Optimization on the Crosswind Stability of Trains Using Neural Network Surrogate Model

Abstract Under the influence of crosswinds, running safety trains will decrease sharply, so it is necessary to optimize suspension parameters trains. This paper studies dynamic performance high-speed under crosswind conditions, and optimizes train. A computational fluid dynamics simulation was used determine aerodynamic loads moments experienced by a series models train, with different were constructed, analyzed, metrics for these being determined. Finally, surrogate model built an optimization algorithm upon this model, find minimum possible values for: derailment coefficient, vertical wheel-rail contact force, wheel load reduction ratio, lateral force overturning coefficient. There 9 design variables, all associated bogie. When train speed 350 km/h, 15 m/s, benchmark performed poorly. The coefficient 1.31. 133.30 kN. rate 0.643. 85.67 kN, 0.425. After optimization, same 0.268, 100.44 0.474, 34.36 0.421, respectively. show that combining aerodynamics, vehicle system many-objective theory, train’s stability can be more comprehensively considered.