Development of fuzzy logic modeling method in probabilistic risk assessments

This dissertation provides a more realistic approach to the probabilistic risk assessment (PRA) by considering the effects of partial failure with fuzzy logic modeling rather than the traditional PRA binary logic modeling. This is accomplished by using an integrated methodology of thermal-hydraulic analysis and probability-possibility modeling with using the Dynamic Master Logic Diagram (DMLD). This dissertation documents the methodology, the computer codes required for it, partial failure data analysis, and application of the methodology to the most significant sequence from the NUREG-1150 (1) PRA for the Zion Nuclear Power Station, Unit One.; The approach taken utilizes a screening process to determine dominant accident sequences defined by traditional PRAs. The thermal-hydraulic analysis model satisfied by the use of the RELAP5/MOD3 (21) verifies the effects of the partial failure of system functions in the dominant accident sequence. The DMLD modeling measures the "possibility" of certain performance level of the system through the evaluation of dominant combinations of mutually exclusive cut sets of systems with partial failure state components. The application, using the results of partial failure data analysis for pumps and motor operated valves, has demonstrated that partial failures caused by component degration contributes to a high percentage of unavailability of system functions required by sequences. The methodology developed in this dissertation can be applied to current PRAs to provide more realistic results without fundamental changes of the PRA technology.; This dissertation proposes a method using DMLD for component partial failure analysis on the basis of the fuzzy state assumption. DMLD provides an effective tool for presenting and manipulating linguistic knowledge of component failure modes and severity. One example of using DMLD to predict and estimate the performance of a centrifugal pump is presented.