Unexpected disturbance and ever-changing passengers are unfavorable factors that always accompany maglev trains. If not considered or handled properly, they would deteriorate the control system performance significantly and even cause instability. This paper proposes a neural network-based adaptive control approach to stabilize the airgap of the nonlinear maglev vehicle. Meanwhile, the time-varying mass and external disturbance can be estimated accurately. Specifically, to ensure the asymptotic stability of the maglev system, a nonlinear basic control law is developed first. To tackle the uncertainty, a radial basis function neural network is fused into the basic controller, which can recover the unknown mass and disturbance more quickly and accurately. Lyapunov stability techniques are utilized to prove the stability of the whole maglev control system without any linear approximation. The sufficient comparative simulation results are provided to demonstrate that the established control scheme can obtain better levitation performance and achieve a precise estimation of time-varying and disturbance simultaneously. Finally, we build a dSPACE-based single electromagnet suspension test bed to examine its efficacy and practical applicability as well.

意外的干扰和不断变化的乘客是磁悬浮列车始终伴随的不利因素。如果考虑不当或处理不当，它们将严重降低控制系统的性能，甚至导致不稳定。本文提出了一种基于神经网络的自适应控制方法来稳定非线性磁悬浮飞行器的气隙。同时，可以准确估计随时间变化的质量和外部干扰。具体而言，为确保磁悬浮系统的渐近稳定性，首先开发了非线性基本控制律。为了解决不确定性，将径向基函数神经网络融合到基本控制器中，该神经网络可以更快速，准确地恢复未知质量和扰动。利用Lyapunov稳定性技术来证明整个磁悬浮控制系统的稳定性，而无需任何线性逼近。提供了足够的比较仿真结果，以证明所建立的控制方案可以获得更好的悬浮性能，并同时实现时变和扰动的精确估计。最后，我们建立了一个基于dSPACE的单电磁铁悬挂测试台，以检验其功效和实际​​适用性。