图神经网络移动/变形网格非定常流场多步推进与未见边界运动及网格的适应性:振荡圆柱、俯仰翼型、往复运动物体、双向流固耦合
判定框架与检索逻辑:物理边界运动、网格随动变形、重网格与自适应加密的区分
承接两份初始化分组的“研究范围与判定框架”设想,作为全文分类坐标与检索说明,不绑定文献。明确四条区分轴:(1) 物理边界运动 vs 网格随边界变形 vs 重网格导致连接关系变化 vs 固定边界下自适应加密;(2) 预测对象(流场、流体载荷、网格位置/加密方案);(3) 推进模式(一步预测、真值逐步校正、仅由初始场出发的自由多步);(4) 测试时未来边界运动与网格的来源(是否取自参考CFD的未来网格)。并说明以 zhao2026ale、lange2026spatio 为起点的前序/后序追踪路径与文献纳排标准。
移动/变形网格上GNN时空推进的主线:显式边界运动信息、ALE一致更新与多步误差累积
两份分组在此处高度重叠,合并为主线段:均为在非结构/动网格上用图神经网络(或其算子化变体)做非定常流场时空推进,输入逐帧节点坐标、连接关系与壁面/网格运动信息,关注自由多步推进的误差累积与未见运动轨迹/网格的泛化。dynamic_mesh_boundary_aware_temporal_graph_neural_operator(lange2026spatio 起点)与 a_mesh_based_graph_neural_network_approach(zhao2026ale 方向的ALE动网格GNN)为两条追踪起点;a_physics_informed_graph_neural_network_method 提供物理约束版对照;spatio_temporal_graph_neural_network_for_multi_step 给出入水动边界算例的多步预测、插值/外推泛化与消融(是否显式加入运动信息)。
- Dynamic-Mesh Boundary-Aware Temporal Graph Neural Operator for Complex Flow-Field Prediction in Moving-Boundary Systems(Jin-Heng Yang, Ying-Kun Li, Shao-Peng Zhang, Ming Zhu, Ze-Min Zhao, 2026, Aerospace Science and Technology)
- A mesh-based Graph Neural Network approach for surrogate modeling of Lagrangian free surface fluid flows(Federico Lanteri, M. Cremonesi, 2025, Computers & Fluids)
- A Physics-Informed Graph Neural Network Method for Unsteady Flow Field Prediction(Haowen Tian, Junzhe Cao, 2025, ACM Cloud and Autonomic Computing Conference)
- Spatio-temporal graph neural network for multi-step prediction of cylinder water entry flow fields(Chongbin Shi, Gui-Yong Zhang, Teizhi Sun, Zi-Hao Wang, Heng Wang, 2025, The Physics of Fluids)
面向变几何与跨网格泛化的GNN代理:旋转不变输入、ghost cell时变边界条件与物理损失
合并“几何/网格泛化型GNN代理”与初始化分组中的泛化评测意图,但与主线区分:重点不在ALE推进,而在训练集外几何/网格的可迁移性与时变边界条件的注入方式。含变几何叶轮机非定常/时均场预测、非结构网格跨网格泛化、旋转不变输入设计、ghost cell施加时变边界条件以降低对求解器初值依赖,以及CO2运移中引入质量守恒、相平衡与Darcy渗流损失提升未见地质构型泛化。为“网格泛化评测协议”和‘边界条件如何进入模型’提供对照来源。
- Prediction of Steady and Unsteady Flow Quantities Using Multiscale Graph Neural Networks(Sebastian Strönisch, Maximilian Sander, M. Meyer, Andreas Knüpfer, 2024, Volume 12D: Turbomachinery — Multidisciplinary Design Approaches, Optimization, and Uncertainty Quantification; Radial Turbomachinery Aerodynamics; Unsteady Flows in Turbomachinery)
- Multi-scale hydraulic graph neural networks for flood modelling(R. Bentivoglio, E. Isufi, S. Jonkman, R. Taormina, 2025, Natural Hazards and Earth System Sciences)
- Advancing Graph Neural Networks for CO2 Plume Migration in Complex Geological Formations(Adriano M. A. Côrtes, R. Velho, Rodrigo de S. Luna, Thiago H. N. Coelho, R. Elias, A. Evsukoff, H. Gross, Mauricio Araya, F. Rochinha, A. Coutinho, 2026, Ibero-Latin American Congress on Computational Methods in Engineering (CILAMCE))
非图时序深度模型基线:振荡翼型与振荡圆柱的ConvLSTM与混合网络
保留为独立对照组,是判断“收益来自图结构还是时序建模本身”的必要对照。以卷积LSTM及动力学特征增强变体、混合深度网络在振荡翼型/振荡圆柱等移动边界算例上做时空预测,并在相同算例、相似输入长度下比较预测与泛化误差;同时充当俯仰翼型/振荡圆柱代表算例的数据与指标(力与力矩幅值/相位、可用推进时长)来源。
- Novel convolutional long short-term memory networks for spatiotemporal prediction of unsteady flow with moving boundary(Rong Han, Xiao-Liang Yang, Wei Liu, 2026, The Physics of Fluids)
- Hybrid deep neural network based prediction method for unsteady flows with moving boundary(Ren-Kun Han, Zhong Zhang, Yi-Xing Wang, Ziyang Liu, Yang Zhang, Gang Chen, 2021, Acta Mechanica Sinica)
双向流固耦合:网格—流场—结构联合演化与闭环稳定性
两份分组仅在一侧出现,但属任务定义上的独立方向(结构运动由流体力反馈决定而非外部给定)。评价重点从单步精度转向闭环稳定性、载荷幅值与相位、长时间耦合推进能力;同时用于区分“给定结构运动下的预测”与“载荷反馈后的闭环验证”,避免把任务难度差异误判为方法收益。含GNN代理与超图神经网络代理两支。本批文献对该方向的覆盖有限,需后续单独追踪补充。
- A Graph Neural Network Based Surrogate Model for Flapping-Driven Ocean Energy Harvesting(Aarshana R. Parekh, Rui Gao, R. Jaiman, 2025, Volume 4: CFD, FSI and AI)
- Hypergraph Neural Network Surrogate Modeling for Prediction of Vortex- and Wake-Induced Vibrations(Shayan Heydari, Rui Gao, R. Jaiman, 2025, Volume 4: CFD, FSI and AI)
固定网格时空GNN与物理约束注入:多尺度、等变与物理引导节点特征
合并初始化分组中“多尺度/等变/高效GNN架构”与“固定网格时空GNN与物理约束”两组:共同点是网格静止或未显式移动,聚焦GNN架构与物理先验本身——多尺度图粗化与消息传递、旋转等变、因果自注意力、GCN/GAT-LSTM、物理引导节点特征、物理约束消息传递与Roe格式增强、多处理器与混合精度训练。它们不回答移动网格问题,但提供可迁移训练组件,并构成“普通图更新 vs 物理/ALE一致更新”和“GNN vs 非图模型”的架构基线。
