Research On DUCG Theory And Application For Fault Diagnosis And Procedure Improvement Of Nuclear Power Plant

Since Three Mile Island accident in the United States,the economic,reliability,safety and operational stability of nuclear power plants have been paid more attention by the international nuclear industry to the monitoring and fault diagnosis of nuclear power plants.Most of the current fault diagnosis research for nuclear power plants is still in the theoretical research stage,and a small number of developed systems still remain in the prototype stage and have many defects.Therefore,under the premise of establishing a complete and accurate nuclear power plant system level model as far as possible,it is of utmost importance to apply an advanced theory to the fault diagnosis of plant system,and based on which to design and develop an efficient and accurate multi-function nuclear power plant fault diagnosis software to assist and guide the operation of the procedures under fault or accident conditions.This paper choose to apply and develop the theory of Dynamic Uncertai n Causality Graph(DUCG)to solve the problems above.Based on the analysis of the problems in the construction of DUCG knowledge base of nuclear power plant,this paper puts forward a knowledge base construction method based on fuzzy decision tree and genetic algorithm to improve the efficiency to represent expert knowledge in the knowledge base construction,enhance the completeness and accuracy of the knowledge base so as to avoid the misdiagnosis or inability to diagnose due to knowledge base errors.Cubic-DUCG is the most new development of DUCG,which is proposed for the complex situation such as dynamic negative feedback.Based on the study of the evidence event expansion algorithm of Cubic-DUCG,the problem of the combination explosion is revealed,and then a time slice recursive algorithm is presented,which can greatly reduce the number of items to be expanded and the amount of absorption.On the basis of this high efficient method,the computer software for the safe operation and maintenance of nuclear power plant based on the improved Cubic-DUCG is designed and developed.The software can carry out accurate and efficient condition monitoring,fault forcast,fault diagnosis,and the prediction of the fault development.As a platform,the software is versatile for different types of nuclear power plants.After the analysis of the weaknesses of existing fault procedures and State Oriented Procedures(SOP)for accident,the improved strategy for fault procedures and SOP is proposed based on the improved Cubic-DUCG algorithm and software.Compared with the original procedures,the improved strategy has a significant improvement in the operation effect for the failure or accident operation over the orginal procedures.The improved stratagem as depicted above can provide effective decision support for the safe operation and maintenance of nuclear power plants,and significantly enhance the operational sustainability,safety and economical benefit of plants.At the same time,the improved strategy is also applicable to other nuclear power plants using éLECTRICITé DE FRANCE(EDF)procedures and other types of nuclear power plants.