基于信息融合的滚动轴承故障诊断

Fault Diagnosis for Rolling Bearings Based on Information Fusion

针对传统的单一传感器检测准确率不高,诊断系统不稳定等问题,将振动和声发射2种检测方法进行融合。首先对采集到的2种信号进行小波降噪及Hilbert解调,得到故障信号的频域包络谱,计算其频段能量值并组成特征向量;然后利用BP神经网络建立多传感器的信息融合系统,选取合适的样本输入网络进行训练,直至达到所要求的误差范围;最后实现对样本轴承的故障诊断,达到了相对较高的诊断正确率。

Aiming at problems about low detection accuracy and unstable diagnosis system,the vibration and acoustic emission detecting methods are fused.Firstly,the wavelet denoising and Hilbert demodulation are carried out for two signals,the envelope spectrum in frequency domain of fault signals are obtained,and the energy value for frequency band is calculated to compose feature vector.Then the BP neural network is used to establish information fusion system of multi -sensor,the appropriate sample is selected to input network for training until required error range is reached. Finally,the fault diagnosis is realized for sample bearings,and a relatively high diagnostic accuracy is reached.