The 'Neural' Phonetic Typewriter
Computer
Self-Organizing Maps
Application of Self-Organizing Networks to Signal Processing
Proceedings of the EURASIP Workshop 1990 on Neural Networks
Underwater target classification using wavelet packets and neural networks
IEEE Transactions on Neural Networks
RBFNN-based hole identification system in conducting plates
IEEE Transactions on Neural Networks
Hi-index | 0.00 |
A new classification method, for isolating steam generator tube defects in nuclear power plants using Eddy Current Test (ECT) signals, has been developed. The method uses Self-Organizing maps (SOM) with different data signatures to identify and classify these defects. A multiple inference system is proposed which evaluates different extracted characteristic SOMs to infer the defect type. Wavelet zero-crossing representation, a linear predictive coding (LPC), and other basic signal representations, such as magnitude and phase, are used to construct characteristic vectors that combine one or more of these features. These vectors are evaluated for their ability to classify tube defects and the ones with the best performance are used in the multiple inference system. The effectiveness of the method is demonstrated by applications of the characteristic maps to ECT data from various cases of tube defects in pressurized water reactor plant steam generators. The developed algorithm enables real-time applications such as fast tube defects classification systems and visualization of ECT signal feature prototypes, which may improve the speed of time-critical decision making during power plant maintenance outages.