NeuralCFD: Deep Learning on High-Fidelity Automotive Aerodynamics Simulations

2025-02-13

AI Summary

This work addresses two key challenges in neural surrogate modeling for automotive aerodynamics: (i) poor scalability to geometric inputs—particularly with tens-of-millions of volumetric mesh cells—and (ii) data scarcity of high-fidelity samples hindering robust training. To this end, we propose the Geometry-Preserving Universal Physics Transformer (GP-UPT), a novel architecture that decouples geometric encoding from physical prediction. GP-UPT enables direct ingestion of raw CAD geometry, facilitates mesh-free modeling, and supports arbitrary sampling strategies. It integrates neural operators with multi-fidelity transfer learning to enhance generalization. Experiments on 3D velocity field prediction over 20-million-cell domains demonstrate that GP-UPT achieves full high-fidelity training accuracy using only 47% of the high-fidelity samples. Moreover, it completely bypasses conventional mesh generation, significantly improving both simulation efficiency and cross-geometry generalizability.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Search and Optimization: Sampling/Simulation-based Search
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Recent advancements in neural operator learning are paving the way for transformative innovations in fields such as automotive aerodynamics. However, key challenges must be overcome before neural network-based simulation surrogates can be implemented at an industry scale. First, surrogates must become scalable to large surface and volume meshes, especially when using raw geometry inputs only, i.e., without relying on the simulation mesh. Second, surrogates must be trainable with a limited number of high-fidelity numerical simulation samples while still reaching the required performance levels. To this end, we introduce Geometry-preserving Universal Physics Transformer (GP-UPT), which separates geometry encoding and physics predictions, ensuring flexibility with respect to geometry representations and surface sampling strategies. GP-UPT enables independent scaling of the respective parts of the model according to practical requirements, offering scalable solutions to open challenges. GP-UPT circumvents the creation of high-quality simulation meshes, enables accurate 3D velocity field predictions at 20 million mesh cells, and excels in transfer learning from low-fidelity to high-fidelity simulation datasets, requiring less than half of the high-fidelity data to match the performance of models trained from scratch.

Problem

• Scalable neural surrogates for large meshes
• Training with limited high-fidelity simulation data
• Accurate 3D velocity field predictions

Innovation

• Geometry-preserving Universal Physics Transformer
• Scalable large surface and volume meshes
• Accurate 3D velocity field predictions