Separation of saliency components for speed sensorless detection of flux and rotor position of induction machines

Sensorless field oriented control schemes of induction machines have been of high interest in recent years since the mechanical sensor considerably increases the cost of a drive and decreases its reliability. Methods which do not fail at zero fundamental frequency have to rely on non-fundamental effects of the motor. In the non fundamental wave model based mechanical sensorless control of AC machines the flux/rotor position can be estimated by evaluating the current response to voltage pulses. This current slope is modulated by all spatial saliencies in the machine. Sources of these saliencies can be various, for instance the saturation of the machine by the main flux, the slotting, or anisotropy, as well as inter-modulation effects. In this paper the separation of the saturation signal, the slotting signal, and the inter-modulation signal components in squirrel cage induction machines operating at low and zero frequency using artificial neural networks (ANN) has been experimentally implemented and measurement results are given to show the advantages of the application of the proposed technique at low and zero speed sensorless control even at high load levels.