Research Status and Challenges of Deep Learning-Based Remaining Useful Life Prediction of Equipment

Abstract

With the development of modern science and technology, aviation, aerospace, satellite and other equipment are developing in the direction of high reliability, safety and stability, which puts forward higher requirements for the performance of components. In order to ensure the normal operation of complex equipment, the remaining useful life (RUL) prediction technology has been widely concerned by researchers. Deep learning emerging in recent years has powerful data processing capabilities and feature expression capabilities, realizing autonomous learning of model parameters and providing accurate RUL prediction results. In view of this, four typical deep learning methods applied to RUL prediction are analyzed and elaborated in detail, and the research status of each method is combed. Then the corresponding advantages and disadvantages are sorted out through experiments. Finally, the future research directions of deep learning-based RUL prediction are discussed.