PMRT: A Training Recipe for Fast, 3D High-Resolution Aerodynamic Prediction

• URL: http://arxiv.org/abs/2509.17182v1
• Date: Sun, 21 Sep 2025 18:05:50 GMT
• Title: PMRT: A Training Recipe for Fast, 3D High-Resolution Aerodynamic Prediction
• Authors: Sam Jacob Jacob, Markus Mrosek, Carsten Othmer, Harald Köstler,
• Abstract summary: Surrogate models have been shown to accurately predict aerodynamics within the design space for which they were trained.<n>We propose Progressive Multi-Resolution Training (PMRT), a probabilistic multi-resolution training schedule.<n>PMRT samples batches from three resolutions based on probabilities that change during training.<n>We show that a single model can be trained across five datasets from different solvers, including a real-world dataset, by conditioning on the simulation parameters.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The aerodynamic optimization of cars requires close collaboration between aerodynamicists and stylists, while slow, expensive simulations remain a bottleneck. Surrogate models have been shown to accurately predict aerodynamics within the design space for which they were trained. However, many of these models struggle to scale to higher resolutions because of the 3D nature of the problem and data scarcity. We propose Progressive Multi-Resolution Training (PMRT), a probabilistic multi-resolution training schedule that enables training a U-Net to predict the drag coefficient ($c_d$) and high-resolution velocity fields (512 x 128 x 128) in 24 hours on a single NVIDIA H100 GPU, 7x cheaper than the high-resolution-only baseline, with similar accuracy. PMRT samples batches from three resolutions based on probabilities that change during training, starting with an emphasis on lower resolutions and gradually shifting toward higher resolutions. Since this is a training methodology, it can be adapted to other high-resolution-focused backbones. We also show that a single model can be trained across five datasets from different solvers, including a real-world dataset, by conditioning on the simulation parameters. In the DrivAerML dataset, our models achieve a $c_d$ $R^2$ of 0.975, matching literature baselines at a fraction of the training cost.

