摘要 :
System reliability is important for systems engineers, since it is directly related to a company's reputation, customer satisfaction, and system design costs. Improving system reliability has been an important task for the system ...
展开
System reliability is important for systems engineers, since it is directly related to a company's reputation, customer satisfaction, and system design costs. Improving system reliability has been an important task for the system engineers and a number of studies have been published to discuss methods for improving system reliability. For this purpose sensitivity analysis and fault diagnosis has been used in various studies, where identification of significant and problematic components plays an important role. Both sensitivity analysis and fault diagnosis require understanding the system structure and component relationships; and Bayesian networks (BN) have been shown to be an effective tool for modeling the systems and quantifying the component interactions. In this study, we use BN for sensitivity analysis and fault diagnosis to improve system reliability. We focus on the complex systems, where the number of components and component interactions can be very large. In this study, we first discuss sensitivity analysis in complex systems using BN, which can be used for identification of significant system components. Sensitivity analysis using BN is concerned with the question of how sensitive system reliability is to possible changes in the nodes in BN. In this paper we demonstrate that BN can be efficiently and effectively used for sensitivity analysis in complex system reliability. This study is the first that considers component reliabilities and uses BN for sensitivity analysis in complex systems. In this paper as a part of our method for sensitivity analysis, an efficient algorithm (SA) is introduced to perform sensitivity analysis in complex systems. Our SA algorithm is based on a graph traversal algorithm that can be effectively used in BN. The SA algorithm traverses the BN through the connected nodes and evaluates the reliabilities to perform sensitivity analysis. Our method helps the systems engineers understand the cause and effect relationships between system components and their reliability and discover the key components that have significant effects on system reliability. Once the key components are identified, system structure can be revised to improve the overall system reliability. Next we discuss fault diagnosis in complex systems and show how fault diagnosis can be used to improve complex system reliability' Due to component aging and environmental factors, the system components in real-life complex systems may fail or not function as expected. Such failures may cause unprecedented changes in the system reliability values and affect the reliability of not only the failed component, but also the overall system. One important issue in complex systems is that, the system engineers must process large amounts of information before making operational decisions. Since BN combine expert knowledge of the system with probabilistic theory for construction of effective diagnosis methodologies, they have been applied to fault diagnosis in various studies. In this paper, we present a new method for fault diagnosis in complex systems. Our method uses the complex system reliability to detect the faulty components. We continuously monitor the overall system reliability value, and our fault diagnosis mechanism is only triggered when significant changes to system reliability are detected. As part of our method, an efficient search algorithm is designed specifically for BN. This algorithm is empowered with popular heuristics. In this paper, we discuss how our method can be efficiently applied to complex systems since our search algorithm needs to check only a small portion of the system's components before detecting the failed one. We believe that our method provides system engineers with invaluable information to diagnose the faulty component and improve reliability in complex systems.
收起
原文获取