FoilDiff: A Hybrid Transformer Backbone for Diffusion-based Modelling of 2D Airfoil Flow Fields

2025-10-05

AI Summary

To address the challenge of rapid and high-fidelity prediction of two-dimensional airfoil flow fields, this paper proposes a general-purpose surrogate model based on diffusion modeling. Methodologically, we design a hybrid convolutional-Transformer denoising backbone that jointly leverages local feature extraction and global attention mechanisms, coupled with DDIM sampling for accelerated generation. The input encoding incorporates Reynolds number, angle of attack, and parametric airfoil geometry. Evaluated on a unified benchmark dataset, our model achieves up to 85% reduction in mean prediction error compared to state-of-the-art surrogates, significantly improving flow-field reconstruction accuracy and uncertainty calibration. Moreover, it demonstrates strong generalization across diverse aerodynamic conditions—including varying Reynolds numbers, angles of attack, and airfoil shapes—thereby establishing a new paradigm for efficient, reliable surrogate modeling in aerodynamic design.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

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

Abstract

The accurate prediction of flow fields around airfoils is crucial for aerodynamic design and optimisation. Computational Fluid Dynamics (CFD) models are effective but computationally expensive, thus inspiring the development of surrogate models to enable quicker predictions. These surrogate models can be based on deep learning architectures, such as Convolutional Neural Networks (CNNs), Graph Neural Networks (GNNs), and Diffusion Models (DMs). Diffusion models have shown significant promise in predicting complex flow fields. In this work, we propose FoilDiff, a diffusion-based surrogate model with a hybrid-backbone denoising network. This hybrid design combines the power of convolutional feature extraction and transformer-based global attention to generate more adaptable and accurate representations of flow structures. FoilDiff takes advantage of Denoising Diffusion Implicit Model (DDIM) sampling to optimise the efficiency of the sampling process at no additional cost to model generalisation. We used encoded representations of Reynolds number, angle of attack, and airfoil geometry to define the input space for generalisation across a wide range of aerodynamic conditions. When evaluated against state-of-the-art models, FoilDiff shows significant performance improvements, with mean prediction errors reducing by up to 85% on the same datasets. The results have demonstrated that FoilDiff can provide both more accurate predictions and better-calibrated predictive uncertainty than existing diffusion-based models.

Problem

• Accurately predicting airfoil flow fields for aerodynamic design
• Developing efficient deep learning surrogate models for CFD
• Improving flow field prediction accuracy with hybrid architecture

Innovation

• Hybrid transformer backbone combines CNN and transformer attention
• DDIM sampling optimizes efficiency without generalization loss
• Encoded Reynolds number, angle, and geometry inputs generalize conditions