Metamodel of the two-dimensional magnet-rail relationship based on a BP neural network

A finite-element simulation that solves the electromagnetic force of a maglev train when the magnetically permeable medium is saturated is very time-consuming. The existing analytical formula for the two-dimensional magnet-rail relationship assumes that the magnetic medium operates in the linear regime of the B-H curve, which does not apply in the case of magnetic saturation. To address these shortcomings, this study combines the uniform design method and BP neural-network modeling to obtain a metamodel that applies even when the magnetically permeable medium is saturated. The data fitting and prediction accuracy of the model are then assessed. The results show that the prediction accuracy of this model meets the requirements for system simulation and research on the dynamic characteristics of maglev trains.