Reliability Analysis for Industrial Systems with Interdependency

这笔赠款支持为相互依存的系统开发统计模型的研究。正在调查的特定系统是指系统中某个组件的故障会影响其余组件的可靠性的系统。这项研究的主要目标是用尽可能少的关于组件寿命分布的假设来统计地描述这种相关性。使用几种工具来识别系统的动态性质，包括可靠性估计、相应的可靠性不确定性测量以及相互依赖的统计测试。负荷分担模型最早是在纺织工程中推导出来的，用来描述纤维束中单个纤维的失效，它被用作描述部件相互依赖关系的基础。负载共享规则的统计测试包括查看负载共享环境是否存在的测试，以及系统压力是否会导致组件在其他组件发生故障时以更高的速度发生故障的测试。在本研究项目中，强调了与特定应用相对应的感兴趣的组件类别(例如，故障率不断增加的组件类别)。统计方法的应用针对不同的行业而设计。这项研究的动机是核工业中的问题，如果附近的部件降级或故障，发电厂中的部件可能会降低可靠性。在软件可靠性领域，软件中发现程序错误的时间取决于软件中还存在哪些其他错误。本文推导的方法可以帮助估计像这样的动态软件可靠性问题。其他例子包括材料测试。例如，金属疲劳通常以裂纹扩展为特征。在给定的试件中，材料中可能存在多个裂纹，但最大的一个继承了大部分施加的应力，因此将以比其他裂纹更快的速度扩展。这为将载荷分担模型推广到材料工程中的退化模型提供了平台。

This grant supports research for the development of statistical models for interdependent systems. Specific systems under investigation are those for which the failure of a component within the system affects the reliability of the remaining components. The main goal of this research is to characterize this dependence statistically, using as few assumptions as needed about the component lifetime distributions. Several tools are used to identify the dynamic nature of a system, including reliability estimation, correspondent reliability uncertainty measurement, and statistical tests for interdependence. The load share model, first derived in textile engineering to describe the failure of individual fibers within a bundle, is used as a basis for characterizing component interdependencies. Statistical tests for load sharing rules include a test to see if the load-sharing environment exists, and a test of whether the system stress causes components to fail at an increasing rate as other components fail. Component classes of interest (e.g., those with increasing failure rate) corresponding to specific applications are emphasized in this research project. Applications of the statistical methods are designed for various industries. The research is motivated by problems in the nuclear industry where components in a power plant can suffer decreased reliability if nearby components are degraded or failed. In the field of software reliability, the time until a program bug is found in the software depends on what other bugs exist in the software. Methods derived here can help in the estimation of dynamic software reliability problems like this one. Other examples include materials testing. Metal fatigue, for example, is often characterized by crack growth. In a given specimen, several cracks can exist in the material, but the largest one inherits most of the applied stress, and thus will grow at a faster rate than the other cracks. This provides a platform for extending the load-share model to degradation models in materials engineering.