---
title: 'WebWing: Rapid Aerodynamic Predictions'
url: https://www.emergentmind.com/topics/webwing
type: topic
---

# WebWing: Rapid Aerodynamic Predictions

WebWing is an interactive, browser-based design tool for rapid aerodynamic prediction of three-dimensional transonic wings. It is built on AeroTransformer, a large-scale transformer surrogate model pre-trained on the diverse “SuperWing” dataset and fine-tuned to detailed perturbations of the NASA Common Research Model (CRM). WebWing enables users to modify wing geometry and operating conditions via an intuitive web interface, performing real-time predictions of surface flow and aerodynamic coefficients, thus replacing computationally expensive CFD calculations during early-stage design exploration. The system exemplifies the foundation-model paradigm in engineering surrogate modeling, combining broad pre-training with targeted fine-tuning to achieve high accuracy and generalization [2604.18062].

## 1. AeroTransformer Surrogate Model Architecture

AeroTransformer extends the hierarchical PDE-Transformer backbone to map three-dimensional wing geometries and operating conditions to either surface-flow fields or integrated aerodynamic coefficients. Input wing geometries $g \in \mathbb{R}^{H \times W \times 3}$ are discretized on a structured mesh ($H=256$, circumferential; $W=128$, spanwise), with each mesh cell containing $(x, y, z)$ coordinates. These are embedded using a shared $4 \times 4$ convolutional “patching” layer, yielding $M = (H/4) \times (W/4)$ tokens.

Operating conditions $c \in \mathbb{R}^{N_c=2}$, specifically Mach number ($Ma$) and angle of attack ($\alpha$), are injected into every transformer block using adaptive layer normalization conditioning (“adaLN-Zero”). A small MLP projects $c$ to $e_c$, from which scale-and-shift vectors $(\gamma, \beta)$ are regressed to modulate activations.

The transformer backbone is organized in a U-shaped, hierarchical structure—with two down-sampling stages, a latent stage, and two up-sampling stages—and incorporates skip connections. Down-sampling uses PixelUnshuffle convolutions to decrease resolution while doubling the hidden dimension; up-sampling reverses this with PixelShuffle. Windowed (8×8) multi-head self-attention (W-MSA) and shifted window-attention (SW-MSA) limit global attention for computational efficiency. Positional information is encoded via log-spaced relative position embeddings.

Surface-flow predictions ($Cp(x)$, $C_f(x)$) are output via a final convolutional expansion and reshaping layer. For integrated coefficients ($k = [C_L, C_D, C_{M,z}]$), latent tokens are aggregated by attention pooling and projected via a fully connected layer. The multi-task training objective is
\[
L = L_{surf} + \lambda L_{coef},
\]
with $\lambda=0.1$ balancing field and coefficient accuracy. $L_{surf}$ is a mean squared error (MSE) loss on predicted surface pressure and friction vectors; $L_{coef}$ is an MSE on integrated coefficients obtained from the predicted surface fields [2604.18062].

## 2. Training Strategy: SuperWing Pre-training and CRM Fine-tuning

WebWing’s surrogate accuracy is enabled by a two-stage training paradigm:

**A. SuperWing Pre-training:**  
The model is first pre-trained on SuperWing, comprising 4,239 distinct transonic wing shapes (planform parameters sampled uniformly across $Λ_{LE} \in [25°, 40°]$, $AR \in [8,11]$, $TR \in [0.15, 0.40]$, $η_k \in [36\%, 42\%]$, $\kappa \in [10\%, 110\%]$). Sectional airfoils are parameterized via B-splines; dihedral and twist via control points. Each shape is simulated at eight operating points ($Ma \in [0.75, 0.90]$, $\alpha \in [2°, 12°]$), yielding 28,856 RANS flow fields. Model sizes S/M/L ($N_{h0}$=$16/32/64$) span $1.0M/3.8M/14.5M$ parameters [2604.18062].

**B. CRM Fine-tuning:**  
Fine-tuning uses 288 CRM wing perturbations around the NASA CRM baseline. Each shape (parameterized by 7 spanwise sections, airfoils varied with 20 CST coefficients, dihedral $\pm0.05\,c_{root}$, twist $U(-3°, 0°)$) is simulated at eight operating conditions. Subsets of 450 samples serve as fine-tuning budgets. Fine-tuning the L-size model for 5.6k steps (49 minutes) achieves substantial error reduction versus scratch training. Ten-fold cross-validation is conducted on the 450 samples [2604.18062].

## 3. Surrogate Modeling Formulation and Metrics

Given inputs $(g, c)$, AeroTransformer predicts either:

- Surface flow $u = [C_p(x), C_f(x)]$ on the mesh, enabling downstream integration to coefficients.
- Directly, aerodynamic coefficients $k = [C_L, C_D, C_{M,z}]$.

