滚动轴承时域新指标的WNN状态退化预测研究

Research of Wavelet Neural Network State Degradation Prediction of Rolling Bearing New Time Domain Index

针对传统时域指标在滚动轴承信号特征提取时状态预测精度不高的问题,首先,选取适合在线简单快速判别的时域指标,并根据轴承疲劳损伤大小和局部损伤数量增加,分析时域指标特征对状态变化的敏感性;其次,基于传统时域指标,寻求两个更为敏感的时域组合指标TALAF和THIKAT;最后,利用小波神经训练和测试两个新指标的数据样本,并与传统时域指标峭度及BP神经网络预测方法进行比较,仿真结果验证了TALAF和THIKAT指标,有效提高了轴承故障预测的准确性。

Aiming at lower accuracy of classification for signal feature extraction of rolling bearing,firstly,some time domain indexes for online simple rapid discrimination are selected. The sensitivity of time domain index of fault is analyzed based on size of bearing fatigue damage and number of local damage. Secondly,based on the traditional time domain index,two more sensitive time domain index ‘TALAF ＇and ‘THIKAT ＇is searched. Lastly,the data set including two new indicators are trained and tested based on wavelet neural network which has a good real-time. The training and testing results for the traditional time domain indexes kurtosis and BP neural network are compared with results of the data. The simulation results show that TALAF and THIKAT can effectively improve the accuracy of prediction index state bearing.