A Geometry-Aware Triplane Field Network for Vehicle Aerodynamic Prediction

2026-06-05

AI Summary

This study addresses the high computational cost of high-fidelity CFD simulations in vehicle aerodynamic analysis, which hinders early-stage design exploration. To overcome this limitation, the authors propose the Geometry-aware Tri-plane Field Network (GTF-Net), a novel deep learning framework that constructs tri-plane features from surface points via shared MLPs and bilinear rasterization. The architecture integrates a dual-stream backbone combining Adaptive Fourier Neural Operators (AFNO) and CNNs to jointly model long-range aerodynamic couplings and fine-grained geometric details within a unified representation. The method innovatively fuses structured tri-plane decomposition, explicit geometric cues, and a hybrid spectral-spatial mechanism. Experiments demonstrate that GTF-Net achieves state-of-the-art performance, with relative L2 errors of 0.145 for pressure and 0.226 for wall shear stress predictions, significantly outperforming existing approaches, while ablation studies confirm the contribution of each component.

Abstract

High-fidelity computational fluid dynamics (CFD) is crucial to vehicle aerodynamic analysis, but its cost still constrains early-stage design exploration. Machine-learning-based surface-field prediction offers a faster alternative if the model can efficiently capture both global flow context and local geometric detail. This work proposes a machine-learning-based method, named the geometry-aware triplane field network (GTF-Net), for vehicle aerodynamic pressure and wall shear stress prediction. GTF-Net constructs triplane features directly from sampled surface points through a shared multilayer perceptron (MLP) and smooth bilinear rasterization. The planes are then processed by a dual-stream backbone that combines adaptive Fourier neural operator (AFNO) spectral mixing with convolutional neural network (CNN) refinement, so long-range aerodynamic coupling and local geometry-induced variations are modeled in the same representation. At query stage, sampled triplane features are combined with vehicle-aligned directional coordinates, normal-projection features, and a voxel-based curvature proxy. GTF-Net is compared with Transolver, geometry-informed neural operator (GINO), and TripNet, a triplane-based surrogate model. GTF-Net improves the relative L2 error from the strongest baseline value of 0.157 to 0.145 for pressure prediction and from 0.237 to 0.226 for wall shear stress prediction. Ablation results show that AFNO mixing, local CNN refinement, and query-side geometric encoding each contribute to accuracy, supporting the proposed mechanism of combining structured triplane representation with explicit aerodynamic geometry cues.

Problem

• vehicle aerodynamics
• computational fluid dynamics
• machine learning
• surface-field prediction
• geometry representation

Innovation

• triplane representation
• adaptive Fourier neural operator
• geometry-aware learning
• aerodynamic field prediction
• dual-stream backbone