对金属跌落零件报警的人工神经网络方法

The Artificial Neural Network Method for Alarm of Loose Parts in Nuclear Power Stations

针对在核电站一回路系统中检测金属跌落零件误报警率高的问题，提出基于信号时频域二维特征的人工种经网络报警模型，并对人工种经网络的学习算法进行改进。在模拟跌落零件试验中应用，表明该报警方法是可行的。

On the problem of alarm whenparts are falling in nuclear power station,the ar-tificial neural network (ANN) method has beenfirst applied in realizing the alarm of falling partsbased on the time-frequency characteristics ofimpact signal. The results from experiments jus-tify the method to be feasible.