- Multi-scale rotation-equivariant graph neural networks for unsteady Eulerian fluid dynamics(M. Lino, Stathi Fotiadis, A. Bharath, C. Cantwell, 2022, The Physics of Fluids)
- Towards Accelerating Physics Informed Graph Neural Network for Fluid Simulation(Yi-Di Wang, F. Guan, Malcolm Yoke Hean Low, Zheng-Kui Wang, Aik Beng Ng, Simon See, 2024, SIGGRAPH Asia 2024 Posters)
- MeshGIN: a graph neural network framework for transonic aerodynamic predictions(Saad Hussain, Yang Pei, Bagh Hussain, Fahad Nawaz, 2026, Journal of the Brazilian Society of Mechanical Sciences and Engineering)
- Physics-guided node features for mesh-native graph neural network prediction of two-dimensional laminar flow around arbitrary shapes(Parham Takyar, Vahid Bazargan, 2026, Results in Engineering)
- Spatiotemporal reconstruction of unsteady bridge flow field via hierarchical graph neural networks with causal attention(C.Z. Cai, Jun Xiao, Yun-Feng Zou, Xu-Hui He, 2025, The Physics of Fluids)
- A Spatio-Temporal Graph Neural Network for Predicting Flow Fields on Unstructured Grids with the SUBOFF Benchmark(Wei Guo, Cheng Cheng, C. Huang, Zhiqing Lu, Kang Chen, Jun Ding, 2025, Journal of Marine Science and Engineering)
- Physics-constrained and flow-field-message-informed graph neural network for solving unsteady compressible flows(Si-Ye Li, Zhen-Sheng Sun, Yujie Zhu, Chi Zhang, 2024, The Physics of Fluids)
- ROE-Scheme Physics-Augmented Graph Neural Networks in Solving Eulerian and Laminar Flow Incompressible NS Equations(Shangxia Song, Longxiang Jiang, Liyuan Wang, Xinkun Chu, Hao Zhang, 2025, 应用数学和力学)
学习驱动的网格生成、重网格、平滑与自适应加密(网格位置与加密方案预测)
合并两份分组中同属“预测对象为网格本身而非流场”的工作:初始网格生成、基于Hessian度量的各向异性自适应加密、结构网格光顺与顶点质量参数、ALE模拟中网格松弛参数的学习式替代,以及GNN增强的虚拟元/有限体积混合以嵌入守恒约束。它们回答“重网格后新网格从何而来、场量如何迁移守恒”的前置问题,并界定了“网格随动变形 vs 重网格 vs 固定边界自适应加密”的任务边界,提供跨分辨率/拓扑训练的对照。
- Generalization-Aware Structured Mesh Smoothing via Graph Neural Networks(Jingzhu Pei, Xin-Hai Chen, Yang Liu, Zhi-Chao Wang, Qingyang Zhang, Qinglin Wang, Xiang Gao, Jie Liu, 2025, Computers & Fluids)
- Graph Neural Networks for Mesh Generation and Adaptation in Structural and Fluid Mechanics(Ugo Pelissier, A. Parret-Fréaud, F. Bordeu, Y. Mesri, 2024, Mathematics)
- A deep learning framework for mesh relaxation in arbitrary Lagrangian-Eulerian simulations(Ming Jiang, Brian Gallagher, N. Mandell, A. Maguire, Keith W. Henderson, G. Weinert, 2019, No journal)
- A Vertex Quality Parameter Moving Mesh Method(J. Felcman, P. Kubera, T. Simos, G. Psihoyios, Ch. Tsitouras, 2007, No journal)
- An intelligent mesh-smoothing method with graph neural networks(Zhi-Chao Wang, Xin-Hai Chen, Junjun Yan, Jie Liu, 2025, Frontiers of Information Technology & Electronic Engineering)
- Learning-Enhanced Finite Volume Methods for Nonlinear Convection-Diffusion Problems(Mustafa Alasadi, 2026, Dijlah Journal of Engineering Sciences ISSN: 3078-9664, e-ISSN: 3078-9656)
- Graph neural network enhanced virtual element method: a novel framework for improved stability(Ru-Yi Zhou, Xing-Yu Hu, Zhi-Xin Zeng, Xiong Zhang, 2026, Computational Mechanics)
ALE描述与几何守恒律(GCL)的数值理论基准
两份分组此部分文献高度重合,合并为一组纯数值基准(不含机器学习):满足GCL的有限元格式、移动壁面条件与相对网格输运、高阶有限差分的度量评估与对称守恒度量、以及GCL重要性的经典讨论。为判断GNN更新是否ALE一致、移动壁面条件残差是否收敛、重网格插值是否守恒提供判据;需强调GCL为必要非充分条件,不能替代跨网格插值守恒核查。
- On the Geometric Conservation Law for Unsteady Flow Simulations on Moving Mesh(R. Ma, X. Chang, Lai-Ping Zhang, Xin He, Ming Li, 2015, Procedia Engineering)
- Geometric conservation laws for flow problems with moving boundaries and deformable meshes, and their impact on aeroelastic computations(M. Lesoinne, C. Farhat, 1996, Computer Methods in Applied Mechanics and Engineering)
- Geometric conservation laws for aeroelastic computations using unstructured dynamic meshes(M. Lesoinne, C. Farhat, 1995, No journal)
