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

Abstract

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

Keywords

• Safety
• Surrogate model
• Optimization
• High-speed train
• Crosswind