Intelligent systems applied on the estimation of bearing faults in inverter-fed induction motors

This paper proposes an approach based on intelligent systems for the classification and diagnosis of bearing fault evolution in inverter-fed induction motors, operating at steady state under a wide range of frequencies and load torque. Due to its robustness and low cost, induction motors are used in various industrial applications. In this work, the classifiers Fuzzy Artmap (FAM), Support Vector Machine - Sequential Minimal Optimization (SVM/SMO), k-Nearest Neighbours (k-NN) and Multilayer Perceptron (MLP) are used for the diagnosis and classification of bearing faults. Results obtained from 1173 experimental tests collected in the laboratory are presented to validate this proposal. The obtained results shows that this approach can accurately classify healthy and bearing defects in inverter-fed induction motors.