- High-order ALE method for the Navier–Stokes equations on a moving hybrid unstructured mesh using flux reconstruction method(Kai Liu, Yi Lu, Chang-Fu You, 2013, International Journal of Computational Fluid Dynamics)
- Geometric Conservation Law in ALE Formulations(L. Garelli, R. Paz, Mario Alberto Storti, 2009, No journal)
- A finite element formulation satisfying the discrete geometric conservation law based on averaged Jacobians(Mario Alberto Storti, L. Garelli, R. Paz, 2012, International Journal for Numerical Methods in Fluids)
- Metric evaluation for higher-order finite difference scheme with geometric conservation law on three-dimensional moving and deforming mesh(2011, No journal)
- Symmetric-conservative metric evaluations for higher-order finite difference scheme with the GCL identities on three-dimensional moving and deforming mesh(Yoshiaki Abe, N. Iizuka, T. Nonomura, K. Fujii, 2012, No journal)
- On the significance of the geometric conservation law for flow computations on moving meshes(H. Guillard, C. Farhat, 2000, Computer Methods in Applied Mechanics and Engineering)
高保真动网格数值方法与往复运动代表算例(真值与数据来源)
保留为独立的数据/真值生成侧:ALE移动网格生成与高阶格式、二维Riemann求解器两相移动网格格式(含节点速度与守恒处理)、移动网格WENO、基于移动网格的格式,以及液舱晃荡等往复/振荡边界典型算例。它们提供连续流场、逐帧网格与边界运动轨迹,是核查“测试时未来网格是否取自参考CFD”、“力/力矩幅值与相位”与“可用多步推进时长”的关键数据来源,也与学习式网格生成组形成“真值—学习”对照。
- ALE moving mesh generation and high performance implementation using OpenMP and MPI Libraries for FSI and Darcy flow problems(Masroor Hussain, 2011, No journal)
- AnALE-basedNumericalSchemewith a Two-DimensionalRiemann Solver for Two-Phase Flows with Non-Equilibrium Pressure(Jian-Qiao Zhang, W. Yan, Xiang-Gui Li, 2026, Communications in Mathematical Research)
- A Moving Mesh WENO Method for One-Dimensional Conservation Laws(Xiao-Bo Yang, Wei-Zhang Huang, Jian-Xian Qiu, 2012, SIAM Journal on Scientific Computing)
- A Moving Mesh Method Based on the Geometric Conservation Law(Wei-Ming Cao, Wei-Zhang Huang, R. Russell, 2002, SIAM Journal on Scientific Computing)
- Numerical Simulation of Liquid Tank Sloshing Based on Graph Neural Network with Self-Attention Mechanism(Wen-Kang Zhang, Xiao-Feng Sun, Yiping Zhong, Hai-Tao Wei, 2025, International Conference on Computing and Artificial Intelligence)
泛化与归因审计:跨轨迹/跨网格实验设计的统一核查表
合并两份分组的“泛化证据审计”与“归因审计”为一节方法学审计,覆盖全部分组文献而不新增文献:(1) 运动泛化——是否固定几何与雷诺数、留出整套振幅/频率/轨迹,属训练范围内插、范围外外推还是微调迁移;(2) 网格泛化——是否保持物理运动相同而仅改布点、分辨率或拓扑;(3) 归因——仅瞬时坐标 vs 显式加入壁面/网格速度、普通图更新 vs ALE一致更新、同等输入下GNN vs 非图模型,并核对初始场、运动信息、网格、时间步与计算成本是否一致;(4) 双向FSI区分给定运动预测与载荷反馈闭环验证。是全文结论的核心判据。
证据缺口与优先实验设计:未见轨迹自由多步推进、跨拓扑迁移与运动信息消融
全文收束节,不新增文献。基于各分组核查结果总结:已有工作是否在整套未见运动轨迹上验证自由多步推进(而非真值校正)、跨网格测试是否只改变布点/分辨率/拓扑、受控对照能否把收益归于图结构与明确运动信息;评估振荡圆柱、俯仰翼型、往复运动物体等代表算例是否开放连续流场、逐帧网格、边界运动、载荷与可运行代码;据此提出优先实验:未见轨迹上的自由多步推进基准、跨拓扑网格零样本迁移与微调对照、以及“显式运动信息/ALE一致性约束”的必要性消融,并标注双向FSI闭环稳定性证据缺失。
合并后形成“判定框架—主线—对照—前置—基准—数据—审计”的统一结构。主线为动/变形网格上GNN的时空推进与显式边界运动信息、ALE一致更新(以 zhao2026ale、lange2026spatio 为追踪起点,含物理约束版与入水动边界多步算例);相邻对照为变几何/跨网格泛化的GNN代理(旋转不变输入、ghost cell时变边界条件、物理损失)与非图时序模型基线(振荡翼型/振荡圆柱的ConvLSTM与混合网络);相邻方向为双向流固耦合的联合演化与闭环稳定性;可迁移组件与架构基线为固定网格时空GNN(多尺度、等变、因果注意力、物理约束消息传递、Roe增强)。前置问题由学习式网格生成/重网格/平滑/自适应加密与GNN增强的虚拟元-有限体积混合界定;数值理论基准由ALE-GCL文献提供(GCL为必要非充分条件);高保真动网格格式与液舱晃荡等往复运动算例提供连续流场、逐帧网格与边界运动真值。最后以统一核查表审计三点:是否在整套未见运动轨迹上验证多步推进、跨网格是否只改布点/分辨率/拓扑、受控对照能否归因于图结构/显式运动信息/一致性约束,并指出双向FSI闭环稳定性与移动边界GNN推进两处证据缺口及优先实验设计。
承接两份初始化分组的“研究范围与判定框架”设想,作为全文分类坐标与检索说明,不绑定文献。明确四条区分轴:(1) 物理边界运动 vs 网格随边界变形 vs 重网格导致连接关系变化 vs 固定边界下自适应加密;(2) 预测对象(流场、流体载荷、网格位置/加密方案);(3) 推进模式(一步预测、真值逐步校正、仅由初始场出发的自由多步);(4) 测试时未来边界运动与网格的来源(是否取自参考CFD的未来网格)。并说明以 zhao2026ale、lange2026spatio 为起点的前序/后序追踪路径与文献纳排标准。
两份分组在此处高度重叠,合并为主线段:均为在非结构/动网格上用图神经网络(或其算子化变体)做非定常流场时空推进,输入逐帧节点坐标、连接关系与壁面/网格运动信息,关注自由多步推进的误差累积与未见运动轨迹/网格的泛化。dynamic_mesh_boundary_aware_temporal_graph_neural_operator(lange2026spatio 起点)与 a_mesh_based_graph_neural_network_approach(zhao2026ale 方向的ALE动网格GNN)为两条追踪起点;a_physics_informed_graph_neural_network_method 提供物理约束版对照;spatio_temporal_graph_neural_network_for_multi_step 给出入水动边界算例的多步预测、插值/外推泛化与消融(是否显式加入运动信息)。相关文献:Jin-Heng Yang et. al, 2026 等 4 篇文献
