小波包和模糊熵特征融合的轴承故障诊断

Bearing Fault Diagnosis Based on Feature Fusion of Wavelet Packet and Fuzzy Entropy

进行轴承多种类型裂纹故障诊断时,为解决单一特征量诊断效率低的问题,提出了基于信号小波包分解的精细时频域分析和模糊熵的特征融合方法。首先对轴承振动信号进行小波包4层分解重构,确定小波包系数模糊熵和频带能量,精细提取振动信号的高低频故障信息特征;然后基于权重指标对模糊熵和频带能量进行融合,构造多种故障状态下轴承信号的特征向量;最后选择适合小样本分类的支持向量机对轴承裂纹故障进行诊断。试验数据处理结果表明,轴承不同裂纹故障状态下,融合特征的方法诊断效率更高,相较于单一特征量识别准确率提高5.0%以上,对10种裂纹故障诊断正确率达到98.0%。

The efficiency of fault diagnosis is low by single feature when multiple crack faults occur in bearings.In order to solve the problem,a feature fusion method is proposed.The fusion feature is composed of frequency band energy and fuzzy entropy based on signal wavelet packet decomposition.Firstly,the signal was decomposed and reconstructed into 4 layers by wavelet packet.The frequency band energy and fuzzy entropy of coefficient were calculated.The fault information was finely extracted from vibration signal at high and low frequency band.Secondly,both extracted features were fused to construct the feature vector of the signal based on the weight.Finally,support vector machine was used to diagnose the bearing fault state.The experimental results proved that the method has higher efficiency than single feature under different bearing conditions and different crack fault states.The method of fusion feature has at least 5.0%higher accuracy than single feature,and the diagnosis accuracy rate reaches 98.0%of 10 kinds of crack fault.