In railway transportation, the evaluation of track irregularities is an indispensable requirement to ensure the safety and comfort of railway vehicles. A promising approach is to directly use vehicle dynamic responses to assess the impact of track irregularities. However, the computational cost of obtaining the dynamic response of the vehicle body using dynamics simulation methods is large. To this end, this study proposes a physics‐informed neural operator framework for vehicle–track spatially coupled dynamics (PINO‐VTSCD) calculation, which can effectively acquire the vehicle dynamic response. The backbone structure of PINO‐VTSCD is established by the branched Fourier neural operator, which features one branch for outputting car body responses and the other branch for estimating the responses of bogie frames, wheelsets, and rails. The relative L2 loss (rLSE) of PINO‐VTSCD under the optimal hyperparameter combination is 4.96%, which is 57% lower than the convolutional neural network‐gated recurrent unit model. Evaluation cases from large‐scale simulations and real‐world track irregularities show that the proposed framework can achieve fast solution in scenarios such as different wavelength‐depth combinations and different wavelength ranges. Compared with the traditional vehicle–track coupled model, the speedup of the PINO‐VTSCD model is 32x. The improved computational efficiency of the proposed model can support many railway engineering tasks that require repetitive calculations.

中文翻译：

---

用于高效计算车辆-轨道空间耦合动力学的分支傅里叶神经算子


在铁路运输中，对轨道不平顺的评估是确保铁路车辆安全和舒适不可或缺的要求。一种很有前途的方法是直接使用车辆动力学响应来评估轨道不规则性的影响。然而，使用动力学仿真方法获得车身动力学响应的计算成本很大。为此，本研究提出了一种用于车辆-轨道空间耦合动力学 （PINO-VTSCD） 计算的物理信息神经算子框架，该框架可以有效地获取车辆动力学响应。PINO-VTSCD 的骨架结构由分支傅里叶神经算子建立，该算子具有一个分支用于输出车身响应，另一个分支用于估计转向架框架、轮对和轨道的响应。在最佳超参数组合下，PINO-VTSCD 的相对 L2 损失 （rLSE） 为 4.96%，比卷积神经网络门控循环单元模型低 57%。来自大规模仿真和真实轨道不规则性的评估案例表明，所提框架可以在不同波长-深度组合和不同波长范围等场景下实现快速求解。与传统的车轨耦合模型相比，PINO-VTSCD 模型的加速提高了 32 倍。所提模型的改进计算效率可以支持许多需要重复计算的铁路工程任务。