| When an accident occurs in Nuclear Power Plant (NPP), it will trigger numerous alarms on the board in the main control room, which will easily confuse the operators and cause them to make more or less human errors due to the heavy workload and mental press. In this thesis, an alarm analysis method based on Causal Dependency Graph (CDG) is proposed, which can be used for analyzing the complex cause-consequence relations between alarms, and is helpful for the operators' decision-making.Aiming at solving the problems of traditional alarm analysis method that the relations between system functions are impossible to be enumerated, in this study, the basic idea of Multilevel Flow Models (MFM) is applied to decompose the system into mass flow, energy flow and information flow, which not only is helpful for reducing the working effort during system modeling, but also can simplify the alarm analysis process.In this thesis, the modeling theory of CDG and its qualitative reasoning method is introduced. A causal dependency graph for a target system of a two-loop Pressurized Water Reactor (PWR) NPP is established, which includes main devices and legs in the primary coolant system. The proposed model is highly abstracted, and some valves and auxiliary devices are simplified in order for simplifying the procedures of alarm analysis during some Design-Base Accident (DBA), such as Loss Of Coolant Accident (LOCA), Steam Generator Tube Rupture Accident (SGTR) and Main Steam Line Break Accident (MSLB).C++, an Object-Oriented program language, is used for making the relative algorithms. The performance of this model-based alarm analysis is validated by a simulation experiment using the data from RELAP5/MOD2. The simulation experiment results show that the proposed CDG model and reasoning method are effective for analyzing and diagnosing incidents before reactor trip.
|
PDF Full Text Request