Bearing residual life prediction method adopting feature screening and loss function improvement

The invention discloses a bearing residual life prediction method adopting feature screening and a loss function improvement, and belongs to the field of traction motor bearing maintenance. The method comprises the following steps: firstly, extracting time domain and frequency domain features from a bearing vibration signal, calculating to obtain a comprehensive feature screening index, and screening out features capable of reflecting a bearing degradation trend according to the comprehensive feature screening index; secondly, training an encoder-decoder model based on the attention mechanism by using the improved loss function and the screened features; and finally, fitting a prediction result by using a linear function to obtain an RUL prediction value of the rolling bearing. Experimental results show that compared with other literature methods and comparison methods, the method provided by the invention shows better performance on a PHM2012 data set. 本发明一种采用特征筛选与改进损失函数的轴承剩余寿命预测方法，属于牵引电机轴承的维护领域；其步骤如下：首先，从轴承