The Application of New RBF Neural Network in Fault Diagnosis of Steam Passage Part of Steam Turbine

In this paper, the author uses the theory of regulation of immune genetic system to choose the number and location of the hidden layer centers of the RBF neural network, while using recursive least squares method to determine the weights of output layer, and thus creates a new type of RBF neural network model. By calculating the characteristic parameters-information entropy, the correlation between the fault source of flow section and the characteristic parameters can be determined. Thus we can quickly and accurately determine the fault diagnosis knowledge base. And the model and the knowledge base can be applied to the fault diagnosis of the flow part of turbine. According to the diagnostic results, the model has high convergence speed, high accuracy and good generalization ability, provides a new way to condition monitoring and fault diagnosis of steam turbine generator.