Remaining useful life prediction for machinery using multimodal interactive attention spatial–temporal networks with deep ensembles

Abstract

Remaining useful life (RUL) prediction is essential for the health management of rotating machinery systems (RMSs). Because of the complex and variable service conditions of RMSs, degraded data over the entire service cycle frequently feature strong noise and data imbalance, which increase the difficulty and instability of RUL prediction. Therefore, an RUL prediction method based on a multimodal interactive attention spatial–temporal network (MIASTN) with a deep ensemble is proposed to improve the reliability and generalizability of intelligent models. First, the MIASTN is constructed as a deep base learner (DBL), and multiple DBLs are integrated to construct a deep ensemble prediction system. Second, a bidirectional multiscale degradation indicator space is constructed using signal processing decomposition theory to transform the original vibration data into a more interpretable form to improve model interpretability. Finally, a learning method ensemble strategy is employed to achieve the final decision using a DBL as a deep integrator. The proposed RUL prediction method is validated through two case studies. The experimental analysis results show that the proposed method offers significant advantages.