不完美维修情形下部分可观测系统的最优维修策略

针对监测费用昂贵导致难以实施连续监测和维修效果通常不完美的问题,研究了维修效果不完美情形下部分可观测系统的最优维修策略.在假设系统的性能退化状态变化规律可以用马尔可夫链进行描述的基础上,利用部分可观测马氏决策过程对该系统进行维修决策建模,并通过引入信息状态将其转化为完全可观测马氏决策过程;引入维修效果矩阵用以刻画维修效果对最优维修策略的影响;针对所建立的最优方程以单位时间期望维修费用最小为目标设计了最优维修决策算法.通过数值仿真验证了所提算法的有效性.

Optimal Maintenance Policies for Partially Observed Systems with Imperfect Maintenance

Considering the issues such as high costs of performing successive observation and imperfect effects of maintenance, the optimal maintenance policies for partially observed systems with imperfect maintenance were studied. Based on the assumption that the deterioration status evolved as a Markov chain, the maintenance was formulated as a partially observed Markov decision process which was then reformulated as a completely observed Markov decision process through choosing a proper state variable. The maintenance effect matrix was introduced to characterize the imperfect maintenance effect on the optimal policy, which was followed by developing the optimal maintenance decision algorithm to minimize the expected cost per unit time. A numerical example was presented to verify the effectiveness of the proposed algorithm.