【作者】 刘宇；

【导师】 黄洪钟；

【作者基本信息】 电子科技大学 ， 机械电子工程， 2010， 博士

【摘要】 随着现代系统和设备日益朝着大型化、复杂化、精密化方向发展以及对系统的失效机理和潜在规律的逐渐深入的探究,人们发现系统及组成单元在寿命周期内的失效演化过程中往往呈现出多状态的特征,并且各个状态的失效规律和机理、工作性能和效率不尽相同。在这种情况下,若采用常规的二态可靠性理论将系统粗略地划分为“正常”和“失效”两个状态显然不符合实际情况,而且忽略系统本身所表现出的多状态特征是不能准确地描述系统和单元的复杂失效过程的,这就迫切需要开展多状态可靠性理论的研究,以解决现代工程中大型复杂装备和系统的可靠性问题。近年来,随着工程需求的不断增加,多状态可靠性理论成为了学术界和工业界所共同关注的热点问题,并在众多领域得到了迅速发展,如：机械工程、计算机和网络系统、网格、通讯系统、能源系统、供给系统、城市基础设施、战略和防御等。本论文针对多状态复杂系统可靠性理论发展中亟待解决的难题和挑战,以解决复杂性和不确定性这两个核心问题为出发点,在多状态复杂系统可靠性建模与维修决策方面开展研究,其主要内容和创新性成果如下：(1)提出了多状态复杂系统的选择性维修决策优化模型。在多状态复杂系统可靠性理论基础上开展了在有限维修资源约束下多状态复杂系统的选择性维修决策优化的研究。通过将二态元件的非完好维修模型引入到选择性维修决策模型中,并构造维修费用—维修效果的模型以反映非完好维修与其维修成本的函数映射关系,该方法可为选择性维修决策的制定提供更大的灵活度。研究结果表明,本论文所提出的选择性维修方案较以往的方法具有明显的优势,系统经维修后的任务完成概率显著提高。(2)提出了多状态元件的非完好维修模型,建立了多状态复杂系统的“系统级”更换维修策略。在实际生产中,复杂系统维修计划的预测和制定往往是从“系统级角度”出发的。本论文提出了一种多状态元件非完好维修模型,用以描述维修措施对元件衰退规律的影响,弥补了目前非完好维修模型仅能适用于二状态元件的缺陷,并在该非完好维修模型基础之上提出了一种新的多状态系统“系统级角度”的更换维修策略,可保证系统在单位时间的平均效益最大。(3)系统地研究了多状态复杂系统的“元件级”维修建模与更换策略。本论文以“元件级角度”的维修策略作为研究的出发点,系统地开展了在完好维修、非完好维修下的多状态复杂系统的元件更换维修决策及全寿命周期内的系统冗余优化等问题的研究。通过提出更具一般性的多状态元件的非完好维修模型,建立元件维修费用和非完好维修效果之间的映射关系,有效地实现了各元件维修费用的合理分配。研究结果显示,本论文提出的方法比传统的维修方案更有效。同时,本论文还提出系统冗余分配和元件更换策略的联合优化问题,通过集成系统的冗余设计优化和维修决策,实现了系统全寿命周期内的可靠性优化设计。(4)提出了模糊多状态复杂系统可靠性建模理论以及模糊多状态元件的更换维修决策方法。本论文开展了模糊不确定性下的多状态复杂系统可靠性建模与性能评估理论的研究,提出了模糊马尔可夫模型和模糊马尔可夫报酬模型以计算在模糊不确定性存在情况下多状态元件的动态状态分布与累积性能,并且通过提出不同类型系统的模糊通用生成函数近似算法,有效地提高了计算模糊多状态系统的性能状态分布的速度。同时,本论文还提出了更为完备的模糊多状态系统可用度计算方法,克服了现有方法的不足。在上述模糊多状态可靠性理论基础之上,本论文进一步深入研究了模糊多状态元件的更换维修决策理论,通过构造模糊不确定性下的元件更换决策模型并利用现有的几种模糊决策方法,为模糊多状态元件的维修方案的选择和制定提供了理论指导。(5)提出了多级层次型复杂系统的统计灵敏度分析方法,并能广泛应用于复杂系统可靠性分析和设计中,以辅助系统不确定分析和降低模型复杂度。统计灵敏度分析是研究系统输入变量的变化对系统输出量变化的影响程度的一种不确定性分析方法,它通过对系统的输入变量的重要度排序可有效地降低复杂系统不确定性下分析、可靠性建模和设计优化等问题的复杂度。本论文提出了共享变量下的层次型统计灵敏度分析方法,有效地克服了现有的层次型统计灵敏度分析法不能处理子模型存在共享输入变量的局限,使复杂系统的统计灵敏度分析法具有更普遍的适用性。由于该方法遵循了自上而下的基本分析方式,符合复杂产品或系统的典型设计过程,并能通过并行计算的方式大幅度降低复杂系统统计灵敏度分析的总时间。更多还原

