Physics-Informed Neural Networks for RUL Prediction

With the complexity and intelligence of industrial equipment continue to increase, Prognostics and Health Management (PHM) of such equipment have become increasingly important, with the accurate prediction of Remaining Useful Life (RUL) emerging as a critical industry need. This paper proposes an RUL prediction method that integrates the Attention-Mamba network with a Physics-Informed Neural Network (PINN) framework. The Mamba network, as an efficient sequence modeling structure, is capable of capturing temporal dependencies. By incorporating the Attention mechanism, the Mamba network can effectively identify key features and changing trends in the operational state of the equipment, highlighting these features through weight allocation to enhance prediction accuracy. Furthermore, we introduce the PINN framework into the model to further improve its predictive performance. We validate the superiority of this method using the C-MAPSS dataset. The proposed approach enhances the accuracy of RUL predictions and provides robust support for intelligent maintenance and reliability management of industrial equipment.