Transforming Fault Tree Analysis into Bayesian Network-Case Study: Intake Point of Hydraulic Power System

• Main Authors: Chen-Ting Chen, 陳甄婷
• Other Authors: Sou-Sen Leu
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/36711646732814890077

碩士 === 國立臺灣科技大學 === 營建工程系 === 98 === Fault Tree Analysis (FTA) has been commonly used in reliability, security and troubleshooting tools for decision making in the field of engineering. For the last decade, the rapid development of Bayesian network (BN) makes it one of the most effective theoretical models to represent the knowledge of uncertainty. Many researchers have developed methods to transform FTA into Bayesian network; however, their results still have some limitations such as: two-state events, deterministic logic gates, independent event interrelationships, and so on. This thesis aims to propose a systematic approach to establish a complete Bayesian network. Generally, it may be difficult to directly establish the architecture of a BN with quite complex relations. To solve this problem, the proposed approach will build a BN from FTA and add horizontal links to connect events so that the BN structure may have strong integrities. Using the proposed approach, a practical case study shows that the BN overcome some limitations of FTA. In addition, without considering the variability, FTA and BN have similar results with the statistics of historical data. Hence, the proposed BN includes the features of FTA. Conversely, if the variability and event interrelationship are considered, Bayesian network can be more objective and practical for applications.