Global Optimization of Medium Low-Speed Maglev Train-Bridge Dynamic System Based on Multi-Objective Evolutionary Algorithm

Medium low-speed maglev trains cause coupling vibration when moving over flexible bridges, which has a detrimental effect on the overall system. To effectively improve the global performance of the train-bridge system, this study proposes a parameter optimization approach that integrates a numerical model, a neural network, and a multi-objective evolutionary algorithm. A three-car maglev train-bridge coupling system is first modeled based on finite element, multi-body dynamics, and the levitation control theory. Based on this, the dynamic response and parameter sensitivity of the system is investigated using simulation analysis and the Sobol method. To enhance the optimization efficiency, an improved neural network is employed to simulate the nonlinear relationship between key parameters and dynamic performance, thereby surrogating the numerical model. The NSGA-III algorithm with a reference point mechanism is used to search for the optimal solution of the key parameters. Finally, simulation experiments verify the validity and accuracy of the neural network and the optimization results. This approach takes into account the coupling effect between multiple parameters and significantly enhances the computational efficiency compared with traditional rail transportation optimization methods. The dynamic response of the maglev system, considering the car-body flexibility, demonstrates that the optimization approach effectively improves the safety and stability of the train and further reduces the negative effect of the car-body’s elastic vibration on the operation quality.