Reliability Analysis And Fault Diagnosis By Multilevel Flow Models For Nuclear Power Plant

To realize good economy on the basis of ensuring nuclear safety is a driving force for the long-term and sustainable development of nuclear power in the world. Reliability analysis technology is one of the most important means to guarantee the safety, reliability and economy of nuclear power plant. Reliability analysis technology can be applied for the quantitative assessment of the reliability of equipments and their contributions to the safety and availability of the whole system, which can support various tasks, such as the optimization of periodic test and in-service inspection period, and making best maintenance strategies as well. Reliability analysis technology can also be applied for revealing equipment faults and their effects on the system operation and safety, which will lay a solid basis for improving the inherent system reliability and safety by revising design, assisting operators in their tasks of fault identification, making emergency operating procedures, or making preventive measures to avoid and release the fault consequence._This thesis, partially supported by National Natural Science Foundation of China (NSFC) and Chinese Nuclear Development Project in the National12th Five-year Plan, aims at developing a new reliability analysis method and a comprehensive fault diagnosis technology by improving the methodology of Multilevel Flow Models (MFM). The main works of this thesis are summarized as follows.For solving the problem that the traditional MFM can only describe static characteristics of a system, this thesis improves the MFM methodology by introducing some new concepts including logic gate, logic attribute, basic goal, time point, and signal. The improved MFM can clearly describe not only the reliability logical relations, but also the time consequence relations of the system and equipments with phased missions.This thesis presents a fundamental method and characteristic quantities for reliability analysis by utilizing the MFM hierarchical flow structure. The algorithms for calculating the reliability of a system with two states and multiple-states are presented, respectively. Especially, a method for solving dynamic reliability problems by mapping MFMs into GO-FLOW models is given and discussed in detailed.For the qualitative reliability analysis, this thesis presents a Failure Mode and Effect Analysis (FMEA) method based on MFM which can be used for analyzing the causes of the functional faults and their effects along MFM flow structures using conservation principles. The proposed FMEA method is easy to be implemented and can avoid overlooking important failure modes of the system. On the basis of FMEA, a fault tree generation method for two-state system is proposed by mapping MFM elements into mini fault trees, selecting a top event, connecting the relevant mini fault trees and breaking the logic loops. In this way, the reliability indexes including the minimal cut sets, element importance and sensitivity can be qualitatively analyzed.Finally, this thesis applies above qualitative reliability analysis methods into the fault diagnosis field of nuclear power plant. A comprehensive fault diagnosis technology consisting of three fault diagnosis methods based on MFM is proposed. An alarm analysis method is proposed to analyze the causalities between alarm states of MFM functions. The alarm reduction can be realized by excluding the consequential alarms and a list of possible causes will be given. A minimal cut set method is presented to offer minimal failure modes of the current abnormal system state by mapping the MFMs with multiple states into fault trees. For solving the common limitation of qualitative reasoning methods that the alarm thresholds may greatly affect diagnosis results, this thesis presents an uncertainty diagnosis method based on Bayesian theory which can provide approximate, but reasonable explanations to the current abnormal system state based on the analysis of a fault tree model mapped from the MFM with uncertainty causalities.A graphical modeling platform, a program for the reliability analysis and a fault diagnosis system for nuclear power plant are designed and developed. The system MFM can be easily built by simple mouse and keyboard operations. Multiple system goals and configurations can be exactly evaluated through one computer calculation. The proposed technologies can be used for analyzing the reliability issues of a system with multiple-state, time delay, time-consequence and phased mission characteristics. In addition, by combining the advantages of goal-based, rule-based, risk-based and expert experience based methodologies into one framework, the proposed technologies can help operators have a comprehensive understanding to the system requirements, perform fault diagnosis and have an insight of the risk in system operation, which shows a good foreground in engineering application.