基于混合神经网络的核动力装置并发故障诊断研究  

Intercurrent Fault Diagnosis of Nuclear Power Plants Based on Hybrid Artificial Neural Network

作　　者：彭俏[1] 郭立峰[1] 马杰[1] 

机构地区：[1]海军工程大学船舶与动力学院,武汉430033

出　　处：《核动力工程》2012年第6期30-34,共5页Nuclear Power Engineering

摘　　要：在对ART-2网络和并行前馈反向神经网络(BP)模型进行分析的基础上,提出了一种针对核动力装置并发故障进行诊断的复合神经网络。先运用ART-2网络识别单一故障,再使用并行BP实现并发故障和新故障的分离。仿真实验表明,所构建的复合神经网络,克服了单一神经网络难于识别并发故障和新故障的缺陷,能够有效地实现并发故障和新故障的诊断识别。Based on the analysis of the structure of ART-2 and parallel BP neural network, a hybrid artificial neural network is proposed aiming at the intercurrent faults diagnosis of nuclear power plants. Firstly the ART-2 net is used to identify the single fault, then the parallel BP net is used to distinguish intercurrent faults from new fault. The simulation shows that, the hybrid artificial neural network resolves the problem of single neural network in distinguishing intercurrent faults from new fault, and can diagnose the intercurrent fault and new fault efficiently.

关 键 词：核动力装置 并发故障 混合神经网 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]