Automatic and robust diagnosis of broken rotor bars fault in induction motor

A non invasive method to detect and diagnosis automatically in an early stage the broken bars fault in the rotor of the induction motor is presented. This method is based on monitoring suitable features by a feedforward Multi Layer Perceptron Neural Network. These features are extracted from the spectral component of the residual “d” current. A diagnostic robustness towards parametric variations (temperature, magnetic state) is realised by the use of a parametric identification technique. The simulated results show the efficiency of this method. A series of data collected, from a 1.1 kW three phase induction motor performed with different loads and different rotor faults, are used to verify and validate experimentally the performance of this method. The obtained experimental results prove the effectiveness of the proposed method.