基于小波包和梯度提升决策树的轴承故障诊断

Bearing fault diagnosis based on Wavelet Packet and Gradient Boosting Decision Tree

针对轴承信号特征难提取,故障诊断准确率不高的问题,提出了一种利用小波包分解和梯度提升决策树(GBDT)进行轴承故障诊断的方法.首先采用小波包分解方法对轴承振动信号进行分解得到小波包系数,再计算每一频带的小波包能量作为轴承故障特征,构成轴承故障特征数据集,最后将故障特征数据集输入梯度提升决策树分类模型进行故障诊断.使用凯斯西储大学轴承测试数据对该方法进行验证,结果表明轴承故障诊断准确率达99.26%,具有良好的轴承故障诊断能力.

In view of the difficulty in extracting the fault signal characteristics of bearings and the low fault diagnosis accuracy,this paper presents a bearing fault diagnosis method combining Wavelet Packet Decomposition and Gradient Boosting Decision Tree(GBDT).Firstly,wavelet packet coefficients are obtained by wavelet packet decomposition of the original vibration signal.Then,the energy of wavelet packet in each frequency band is calculated as the fault characteristic of bearing.Finally,the fault features are input into the Gradient Boosting Decision Tree model for training and testing.The bearing test data from Case Western Reserve University be used to verify the method.The result shows that the accuracy of this method is 99.26%and it has a good bearing fault diagnosis ability.