合并“几何/网格泛化型GNN代理”与初始化分组中的泛化评测意图,但与主线区分:重点不在ALE推进,而在训练集外几何/网格的可迁移性与时变边界条件的注入方式。含变几何叶轮机非定常/时均场预测、非结构网格跨网格泛化、旋转不变输入设计、ghost cell施加时变边界条件以降低对求解器初值依赖,以及CO2运移中引入质量守恒、相平衡与Darcy渗流损失提升未见地质构型泛化。为“网格泛化评测协议”和‘边界条件如何进入模型’提供对照来源。相关文献:Sebastian Strönisch et. al, 2024 等 3 篇文献
保留为独立对照组,是判断“收益来自图结构还是时序建模本身”的必要对照。以卷积LSTM及动力学特征增强变体、混合深度网络在振荡翼型/振荡圆柱等移动边界算例上做时空预测,并在相同算例、相似输入长度下比较预测与泛化误差;同时充当俯仰翼型/振荡圆柱代表算例的数据与指标(力与力矩幅值/相位、可用推进时长)来源。相关文献:Rong Han et. al, 2026 等 2 篇文献
两份分组仅在一侧出现,但属任务定义上的独立方向(结构运动由流体力反馈决定而非外部给定)。评价重点从单步精度转向闭环稳定性、载荷幅值与相位、长时间耦合推进能力;同时用于区分“给定结构运动下的预测”与“载荷反馈后的闭环验证”,避免把任务难度差异误判为方法收益。含GNN代理与超图神经网络代理两支。本批文献对该方向的覆盖有限,需后续单独追踪补充。相关文献:Aarshana R. Parekh et. al, 2025 等 2 篇文献
合并初始化分组中“多尺度/等变/高效GNN架构”与“固定网格时空GNN与物理约束”两组:共同点是网格静止或未显式移动,聚焦GNN架构与物理先验本身——多尺度图粗化与消息传递、旋转等变、因果自注意力、GCN/GAT-LSTM、物理引导节点特征、物理约束消息传递与Roe格式增强、多处理器与混合精度训练。它们不回答移动网格问题,但提供可迁移训练组件,并构成“普通图更新 vs 物理/ALE一致更新”和“GNN vs 非图模型”的架构基线。相关文献:M. Lino et. al, 2022 等 8 篇文献
合并两份分组中同属“预测对象为网格本身而非流场”的工作:初始网格生成、基于Hessian度量的各向异性自适应加密、结构网格光顺与顶点质量参数、ALE模拟中网格松弛参数的学习式替代,以及GNN增强的虚拟元/有限体积混合以嵌入守恒约束。它们回答“重网格后新网格从何而来、场量如何迁移守恒”的前置问题,并界定了“网格随动变形 vs 重网格 vs 固定边界自适应加密”的任务边界,提供跨分辨率/拓扑训练的对照。相关文献:Jingzhu Pei et. al, 2025 等 7 篇文献
两份分组此部分文献高度重合,合并为一组纯数值基准(不含机器学习):满足GCL的有限元格式、移动壁面条件与相对网格输运、高阶有限差分的度量评估与对称守恒度量、以及GCL重要性的经典讨论。为判断GNN更新是否ALE一致、移动壁面条件残差是否收敛、重网格插值是否守恒提供判据;需强调GCL为必要非充分条件,不能替代跨网格插值守恒核查。相关文献:R. Ma et. al, 2015 等 9 篇文献
保留为独立的数据/真值生成侧:ALE移动网格生成与高阶格式、二维Riemann求解器两相移动网格格式(含节点速度与守恒处理)、移动网格WENO、基于移动网格的格式,以及液舱晃荡等往复/振荡边界典型算例。它们提供连续流场、逐帧网格与边界运动轨迹,是核查“测试时未来网格是否取自参考CFD”、“力/力矩幅值与相位”与“可用多步推进时长”的关键数据来源,也与学习式网格生成组形成“真值—学习”对照。相关文献:Masroor Hussain et. al, 2011 等 5 篇文献
合并两份分组的“泛化证据审计”与“归因审计”为一节方法学审计,覆盖全部分组文献而不新增文献:(1) 运动泛化——是否固定几何与雷诺数、留出整套振幅/频率/轨迹,属训练范围内插、范围外外推还是微调迁移;(2) 网格泛化——是否保持物理运动相同而仅改布点、分辨率或拓扑;(3) 归因——仅瞬时坐标 vs 显式加入壁面/网格速度、普通图更新 vs ALE一致更新、同等输入下GNN vs 非图模型,并核对初始场、运动信息、网格、时间步与计算成本是否一致;(4) 双向FSI区分给定运动预测与载荷反馈闭环验证。是全文结论的核心判据。
全文收束节,不新增文献。基于各分组核查结果总结:已有工作是否在整套未见运动轨迹上验证自由多步推进(而非真值校正)、跨网格测试是否只改变布点/分辨率/拓扑、受控对照能否把收益归于图结构与明确运动信息;评估振荡圆柱、俯仰翼型、往复运动物体等代表算例是否开放连续流场、逐帧网格、边界运动、载荷与可运行代码;据此提出优先实验:未见轨迹上的自由多步推进基准、跨拓扑网格零样本迁移与微调对照、以及“显式运动信息/ALE一致性约束”的必要性消融,并标注双向FSI闭环稳定性证据缺失。
总计 44 篇相关文献
No abstract available
With the increasing use of deep neural networks as surrogate models for accelerating computational simulations in mechanics, the application of artificial intelligence in computational fluid dynamics has seen renewed interest in recent years. However, the application of deep neural networks for flow simulations has mainly concentrated on relatively simple cases of incompressible flows. The strongly discontinuous structures that appear in compressible flows dominated by convection, such as shock waves, introduce significant challenges when approximating the nonlinear solutions or governing equations. In this work, we propose a novel physics-constrained, flow-field-message-informed (FFMI) graph neural network for spatiotemporal flow simulations of compressible flows involving strong discontinuities. To enhance the nonlinear approximation capability of strong discontinuities, a shock detector method is leveraged to extract the local flow-field messages. These messages are embedded into the graph representation to resolve the discontinuous solutions accurately. A new FFMI sample-and-aggregate-based message-passing layer, which aggregates the edge-weighted attributes with node features on different hop layers, is then developed to diffuse and process the flow-field messages. Furthermore, an end-to-end paradigm is established within the encoder–decoder framework to transform the extracted information from the flow field into latent knowledge about the underlying fluid mechanics. Finally, a variety of one- and two-dimensional cases involving strong shock waves are considered to demonstrate the effectiveness and generalizability of the proposed FFMI graph neural network.
