SNN-Based Surrogate Modeling of Electromagnetic Force and Its Application in Maglev Vehicle Dynamics Simulation

Abstract

The majority of electromagnetic force calculation models employed in maglev vehicle system dynamics focus exclusively on vertical and lateral movement while neglecting the nonlinear magnetization properties of ferromagnetic materials. This oversight leads to discrepancies between the dynamics simulations and actual conditions. To enhance the accuracy of dynamics simulations and evaluate the performance of maglev vehicle systems under various operational conditions, it is imperative to identify an electromagnetic force calculation model that combines accuracy and applicability. To address this objective, this paper examines a U-shaped electromagnet in medium–low-speed maglev vehicles as a case study. It constructs a spatial electromagnetic force calculation surrogate model using a Shallow Neural Network. The surrogate model is capable of accurately calculating electromagnetic forces considering relative position deviations in the lateral, vertical, rolling, pitching, and shaking directions. Moreover, it can be integrated into vehicle system dynamics simulations. The accuracy of the electromagnetic force calculation surrogate model is confirmed by extensive comparisons with finite element simulation results across various conditions, achieving an impressive concordance rate of up to 95%. An illustrative application of the electromagnetic force calculation surrogate model in maglev vehicle system dynamics simulation is provided to showcase its practical utility.

Keywords

• electromagnetic levitation system
• electromagnetic force calculation
• nonlinear magnetization characteristics
• surrogate model
• maglev system dynamics simulation