Research On Attention Mechanism Based Remaining Useful Life Prognostics Of Industrial Products

Since the 21st century,Chinese aerospace industry has developed rapidly.The safety of spacecraft and plane has received more and more attention.In order to ensure flight safety and monitor equipment’s running state,many researchers start to predict industrial equipment’s remaining useful life(RUL).The current mainstream methods for predicting RUL include model-based methods,data-driven methods and hybrid methods.In recent years,deep learning has been used as a data driven methods to predict the RUL of industrial equipment.In a real production environment,the method of predicting the remaining service life of industrial products based on the attention mechanism collects the status data and usage behavior data of different sensors of the equipment,and extracts high-level features through different neural network structures,thereby discovering different sensors of complex equipment.The information implicit in the data provides an end-to-end high-accuracy remaining service life prediction capability.This paper proposes an attention mechanism based method for predicting the RUL of industrial products.To extract spatiotemporal data of different scales,this method imports a multi-layer convolutional neural network,and abstract the RUL prediction problem into a time series problem.Our method could catch time series information and predict whether the equipment will fail within a fixed time.In addition,this paper also introduces the attention mechanism into the RUL prediction problem for the first time,which could discover more relevant information in the high-level features.This paper compares our method’s performance on the C-MAPSS data set with several baselines.In order to verify the benefits of different structures,this article also sets up multiple experiments to verify the effects of different changes.The final results show that our method can effectively improve the prediction effect of the remaining useful life.