.png){kind=icon}
Return
A system-centred predictive maintenance re-optimization method based on multi-agent deep reinforcement learning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2025.127034.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Predictive maintenance (PdM) is a proactive maintenance strategy that utilises data monitoring and analysis to forecast potential failures and perform maintenance activity in advance, which can significantly improve system availability and reduce maintenance costs. However, challenges such as the discrepancy amongst components and the complexity of the multi-component system are difficult to address through current PdM methods. A system-centred PdM re-optimization method based on multi-agent deep reinforcement learning (DRL) is proposed. The existing challenges can be effectively addressed through phased decision-making and collaborative learning amongst multiple agents. Considering the nonlinear characteristics and uncertainty in the degradation process of the engineering system, a synergistic data-model framework is implemented to predict the cumulative degradation value and remaining useful life (RUL). The predicted values of cumulative degradation and RUL serve as multiple environmental observations to balance the environmental dynamics in multi-agent DRL. Considering the differences amongst components, a component-centred PdM decision is initially made according to the optimal component performance and maintenance cost. Given the complexity of the multi-component system, a system-centred PdM decision is re-optimized on the basis of the component-level decision results and the system performance constraint. The two-stage PdM re-optimization method is implemented for a subsea production control system to demonstrate the application of the proposed method.
Keywords:
Predictive maintenance
Multi-agent
Deep reinforcement learning
Subsea production control system
Journal
IF:
7.5
Papers:
3.0W
Citations:
10.2W
Organization
Cited Papers
A reinforcement learning framework for optimal operation and maintenance of power grids.png)
APPLIED ENERGY
OAAI
APPLIED ENERGYIF11
OAAIShareSaveShareSaveShareSaveDeep learning models for predictive maintenance: a survey, comparison, challenges and prospectsAPPLIED INTELLIGENCE
APPLIED INTELLIGENCEIF3.5
PREAI
ShareSave
View more