ABSTRACT Reliable and effective aero‐engine remaining useful life (RUL) prediction plays a vital part in predictive maintenance strategy. However, the inherent complexity of the aero‐engine operating environment makes it difficult to achieve effective maintenance with traditional point estimation methods. To enhance the reliability of the maintenance strategy, this study proposes a novel method for RUL prediction interval (PI), which quantifies the uncertainty in RUL forecasts and integrates the resulting RUL interval predictions into predictive maintenance. First, RUL PI is estimated using a hybrid deep learning model combined with Monte Carlo dropout. The predicted RUL PI is then transformed into a probability distribution to facilitate real‐time monitoring of the aero‐engine's operational status. Meanwhile, an effective maintenance threshold is introduced to ensure timely aero‐engine maintenance. In addition, the constructed RUL distribution is linked to the predictive maintenance strategy by considering the maintenance cost rate (MCR). Optimizing the MCR for the maintenance economy requirement determines the best maintenance strategy. Finally, the proposed method is demonstrated using NASA's aero‐engine C‐MAPSS dataset. The results show that the optimal maintenance strategy not only significantly reduces maintenance costs but also leads to system operational reliability of more than 90%.
Zhu et al. (Tue,) studied this question.