Study On The Fault Diagnosis Method Integrated By Neural Network And Expert System For The Marine Nuclear Power Plant

The Marine Nuclear Power Plant (MNPP) is a very complex system, in order to guarantee its security and dependability when operating , it is very significant to set up a intelligence system which can monitor the state of the MNPP in real time and predict, alarm, diagnose to the fault of the MNPP in time.At present, Artificial Neural Networks (ANN) and Expert System (ES) have already become two studied focuses in the Artificial Intelligence (AI) field and they are widely applied to the fault diagnosis field. Nevertheless, there are certain defects in each independently practical application. Though the analysis and comparison between their merits and demerits, it is educed that there are strong complementarities between them and that their deficiencies were made up by combining in this paper. Therefore, they have been combined together in this paper. A fault diagnosis system model has been proposed in which fuzzy neural network and expert system are integrated. Regarding the condenser of the MNPP as the study target, a condenser fault diagnosis system has been set up and realized by VB6.0 and Access2000 in this paper.Based on consulting the condenser information, in this system the corresponding relation of fault types and fault symptoms are translated to fault training samples by studying operation principle and fault mechanism of the condenser, the fuzzy neural network are trained by Improving Genetic Algorithm (IGA), the training result and training samples make up the knowledge base together, the online and off-line diagnosing are realized by automating and human-computer interacting way, the diagnosing results are presented to users by explaining mechanism in the form of report table.In this system, interfaces are friendly and simple, diagnosing speed is quick, operation and maintenance is easy, etc.