基于安全风险的长贮系统预测维修动态决策技术研究

Predictive maintenance quantitatively analyzes the deterioration process and predicts the future trend of the equipment’s condition based on the data from on-line condition monitoring. Then the maintenance strategy is optimized in terms of the given aim and residual lifetime prediction from above. Adoption of Predictive maintenance can result in substantial cost savings and higher system reliability, which is the integrity of prediction and decision analysis. However, when it concerns with the long-storage system such as missile which has higher requirement on safety, predictive maintenance should be combined with risk evaluation to get the optimal maintenance policies which simultaneously satisfy the requirement of economy and safety. The proposed research proposal aims to investigate the characteristic of long-storage system in application field with higher safety requirements and solve the maintenance decision problem with the combination of real-time condition predictive, risk evaluation and maintenance decision. Firstly, the degradation process of equipment will be proposed by considering the maintenance effect，which is studied based on data from on-line condition monitoring or off-line inspection.. Then the relationship among the deterioration process of equipment and the risk level of system is investigated by analyzing the structure and function of system. Simultaneously the quantitative evaluation model of system risk will be built up to determine the system risk level and unit relative importance. The influence of risk and equipment condition will be analyzed by referring to the two kinds of model given above and the configuration of system. Furthermore, comparing the risk from predictive model and the acceptance level from a certain criterion, the dynamic maintenance decision model will be investigated with several constraints. This research will improve the application of maintenance model in high risk field, promote theoretical research, and provide the technical support for the control, decision and maintenance management of complex engineering system.

预测维修以现场获取的状态信息为基础，定量分析当前系统的退化过程并预测其未来的变化趋势与剩余寿命, 进而按照指定的目标进行维修策略优化，是预测与决策的联合研究，但为了满足导弹等长贮系统的安全性要求，需要引入风险约束，将状态预测、风险与维修决策三方面研究结合，制定达到系统安全性和经济性统一的最佳维修策略。本项目针对导弹等长贮系统的特点，首先从现场获取的在线监测或离线检测状态数据出发，建立考虑维修效果的状态实时预测模型。然后结合系统组成结构和功能交互关系，研究系统风险水平的量化方法。据此确定当前预测结果下的系统风险水平及部件重要度，在此基础上与一定准则下的风险可接受值比较，研究建立基于风险的系统预测维修动态决策建模和求解方法，为复杂工程系统的控制、决策、管理一体化提供重要的技术支持。

导弹等安全要求极高的复杂长贮系统，由于环境等因素的作用，贮存期间性能退化引发故障，造成严重的安全事故。传统以定时检修和任务前突击检修为主的方法，无法根据状态变化动态调整，难以保证系统的安全贮存和有效使用。为此，项目以现场状态监测或检测数据为基础，分析单机状态、维修过程以及系统风险之间的相互关系，研究基于安全风险的长贮系统预测维修动态决策技术，主要解决以下问题：.（1）围绕长贮系统的安全性要求及贮存期特点，改进现有基于风险的维修决策框架RBMI，建立基于安全风险的长贮系统预测维修动态决策分析框架，为实际工作中长贮系统的维修策略制定提供整体流程与框架参考。.（2）提出基于事故链的系统安全风险分析方法，综合运用主逻辑图、动态故障树及证据网络等理论，确定影响长贮系统安全的关键单机，结合系统风险的概率重要度与单机维修特征，建立基于故障回溯的迭代风险分析方法，确定综合重要度，为后续维修决策提供输入控制。.（3）针对可修单机贮存期状态受检测与维修影响的问题，建立维修影响描述模型，分别从数据和机理出发，结合灰色理论、相关向量机、随机过程等提出多阶段的可修单机状态实时预测模型。针对不可修单机特点，仅考虑检测影响，综合历史、专家和现场检测信息建立相应的剩余寿命实时预测模型，关键是解决不同分布参数的估计问题。.（4）在单机状态实时预测模型基础上，考虑系统设定的概率风险水平约束、维修效果等因素，利用更新过程、半更新过程、马氏决策过程理论，建立单机的序贯检测与维修优化模型及求解算法，结合弹上设备的实例验证模型与算法的有效性。.（5）针对单机间复杂相关性、可修与不可修单机特点以及系统安全事故中的逻辑关系，综合考虑系统定期检测、不可修单机定时切换、可修单机视情更换等不同维修策略，提出基于风险回溯评估与综合重要度的长贮系统维修决策模型，以及基于Petri网的仿真优化模型，结合算例分析及方法对比，验证模型的可行性及适用条件。