Design And Research On Distributed Fault Diagnosis System For Nuclear Power Plant

In order to find and locate the fault timely and accurately, an effective distributed fault diagnosis system has been designed and established to assist operator in judging and making decisions. So it has great significance for the security and economy of the nuclear power plant.This paper aims to study the key technology of distributed fault diagnosis, design and build an effective nuclear power plant fault diagnosis system. The main contents are as follows:(1) By taking the two-loop PWR as diagnosis object, this paper divides the whole task of fault diagnosis into some sub-systems on the basis of macro-analysis, and then some detailed analysis of each sub-system was given respectively. Some typical faults and their symptom were collected in this paper,which has also discussed extraction of simultaneous fault's feature. Compared with the monitoring parameters, fault which can be diagnosed by the distributed fault diagnosis system would be determined, and the knowledge base for diagnosis was built.(2) By comparing several theories and analyzing the characteristics of distributed fault diagnosis system of nuclear power plant, a diagnosis method, which using BP neural network for local diagnosis and evidence theory for global decision, was researched in this paper.Furthermore, some key problems on training of neural network and dynamic framework of evidence theory were studied in detail.(3) A distributed fault diagnosis system of nuclear power plant has been designed in the light of integration, modularization and expansibility, and built by using language of C# to achieve different designed functions.(4)Some application research of distributed fault diagnosis system, which has been built,would be given by using the simulator of nuclear power plant.The result shows that distributed fault diagnosis system developed can diagnose no matter single fault and simultaneous faults under steady state condition. It can also apply for diagnosis in variable condition. By using the offline diagnosis model, the distributed fault diagnosis system could distinguish the degree of faults, such as break area or rates of rod withdraw. Overall, this distributed fault diagnosis has a good performance and could be reference for future study.