基于概率神经网络的船用核动力装置故障诊断方法研究

船用核动力装置是一个复杂的大系统，并且其多数设备及系统具有非线性、时变性、耦合性及不确定性，而神经网络能够逼近任意的非线性映射，因此在核动力故障诊断中得到广泛的应用。目前，应用较广泛的是BP神经网络，但其网络层数及每层神经元的个数不易确定，而且在训练过程中网络容易陷入局部最小点。鉴于此，文章把概率神经网络（PNN）应用到船用核动力装置故障诊断中去，结果表明该方法是可行有效的。

Research on diagnosis of faults for nuclear marine apparatus based on probability neural net

Nuclear marine apparatus is a huge complicated system, most of which equipments are of nonlinearity, time varying, coupling and inexactness. Neural net is widely applied in nuclear fault diagnosis for its approaching any kinds of nonlinearity mapping. At present, BP neural net is used more widely, but the layers of the net and the neurones on each layer can not be delimited easily; such net may fall into the minimum point in the course of training. In this essay, PNN proves effective in diagnosing faults on nuclear marine apparatus.