SVM-Based System For Broken Rotor Bar Detection In Induction Motors

C. M. PEZZANI; J. M. FONTANA; P. D. DONOLO; C. H. DE ANGELO; G. R. BOSSIO; L. SILVA

Cali

Conferencia; IEEE Andean Conference (ANDESCON 2018); 2018

IEEE

In this paper we propose a system for the detection of broken bars (BB) in induction motors (IM) based on a binary SVM classifier. Statistical features obtained from a IM current are used as inputs of the classifier. We train the classifier with simulation data, using a qd model of IM. Through simulation, 360 records of 120 cycles were generated in which different IM conditions were considered, healthy, rotor fault and disturbances in torque and voltage. Finally, the classifier was validated with experimental laboratory data, obtained from an IM of 5.5 kW in healthy, 2 and 3 BB, with different load and supply voltage conditions. The results show that the proposed classifier allows detection with 91.67% accuracy, and without false negatives (faulty IMs classified as heathy).