LEARNING CONTROL OF CURRENT-FED INDUCTION MOTOR WITH MECHANICAL UNCERTAINTIES M. Montanari * A. Tilli * * Center for Research on Complex Automated Systems “G. Evangelisti” (CASY) Dept. of Electronics, Computer Science and Systems (DEIS) University of Bologna Viale Risorgimento 2, 40136 Bologna, Italy e–mail: {mmontanari, atilli}@deis.unibo.it Abstract: In this paper repetitive learning control technique has been applied to the position/flux tracking control of an Induction Motor (IM) under hypothesis of periodic reference trajectory and uncertainties on the mechanical model. The electro-magnetic IM model has been directly taken into account in the control development. Indirect Field Oriented approach has been exploited and combined with control actions derived from Lyapunov-like design. In order to compensate the periodic disturbances, the model of a generic periodic signal with known period has been embedded in the controller with a suitable update rule. The convergence properties of the overall solution proposed have been formally proven. Simulation results confirm the validity of the approach presented. Copyright 2005 IFAC Keywords: Induction motor, learning control, position control, system uncertainties, tracking error convergence 1. INTRODUCTION Periodic reference trajectories and mechanical uncer- tainties cause unknown periodic disturbances in servo- drives dynamics. The Internal Model Principle represents the basic idea to solve the control problem of asymptotic tracking under condition of unknown disturbances/trajectories with known dynamic model (exosystem), without us- ing high gain/large bandwidth approaches. The Repet- itive Learning Control (RLC) can be interpreted as a formalization of the above-mentioned principle in case of generic periodic references/disturbances with known period. In particular, as reported in (Hara et al., 1988), the adopted internal model is a closed-loop time-delay system with delay T (in the continuous- time framework) which is able to generate any peri- odic signal with period T. The RLC approach (or similar solutions as Betterment Learning Control (Arimoto et al., 1984) or Iterative Learning Control (Moore, 1999), (Ham et al., 2001), (Xu and Tan, 2002)) has been widely used in ro- botic applications to cope with mechanical uncer- tainties leading to periodic disturbances (Horowitz et al., 1991), (Dixon et al., 2001). Nevertheless, the elec- tromagnetic dynamics of the adopted servo-drive has been usually neglected. In this paper, the case of Induction Motor (IM) servo drives with mechanical periodic disturbances is con- sidered. The IM electromagnetic dynamics has been taken into account in the control design and the stan- dard Indirect Field Oriented (IFO) solution has been adopted as starting point (see for instance (Taylor, 1994), (Novotny and Lipo, 1996), (Leonhard, 2001) as references on speed/flux control techniques for IM). A position controller designed according to back- stepping and robust control techniques has been de-