Dynamic Separation Model-Based Sliding Mode Control with Adaptive Neural Network Compensators for a Reluctance Actuator Motion System

The maglev technology has been recently used for advanced semiconductor equipment. The stringent accuracy requirement of the semiconductor manufacturing processes has posed new challenges about modeling and control of maglev systems (MLSs). This paper presents a new sliding mode control (SMC) scheme, named as SMCLFF, to tackle the impacts of inherent non-linearities caused by leakage and fringing fluxes (LFF), and external disturbances caused by the gap measurement mismatch (GMM) and non-orthogonal force (NOF) on the control of the MLS. A dynamic separation model (DSM) is designed to model the LFF effects in both the current–flux density (I–B) relationship and the flux density–force (B–F) relationship. The system is linearized by the DSM firstly, and the residual LFF effects and the external disturbances are suppressed by adaptive RBF neural networks (NNs) in SMCLFF respectively. The stability of the closed-loop control system was analyzed. Experiments were performed on a one-dimensional MLS plant. Results show that the DSM can effectively compensate for the LFF effects, and SMCLFF can enable the MLS to obtain high performance in a closed-loop control system.