.png){kind=icon}
.png){kind=icon}
Return
A data-driven intelligent predictive maintenance decision framework for mechanical systems integrating transformer and kernel density estimation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.cie.2025.110868.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Mechanical systems are crucial for the safety of complex equipment. As they often operate in harsh environments, the patterns of their failures are becoming increasingly difficult to grasp. Rich condition data generated by sensor monitoring robustly supports the prediction of remaining useful life and maintenance decision-making of mechanical systems. However, current research typically investigates the prediction of remaining useful life and maintenance decision-making independently, which impedes the effective application of condition data. A novel predictive maintenance decision-making method based on deep learning is proposed. This method trains the Transformer network with historical data to determine the system's degradation category, and the probability distribution of remaining useful life is dynamically obtained using the Kernel Density Estimation method. Based on this, a comprehensive optimization objective that considers maintenance cost rate, availability, and reliability is introduced to determine the optimal maintenance time. The optimal timing for spare parts ordering is determined with the goal of minimizing costs. Therefore, this study has realized a complete framework from condition monitoring data acquisition, through real-time remaining useful life prediction, to maintenance and spare parts ordering decisions. The proposed strategy is validated through case studies of bearings and gearboxes. It has achieved higher classification accuracy and constructed high-quality prediction intervals. Compared with periodic maintenance strategies, the dynamic predictive maintenance strategy can improve the availability of bearings by 11.1%, reliability by 15%, and save 55% of maintenance costs; it can increase the availability of gearboxes by 1.2%, reliability by 67%, and save 79% of maintenance costs.
Keywords:
Predictive maintenance
Transformer
Remaining useful life
Failure probability density
Spare parts ordering
Journal
IF:
6.5
Papers:
1.0W
Citations:
3.8W