Research on Feature Extraction Method of Loss-of-Coolant Accident of Nuclear Power System
Duan Meng-qian · Ship Engineering · 2014
Nuclear Power system has involute functions and lots of parameters. When the fault occurs, how to choose a proper method to pickup features of the system is the key precondition of orientation, analysis and forecast. This paper discusses four methods of feature extraction applied in the Loss of Coolant Accident(LOCA), including direct feature extraction and PCA belong to liner method and manifold learning method LLE and ISOMAP belong to nonlinear ones. The result provides evidence to application of the fault diagnosis.