Ensemble of neural networks for detection and classification of faults in nuclear power systems

Authors

Roozbeh Razavi-Far, Politecnico di Milano
Piero Baraldi, Politecnico di Milano
Enrico Zio, Politecnico di Milano

Document Type

Conference Proceeding

Publication Date

1-1-2012

Publication Title

World Scientific Proc. Series on Computer Engineering and Information Science 7; Uncertainty Modeling in Knowledge Engineering and Decision Making - Proceedings of the 10th International FLINS Conf.

Volume

7

First Page

1202

Last Page

1207

Abstract

In this work, an ensemble of neural networks is built by an algorithm called Learn++.NC and applied for fault detection and diagnosis. The algorithm is capable of incrementally learning new classes of faults, while retaining the previously acquired knowledge. The detection of new classes in subsequent data is achieved by thresholding the normalized weighted average of outputs (NWAO) of the neural networks in the ensemble. The unknown classes detected remain unlabeled until their correct labels can be assigned. The proposed method is applied to the identification of simulated faults in the feedwater system of a Boiling Water Reactor (BWR).

DOI

10.1142/9789814417747_0193

ISBN

9789814417730

Recommended Citation

Razavi-Far, Roozbeh; Baraldi, Piero; and Zio, Enrico. (2012). Ensemble of neural networks for detection and classification of faults in nuclear power systems. World Scientific Proc. Series on Computer Engineering and Information Science 7; Uncertainty Modeling in Knowledge Engineering and Decision Making - Proceedings of the 10th International FLINS Conf., 7, 1202-1207.
https://scholar.uwindsor.ca/electricalengpub/173