多重人工神经网络在10MW高温气冷堆故障诊断中的应用研究

Application of Artificial Neural Networks in Fault Diagnosis for 10MW High-Temperature Das-Cooled Reactor

对人工神经网络技术在10MW高温气冷堆故障诊断中的应用进行了可行性研究,并用事故工况下反应堆监控参数的实际值和趋势变化值分别对2个BP网络进行训练和检验, 综合2个网络诊断结果得出最终诊断结果。经检验证明,采用神经网络具有较好的容错能力,对噪音信号有良好的鲁棒性。

This paper makes researches on 10MW High-Temperature Gas-Cooled Reactor fault diagnosis system using Artificial Neural Network, and uses the tendency value and real value of the data under the accidents to train and test two BP networks respectively. The final diagnostic result is the combination of the results of the two networks. The compound system can enhance the accuracy and adaptability of the diagnosis compared to the single network system.