Search

Different types of faults could occur in a nuclear power plant, and there was no direct correspondence between a specific fault and its symptoms. So, a hybrid intelligence approach is proposed for the fault diagnosis at a nuclear power plant. Depending up the symptoms observed and the progress of fault diagnosis process, different fault diagnosis technologies, such as artificial neural network, data fusion and signed directed graph, could be combined as appropriate to detect and identify different faults at local or global level in nuclear power plants. The effectiveness of hybrid intelligence approach in improving the fault diagnosis efficiency in nuclear power plants was verified through simulation experiments. [ABSTRACT FROM AUTHOR]

A dynamic model was developed using two back-propagation neural networks of the same structure, one for online training and the other for prediction, and proposed for continuous dynamic prediction of the time series of NPP operating parameters. The proposed prediction model was validated by predicting such time series of NPP operating parameters as coolant void fraction, water level in SG and pressurizer. Validation results indicated the proposed model could be used to achieve a stable prediction effect with high prediction accuracy for the prediction of fluctuating data. [ABSTRACT FROM AUTHOR]