基于小波和信息粒化的BP神经网络的轴承故障诊断

The BP Neural Network Fault Diagnosis of Bearings Based on Wavelet and Information Granulation

通过对圆锥滚子轴承轴向故障振动信号的预处理,得到响应的特征,从而利用BP神经网络进行故障诊断。首先利用一种新的小波消噪算法对监测信号进行预处理,该算法是基于最佳正交小波基的选择,使熵在小波收缩过程中的作用最小;文章重点在于利用模糊信息粒化对消噪后信号进行模糊粒化,从而更好的特征提取;最后将特征向量作为输入,运用BP神经网络进行故障诊断。通过实验故障信号验证了,消噪后的信号能更好地进行特征提取;同时,模糊粒化后能更准确的进行故障诊断。

Through using the tapered roller-bearing axial fault vibration signal,we get the response characteristics and take advantage of BP neural network for fault diagnosis.Firstly,a new wavelet-denoising algorithm is used to pre-monitor signals,which is based on the best selection of orthogonal wavelet bases,aiming at the minimum entropy effect in the process of Wavelet Shrinkage.Then use the fuzzy information granulation to granulate the noised signal;followed signal of fuzzy granular layer wavelet packet decomposition and reconstruction,by no fault and fault feature vectors.Finally,take the feature vectors as input and use BP neural network for fault diagnosis.Fault signal is verified by experiments and the signal after denoising is better for feature extraction;meanwhile,the signal after fuzzy granulation can be used for more accurate fault diagnosis.