Research Of Nuclear Power Plant Intelligent Fault Diagnosis Method Based On Data Fusion

Nowadays,researchers expect that more advanced technology could be applied to nuclear power plant running service,for the reason that nuclear power plant is a high complex nonlinear system which needs very high safety,and the society has paid a lot of attention to the safety of nuclear power plant.Therefore,a more advanced fault diagnosis technology is demanded for keeping nuclear power plant running safely.With the development of artificial intelligence technology(AIT),intelligence diagnosis technology has been applied in many industries gradually.In this paper,intelligence method data fusion is applied into nuclear power plant fault diagnosis,and researched the application of the combining method of backpropagation neural network and D-S evidence theory(BP-DS)and deep learning method to nuclear power plant fault diagnosis.As a common intelligent fault diagnosis method,BP neural network is easy to fall into local minimum during training procession and slower convergence speed which cause a lot shortcomings.However,D-S evidence theory has some advantages of making a precise conclusion through comprehensive analyzing many uncertain evidences.To avoid the shortcomings of BP neural network and improve the accuracy of fault diagnosis,D-S evidence theory was introduced.As a new artificial intelligence network,the deep learning has strong feature learning ability,and has made a series of successes in the field of pattern recognition.In this paper,deep learning is applied into nuclear power plant fault diagnosis with complex faults knowledge,for which reason fault diagnosis is one of pattern recognition problems.In this paper,the concept of equipment intelligent fault diagnosis,the present research situation of nuclear power plant intelligent fault diagnosis and the concept of data fusion are first introduced in detail.Then evidence theory,BP neural network and deep learning are expounded.After that,two fault diagnosis models are built based on BP-DS method and deep learning method,which are used for further research.In the procession of model validation,BP neural network method,support vector machine(SVM)method and K-nearest neighbor(K-NN)method are used to compare with proposed method,and results show that BP-DS method and deep believe network(DBN)method can achieve better results in nuclear power plant fault diagnosis.Finally,in WINDOWS 7 platform,a nuclear power plant intelligent diagnosis fault system is built by using Matlab programming language,and verify the effectiveness of the system.