Fault Diagnosis Study Based On Wavelet Neural Network

It is out of question that default diagnosis is very important both in safety production and in national economy In the course of my study , because wavelet neural network have the following merits : high precision ,leaming rate fast etc. the final experiment proved that we can well classify the fault signals with wavelet neural network and the constringent speed of wavelet neural network is very fast ,we use wa.velet neural network in the field of fault diagnosis.In the basis of one dimension wavelet neural network , we put forward two different structures wavelet neural networks. And we used them into the fault diagnosis. We used them because the constringent speed of wavelet neural network is very fast, they are not sensitive to the inputs of the network and they have the characteristic that they also can effectively approach the functions or signals.Classical neural networks mostly train the network with back propagation arithmetic. But the back propagation arithmetic often get into the minimum value and its constringent speed is slow. Aim at these fault, we design a optimum back propagation arithmetic. We improve the Classical back propagation arithmetic on two points : one is that we adopt a method of self-adaptive learning rate arithmetic which is based on the sign change of the grads ,second is that we adopt momentum item. Finally we contrast the result of the experiment ,we find that the optimum arithmetic can effectively improve the training speed of wavelet neural network and avoid that arithmetic get into minimum values. The fault diagnosis result is satisfying.The two different structures wavelet neural network finally are used into fault diagnosis of the major motor cooling system of ship , the final diagnosis results are list at the text. Compared them to the classical back propagation network, it can find that the speed of wavelet neural network (WNN) is very fast.