When calculating the transient flow around a bridge structure, traditional computational fluid dynamics methods are extremely time-consuming, especially for multiparameter optimization analyses. Inspired by the development of deep graph neural networks with a mesh structure, this paper describes a spatiotemporal prediction framework for the rapid reconstruction and prediction of transient flows on large-scale unstructured grids. To ensure stability and reliability during self-supervised training, a causal self-attention mechanism is employed in the temporal model. The framework is trained and tested on a dataset containing 40 000 snapshots of bridge flow fields with Reynolds numbers ranging from 104 to 105. The relative mean square error of the model in predicting the velocity and pressure fields is found to be in the order of 10−3 and the relative error does not exceed 10%. This demonstrates that the model is capable of reconstructing high-dimensional flow field information from low-dimensional data. Furthermore, the proposed model achieves a computational speedup by two orders of magnitude compared with traditional computational fluid dynamics methods with respect to the temporal inference. To validate its ability to infer bridge aerodynamic characteristics, the model is used to predict the bridge surface pressure, aerodynamic coefficients, streamlines, and vorticity. The results demonstrate that the proposed model has reliable accuracy, representation, and stability in predicting bridge flow fields and identifying multiscale characteristics.
No abstract available
No abstract available
Cantilevered elastic foils or panels can undergo self-induced large-amplitude flapping oscillations when submerged in a flowing fluid. This canonical phenomenon of fluid-structure interaction, commonly observed in nature, such as in fluttering leaves or fish fins, offers a promising mechanism for ocean energy harvesting. By leveraging the repetitive flapping motion, elastic foils can harness the abundant kinetic energy present in marine environments, including steady ocean currents and unsteady wave-driven flows. The efficiency of energy harvested by these foils is dictated by the amplitude and frequency of the oscillation, which are influenced by interactions with the surrounding flow. Optimizing these parameters is crucial for maximizing energy conversion efficiency and designing robust systems. However, high-fidelity simulations required for optimization over a wide parameter space are computationally expensive. To address this challenge, we propose a graph neural network-based surrogate model (GNN-ROM) tailored for the inverted foil problem. This model effectively handles mesh-structured data from full-order simulations without modifications, enabling efficient and accurate fluid-structure interaction simulations. The inverted foil is modeled as an elastically mounted rigid foil that undergoes a pitching motion around its trailing edge in uniform flow. The coupled fluid-structure system is simulated using a high-fidelity Petrov-Galerkin finite element approach, leveraging the arbitrary Lagrangian-Eulerian (ALE) formulation for precise interface tracking. The high-fidelity simulations serve as ground truth data for the training and evaluation of neural networks. Our deep learning model utilizes a rotation equivariant, quasi-monolithic GNN architecture based on the ALE formulation, wherein the coupled system dynamics are predicted with two sub-networks. First, essential coefficients describing mesh motion are extracted by proper orthogonal decomposition, which are predicted over time using a single multilayer perceptron. Moreover, the GNN-ROM evolves the flow field according to the state of the system. This approach effectively simulates coupled FSI dynamics, offering a robust surrogate model for rapid parametric optimization and control of energy harvesting devices based on inverted foil configurations.
No abstract available
This study introduces a Spatio-Temporal Graph Neural Network (ST-GNN) for the multi-step prediction of highly dynamic cylinder water entry flow fields. The proposed model integrates a Graph Neural Network (GNN) to capture complex spatial dependencies on unstructured meshes with a Long Short-Term Memory (LSTM) network to model temporal evolution. We generate a dataset from numerical simulations across different cases, varying both Froude numbers and initial entry angles. Extensive ablation studies are conducted to systematically optimize the model's architecture and hyperparameters, revealing that a single-step prediction strategy with an input sequence of eight timesteps yields the most robust performance, significantly reducing error accumulation in long-term prediction. The optimized model achieves high accuracy (R2>0.97) and successfully predicts key physical phenomena during water entry. Furthermore, generalization tests on unseen cases demonstrate the model's excellent interpolation capability within the training data range and reveal the challenges of extrapolation, particularly for the complex phase interface. Crucially, while the multi-step prediction of the velocity field degrades due to error accumulation, the inferred prediction of the water volume fraction remains remarkably stable. This indicates that the model has learned not just the instantaneous spatial correlations between the fields, but more importantly, has captured the underlying temporal dynamics that govern the evolution of the phase interface.
No abstract available
Accurate simulations of unsteady flow with a moving boundary demand sophisticated numerical schemes and dynamic mesh techniques, which incur high computational costs and remain a challenge in engineering applications. To mitigate this realistic challenge, this paper introduces two novel convolutional long short-term memory (ConvLSTM) networks: (1) K-ConvLSTM (kinetic-convolutional LSTM network), which integrates the kinetic features of the unsteady flow, and (2) EK-ConvLSTM (enhanced kinetic-convolutional LSTM network), which enables further interaction between the kinetic module and the original structure of the vanilla ConvLSTM. Furthermore, we construct a spatiotemporal prediction architecture consisting of the proposed neural networks for unsteady flow prediction. In addition, we utilize two representative unsteady cases, the flow around the oscillating airfoil and the oscillating circular cylinder, to evaluate the predictive accuracy and generalization ability of the proposed models. The numerical results show that K-ConvLSTM is effective when the input length is less than eight. Moreover, compared to the baseline ConvLSTM method, the EK-ConvLSTM model significantly reduces predictive and generalized errors, regardless of the input length.
No abstract available
No abstract available
The simulation of fluid dynamics, typically by numerically solving partial differential equations, is an essential tool in many areas of science and engineering. However, the high computational cost can limit application in practice and may prohibit exploring large parameter spaces. Recent deep-learning approaches have demonstrated the potential to yield surrogate models for the simulation of fluid dynamics. While such models exhibit lower accuracy in comparison, their low runtime makes them appealing for design-space exploration. We introduce two novel graph neural network (GNN) models, multi-scale GNN (MuS-GNN) and rotation-equivariant multi-scale GNN (REMuS-GNN), for extrapolating the time evolution of fluid flow. In both models, previous states are processed through multiple coarsenings of the graph, which enables faster information propagation through the network and improves the capture and forecast of the system state, particularly in problems encompassing phenomena spanning a range of length scales. Additionally, REMuS-GNN is architecturally equivariant to rotations, which allows the network to learn the underlying physics more efficiently, leading to improved accuracy and generalisation. We analyse these models using two canonical fluid models: advection and incompressible fluid dynamics. Our results show that the proposed GNN models can generalise from uniform advection fields to high-gradient fields on complex domains. The multi-scale graph architecture allows for inference of incompressible Navier-Stokes solutions, within a range of Reynolds numbers and design parameters, more effectively than a baseline single-scale GNN. Simulations obtained with MuS-GNN and REMuS-GNN are between two and four orders of magnitude faster than the numerical solutions on which they were trained.
