【作者】 徐玉国；

【导师】 邱静；

【作者基本信息】 国防科学技术大学 ， 机械工程， 2012， 博士

【摘要】 在装备技术发展与作战样式演变的推动与牵引下，装备维修保障朝着“科学预知，精确保障”的目标发展。装备自主维修保障系统（Autonomic Logistics System，ALS）是一种能够降低装备寿命周期成本、提高装备使用可用度的智能型、预知型、先导式、网络化的维修保障体系，是未来装备维修保障的必然发展模式。ALS的构成要素包括：故障预测与健康管理（Prognostics and Health Management，PHM）系统、联合分布式信息系统（Joint Distributed Information System，JDIS）以及维修保障资源设施。对于ALS而言，为提升复杂动态环境下的装备维修保障效能，如何构建柔性维修保障网络体系、合理有效利用装备实时状态信息、充分调度维修保障资源是亟待解决的问题。鉴于装备维修保障在精确化、动态性、网络化等方面的需求，本文针对ALS中维修保障组织柔性化、维修决策精确化、资源供应协同化的问题，系统分析了维修保障网络的特点以及装备剩余使用寿命（Remaining Useful Life，RUL）评估的规律，并对RUL评估驱动的维修保障策略进行了深入研究。论文的主要研究内容包括：1. ALS组织体系的优化设计在深入分析ALS组织体系的任务、实体与网络化特性基础上，应用复杂网络与超网络描述与分析方法提出了维修保障体系结构的多元加权超网络模型。基于该模型定义了装备维修保障组织体系网络的性能指标，在ALS组织体系动态演化（集成）规则的基础上提出了两种动态演化（集成）模型——随机演化模型与适应演化模型，深入分析了演化模型的参数对ALS组织体系效能的影响，为ALS组织体系的设计提供了依据。2.基于RUL评估的预测维修决策优化将装备RUL评估分为离散型与累积型，分析了RUL评估过程的动态性与RUL评估结果的随机性，给出了刻画RUL评估结果不确定性的指标。(1)针对离散型RUL评估方法，在连续监控与完美维修的假设下，以单位时间成本最小与平均使用可用度最大为目标，基于维修过程的再生与更新特性提出了周期预测/非周期预测模式下RUL评估驱动的维修决策模型，设计了维修决策优化的算法。应用Monte Carlo仿真方法研究了周期/非周期评估调度函数、预防性维修阈值对维修决策优化的影响。(2)针对累积型RUL评估方法（损伤标尺法），在完美维修的假设下，以单位时间成本最小、平均使用可用度最大与平均效费比最大为目标，基于更新过程理论提出了损伤标尺驱动的预测维修决策模型。对于配置寿命独立型损伤标尺的系统，选择预测距离与威布尔分布的形状参数为决策变量，对于配置寿命相关型损伤标尺的系统，选择累积损伤因子与随机标准差为决策变量，应用MonteCarlo仿真研究了各个决策变量对维修性能的影响，并将维修效果与事后维修策略、年龄换件策略进行了对比，给出了最优维修决策的准则。3.自主维修保障体系的维修与库存联合优化综合考虑自主维修保障行为触发与响应两方面的因素，分析了维修保障体系网络中的信息流与物流，提出了ALS的信息集成与资源共享模式。分析了ALS中装备维修、备件供应等各类成本的产生机制，以单位时间成本最小为目标提出了由RUL评估驱动的维修与库存联合优化模型，应用Monte Carlo仿真方法设计了求解最优维修保障策略的算法，分析了漏检率与虚警率对维修保障效能的影响。总之，论文从ALS体系建立与优化的现实需求入手，建立了ALS体系的加权超网络模型与RUL评估的描述方法，提出并解决了ALS体系的动态演化（集成）与设计、装备预测维修优化、ALS资源协同配置等关键技术问题，为系统科学地解决ALS的构建与决策问题进行了广泛与深入地探索，具有重要的学术与工程实际应用价值。更多还原

