Alternator Rotor Inter-turn Short-Circuit Identification using FEM Based Learning

Alexandre Bacchus (1, 2) , Melisande Biet (1) , Ludovic Macaire (3, 2) , Abdelmounaim Tounzi (1) , Yvonnick Le Menach (1)

1 L2EP - Laboratoire d’Électrotechnique et d’Électronique de Puissance - ULR 2697
2 CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189
3 LAGIS-SI

The goal of this paper is to identify rotor inter-turn short-circuits of an alternator. The detection method is based on the analysis of a flux probe signal located in the air gap of the machine. Previous works have shown that pattern recognition can be applied to detect such a fault by using the experimental data as prototypes. A new method is developed here by considering a learning step based on simulation. Therefore the machine is modeled and validated in that purpose. A feature selection is made by considering feature correlation and disparity. Finally, k nearest neighbors is used to classified experimental test data.

Mots clés: data simulation; diagnostic; Classification; synchronous machine; generator

Domaines: Informatique [cs]

Alexandre Bacchus, Melisande Biet, Ludovic Macaire, Abdelmounaim Tounzi, Yvonnick Le Menach. Alternator Rotor Inter-turn Short-Circuit Identification using FEM Based Learning. SAFEPROCESS 2015 - 9th \{IFAC\} Symposium on Fault Detection, Supervision and Safety for Technical Processes , Sep 2015, paris, France. pp.1432 - 1437, ⟨10.1016/j.ifacol.2015.09.726⟩. ⟨hal-01242421⟩