Related papers

• Transolver-3: Scaling Up Transformer Solvers to Industrial-Scale Geometries [51.028432812178266]
  Transolver-3 is a new member of the Transolver family designed for high-fidelity physics simulations.<n>We show that Transolver-3 is capable of handling meshes with over 160 million cells, achieving impressive performance across three challenging simulation benchmarks.
  arXiv  Detail & Related papers  (2026-02-04T16:52:44Z)
• MeanFlow Transformers with Representation Autoencoders [71.45823902973349]
  MeanFlow (MF) is a diffusion-motivated generative model that enables efficient few-step generation by learning long jumps directly from noise to data.<n>We develop an efficient training and sampling scheme for MF in the latent space of a Representation Autoencoder (RAE)<n>We achieve a 1-step FID of 2.03, outperforming vanilla MF's 3.43, while reducing sampling GFLOPS by 38% and total training cost by 83% on ImageNet 256.
  arXiv  Detail & Related papers  (2025-11-17T06:17:08Z)
• Predictive Feature Caching for Training-free Acceleration of Molecular Geometry Generation [67.20779609022108]
  Flow matching models generate high-fidelity molecular geometries but incur significant computational costs during inference.<n>This work discusses a training-free caching strategy that accelerates molecular geometry generation.<n> Experiments on the GEOM-Drugs dataset demonstrate that caching achieves a twofold reduction in wall-clock inference time.
  arXiv  Detail & Related papers  (2025-10-06T09:49:14Z)
• P3D: Scalable Neural Surrogates for High-Resolution 3D Physics Simulations with Global Context [22.305840640878433]
  We introduce a hybrid CNN-Transformer backbone architecture targeted for 3D physics simulations.<n>Our proposed network can be pretrained on small patches of the simulation domain, which can be fused to obtain a global solution.<n>We demonstrate how to scale our model to high-resolution isotropic turbulence with spatial resolutions of up to $5123$.
  arXiv  Detail & Related papers  (2025-09-12T12:26:06Z)
• Learning to Track Any Points from Human Motion [55.831218129679144]
  We propose an automated pipeline to generate pseudo-labeled training data for point tracking.<n>A point tracking model trained on AnthroTAP achieves annotated state-of-the-art performance on the TAP-Vid benchmark.
  arXiv  Detail & Related papers  (2025-07-08T17:59:58Z)
• AutoHete: An Automatic and Efficient Heterogeneous Training System for LLMs [68.99086112477565]
  Transformer-based large language models (LLMs) have demonstrated exceptional capabilities in sequence modeling and text generation.<n>Existing heterogeneous training methods significantly expand the scale of trainable models but introduce substantial communication overheads and CPU workloads.<n>We propose AutoHete, an automatic and efficient heterogeneous training system compatible with both single- GPU and multi- GPU environments.
  arXiv  Detail & Related papers  (2025-02-27T14:46:22Z)
• Factorized Implicit Global Convolution for Automotive Computational Fluid Dynamics Prediction [52.32698071488864]
  We propose Factorized Implicit Global Convolution (FIGConv), a novel architecture that efficiently solves CFD problems for very large 3D meshes.<n>FIGConv achieves quadratic complexity $O(N2)$, a significant improvement over existing 3D neural CFD models.<n>We validate our approach on the industry-standard Ahmed body dataset and the large-scale DrivAerNet dataset.
  arXiv  Detail & Related papers  (2025-02-06T18:57:57Z)
• Aero-Nef: Neural Fields for Rapid Aircraft Aerodynamics Simulations [1.1932047172700866]
  This paper presents a methodology to learn surrogate models of steady state fluid dynamics simulations on meshed domains. The proposed models can be applied directly to unstructured domains for different flow conditions. Remarkably, the method can perform inference five order of magnitude faster than the high fidelity solver on the RANS transonic airfoil dataset.
  arXiv  Detail & Related papers  (2024-07-29T11:48:44Z)
• Geometry-Informed Neural Operator for Large-Scale 3D PDEs [76.06115572844882]
  We propose the geometry-informed neural operator (GINO) to learn the solution operator of large-scale partial differential equations. We successfully trained GINO to predict the pressure on car surfaces using only five hundred data points.
  arXiv  Detail & Related papers  (2023-09-01T16:59:21Z)
• Learning Controllable Adaptive Simulation for Multi-resolution Physics [86.8993558124143]
  We introduce Learning controllable Adaptive simulation for Multi-resolution Physics (LAMP) as the first full deep learning-based surrogate model. LAMP consists of a Graph Neural Network (GNN) for learning the forward evolution, and a GNN-based actor-critic for learning the policy of spatial refinement and coarsening. We demonstrate that our LAMP outperforms state-of-the-art deep learning surrogate models, and can adaptively trade-off computation to improve long-term prediction error.
  arXiv  Detail & Related papers  (2023-05-01T23:20:27Z)
• Physics-enhanced deep surrogates for partial differential equations [30.731686639510517]
  We present a "physics-enhanced deep-surrogate" ("PEDS") approach towards developing fast surrogate models for complex physical systems. Specifically, a combination of a low-fidelity, explainable physics simulator and a neural network generator is proposed, which is trained end-to-end to globally match the output of an expensive high-fidelity numerical solver.
  arXiv  Detail & Related papers  (2021-11-10T18:43:18Z)
• SURFNet: Super-resolution of Turbulent Flows with Transfer Learning using Small Datasets [0.9297355862757838]
  This paper introduces SURFNet, a transfer learning-based super-resolution flow network. SURFNet primarily trains the DL model on low-resolution datasets and transfer learns the model on a handful of high-resolution flow problems. We empirically evaluate SURFNet's performance by solving the Navier-Stokes equations in the turbulent regime on input resolutions up to 256x larger than the coarse model.
  arXiv  Detail & Related papers  (2021-08-17T14:53:04Z)
• Large Batch Simulation for Deep Reinforcement Learning [101.01408262583378]
  We accelerate deep reinforcement learning-based training in visually complex 3D environments by two orders of magnitude over prior work. We realize end-to-end training speeds of over 19,000 frames of experience per second on a single and up to 72,000 frames per second on a single eight- GPU machine. By combining batch simulation and performance optimizations, we demonstrate that Point navigation agents can be trained in complex 3D environments on a single GPU in 1.5 days to 97% of the accuracy of agents trained on a prior state-of-the-art system.
  arXiv  Detail & Related papers  (2021-03-12T00:22:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.