Back to Search Start Over

Optimal Fault Classification Using Fisher Discriminant Analysis in the Parity Space for Applications to NPPs.

Authors :
Cho, Sungwhan
Jiang, Jin
Source :
IEEE Transactions on Nuclear Science; Mar2018, Vol. 65 Issue 3, p856-865, 10p
Publication Year :
2018

Abstract

A parity space approach to monitoring and fault detection and identification of systems in nuclear power plants (NPPs) can be beneficial. However, if the number of fault classes exceeds the total independent residual signatures, the parity space method needs to be further enhanced to achieve the optimal fault classification. This situation happens frequently in NPP applications, where the safety and reliability are paramount. A possible enhancement proposed in this paper is to combine Fisher discriminant analysis with the parity space method to maximize the scatter among different fault classes, while minimizing the scatter within each class. Under identical conditions, the proposed technique can achieve optimal separation among different fault classes. Design, real-time implementation, and experimental evaluation of the proposed method are detailed in this paper. The implemented system has been validated on the Nuclear Power Control Test Facility to demonstrate the feasibility. The test results have revealed many salient features of the proposed method with potential applications in NPPs. [ABSTRACT FROM AUTHOR]

Details

Language :
English
ISSN :
00189499
Volume :
65
Issue :
3
Database :
Complementary Index
Journal :
IEEE Transactions on Nuclear Science
Publication Type :
Academic Journal
Accession number :
128484474
Full Text :
https://doi.org/10.1109/TNS.2018.2803658