结合CNN与TCN神经网络的滚动轴承寿命预测

Rolling Bearing Life Prediction Combined with CNN and TCN Neural Network

针对滚动轴承振动信号特征提取不充分、过于依赖人工特征提取及预测精度低等问题,提出了CNN-TCN-Attention网络模型预测方法。该方法选取滚动轴承振动信号作为输入通过增强顶帽算子(EAVGH)对信号进行特征增强,运用卷积神经网络(CNN)来提取信号中的深层特征,并构建TCN-Attention模型对滚动轴承剩余寿命进行预测。将注意力机制与时间卷积网络相结合可以有效的提高模型预测精度,通过轴承寿命实验数据进行验证,CNN-TCN-Attention预测模型能有效的提取滚动轴承振动信号中的深层特征,并且具有较高的预测精度。

Aiming at the problems of insufficient feature extraction of rolling bearing vibration signals,over-reliance on manual feature extraction and low prediction accuracy.A prediction method based on CNN-TCN-Attention network model is proposed.In this method,the vibration signals of rolling bearings were selected as the input to enhance the features of the signals by EAVGH,and the deep features of the signals were extracted by using the convolutional neural network(CNN),and the TCN-Attention model was built to predict the remaining life of rolling bearings.Combining the Attention mechanism with the time convolutional network can effectively improve the prediction accuracy of the model.Through the bearing life experimental data verification,the CNN-TCN-ATTENTION prediction model can effectively extract the deep features in the vibration signals of rolling bearings,and has a high prediction accuracy.