Adaptive parameter learning and neural network control for uncertain permanent magnet linear synchronous motors

Abstract

This work aims to achieve accurate parameter estimation and intelligent control of a permanent magnet linear synchronous motor (PMLSM) with unknown parameters, nonlinearities, and disturbances. To address it, we propose a novel adaptive control method composed of a new estimation error-based learning strategy and a neural network (NN)-based adaptive sliding mode controller. An augmented NN with a friction model and a force ripple model is constructed to handle the uncertainties in PMLSM. The proposed learning law that contains leakage terms driven by estimation errors calculates the unknown parameter and NN’s weights online. Especially the gain of the discontinuous term in the learning law is adjusted by the presented update law to reduce the chattering of learned parameters. The controller handles the residual error and external disturbance. Different from the existing methods, the proposed one needs no boundary information of uncertainties in both the controller and parameter-learning strategy. The proposed method is finite-time semi-global uniformly and ultimately bounded (FTSGUUB), which is analyzed by designing a Lyapunov function. Finally, numerical simulations are carried out to validate the parameter learning and control accuracy of the proposed method.