Skip to Main Content
This study develops an adaptive supervisory Gaussian-cerebellar model articulation controller (ASGCMAC) and implements it in a sensor-less direct torque control (DTC) system for controlling induction motor speed. The inherent uncertainties of a DTC system, including parametric uncertainties and the uncertainties in external load torque, make control tasks difficult. A model-free approach, ASGCMAC, is adopted to build a high-performance DTC induction motor drive. The proposed method comprises two parts - a supervisory controller and a Gaussian-CMAC (GCMAC) subsystem. The supervisory controller monitors the control process to maintain the tracking error within a pre-defined range; the GCMAC sub-system learns and approximates system dynamics. The parameters of ASGCMAC are adjusted online according to adaptive rules, which are derived from Lyapunov stability theory, guarantee system stability. Four control schemes, ASGCMAC, supervisory controller, proportional-integral (PI) control and conventional CMAC, are experimentally investigated and the performance index, root-mean-square error (RMSE), is evaluated in each scheme. The results reveal that ASGCMAC outperforms the other comparison schemes. In addition, the robustness of the proposed scheme to the parameter variation and external load torque disturbance has been verified via simulation and experiments.