Research On RUL Prediction Method Of A Rolling Bearing Based On Improved SAE And Bi-LSTM

As important components of rotating machinery equipment,the working conditions of rolling bearings often have a direct influence on the performance of the entire equipment.If the remaining useful life(RUL)can be accurately predicted before the bearing fails,certain measures can be taken in time to avoid causing major economic losses and even casualties.The data-driven method is one of the current mainstream methods for RUL prediction of rolling bearings.Among them,feature extraction and RUL prediction are two key steps of the data-driven method.The paper takes realizing RUL prediction of a rolling bearing as target,and conducting research from feature extraction and RUL prediction progressively.In terms of feature extraction,for the sigmoid activation function of sparse auto encoder(SAE)is easy to cause the gradient to disappear,a new Tan function is used to replace the original sigmoid function.In SAE,for the limitations in regression prediction when KL divergence is used for sparseness constraints,KL divergence is replaced with the dropout mechanism to achieve network sparsity.And the improved SAE is used to perform unsupervised adaptive deep feature extraction for the vibration signals of rolling bearings,without designing labels manually for supervised fine adjustment,and deep features with certain trends that can characterize the trend of bearing degradation are obtained.In terms of RUL prediction,for the long short-term memory(LSTM)generally only the past information is considered and the future information is ignored in processing sequences chronologically,the bi-directional long short-term memory(Bi-LSTM)is proposed as the prediction model in this paper.In addition,in order to obtain better prediction results,the forward propagation training process and the back propagation training process of the Bi-LSTM model are optimized respectively.Using two bearing data sets,experimental results both show that the proposed RUL prediction method of a rolling bearing based on improved SAE and Bi-LSTM can improve the convergence speed of the model and has lower prediction error.