Evaluation of the performance of various artificial neural networks to the signal fault diagnosis in nuclear reactor systems

A study on the evaluation of the performance and comparison between various paradigms of artificial neural networks in nuclear reactor signal analysis is presented for the purpose of developing a diagnostic monitoring system. Reactor signals from the Experimental Breeder Reactor II (EBR-II) are analysed. The signals are both measured signals collected by a data acquisition system (DAS), as well as simulated signals, ART2, ART2-A, fuzzy adaptive resonance theory (Fuzzy ART), and fuzzy ARTMAP paradigms of the adaptive resonance theory (ART) family, standard backpropagation, cascade correlation, and RCE networks are examined and compared.<
>