Application of BP-RBF Neural Network to Fault Diagnosis of Nuclear Power Plant

The paper introduces a mutual mixture of the back propagation (BP) neural network and the radial basis function (RBF) neural network, and applies it in the condition monitoring and fault diagnosis system of the nuclear power plant. By analyzing the typical fault characteristic of nuclear power plant, the corresponding network architecture was established. In order to confirm the validity of this mixture network, the simulation experiment was carried out on the nuclear power plant simulator and the codes of network program were written with Visual Basic 6.0. The experiment results show that this mixture network has the good diagnosis accuracy and the real-time extendibility.