Neural identification and control for linear induction motors

We propose a new adaptive control scheme, composed of a neural identifier and a nonlinear controller and applied it to a linear induction motor (LIM). In order to compare the performance of LIM, we use α - β and d - q models. A neural identifier of triangular form is proposed for both models as a nonlinear block controllable form (NBC). Then, a reduced order observer is designed in order to estimate no measured variables. Learning law for neural network weights ensure that the identification error converges to zero exponentially. Sliding mode control is developed to track velocity and flux magnitude. Simulations are presented to compare the behaviour of both models of LIM and the applicability of the proposed identification and control scheme.