Evaluation of Hysteresis Loss Curve on 3 Phase Induction Motor by Using Cascade Feed Forward Neural Network

Hysteresis loss under rotating stator flux density and magnetizing current conditions is one of the factors causing decreased performance of induction motor (IM). Unfortunately, the limited air gap space makes it costly to install additional instrument. Moreover, the non-linearity and uncertainty of these material properties cause the stator flux density and magnetizing current properties to be difficult to evaluate when using direct measurement technique. Based on this problem, this paper discusses a novel non-destructive method for evaluating hysteresis loss on IM. We recorded the speed, voltage, and current signal from IM supplied by SVPWM voltage source inverter. This data was then transformed to become stator magnetizing current model for determining the magnetic field intensity. The multilayer perceptron Cascade Feed Forward Neural Network (CFNN) estimated the stator flux density from data observer. By using this method, stator magnetizing current and the stator flux density could be plotted as hysteresis loss curve in the direct axis and quadrature axis including its magnitude and phase angle. This method is powerful and effective because it is a non-destructive method because it needs no changes to the motor construction nor any additional instrument, and it can be analyzed for a non-linearity function. The experiment result shows that the CFNN estimator response is satisfied with the accurate estimation which has been compared and verified.