A Global Bivalued-Observer for the Sensorless Induction Motor

Sensorless control of Induction Motors has received a lot of attention both from the applied as well as from the theoretical perspectives. In its more formal approach, a complete observability analysis of this machine, assuming the load torque is known, has been developed and several nonlinear observers have been reported in the literature. Since the system has indistinguishable trajectories, none of the existing observers is able to operate correctly under all operating conditions of the machine. In this paper we present a novel observer that is able to estimate all (non trivial) distinguishable or indistinguishable trajectories exhibited by the induction motor model under sensorless operation. The structure of the estimator comes from a detailed observability analysis under assumption that the load torque is unknown, that shows that the non trivial indistinguishable trajectories always appear in pairs, i.e. for each actual trajectory there exists at most another indistinguishable one. Hence we propose a Bi-valued observer which is always well-behaved and its convergence is guaranteed for any operation regime. The usefulness of the proposed scheme is validated by numerical simulations.