【Abstract】 With the development of equipment technology and the evolution of combatstyle, the goal of equipment maintenance and support is to be “scientific predictionand precise logistics”. The Autonomic Logistics System (ALS) is an intelligent,predictive, proactive and networked logistics architecture with the employment oftechnologies such as prognostics and health management (PHM) system and jointdistributed information system (JDIS). This new approach shows the potential for costsavings, increased operational availability and better system performance. There is anurgent need to build a flexible networked logistics architecture, to utilize theequipment’s real-time health information efficiently and to schedule logisticsresources successfully in a complex and dynamic logistics environment for betterALS efficiency.In order to establish a precise, dynamic and networked ALS, the characteristicsof logistics networks and the rules of equipment’s remaining useful life (RUL)estimation are analyzed entirely to improve logistics organization flexibility. Andthe optimal maintenance and support decision based on RUL estimation is studieddeeply to improve maintenance decision precision and resource supplycooperativeness.The main research contents include:1. Optimization design on ALS organizational architectureBased on the analysis of the tasks, entities and network characteristics of ALSorganizational architecture, a multi-element weighted super-network model oflogistics organizational architecture is presented using description and analysismethod of complex network and super-network. The performance indices whichdescribe the logistics organizational structure are defined, and two kinds of dynamicevolution (integration) models--stochastic evolution model and adaptive evolutionmodel are respectively proposed based on the evolution (integration) rules of ALSorganizational architecture. And then the effects of evolution modes and parameterson ALS organizational network performance are analyzed deeply to help designlogistics networks.2. Optimal predictive maintenance decision based on RUL estimationSince the RUL estimation process can be divided into discrete type andcumulative type, the dynamic characteristics of RUL estimation processes and therandomicity of RUL estimation results are thus analyzed, and the uncertainty indices,including false alarm rate and missed detection rate which depict RUL estimationperformance, are also proposed.(1) For discrete type of RUL estimation, under the assumption of continuous monitoring and perfect maintenance, a periodic/aperiodic maintenance decision modeland arithmetic are proposed with the aim of minimizing the long-term average costrate or maximizing average operational availability based on regeneration and renewalproperties of maintenance process. This model is demonstrated with a numericalimplementation example using Monte Carlo simulation to analyze the effects ofperiodic/aperiodic RUL estimation schedule function and preventive maintenancethreshold on optimal maintenance decision.(2) For cumulative type of RUL assessment method such as canary in RULestimation of electronic system, and under perfect maintenance assumption, acanary-driven predictive maintenance decision model is proposed with the aim ofminimizing the long-term average cost rate or maximizing average operationalavailability or maximizing average cost-effectiveness ratio, based on renewalproperties of maintenance process. For LRU-independent and LRU-dependentcanaries, the decision variables are respectively as prognostic distance, shapeparameter of life distribution (Weibull distribution) and as accumulated damage ratioand stochastic standard deviation of life distribution (Weibull distribution). Thismodel is demonstrated with a numerical implementation example using Monte Carlosimulation to analyze the effects of all decision variables on maintenance performance,and the optimal maintenance decision criterion is established thereafter in comparisonwith unscheduled maintenance and age replacement policy.3. Joint optimization of maintenance and inventory of ALSThe information integration and resources sharing modes of ALS are proposedbased on a comprehensively considering sense and respond characteristics and on theinformation flow and material flow of ALS. The mechanisam of costs generation onequipment maintenance and on spare part replenishment under the respond resourcessharing mode are analyzed. And a joint optimization model of maintenance andinventory based on RUL estimation is proposed to minimize the long-term averagecost rate. The arithmetic for optimal logistics strategies is designed using Monte Carlosimulation methods which can analyze the effect of missed detection rate and falsealarm rate on ALS performance.In summary, the weighted super-network model of ALS architecture and thedescription method of RUL estimation are proposed, and the key problems such as thedynamic evolution and design of ALS architecture, the optimal equipment predictivemaintenance decision and the cooperative resource supply of ALS are all presentedand resolved after the requirement analysis of establishment and optimization of ALS.This dissertation aims at exploring the construct and decision of ALS in a thoroughand deep way, with the hope that these researches may have some academic andpractical significance.更多还原