TripNet: Learning Large-scale High-fidelity 3D Car Aerodynamics with Triplane Networks

2025-03-19

AI Summary

High-fidelity CFD simulations incur prohibitive computational costs, severely limiting the iteration efficiency of automotive aerodynamic design. To address this, we propose TripNet—the first end-to-end CFD surrogate model based on triplane implicit representation. TripNet innovatively introduces triplane neural radiance field encoding into flow field modeling, enabling continuous querying at arbitrary 3D spatial locations. It employs a multi-task shared backbone network coupled with an implicit flow field decoder to jointly predict drag coefficient, surface pressure/velocity fields, and the full 3D voxelized flow field. Evaluated on the DrivAerNet dataset series, TripNet achieves a drag coefficient error <0.8%, surface field PSNR >32 dB, and accelerates full-flow-field prediction by over 2000× compared to conventional CFD—while matching the accuracy of industrial-grade solvers. Moreover, it significantly enhances geometric fidelity and cross-configuration generalization capability.

Technology Category

• Computer Vision: 3D Computer Vision
• Planning, Routing, and Scheduling: Planning with Language Models
• Application Domains: Mobility, Driving & Flight

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Computational Fluid Dynamics (CFD) simulations are essential in product design, providing insights into fluid behavior around complex geometries in aerospace and automotive applications. However, high-fidelity CFD simulations are computationally expensive, making rapid design iterations challenging. To address this, we propose TripNet, Triplane CFD Network, a machine learning-based framework leveraging triplane representations to predict the outcomes of large-scale, high-fidelity CFD simulations with significantly reduced computation cost. Our method encodes 3D geometry into compact yet information-rich triplane features, maintaining full geometry fidelity and enabling accurate aerodynamic predictions. Unlike graph- and point cloud-based models, which are inherently discrete and provide solutions only at the mesh nodes, TripNet allows the solution to be queried at any point in the 3D space. Validated on high-fidelity DrivAerNet and DrivAerNet++ car aerodynamics datasets, TripNet achieves state-of-the-art performance in drag coefficient prediction, surface field estimation, and full 3D flow field simulations of industry-standard car designs. By utilizing a shared triplane backbone across multiple tasks, our approach offers a scalable, accurate, and efficient alternative to traditional CFD solvers.

Problem

• Reducing high computational cost of large-scale CFD simulations
• Predicting 3D aerodynamic outcomes with triplane representations
• Enabling scalable and accurate alternatives to traditional CFD solvers

Innovation

• Uses triplane networks for 3D CFD predictions
• Encodes geometry into compact triplane features
• Enables scalable accurate aerodynamic simulations