuracy of diagnosis.