LGFNet: Local-Global Fusion Network with Fidelity Gap Delta Learning for Multi-Source Aerodynamics

2026-03-31

AI Summary

This work addresses the challenge in multi-source aerodynamic data fusion, where preserving both high-resolution local details and global long-range dependencies is difficult, often leading to loss of discontinuous features like shock waves or distorted topological relationships. To overcome this, the authors propose the Local-Global Fusion Network (LGFNet), which jointly leverages a spatial-aware layer and a self-attention-based relational reasoning layer to extract multi-scale features. A novel Fidelity Gap Discrepancy Learning (FGDL) strategy is introduced to explicitly model nonlinear deviations using CFD data as a low-frequency prior. Integrated with a sliding window mechanism and multi-scale feature decomposition, LGFNet simultaneously enhances fine local structures and coherent global flow trends, effectively mitigating non-physical smoothing. Experiments demonstrate that the method significantly outperforms existing approaches across diverse aerodynamic scenarios, achieving state-of-the-art performance in both prediction accuracy and uncertainty quantification.

Abstract

The precise fusion of computational fluid dynamic (CFD) data, wind tunnel tests data, and flight tests data in aerodynamic area is essential for obtaining comprehensive knowledge of both localized flow structures and global aerodynamic trends across the entire flight envelope. However, existing methodologies often struggle to balance high-resolution local fidelity with wide-range global dependency, leading to either a loss of sharp discontinuities or an inability to capture long-range topological correlations. We propose Local-Global Fusion Network (LGFNet) for multi-scale feature decomposition to extract this dual-natured aerodynamic knowledge. To this end, LGFNet combines a spatial perception layer that integrates a sliding window mechanism with a relational reasoning layer based on self-attention, simultaneously reinforcing the continuity of fine-grained local features (e.g., shock waves) and capturing long-range flow information. Furthermore, the fidelity gap delta learning (FGDL) strategy is proposed to treat CFD data as a "low-frequency carrier" to explicitly approximate nonlinear discrepancies. This approach prevents unphysical smoothing while inheriting the foundational physical trends from the simulation baseline. Experiments demonstrate that LGFNet achieves state-of-the-art (SOTA) performance in both accuracy and uncertainty reduction across diverse aerodynamic scenarios.

Problem

• multi-source aerodynamics
• local-global fusion
• fidelity gap
• flow structure
• aerodynamic data fusion

Innovation

• Local-Global Fusion
• Fidelity Gap Delta Learning
• Multi-Source Aerodynamics
• Self-Attention
• Sliding Window Mechanism