An Efficient Architecture of Multi-Stage Neural Network for Wound-Rotor Induction Generator Short-Circuit Fault Classification

Abstract : The aim of this paper is to show the efficiency of the time-domain analysis in electrical machines fault diagnosis due to early short-circuits detection in both stator and rotor windings. The contribution is based on a multi-stage artificial neural network which has been shown to be more efficient than a single network due to the problem complexity. This new method is based on the time-domain analysis of digital data directly coming from sensors without any computation but with a dedicated pre-processing process for data and a postprocessing technique for both faults detection and localization. After, its presentation the proposed technique has been applied to a wound rotor induction generator for which sixteen nondestructive windings faults have been analyzed. This new technique has been tested on real data showing clearly its efficiency in term of training time and of errors compared to already proposed techniques.
Type de document :
Communication dans un congrès
XXth International Conference on Electrical Machines (ICEM), Sep 2012, Marseille, France. pp.1565-1571, 2012
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01083099
Contributeur : Laurent Capocchi <>
Soumis le : samedi 15 novembre 2014 - 13:05:30
Dernière modification le : lundi 21 mars 2016 - 17:29:34

Identifiants

  • HAL Id : hal-01083099, version 1

Collections

Citation

Samuel Toma, Laurent Capocchi, Gerard-André Capolino. An Efficient Architecture of Multi-Stage Neural Network for Wound-Rotor Induction Generator Short-Circuit Fault Classification. XXth International Conference on Electrical Machines (ICEM), Sep 2012, Marseille, France. pp.1565-1571, 2012. 〈hal-01083099〉

Partager

Métriques

Consultations de la notice

207