基于RBF神经网络的核动力装置故障诊断方法研究

核动力装置是一个高度复杂并具有高度安全性要求的结构体系,其故障检测方法一般采用传统的阈值方法。为克服阈值方法的不足,提出了基于RBF(radial basis function)神经网络的核动力装置故障诊断方法。该方法选择对核动力装置安全具有重要影响的运行参数作为神经网络的输入,并利用核动力装置正常运行模式及典型故障模式的监测数据作为训练样本,网络训练采用正交最小二乘算法(orthogonal least square,OLS)。为了验证所提方法的可行性,利用核动力装置运行监测数据进行检验。结果表明,RBF神经网络成功地诊断出了故障,具有良好的诊断效果。

Research on fault diagnosis method based on RBF neural networks for nuclear power plant

A nuclear power plant(NPP) is a structural system that is of high complexity with strict safety demands.The conventional threshold method is usually used for fault detecting for nuclear power plant.In order to solve the problem that existed in conventional fault diagnosis method for nuclear power plant,a new fault diagnosis approach based on RBF(radial basis function) neural network was proposed.In the proposed method,the operational parameters,which are important to nuclear safety,were chosen as the input signals of neural networks and the monitoring data of normal operation modes and typical fault modes coming from nuclear power plant were used as the training data of neural networks.The orthogonal least square algorithm was used to train the network.In order to demonstrate the feasibility of proposed method,the monitoring data of nuclear power plant operation was used for testing.The results reveal that the RBF network is successful in fault detecting and is effective for fault diagnosis.