The Arbitrary Lagrangian-Eulerian (ALE) method is used in a variety of engineering and scientific applications for enabling multi-physics simulations. Unfortunately, the ALE method can suffer from failures that require users to adjust a set of parameters to control mesh relaxation. In this paper, we present a deep learning framework for predicting mesh relaxation in ALE simulations. Our framework is designed to train a neural network using data generated from existing ALE simulations developed by expert users. In order to capture the spatial coherence inherent in simulations, we apply convolutional-deconvolutional neural networks to achieve up to 0.99 F1 score in predicting mesh relaxation.
This study introduces a novel physics-informed spatiotemporal graph neural network (PIST-GNN) for rapid, physically consistent, and accurate long-term prediction of complex water entry flow fields. The model integrates a MeshGraphNet-based GNN for spatial feature extraction on unstructured meshes and a long short-term memory (LSTM) network for temporal evolution, critically enhanced by embedding mass conservation and volume of fluid (VOF) transport equations as soft constraints during training. Comprehensive evaluations demonstrate that the combined physics constraints provide an optimal balance, achieving robust predictions with only a slight increase in data loss while significantly reducing physical inconsistencies like velocity divergence and VOF advection residuals. The PIST-GNN exhibits superior long-term stability and generalization capabilities, consistently maintaining the lowest prediction errors and the best physical consistency over 30 autoregressive steps on unseen Froude number and initial angle cases. Furthermore, through both quantitative and qualitative analysis of prediction results, we elucidate the complex influence of initial conditions on prediction accuracy. Ultimately, this study provides a practical and feasible approach to address the persistent challenge of error accumulation in long-term predictions of water entry flow fields by purely data-driven models. Future research will focus on overcoming the limitations of the current two-dimensional slice approximation method by optimizing physical constraints and model architecture, thereby more directly capturing the complete three-dimensional dynamic characteristics of the water entry flow field.
To overcome the limitations of traditional convolutional and recurrent neural networks in capturing spatio-temporal dynamics in flow fields on unstructured grids, this study proposes a novel Spatio-Temporal Graph Neural Network (ST-GNN) model that integrates a Graph Neural Network (GNN) with a Long Short-Term Memory (LSTM) network. The GNN component captures spatial dependencies among irregular grid nodes via message passing, while the LSTM component models temporal evolution through gated memory mechanisms. This hybrid framework enables the joint learning of spatial and temporal features in complex flow systems. Two variants of ST-GNN, namely, GCN-LSTM and GAT-LSTM, were developed and evaluated using the SUBOFF AFF-8 benchmark dataset. The results show that GAT-LSTM achieved higher accuracy than GCN-LSTM, with average relative errors of 2.51% for velocity and 1.43% for pressure at the 1000th time step. Both models achieved substantial speedups over traditional CFD solvers, with GCN-LSTM and GAT-LSTM accelerating predictions by approximately 350 and 150 times, respectively. These findings position ST-GNN as an efficient and accurate alternative for spatio-temporal flow modeling on unstructured grids, advancing data-driven fluid dynamics.
This study employs a finite-element inspired hypergraph neural network surrogate model to investigate the coupled dynamics of a freely oscillating cylinder in fluid flow, a well-established benchmark problem in vortex-induced and wake-induced vibration research. The presented methodology integrates high-fidelity numerical simulations based on variational finite-element methods with graph neural network surrogate modeling to forecast the spatio-temporal dynamics of flow-induced oscillations. The surrogate model utilizes graph-based representations to capture intricate fluid-structure interactions (FSIs), incorporating a multi-layer perceptron for mesh displacement updates and a hypergraph neural network for fluid-state predictions. Quantitative analyses demonstrate that the model could effectively learn from high-fidelity simulation data and provide stable and accurate predictions on unseen test data. These results underscore the potential of graph neural network-based methods for predicting fluid-structure interactions. This research supports the development of physics-based digital twins in marine and offshore engineering applications.
Abstract When using the dynamic mesh method to deal with moving boundary problems, the Geometric Conservation Law (GCL) must be considered carefully. This paper reviews the study on GCL problem of dynamic grid in finite volume framework based on Arbitrary Lagrangian-Eulerian (ALE) methods. Several common approaches to satisfy discretized geometric conservation law (DGCL) are studied and simplified to a uniform form. The uniform flow and the isentropic vortex are tested to validate the geometric conservation property and the temporal-accuracy. Numerical results illustrate that although the violating of GCL does not pollute the original time-accuracy of numerical schemes for the governing equations, it may introduce extra artificial errors. When the “Adding Source Term” methods are adopted, the temporal accuracy order of face velocity scheme must match the temporal accuracy order of numerical schemes for the governing equations, otherwise it will pollute the original temporal accuracy.
No abstract available
No abstract available
No abstract available
No abstract available
No abstract available
No abstract available
No abstract available
No abstract available
No abstract available
No abstract available
No abstract available
No abstract available
In this work, a homogeneous two-phase flow model in two-dimensional space is proposedfor systemswithout a sharp interface between phases. The model allows the two phases to coexist within the same domain with distinct pressures, while their local proportions are represented by volume fractions. Within an arbitrary Lagrangian-Eulerian (ALE) framework, a two-dimensional Riemann solver is developed to construct the numerical scheme with novel formulations for numerical fluxes and nodal velocities, ensuring consistency on moving control volumes. Furthermore, a new moving-mesh strategy is designed, in which the evolution of the volume fraction is obtained directly from the geometric motion of the mesh and mass conservation, thereby avoiding solving the non-conservative convective equation with source terms. This enables the incorporation of physical drag modelsand finite relaxation coefficients, allowing the gradual equilibration of velocity and pressure between the two phases to be calculated. Based on the volume-fraction formulation, a discrete representation of the pressure work term is derived. Finally, a series of numerical experiments are presented to demonstrate the accuracy, stability, and physical consistency of the proposed model and numerical scheme.
