小波包和GA-SVM在轴承故障诊断中的应用

Application of Wavelet Packet and GA-SVM in Bearing Fault Diagnosis

摘要:

为了解决傅里叶变换难以兼顾信号在时域和频域中的全貌和局部化特征以及支持向量机惩罚参数c和核函数参数g选取的问题,提出了基于小波包和GA-SVM的轴承故障诊断方法;首先通过实验采集多种工况下故障轴承和正常轴承的振动信号,从振动信号中提取能够表征轴承运行状态的时频域特征以及基于小波包分析的特征向量来作为GA-SVM的输入,然后在SVM的基础上,针对SVM的惩罚参数和核函数参数在不同应用场景下的取值难以确定的特性,采用了遗传算法对支持向量机进行参数优化的GA-SVM算法进行模式识别;实验结果显示,基于小波包和GA-SVM的轴承故障诊断方法比SVM和BP都具有更高的识别精度。

Abstract:

In order to solve the problem that Fourier transform cannot detect both full view and local feature in time domain and the problem of the action of penalty parameter and kernel function parameter of support vector machine selection.put forward a method of fault diagnosis of bearing based on wavelet packet and GA-SVM.First,Collect vibration signal of fault bearing and normal under various conditions and extract the characteristics which are typical in time domain and wavelet packet domain as the input of GA-SVM input.Then,on the basis of SVM,using the genetic algorithm to optimize the parameter of GA-SVM as the parameters of SVM are hard to put forward..The experimental results show that the fault diagnosis method based on wavelet packet and GA-SVM has higher recognition accuracy than SVM and BP.