梯度引导的高速列车表面传感器布局优化和稀疏压力场重建策略

李桐; 邹俊泓; 孙振旭; 雷震; 朱翔昱

doi:10.7638/kqdlxxb-2025.0255

欢迎来到能源电力期刊网！

English Version 期刊登录

您当前的位置：

首页 >

文章列表页 >

梯度引导的高速列车表面传感器布局优化和稀疏压力场重建策略

研究论文 | 更新时间：2026-10-14

- 梯度引导的高速列车表面传感器布局优化和稀疏压力场重建策略
- G-PROSNet: Gradient-guided pressure-field reconstruction with optimal sensor placement for high-speed trains
- 空气动力学学报 2026年44卷第8期页码：20-28
- 作者机构：
  
  1. 中国科学院自动化研究所,北京,100190
  2. 中国科学院大学前沿交叉科学学院,北京,101408
  3. 中国科学院力学研究所,北京,100190
  4. 中国科学院大学人工智能学院,北京,101408
- 作者简介：
  
  朱翔昱*，研究员，研究方向：深度学习、生成式神经网络、AI4Science（空气动力学）. E-mail：xiangyu.zhu@ia.ac.cn
- 基金信息：
  
  新一代人工智能国家科技重大专项（2025ZD0122000）
- DOI：10.7638/kqdlxxb-2025.0255
  中图分类号： U270.1;TP183
- 纸质出版：2026
- 稿件说明：
移动端阅览
李桐, 邹俊泓, 孙振旭, 等. 梯度引导的高速列车表面传感器布局优化和稀疏压力场重建策略[J]. 空气动力学学报, 2026, 44(8): 20−28.

Li Tong, Zou Junhong, Sun Zhenxu, et al. G-PROSNet: Gradient-guided pressure-field reconstruction with optimal sensor placement for high-speed trains[J]. 2026, 44(8): 20-28.
李桐, 邹俊泓, 孙振旭, 等. 梯度引导的高速列车表面传感器布局优化和稀疏压力场重建策略[J]. 空气动力学学报, 2026, 44(8): 20−28. DOI： 10.7638/kqdlxxb-2025.0255

Li Tong, Zou Junhong, Sun Zhenxu, et al. G-PROSNet: Gradient-guided pressure-field reconstruction with optimal sensor placement for high-speed trains[J]. 2026, 44(8): 20-28. DOI： 10.7638/kqdlxxb-2025.0255.

摘要

准确获取高速列车表面压力场的精确分布，是开展空气动力学外形优化设计的重要数据基础。但受限于车体结构的物理限制，在实际测试中仅能通过布置少量的传感器来获取稀疏压力数据。为此，本文提出一种名为G-PROSNet的梯度引导的传感器布局优化与稀疏压力场重建方法。该方法在深度学习框架中引入基于点云邻域的Laplacian正则化，以强化压力场的空间平滑性与物理一致性。此外，通过两阶段训练策略实现数据驱动与物理约束的协同优化：第一阶段进行预训练，学习压力场稀疏至稠密的映射规律；第二阶段结合可微节点选择模块与梯度约束，实现传感器布局与重建性能的联合提升。在CRH380A高速列车压力数据集上，该方法在RMSE与L2RE指标上分别达到1.18×10

–4

和0.73×10

–3

，较基线模型PROSNet分别降低约4.8%和3.9%。结果表明，该方法在复杂工况下仍能保持高精度与物理合理性，为建立高速列车气动测试与稠密压力场获取之间的桥梁提供了有效方案。

Abstract

Accurate knowledge of the surface pressure distribution is essential for the aerodynamic design and optimization of high-speed trains. However

due to the physical constraints imposed by the train structure

only sparse pressure measurements can be obtained in practical tests. To address this limitation

this study proposes a gradient-guided framework for sensor layout optimization and sparse pressure-field reconstruction. The method integrates a Laplacian regularization based on point-cloud neighborhoods within a deep learning model

to enhance spatial smoothness and physical consistency of the reconstructed pressure field. A two-stage training strategy is employed to achieve collaborative optimization of data-driven learning and physical constraints. In the first stage

the network is pretrained to learn the mapping from sparse sensor measurements to dense pressure fields. In the second stage

a differentiable node-selection module is combined with gradient-based constraints to jointly optimize the sensor configuration and reconstruction performance. Experiments conducted on the CRH380A high-speed train dataset demonstrate that the proposed approach achieves RMSE and L2RE values of 1.18×10

–4

and 0.73×10

–3

representing improvements of approximately 4.8% and 3.9% over the PROSNet baseline. The results indicate that the method maintains high reconstruction accuracy and physical fidelity under complex operating conditions

effectively bridging the gap between sparse aerodynamic testing and dense pressure-field acquisition

and providing a reliable tool for aerodynamic analysis and sensor deployment planning.

