Multi-Kernel Neural Network Sliding Mode Control for Permanent Magnet Linear Synchronous Motors

Abstract

In this paper, a multi-kernel neural network sliding mode control (MNNSMC) method is proposed to enhance the position tracking performance and disturbance suppression capability of permanent magnet linear synchronous motors (PMLSMs). The designed MNNSMC strategy consists of two parts, a sliding mode control with a dynamic boundary layer and a multi-kernel neural network (MNN). In the former, the dynamic boundary layer is utilized to guarantee that the sliding mode variable converges to the sliding manifold asymptotically. In the latter, disturbances are introduced into the design of kernel functions to further approximate and compensate for the internal and external disturbances, such as parameter variations, positioning force, friction, and un-modeled nonlinearities of PMLSMs. The stability of the proposed MNNSMC strategy is analyzed and proved based on Lyapunov Theorem. Experiments are conducted on a PMLSM servo drive system to validate the effectiveness of the presented MNNSMC method, and results show that the proposed scheme has significant improvement on the position tracking performance, disturbance suppression capability, and robustness of the PMLSM system compared with an existing method.