Induction motor inter-turn fault detection using heuristic noninvasive approach by artificial neural network with Levenberg Marquardt algorithm

Induction motors are used extensively for various industrial applications. These motors have to endure and cope with a wide variety of environments and conditions, so with time gradually developing (incipient) faults build up. If the fault is not trace out in embryonic state, it leads degradation which eventually causes potential failure of the motors and huge economic loss in industry. Around two fifth incipient faults happen due to stator faults caused by mainly failure of inter-turn insulation. So this paper proposed an innovative idea to detect induction motor stator´s inter-turn short circuit fault using noninvasive heuristic approach by Artificial Neural Network (ANN). In this fault detection research, a 0.5 hp, single phase 50 Hz induction motor at no load condition is used as an experimental prototype. Generalized Feed forward neural network is used as NN with Levenberg Marquardt gradient descend algorithm for training. Three input parameters (motor current, supply voltage and rotor speed) develop the feature space of NN to detect whether the motor is faulty or not by analyzing performance function. By the experiment, the proposed system is found more efficient, reliable, economical and smart than the existing system.