Fluid simulations in Computational Fluid Dynamics (CFD) often involve solving complex Partial Differential Equations (PDEs) with varying parameters, leading to high computational costs and slow inference times. Graph neural networks (GNNs) offer a promising alternative by efficiently representing mesh spaces and reducing these costs. In this work, we introduce a pioneering Physics-Informed Graph Neural Network (PIGNN) approach, utilizing a multi-GNN processor architecture that assigns each output parameter to a dedicated GNN processor with individual decoders. This innovative method reduces the training epochs for PIGNN by half while preserving accuracy. Additionally, we adopted mixed-precision training and achieved further improvements in reducing the training time to one-quarter of the original network while maintaining VRAM usage.
The finite element discretization of computational physics problems frequently involves the manual generation of an initial mesh and the application of adaptive mesh refinement (AMR). This approach is employed to selectively enhance the accuracy of resolution in regions that encompass significant features throughout the simulation process. In this paper, we introduce Adaptnet, a Graph Neural Networks (GNNs) framework for learning mesh generation and adaptation. The model is composed of two GNNs: the first one, Meshnet, learns mesh parameters commonly used in open-source mesh generators, to generate an initial mesh from a Computer Aided Design (CAD) file; while the second one, Graphnet, learns mesh-based simulations to predict the components of an Hessian-based metric to perform anisotropic mesh adaptation. Our approach is tested on structural (Deforming plate–Linear elasticity) and fluid mechanics (Flow around cylinders–steady-state Stokes) problems. Our findings demonstrate the model’s ability to precisely predict the dynamics of the system and adapt the mesh as needed. The adaptability of the model enables learning resolution-independent mesh-based simulations during training, allowing it to scale effectively to more intricate state spaces during inference.
No abstract available
No abstract available
In computational fluid dynamics (CFD), mesh-smoothing methods are widely used to refine the mesh quality for achieving high-precision numerical simulations. Specifically, optimization-based smoothing is used for high-quality mesh smoothing, but it incurs significant computational overhead. Pioneer works have improved its smoothing efficiency by adopting supervised learning to learn smoothing methods from high-quality meshes. However, they pose difficulties in smoothing the mesh nodes with varying degrees and require data augmentation to address the node input sequence problem. Additionally, the required labeled high-quality meshes further limit the applicability of the proposed method. In this paper, we present graph-based smoothing mesh net (GMSNet), a lightweight neural network model for intelligent mesh smoothing. GMSNet adopts graph neural networks (GNNs) to extract features of the node’s neighbors and outputs the optimal node position. During smoothing, we also introduce a fault-tolerance mechanism to prevent GMSNet from generating negative volume elements. With a lightweight model, GMSNet can effectively smooth mesh nodes with varying degrees and remain unaffected by the order of input data. A novel loss function, MetricLoss, is developed to eliminate the need for high-quality meshes, which provides stable and rapid convergence during training. We compare GMSNet with commonly used mesh-smoothing methods on two-dimensional (2D) triangle meshes. Experimental results show that GMSNet achieves outstanding mesh-smoothing performances with 5% of the model parameters compared to the previous model, but offers a speedup of 13.56 times over the optimization-based smoothing.
No abstract available
No abstract available
Data-driven surrogates are increasingly used to accelerate simulations of unsteady flows with moving boundaries, yet their suitability for long-horizon prediction remains an open question. This work compares two representative learning approaches: motion-conditioned latent dynamical models based on recurrent neural networks and motion-conditioned operator-learning approaches based on Fourier Neural Operators enhanced with autoregressive consistency and physics-based temporal regularization (AR-FNO). Using a heaving square cylinder as a canonical test case, we assess short-horizon reconstruction accuracy, error growth during autoregressive rollout, and spatial flow-field behaviour. A floating wind-turbine configuration is included as a supplementary case to assess performance in a more complex flow environment. Both models achieve comparable short-horizon accuracy. During extended autoregressive rollout, the ConvLSTM exhibits progressive phase drift and amplitude distortion, whereas the AR-FNO maintains improved phase consistency and lower cumulative RMSE. In the square-cylinder case, the AR-FNO reduces the long-horizon cumulative error growth by approximately 15 - 20 % relative to the latent-dynamical surrogate, thereby demonstrating improved performance.
Abstract. Deep-learning-based surrogate models represent a powerful alternative to numerical models for speeding up flood mapping while preserving accuracy. In particular, solutions based on hydraulic-based graph neural networks (SWE-GNNs) enable transferability to domains not used for training and allow the inclusion of physical constraints. However, these models are limited due to four main aspects. First, they cannot model rapid differences in flow propagation speeds; secondly, they can face instabilities during training when using a large number of layers, needed for effective modelling; third, they cannot accommodate time-varying boundary conditions; and fourth, they require initial conditions from a numerical solver. To address these issues, we propose a multi-scale hydraulic-based graph neural network (mSWE-GNN) that models the flood at different resolutions and propagation speeds. We include time-varying boundary conditions via ghost cells, which enforce the solution at the domain’s boundary and drop the need for a numerical solver for the initial conditions. To improve generalization over unseen meshes and reduce the data demand, we use invariance principles and make the inputs independent from coordinates' rotations. Numerical results applied to dike-breach floods show that the model predicts the full spatio-temporal simulation of the flood over unseen irregular meshes, topographies, and time-varying boundary conditions, with mean absolute errors in time of 0.05 m for water depths and 0.003 m2 s−1 for unit discharges. We further corroborate the mSWE-GNN in a realistic case study in the Netherlands and show generalization capabilities with only one fine-tuning sample, with mean absolute errors of 0.12 m for water depth, a critical success index for a water depth threshold of 0.05 m of 87.68 %, and speed-ups of over 700 times. Overall, the approach opens up several avenues for probabilistic analyses of realistic configurations and flood scenarios.
