Remaining Useful Life Prediction Of Rolling Bearings Based On Improved Long And Short-Term Memory And Residual Networks

In industrial production,bearings as the core components of rotating machinery,widely used in transportation,machining,aerospace,instrumentation and many other fields.In the process of use,the reliability of bearings is very important to ensure normal production and the safety of people’s lives and property.The remaining life prediction of bearings can predict the time of possible damage at an early stage so that timely maintenance can be carried out,therefore,the remaining life prediction of bearings is of great importance.In this paper,we take rolling bearings as the research object and use deep learning methods to predict the remaining life of bearings,the main work and research content are as follows.Firstly,in order to solve the long-term dependence problem of the long and short-term memory network and at the same time improve the prediction accuracy,a bearing remaining life prediction model based on the combination of the long and short-term memory network and the Transformer network is proposed.The model uses the long and short-term memory network to extract the degradation information in the time-frequency domain indicators in the vibration signal,and then the Transformer network further processes and predicts the degradation trend.Since the Transformer network can integrate the correlation between the before and after data,it makes up for the shortcomings of the LSTM network to a certain extent.The calculation results show that the model has better prediction accuracy than several commonly used remaining life prediction models.Secondly,in order to further improve the prediction performance,a bearing remaining life prediction method based on wavelet threshold denoising,random forest feature extraction and ordered neuron long and short-term memory network is proposed.The wavelet threshold denoising algorithm can reduce the interference of useless noise,the random forest feature selection algorithm can select features that better characterize the degradation trend,and the ordered neuron long and short-term memory network adds a hierarchical structure to the long and short-term memory network,so that important information at the higher levels can be more easily preserved,which is conducive to improving the model performance.The experimental results show that the model has higher prediction accuracy and computational efficiency than several commonly used prediction models.Finally,in order to improve the prediction accuracy while reducing manual intervention,a method for predicting the remaining life of rolling bearings based on residual networks and ordered neuronal long and short-term memory networks is proposed.The use of residual networks for feature extraction eliminates the need for manual feature extraction while allowing for the mining of deeper degradation information,and then the ordered neuron long and shortterm memory networks are used to further process the degradation information output from the residual networks,with a hierarchical structure that is more conducive to the extraction of useful information.Experiments on two publicly available bearing datasets show that,compared with a variety of commonly used methods,the model not only has a higher prediction performance,but also has a strong generalization property.