METHODOLOGY FOR SYSTEM ANALYSIS USING FAULT TREES, SUCCESS TREES, AND IMPORTANCE EVALUATIONS.

摘要

Reliability, safety, and economic issues of nuclear power plants have always been of great concern. This area of study has received considerable attention, especially in the past two decades. The occurrence of the Three Mile Island accident has motivated further studies and an application of various new techniques in analyzing the nuclear power plants.;Modified versions of the computer codes PREP and KITT-1 were produced. In addition, a code IMP was written in order to provide the capability to analyze fault trees and success trees and compute various measures of importance. This methodology was then applied to various cases. The results generated by the computer codes were verified by comparison with hand calculations.;Two new measures of importance are introduced. The first of these measures, referred to as the Relative measure of importance, is more sensitive to potential changes in component reliability than any previously used measure. The second newly introduced measure, referred to as the Differential Natural Logarithmic measure of importance, is very useful in success-tree analysis because it provides the same results as the Criticality measure with a considerable savings in computation time. A thorough study of six measures of importance (including the two new ones) is reported and guidelines given for selection of importance measures in future system analyses.;This dissertation describes an investigation of certain methods available for reactor safety analyses. These approaches include fault-tree and success-tree approaches. It then focuses on the application, the development and evaluation, of several importance measures for use with fault-tree and success-tree analyses. These developments provide a basis on which to develop a thorough and systematic approach to sensitivity analysis when using fault trees or success trees.