RBF Neural Network-Based Compound Control for High-Speed Maglev Trains
Aiming at the inherent instability, strong nonlinearity, and high dynamic characteristics of normal-conducting maglev suspension systems, this paper adopts a composite supervisory control scheme integrating PD control and an RBF neural network. First, a high-speed maglev train-track coupled dynamics model considering track elasticity is established. On this basis, a phased control strategy is designed: the initial phase employs a PD controller to ensure system stability, after which control is seamlessly handed over to an RBF neural network. The weights of this network are continuously refined online via a gradient descent algorithm, enabling progressive enhancement of control precision. Simulation results validate the effectiveness of this approach, confirming its superior performance in both precise suspension gap regulation and robust disturbance rejection. Consequently, the proposed method not only underpins the stable operation of maglev trains but also constitutes a reliable intelligent control framework for high-speed maglev systems.