BP神经网络改进算法在核电设备故障诊断中的应用

Application of Improved BP Algorithm in Fault Diagnosis of Nuclear Power Equipment

根据训练误差大小自适应调整神经元输入特性参数,并应用改进的遗传算法对神经网络的权值和隐含层数目进行优化,对传统的人工神经网络误差反传算法进行了改进,使训练算法的收敛速度大大提高。将人工神经网络技术和改进的BP网络训练算法应用于核电设备故障诊断,并以核电蒸汽发生器U形管破裂为例,建立了故障诊断模型。仿真结果表明,该算法的应用是可行的。

The error back propagation （BP） training algorithm for artificial neural networks was improved, by adjusting the coefficient of neuron according to the size of the training error, and an improved genetic algorithm used to improve the structure and weight of the traditional BP neural network simultaneously in this paper, which greatly increased the convergence rate of the training algorithm. The artificial neural network technology and the improved BP network training algorithm were applied to the nuclear power plant fault diagnosis. The fault of the break of the steam generator inverted U-tube in the nuclear power plant was taken as the example, and the fault diagnosis model was established. The simulation results showed that the application of this algorithm is feasible.