Methods, systems, equipment and storage media for generating aerodynamic shapes of aircraft
By combining diffusion models and deep learning models, multiple complete aerodynamic parameter matrices are generated and shape representation data is predicted, solving the problems of high computational cost and unstable generation in aircraft aerodynamic shape design, and realizing efficient and diverse aerodynamic shape generation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA ACAD OF AEROSPACE AERODYNAMICS
- Filing Date
- 2025-12-30
- Publication Date
- 2026-05-26
AI Technical Summary
In existing technologies, the aerodynamic shape design of aircraft relies on empirical formulas and parametric modeling, which has high computational costs and unstable and undiversifiable results. Direct mapping is prone to ill-conditioned problems, and existing generative models lack structural awareness.
Multiple sets of complete aerodynamic parameter matrices are generated through a diffusion model. After filtering, the data is input into a deep learning model to predict the shape representation data. The shape representation is performed using structure-aware models such as the Transformer architecture, convolutional neural networks, and graph neural networks, supporting multiple representation methods such as explicit parameters, latent space encoding, and point cloud data.
It achieves efficient and stable aerodynamic shape generation of aircraft, avoids the problems of direct mapping, improves the diversity and physicality of the generated results, and reduces computational costs.
Smart Images

Figure CN122087941A_ABST