On-line Nuclear Reactor Monitoring and Diagnosis in earliest stages using Neural Network

点击翻译按钮获取中文摘要

The object of this research is to develop the on-line nuclear reactor monitoring and diagnosis system using Neural Network theory. The nuclear reactor diagnostic system was developed and examined using reactor simulator and existing reactor data.In 2004 the diagnostic system applying auto-associative neural network theory was developed and examined about the ability of anomaly detection.In 2005 application of Kohonen network theory was investigated selectively. It was thought that in addition to anomaly detection, reactor diagnosis will be realized by using Kohonen network. The operational data of existing nuclear reactor which contain noise and power fluctuations was used for training the diagnosis system and the ability of diagnosis was tested.In 2006 the occurrence of error signal and anomaly diagnosis was examined using PWR simulator. The diagnosis of research reactor data(RSG-GAS, Indonetia) which contain noise was tested and succeeded. The detection of small helium leak which occur in HTTR ( High Temperature Gas Cooled Test Reactor, JAEA ) and power change caused by reactivity change was estimated correctly using this diagnosis system.These results were presented in some international proceedings and journals.