Logo image
模糊故障樹的應用與錯誤診斷專家系統的建立
Thesis

模糊故障樹的應用與錯誤診斷專家系統的建立

廖為銘
Masters, National Tsing Hua University
1992

Abstract

模糊集理論 故障樹分析 專家系統 fuzzy set theory fault tree analysis expert system
在現今世界上,核能成為一種愈來愈重要的能源,如果建立核 能電廠是必要的,則我們最應該注意的是確保核能電廠絕對的安全。此論文的目的是要建立一個專家系統,期望能夠藉此專家系統增進核能電廠的安全。@ 機率風險估計 PRA(Probability Risk Assessment) 在數值上 和機率上估計系統的安全性,其主要的組成分子之一,故障樹分析FTA(Fault Tree Analysis) 是本論文估計系統安全性的工具。在現今的PRA中,由於資料的短缺,存在著相當大的不確定性,本論文以模糊集理論FST(Fuzzy Set Theory) 來描述這樣的不確定性。在FST 中,以鐘形(bell-shaped) 模糊數來替代其他的模糊數,除此之外,重要性分析也在討論的範團之內。@ 在此專家系統中,亦發展對於錯誤診斷的知識庫決策支援系統,藉由前述的故障樹,此專家系統可問使用者一連串的問題而做出診斷。當緊急警報發生時,操作人員必須在不精確,不完整且不全然可靠的環境下做出正確的決定,此專家系統能幫助操作人員更快速而正確地做出決策。In the present world, the nuclear energy becomes a more andmore important energy resource. If the nuclear plants have tobe established, then the most important thing we must take careof is to ensure the complete safety of such plants. The purposeof this thesis is to build an expert system which can increasethe safety of these plants. Probability Risk Assessment (PRA)which assesses the safety of any studied system numerically andprobabilistically is discussed. One of the major components inPRA, the so-called Fault Tree Aanlysis (FTA), is utilized toimplement this assignment. Since there are rather largeuncertainties in the current analysis of PRA due to theshortage of field information, Fuzzy Set Theory (FST) may beused to describe such uncertainties for the studied system. InFST, the bell-shaped fuzzy numbers are used to substitute theother fuzzy numbers. Besides, the importance analysis will beimplemented in FST too. A knowledge-based decision supportsystem for fault diagnosis is developed in the system. In termsof this fault tree, the system can ask the user a sequence ofquestions to make a diagnosis. When the emergency alarmshappen, the operation staffs have to make a correct decision inan imprecise, incomplete and not totally reliable informationenvironment. The software can assist the operation staffs tomake a decision more quickly and correctly.

Metrics

1 Record Views

Details

Logo image