【Abstract】 As modern advanced engineering devices and systems designed towards larger size, more complex, and higher precision, as well as the understanding of system failure mechanism and physics developed continuously, it has been observed that multiple states occur during the deterioration process of engineering systems and components. The associated failure mechanism, performance rate, and efficiency vary across different states. Conventional reliability analysis methods are limited by their fundamental assumption that both systems and components can be characterized as one of only two possible states:working perfectly or completely failed. It is inappropriate to describe the complicated failure process using these two states while ignoring the multiple states that possibly exist in engineering systems. Therefore, there is an urgent need for developing reliability theories and reliability analysis methodologies to facilitate the reliability assessment and enhancement in sophisticated multi-state systems.With an increasing demand in industry in recent years, multi-state reliability theory has become an emerging research topic in both industry and academia. It has been applied to a variety of industrial domains, including mechanical engineering, computer and network, grid systems, communication, energy, supply systems, municipal infrastructure, and defense strategy research, etc. The overall objective of this dissertation is to address key challenges and critical issues in multi-state reliability theory, with a special emphasis on two fundamental aspects:complexity and uncertainty. In particular, this dissertation is focused on the reliability modeling and maintenance decision making of multi-state complex systems. The primary research contributions and innovative outcomes are summarized as follows:(1) Development of a selective maintenance optimization strategy for multi-state systems. With the consideration of the limitation of maintenance resources, a selective maintenance optimization methodology based upon the multi-state system reliability theory is proposed. To take into account the imperfect maintenance quality, an imperfect maintenance model for binary state components is incorporated into the maintenance decision model. In addition, a cost-maintenance quality functional relationship is developed to consider the age reduction factor as a function of allocated maintenance cost. Demonstrated by numerical studies and examples, the proposed methodology provides much more flexibility in assigning the maintenance cost and yields better results than the existing methods in literature.(2) Development of an imperfect maintenance model for multi-state components and a "system perspective" replacement policy for multi-state systems. In practice, the maintenance planning of complex systems is usually carried out from the system-level perspective. The existing imperfect maintenance models are applicable only to binary state components. In this work, a novel imperfect maintenance model is proposed to quantify the impact of maintenance activities on the degradation trend of multi-state components. Based upon the proposed imperfect maintenance model, a new system replacement strategy is developed to achieve the maximum expected profit per unit time for the entire multi-state system.(3) Systematic investigation of the component-level maintenance and replacement optimization strategies for multi-state systems. In this dissertation, the optimized component replacement strategy, and the lifecycle-based redundancy design optimization under perfect and imperfect maintenance are studied from the component-level perspective. Building upon a generalized imperfect maintenance model for multi-state components, a flexible and effective repair cost assignment among components is achieved by proposing a set of functional relationships between the assigned repair cost and the imperfect repair quality. Through the study, it has been demonstrated that the proposed method is more economically efficient than conventional methods. In addition, a joint optimization method for redundancy and component replacement strategy is put forth by considering the component maintenance planning in the design stage of multi-state systems. It enables the system reliability design optimization across the entire lifecycle.(4) Development of fuzzy multi-state system reliability modeling methods and fuzzy multi-state component replacement strategy. The multi-state system reliability modeling and assessment methods are developed based on the fuzzy uncertainty theory. A fuzzy Markov model and a fuzzy Markov reward model are proposed to evaluate the dynamic state distribution and the cumulative performance of fuzzy multi-state components. By developing a set of composition rules for the fuzzy universal generation function of different types of systems, the system state distribution can be derived in a computationally efficient manner. In addition, a modified fuzzy multi-state systems availability assessment approach is developed to overcome the drawbacks of the existing methods in assessing the system availability. A replacement strategy for fuzzy multi-state components that are based on the proposed fuzzy multi-state system reliability theory is further investigated. By constructing the formulation of replacement decision and utilizing the existing fuzzy ranking methods, it provides a general framework and guideline of maintenance planning for fuzzy multi-state components.(5) Development of a hierarchical statistical sensitivity analysis method to facilitate uncertainty analysis and reduce complexity in design and reliability assessment of complex multi-level hierarchical systems. Statistical sensitivity analysis is an effective tool to examine the impact of variation in model inputs on the variations in model outputs, and can relieve the computational burden and manage the complexity in complicated system design and reliability analysis under uncertainty. A hierarchical statistical sensitivity analysis method is developed to deal with the shared variable among submodels, which is not addressed in other existing statistical sensitivity analysis methods. The inherited advantages of the proposed method are that the top-down strategy used in the method matches better with a typical design process of complex systems and products. The global statistical sensitivity index is obtained by aggregating the local sensitivity indices of submodels across multi-level hierarchy. The associated computational time is significantly reduced using the concurrent computing fashion. Therefore, the proposed method has a broader range of applicability to complex engineering systems design than other existing sensitivity analysis methods.更多还原