基于并行多通道卷积长短时记忆网络的轴承寿命预测方法

Bearing Life Prediction Method Based on PMCCNN-LSTM

在预测轴承剩余使用寿命时,数据间的时序特性是一个可以利用的重要隐藏信息。为了更好地提取具有时序信息的特征用于预测,提出了一种基于并行多通道卷积长短时记忆网络(PMCCNN-LSTM)的剩余使用寿命预测模型。该模型主要由两部分组成:前端为并行多通道卷积网络(PMCCNN),提取信号特征,挖掘数据的时序特性,并采用逐层训练和微调的方式提升参数的收敛性;后端为长短时记忆(LSTM)网络,基于特征进行剩余使用寿命预测,并采用加权平均的方法对预测结果进行平滑处理。在一个轴承加速寿命实验的公开数据集上使用留一法验证了该模型的准确性,实验结果表明:所提模型的平均误差与最大误差分别比传统的卷积神经网络(CNN)低23.38%和15.84%,比传统的LSTM低24.14%和19.01%,比卷积长短时记忆网络(CNN-LSTM)低30.32%和23.09%。

Timing characteristics between data was an important hidden information that might be utilized while predicting the remaining life of the bearings.In order to better extract features with timing informations for prediction,a remaining useful life prediction model was proposed based on PMCCNN-LSTM.The model was consist of two parts.The front end was PMCCNN,which extracted the signal features,mined for the timing characteristics of the data,and used layer-by-layer training and fine-tuning to improve the convergence of the parameters.The back end was an LSTM network with remaining useful life prediction based on features,and the weighted average method was used to smooth the prediction results.The accuracy of the model was verified by using leave-one-out method on a public dataset of bearing accelerated life tests.The experimental results show that the mean errors and maximum errors of the proposed model are 23.38%and 15.84%lower than that of CNN,24.14%and 19.01%lower than LSTM,30.32%and 23.09%lower than that of CNN-LSTM respectively.