Sensorless position control of skewed rotor induction machines based on multi saliency extraction

This paper addresses sensorless position and torque control of induction machines around zero frequency using inherent saliencies of a skewed rotor machine. The estimation method employs transient excitation of voltage pulses to detect machine saliencies. The resulting signals from injection based sensorless control methods are composed of components caused by saturation, slotting, as well as inter-modulation effects. As multiple saliencies cause problems in tracking only a particular saliency, decoupling of the different signal components is applied via neural network. This leads to the estimation of both the rotor angle as well as the flux angle, which are used in combination with the stator and the rotor equations. Identifying the disturbances of the different signal components it is possible to establish an adaptive control signal merging to reduce disturbance based torque pulsations. The performance of the control is investigated at loaded operation and zero speed as well as zero fundamental frequency.