Carbon Capture and Storage (CCS) is essential for mitigating climate change by permanently storing industrial CO2 emissions in subsurface formations [1]. As net-zero targets gain urgency, accurately predicting injected CO2 behavior over long timeframes becomes critical for ensuring storage integrity and regulatory compliance. Simulating CO2 plume migration presents significant computational challenges. Conventional numerical simulators, while accurate, require prohibitive computational resources when modeling the centuries-long timescales relevant to CO2 storage, especially for uncertainty quantification or optimization studies requiring multiple simulations. Deep learning surrogate models offer orders-of-magnitude speedup while maintaining reasonable accuracy [2]. Current literature features two main surrogate modeling approaches: Fourier Neural Operators (FNO) [2] and Graph Neural Networks (GNN) [3]. FNO models perform well on structured domains but are limited to Cartesian grids, inadequately representing complex geological features. In contrast, GNN approaches like MeshGraphNet (MGN) [4] can naturally handle unstructured meshes, better representing heterogeneous subsurface environments and geological boundaries critical to CO2 migration. We present our MGN-LSTM implementation for temporal prediction of CO2 plume migration in faulted reservoirs [3], incorporating algorithmic improvements for enhanced prediction stability and efficiency. Our key contribution is integrating physics-aware loss terms from FNO literature [5] into the GNN framework. By incorporating mass conservation, phase behavior, and Darcy flow principles directly into the loss function, we achieve physically consistent predictions even with limited training data. These physics-aware constraints improve model generalization to unseen geological configurations. References [1] IPCC. Special Report on Carbon Dioxide Capture and Storage. CUP, 2005. [2] Wen, G. et al. Towards a predictor for CO2 plume migration using deep neural networks. Int. J. Greenhouse Gas Control, 2021. [3] Ju, X et al Learning CO2 plume migration in faulted reservoirs with Graph Neural Networks. Computers & Geosciences, 193, 2024. [4] Pfaff, T. et al. Learning mesh-based simulation with graph networks. ICML, 2021. [5] Badawi, D., Gildin, E. Neural operator-based proxy for reservoir simulations considering varying well settings, locations, and permeability fields. J. Petroleum Science and Engineering, 2023.
Background: The non-linear convection-diffusion equation is one of the cornerstones of computational fluid dynamics and applied mathematics, particularly in applications such as environmental modeling, heat transfer or reactive flows. The finite volume method (FVM) is known for its good conservation property and robustness, but usually has difficulty in accurately simulating the sharp gradients and discontinuities without high computational cost. PINNs also provide a mesh-free data-driven technique however it may suffer from the slow convergence rate and poor generalization. To address these shortcomings, in this manuscript we introduce Learning-Enhanced Finite Volume Methods (LE-FVM), a new class of approach that bridges the scientific computation and machine learning communities using advanced deep learning models — including PINNs, FVGNs, and KANs — into an FVM framework. Methods: We have developed the physics-constrained hybridization of machine learning models assigned directly within a FVM context to improve flux determination, support in situ adaptive mesh refinement optimization and end-to-end solvers constrained by integral conservation laws. Important contributions include twice-message aggregation in FVGN for irregular mesh, and adaptive loss weighting using the Neural Tangent Kernel to achieve a good balance between PDE residuals and boundary conditions. The methodology was extensively tested on standard 1D and 2D nonlinear convection-diffusion-reaction test-cases (Burgers’, Fisher’s, Burgers–Huxley, Newell–Whitehead–Segel) employing the metrics of error norms (L₁, L₂, L∞), statistical tests (Wilcoxon signed-rank coefficient variation) and generalization tests on unseen geometries. Results: The LE-FVM framework achieved a 95%+ success rate in delivering high-fidelity solutions across all benchmarks. For the 1D Burgers’ equation at Re=1, PINN-based LE-FVM reduced the maximum absolute error by a factor of 173 and the RMSE by 60x compared to GFEM. In 2D unsteady flow simulations, FVGN demonstrated a 77% improvement in prediction accuracy for velocity fields and a 56% reduction in training time compared to purely data-driven graph networks. The framework also exhibited superior generalization, accurately predicting flow around unseen elliptical and airfoil geometries without retraining. Mean residual errors for LE-FVM solutions were consistently on the order of 10⁻⁴ to 10⁻⁵, outperforming traditional methods by one to two orders of magnitude. Conclusions: The proposed LE-FVM framework successfully bridges the gap between physical fidelity and data-driven adaptability, offering a robust, efficient, and highly accurate solver for complex nonlinear systems. By embedding FVM’s conservation principles into the loss functions and architectures of modern neural networks, LE-FVM ensures physically consistent solutions while dramatically accelerating computation for parametric studies and real-time applications.
Accurately, efficiently, and stably computing complex fluid flows and their evolution near solid boundaries over long horizons remains challenging. Conventional numerical solvers require fine grids and small time steps to resolve near-wall dynamics, resulting in high computational costs, while purely data-driven surrogate models accumulate rollout errors and lack robustness under extrapolative conditions. To address these issues, this study extends existing neural PDE solvers by developing a physics-integrated differentiable framework for long-horizon prediction of immersed-boundary flows. A key design aspect of the framework includes an important improvement, namely the structural integration of physical principles into an end-to-end differentiable architecture incorporating a PDE-based intermediate velocity module and a multi-direct forcing immersed boundary module, both adhering to the pressure-projection procedure for incompressible flow computation. The computationally expensive pressure projection step is substituted with a learned implicit correction using ConvResNet blocks to reduce cost, and a sub-iteration strategy is introduced to separate the embedded physics module's stability requirement from the surrogate model's time step, enabling stable coarse-grid autoregressive rollouts with large effective time increments. The framework uses only single-step supervision for training, eliminating long-horizon backpropagation and reducing training time to under one hour on a single GPU. Evaluations on benchmark cases of flow past a stationary cylinder and a rotationally oscillating cylinder at Re=100 show the proposed model consistently outperforms purely data-driven, physics-loss-constrained, and coarse-grid numerical baselines in flow-field fidelity and long-horizon stability, while achieving an approximately 200-fold inference speedup over the high-resolution solver.
Analysis, optimization and uncertainty quantification of the aerodynamic behaviour of turbomachinery components is a fundamental part of the current industrial design process and requires the extensive use of compute-intensive CFD simulations. In this paper we investigate whether graph neural networks can be useful as surrogate models to accelerate the design process, for example in a multi-fidelity framework. Graph neural networks promise to provide good estimates of flow quantities while maintaining the geometric accuracy at a fraction of the computational effort of classical CFD. An application to industrially relevant turbomachinery flows is performed to gain a good understanding of the capabilities and limitations of such methods. We therefore apply a state-of-the-art graph neural network to a turbomachinery setup of industry-relevant mesh size. In particular, a multiscale graph neural network is used to overcome the problems of large information distances when applying message-passing based graph-net blocks to large meshes. The database used to train the network consists of a space-filling DoE of 100 CFD solutions with different geometries. The first use case encompasses the prediction of the flow quantities of the complete fluid domain with 2.5e6 mesh points. The second use case focuses on predicting a single scalar (e.g. pressure or temperature) on surface meshes with up to 30e3 mesh points. In both cases, the networks are employed to predict time-averaged and unsteady flow fields on unstructured meshes of variable point sizes for new geometries not present in the training set. The results demonstrate the proficiency of the approach in predicting time-averaged and unsteady flow quantities on surfaces as well as for full fluid domains for new geometries.