Key evaluation metrics:

- **Pointwise surface errors (normalized MAE):**
  \[
  \delta X = \frac{1}{N_s} \sum_{n=1}^{N_s} \left( \frac{1}{H \cdot W} \sum_{i,j} \frac{|\hat{X}_{n;i,j} - X_{n;i,j}^{\mathrm{CFD}}|}{\max X^{\mathrm{CFD}} - \min X^{\mathrm{CFD}}} \right) \times 100\%.
  \]

- **Aggregate surface flow error (SFE):**
  \[
  SFE = \frac{\delta C_p + \delta C_{f,\tau} + \delta C_{f,z}}{3}.
  \]

- **Coefficient errors (MAE):**
  \[
  \delta X = \frac{1}{N_s} \sum_{n=1}^{N_s} |\hat{X}_n - X_n^{\mathrm{CFD}}|, \quad X \in \{ C_L, C_D, C_{M,z} \}.
  \]

These metrics enable quantitative comparison to CFD and to other ML-based surrogates [2604.18062].

## 4. WebWing Interface and Prediction Pipeline

WebWing utilizes a JavaScript/Three.js frontend and a remote GPU backend hosting the pre-trained L-size AeroTransformer. Users can adjust:

- Global planform (sliders: $Λ_{LE}$, $AR$, $TR$, $η_k$, $\kappa$)
- Spanwise control points (airfoil CST coefficients, dihedral, twist at 7 stations) via drag handles or file upload
- Operating conditions (sliders: $Ma$, $\alpha$)

Upon modification, the frontend serializes $(g, c)$ as JSON and sends it to the backend, which remeshes the geometry to a $256 \times 128$ grid and executes AeroTransformer inference (∼30 ms for 9 conditions, on NVIDIA A5000). Predictions (surface $C_p$, $|C_f|$, $C_L$, $C_D$, $C_{M,z}$) are returned and visualized instantaneously:

- Interactive 3D coloring by $C_p$ or $|C_f|$
- Sectional 2D plots of $C_p, C_f$ vs. chordwise location
- Numeric display of integrated coefficients
- Real-time response on all user actions

This provides a rapid design loop that obviates hours of CFD per geometry [2604.18062].

## 5. Quantitative Performance and Comparative Analysis

The table below summarizes selected surrogate error metrics:

| Method & Data Regime             | SFE (%) | $\delta C_L \ (\times 10^{-3})$ | $\delta C_D \ (\times 10^{-4})$ |
|-----------------------------------|---------|--------------------|--------------------|
| U-Net (SuperWing test, S)         | 0.911   | –                  | –                  |
| ViT (SuperWing test, S)           | 0.353   | –                  | –                  |
| Transolver (SuperWing test, S)    | 0.401   | –                  | –                  |
| AeroTransformer S (SuperWing test)| 0.279→0.264 | 2.35             | 2.11              |
| From scratch (CRM 450, L)         | 1.00    | 38.1               | 27.9               |
| Zero-shot pretrain (CRM 450, L)   | 0.376   | 14.7               | 8.87               |
| Fine-tuned (CRM 450, L)           | 0.159   | 4.06               | 3.36               |
| Attn-only finetune (15% params)   | 0.465   | –                  | –                  |
| LoRA finetune (1.7% params)       | 0.499   | –                  | –                  |

Fine-tuning on 450 CRM samples achieves SFE=0.159% (a reduction of 84.2% vs. training from scratch), with inference times of ∼30 ms per geometry × 9 conditions. This outperforms baseline surrogates and reduces costs relative to RANS CFD, which requires hours per case. Parameter-efficient fine-tuning (attention layers or LoRA) yields only slight accuracy penalties [2604.18062].

## 6. Practical Usage, Limitations, and Case Studies

Recommended deployment involves two workflows: (1) zero-shot mode for rapid broad surveys in geometry/condition space, and (2) fine-tuning with a limited set of high-fidelity CRM-like samples when increased local accuracy ($<$0.2% SFE) is required.

A documented case study involving CRM at $Ma=0.85$, $\alpha \in \{0°, 2°, 4°\}$ demonstrates that both zero-shot and fine-tuned AeroTransformer predictions capture surface shock location, $C_p$ distribution, and polar curves within approximately $5 \times 10^{-3}$ error on $C_L$. This enables rapid cruise and off-design exploration.

Limitations are as follows:

- The SuperWing pre-training dataset is limited to single-airfoil-derived sectional shapes; extreme camber or novel planforms may degrade zero-shot accuracy.
- Fine-tuning is validated near CRM-derived perturbations; significant planform changes require further targeted data.
- The approach models surface flow only; volumetric (off-surface) effects are not addressed.

This suggests that while WebWing offers robust surrogate modeling for standard and moderate wing variations, further generalization to exotic geometries would require additional pre-training or domain adaptation [2604.18062].

## 7. Broader Context and Foundation-Model Paradigm

WebWing exemplifies the application of the foundation-model paradigm to aerodynamic design: large-scale diversity pre-training, followed by rapid, targeted adaptation via fine-tuning. By integrating advanced architectures with efficient surrogate modeling, WebWing provides near-CFD accuracy and substantially reduced response times, directly impacting early-stage design workflows in computational aerodynamics. The adaptive conditioning, hierarchical modeling, and real-time interactivity represent a substantial convergence of machine learning, engineering domain knowledge, and web-based design tools [2604.18062].

Source: https://www.emergentmind.com/topics/webwing