关键词

Keywords

references

Yu Yizheng, Liu Tanghong, Xia Yutao, et al. Development and prospect of aerodynamic drag-reduction technologies for trains at higher speed (400+ km/h)[J]. Acta Aerodynamica Sinica, 2021, 39(5): 83−94, 82 (in Chinese). doi: 10.7638/kqdlxxb-2021.0272

Su Xinchao, Wang Jiabin, Xiong Xiaohui, et al. Aerodynamic layout optimization of high-speed train aerodynamic braking based on physical partitioning[J]. Journal of Railway Science and Engineering, 2025: 1−13 (in Chinese). doi: 10.19713/j.cnki.43-1423/u.T20250354

Wang T T, Yang L, Wang Y, et al. Research on aerodynamic drag reduction for 400 km/h high-speed trains[J]. Advances in Wind Engineering, 2025, 2(2): 100056. doi: 10.1016/j.awe.2025.100056

Gao H R, Liu T H, Chen X D, et al. An accurate and efficient methodology on wind spectra relative to moving trains: field measurements of wind characteristics in complex terrains[J]. Stochastic Environmental Research and Risk Assessment, 2025, 39(4): 1423−1453. doi: 10.1007/s00477-025-02924-2

Liu T H, Chen Z W, Guo Z J, et al. Reasonable pressure tap layout to measure the aerodynamic forces of a train at different yaw angles[J]. Measurement, 2020, 166: 108255. doi: 10.1016/j.measurement.2020.108255

hang Q, Zhao X, Li K, et al. Prediction model of aircraft hinge moment: compressed sensing based on proper orthogonal decomposition[J]. Physics of Fluids, 2024, 36(7): 076102. doi: 10.1063/5.0214653

Zhao X, Du L, Peng X H, et al. Research on refined reconstruction method of airfoil pressure based on compressed sensing[J]. Theoretical and Applied Mechanics Letters, 2021, 11(2): 100223. doi: 10.1016/j.taml.2021.100223

Donoho D L. Compressed sensing[J]. IEEE Transactions on Information Theory, 2006, 52(4): 1289−1306. doi: 10.1109/TIT.2006.871582

Bright I, Lin G, Kutz J N. Compressive sensing based machine learning strategy for characterizing the flow around a cylinder with limited pressure measurements[J]. Physics of Fluids, 2013, 25(12): 127102. doi: 10.1063/1.4836815

Manohar K, Brunton B W, Kutz J N, et al. Data-driven sparse sensor placement for reconstruction[J]. IEEE Control Systems Magazine, 2018, 38(3): 63−86. doi: 10.1109/MCS.2018.2810460

Bai Z, Wimalajeewa T, Berger Z, et al. Low-dimensional approach for reconstruction of airfoil data via compressive sensing [J]. AIAA Journal, 2014, 53(4): 920−933. doi: 10.2514/1.J053287

Qiu W, Sun Z X, Wang J X, et al. A compressed sensing based framework for surface pressure field reconstruction from sparse measurement[J]. Physics of Fluids, 2025, 37(4): 045144. doi: 10.1063/5.0264594

Li Z Y, Han W, Zhang Y, et al. Learning spatiotemporal dynamics with a pretrained generative model[J]. Nature Machine Intelligence, 2024, 6(12): 1566−1579. doi: 10.1038/s42256-024-00938-z

Fukami K, Fukagata K, Taira K. Super-resolution reconstruction of turbulent flows with machine learning[PP]. 2018. arXiv: 1811.11328 [2025−11−04]. doi: 10.48550/arXiv.1811.11328

Wang H B, Wang S B, Zhuang D Y, et al. Surrogate-based aerodynamic shape optimization of high-speed train heads: a review of four key technologies[J]. Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 2024, 238(8): 907−920. doi: 10.1177/09544097241251509

Duthé G, Abdallah I, Barber S, et al. Graph neural networks for aerodynamic flow reconstruction from sparse sensing[PP]. 2023. arXiv: 2301.03228 [2025−11−04]. doi: 10.48550/arXiv.2301.03228

Li Y F, Chang J T, Kong C, et al. Recent progress of machine learning in flow modeling and active flow control[J]. Chinese Journal of Aeronautics, 2022, 35(4): 14−44. doi: 10.1016/j.cja.2021.07.027

Yu C D, Bi X J, Fan Y W. Deep learning for fluid velocity field estimation: a review[J]. Ocean Engineering, 2023, 271: 113693. doi: 10.1016/j.oceaneng.2023.113693

Raissi M, Perdikaris P, Karniadakis G E. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations[J]. Journal of Computational Physics, 2019, 378: 686−707. doi: 10.1016/j.jcp.2018.10.045

Fan J H, Duan W Y, Huang L M, et al. High-fidelity flow field reconstruction model for incompressible fluid with physical constraints [J]. Ocean Engineering, 2023, 280: 114597. doi: 10.1016/j.oceaneng.2023.114597

Xu S F, Sun Z X, Huang R F, et al. A practical approach to flow field reconstruction with sparse or incomplete data through physics informed neural network [J]. Acta Mechanica Sinica, 2023, 39(3): 322302. doi: 10.1007/s10409-022-22302-x

Lai B Z, Liu Y Z, Wen X. Temporal and spatial flow field reconstruction from low-resolution PIV data and pressure probes using physics-informed neural networks [J]. Measurement Science and Technology, 2024, 35(6): 065304. doi: 10.1088/1361-6501/ad3307

Cai S Z, Mao Z P, Wang Z C, et al. Physics-informed neural networks (PINNs) for fluid mechanics: a review[PP]. 2021. arXiv: 2105.09506 [2025−10−24]. doi: 10.48550/arXiv.2105.09506

Zou J H, Qiu W, Sun Z X, et al. Joint optimization of sensor placement and sparse pressure field reconstruction with a two-stage framework for limited data [J]. Physics of Fluids, 2025, 37(7): 075231. doi: 10.1063/5.0278268

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于POD和KAN的三维球水滴收集系数快速预测

基于DDIM混合加速采样的飞行器布局生成式设计方法

关键区域优化对高速列车横风气动特性影响

基于U-Net的磁浮飞行风洞气动力干扰剥离算法

基于条件扩散模型的火箭气动外形快速设计方法