随机服役环境下阈值时变设备剩余寿命预测与维修库存决策

Predictive maintenance is an important means to reduce failure occurrence frequencies and realize the safe and reliable operation of a device, and therefore it falls into the forward and hotspot issue of fault prognostics and health management. The predictive maintenance of a device involves two core problems: residual life prediction and inventory management. However, the existing research for the former lacks the consideration of stochastic serving environment and time-varying threshold, while the latter is usually carried out separately with the maintenance scheduling issue. This may not only result in huge waste of resources, but also increase the operating risk of the device. This program is investigated under the background of the inertial platform of the strategic missile. The main research points of the program include: ① Considering the stochastic serving environment and failure determination rule of the inertial platform, we study a kind of degradation modeling and residual life prediction method associated with the device subject to time-varying threshold and randomly occurring nonlinearities; ② Based on the residual life prediction information obtained in ①, the predictive maintenance and inventory management strategy are jointly determined. The solving of the above issues can not only enrich the life prediction and health management theory but also improve the current logistics support mechanism of the strategic missiles, and is of significance in both theory and military applications.

预测维修是减低故障发生频率，实现设备安全可靠运行的重要手段，是故障预测与健康管理领域的前沿和热点问题。设备的预测维修涉及两个核心问题：剩余寿命预测和库存管理。目前，前者的研究缺乏对随机服役环境和时变阈值的考量，而后者又通常与维修安排问题分开研究，这在实际中将会造成巨大的资源浪费，大大增加设备的运行风险。本项目以战略导弹惯性平台为背景，研究内容主要包括：①.考虑惯性平台的随机服役环境和失效判定准则，研究一类存在随机发生非线性退化的阈值时变设备退化建模与剩余寿命预测方法；②.基于①中得到的剩余寿命预测信息实现设备预测维修与库存管理联合决策。相关问题的解决对于丰富寿命预测和健康管理理论，优化战略导弹后勤保障体制具有重要的理论意义和军事价值。

预测与健康管理是降低故障发生率、提高设备可用性、削减维修保障成本的重要技术手段。项目以某型导弹惯性平台系统为背景，研究了考虑随机服役环境、非线性退化和时变阈值等情形下的设备寿命预测与健康管理问题。（1）寿命预测方面，考虑随机服役环境，提出一种带随机参数的非线性维纳过程模型，给出了退化模型参数离线估计在线更新的联合优化确定法，实现了设备剩余寿命实时预测。构建考虑时变阈值的设备性能退化模型，基于首达时间概念，推导出剩余寿命分布解析形式表达式，给出该类模型参数估计方法。成果发表于IEEE Trans.、MSSP等权威刊物。（2）健康管理方面，考虑维修活动对设备退化量和退化率双重影响，提出一种考虑不完美维修影响的性能退化建模与预测维修决策模型。针对具有两阶段退化模式的随机退化设备，考虑检测过程不完美性，提出一种基于状态监测的预防性更换策略。针对部分可观测设备，基于马尔可夫决策过程模型，提出一种考虑维修次数、备件库存等外界资源约束下的不完美维修策略。考虑维修决策和库存管理间耦合关系，提出一种基于寿命预测的预测维修与库存管理联合决策策略。成果发表于JRR、《自动化学报》等重要刊物。领域内知名学者对相关研究给予广泛关注和正面评述、引用。.项目实施过程中，出版专著2部（1部入选十三五国家重点出版物出版规划），发表论文14篇（含录用），所有论文均被SCI或EI收录。协助培养研究生5名，2人获中国自动化学会优秀博士学位论文提名奖、2人获陕西省优秀博士学位论文。申请国家发明专利5项，获军队科技进步二、三等奖各1项，中国自动化学会自然科学二等奖1项，火箭军优秀科技论文一等奖1项。项目负责人王兆强入选首批中国科协/中央军委军事科技领域青年人才托举工程、陕西省高校科协青年人才托举计划、大学321人才培养对象，评为大学青年科技之星、优秀教员；指导学生参加数学建模和军事建模竞赛，获军队特等奖2项、二等奖2项、三等奖4项，陕西省一等奖1项、二等奖2项。