Markov Decision Process Based Multi-agent System Applied to Aeroengine Maintenance Policy Optimization
Jianrong Wang, Shouming Hou, Yingying Su, Jianwei Du, Wanshan Wang · 2008
On the analysis of failure and deterioration model of aeroengine, Markov decision process (MDP) based multi-agent framework was introduced to overcome the obstacles of conventional maintenance policy optimization of aeroengine. In the design phase, maintenance ontology was defined to smooth the communication among agents. Then the system architecture and agent interaction mechanism were described in detail. For the implementation of MDP, the MDP-based deterioration model and transition probabilities were identified on the analysis of constraint conditions associated with a deterioration process. Based on these, a MDP-based multi-agent prototype system for aeroengine maintenance policy optimization was developed on the JADE (Java Agent DEvelopment Framework) platform. The process of system implementation and a case example of MDP applied to optimize the maintenance plan were depicted in detail. The maintenance optimization results under various objects and were analyzed, which showed that the model realizes the quantitative analysis in maintenance decision and provides a decision support tool for decision makers.