Zero Speed Sensorless Control of Induction Machines Using Rotor Saliencies

Control of induction machines without encoder around zero frequency can only be achieved by evaluating inherent saliencies of the machine. The signals resulting from injection based sensorless control methods are composed of components caused by saturation, slotting, or anisotropy saliencies, as well as inter-modulation effects. In addition there is a strong influence of the load level on some components. Applying neural network technique and Fourier transform it is possible to almost eliminate the load dependence and to clearly reduce inter-modulation. An accurate and reliable separation of the different signal components is thus possible even at high load levels. Measurement results are given to show the performance of the arrangement allowing the determination of the rotor position at low and zero speed sensorless control.