Dynamic Reliability Models for Systems of Interacting Components

9877107许多系统由简单组件的复杂子系统组成，或者可以被视为简单组件的复杂子系统，由于组件的交互和依赖关系，这些简单组件在系统中的行为可能相当复杂。例如，这些依赖关系可能会发生，因为当某些组件失效时，在剩余组件之间发生应力的动态重新分配。动态方法是将故障机制合并到模型中，以解释此类交互和依赖关系，以及使用微观级别（组件）关系来确定宏观级别（系统）故障。在这样的动态环境中，关于系统的可靠性或故障模型，有几个重要但困难的问题。本研究的主要目的是研究以下五个问题：(1)基于故障的物理特性和假设的环境对组件和系统可靠性进行适当的建模；(2)将组件可靠性信息纳入动态环境中系统可靠性的确定；(3)识别导致故障的组件的关键配置；(4)确定用于统计推断和数据分析目的的复杂系统的近似可靠性模型。(5)将多变量失效机制纳入系统可靠性建模。虽然(2)-(5)已被广泛研究用于独立组件的相干系统，但包含组件之间依赖关系的模型将产生更现实的复杂系统失效分析。这种模型可以直接应用于许多当前重要的问题，例如，现代复杂材料和其他结构的失效，以及由大量部件组成的网络的失效。对复杂系统失效机制的研究是非常重要的，例如，在微观层面（组件）和宏观层面（系统）失效尺度之间的建模和理解。要开发的统计方法和概率模型无疑将有助于更好地理解部件级和系统级故障机制的随机性，以及在各种应用中对一般系统故障的相关统计推断。特殊的应用是预测由复合材料（如碳复合材料）或“系统”（如大型计算机网络）制成的结构的可靠性。此外，在本研究中基于累积损伤概念开发的新的非标准统计模型在抽样和从观察到的破坏数据推断方面还没有得到很好的理解。因此，该项目的一个重要目标是研究模型的这些方面。可以获得战术和竞争优势，因为这里开发的理论模型应该导致更好的系统可靠性预测及其对系统设计和系统维护和检查策略的含义。

9877107Many systems are composed of, or can be viewed as, complex subsystems of simple components whose behavior within the system can be quite complicated due to component interactions and dependencies. These dependencies may occur, for example, because dynamic redistribution of stresses occurs among the remaining components when some components fail. The dynamic approach is to incorporate the mechanism of failure into the model to account for such interactions and dependencies, as well as to use micro-level (component) relationships to determine the macro-level (system) failure. There are several important, but difficult, issues regarding reliability, or failure, models for systems in such dynamic environments. The main objectives of this research are to study the following five issues: (1) appropriate modeling of component and system reliability based on the physics of failure and the assumed environment, (2) incorporating component reliability information into the determination of system reliability in a dynamic setting, (3) identifying critical configurations of components that cause failure, (4) determining approximate reliability models for complex systems that are tractable for statistical inference and data analytic purposes, and (5) incorporating multivariate failure mechanisms into modeling system reliability. While (2)-(5) have been studied extensively for coherent systems of independent components, models incorporating dependencies among components would yield more realistic failure analysis of complex systems. Such models would have direct application to many problems of current importance including, for example, failure of modern complex materials and other structures and of networks of large numbers of components.The proposed investigations of the mechanisms for failure of complex systems are important, for example, in modeling and understanding the relationship between the micro-level (component) and macro-level (system) size scales of failure. The statistical methodology and probability models to be developed will undoubtedly contribute to a better understanding of the random nature of failure mechanisms at both the component level and system level, and the associated statistical Inferences for failure of general systems in various applications. Particular applications are in prediction of the reliability of structures made from composite materials (such as carbon composites) or "systems" such as large computer networks. Also, the new, nonstandard statistical models developed in this study based on cumulative damage concepts are not well understood with respect to sampling and inference from observed failure data. Hence, an important goal of the project is to investigate these aspects of the models. Tactical and competitive advantages can be gained since the theoretical models developed here should lead to better system reliability prediction and its implication to system design and the system maintenance and inspection policies.