Skip to main content
Have a personal or library account? Click to login
Rapid Wind-load Prediction for a Standard Tall Building Using Physics-informed Deep Learning and CFD Surrogate Modelling Cover

Rapid Wind-load Prediction for a Standard Tall Building Using Physics-informed Deep Learning and CFD Surrogate Modelling

Open Access
|Nov 2026
Read online

Figures & tables

Table 1:

Comparison of existing studies and proposed framework

StudyCFD dataDeep learning modelPhysics-informed learningRapid predictionMulti-directional wind
Wu & Snaiki (2022)✓CNN✗✓✗
Fernández-Cabán et al. (2018)✓ANN✗✓✗
Nav et al. (2025)✓LSTM✗✓✗
Kim et al. (2024)LES–✗✗✓
Huang et al. (2007)LES–✗✗✓
Cai & Wang (2024)✗–✓✗✗
Proposed framework✓Physics-informed deep learning (PINN)✓✓✓
Figure 1:

Overall framework workflow

Table 2:

Mesh independence analysis based on mean stagnation-point pressure coefficient

Mesh size [106 cells]Cp (Stagnation)Change from previous
1.50.82-
3.20.89+8.5%
5.80.93+4.5%
8.00.95+2.1%
12.50.950.0%
Table 3:

Comparison of predicted mean pressure coefficients with CAARC wind tunnel benchmark data at mid-height (z/H = 0.5) and 0° wind incidence

FacePresent study [Cp]CAARC benchmark [Cp]Difference [%]
Windward0.950.94+1.1%
Leeward−0.52−0.50+4.0%
Side (Suction)−0.78−0.80−2.5%
Figure 2:

Training history of physics-informed neural network: (a) Total training loss, (b) Loss components (Data loss and PDE loss), (c) Loss component ratios

Figure 3:

Prediction versus CFD validation data: (a) U-Velocity, (b) V-Velocity, (c) W-Velocity, (d) Pressure

Table 4:

Performance evaluation of the proposed physics-informed deep learning model using CFD validation data

Output variableMAERMSER2
U-Velocity (m/s)0.3270.4870.986
V-Velocity (m/s)0.2310.3560.942
W-Velocity (m/s)0.0850.1280.918
Pressure (Pa)3.4875.2340.989
Figure 4:

2D Field Comparison at z = 10 m: (a-c) CFD, (d-f) PINN Predictions, (g-i) Difference Fields for U-Velocity, V-Velocity, and Pressure

Figure 5:

Three-Dimensional Velocity Field Predicted by PINN

Figure 6:

Pressure Distribution on Building Surfaces: (a) Windward Surface (x=0), (b) Leeward Surface (x=10 m)

Figure 7:

Vertical wind profile comparison at x = 5 m, y = 0 m: CFD Data (Blue) versus PINN prediction (Red dashed)

Figure 8:

Error distribution histograms: (a) U-Velocity error, (b) V-Velocity error, (c) W-Velocity error, (d) Pressure error

Figure 9:

Rapid prediction example along wind direction (y=0, z=20 m): (a) U-velocity, (b) V-velocity, (c) W-velocity, (d) Pressure

Table 5:

Aerodynamic wind-load coefficients derived from the PINN-predicted pressure field

Aerodynamic parameterSymbolValue
Windward pressure(Pfront)32.85 Pa
Leeward pressure(Pback)−9.17 Pa
Pressure difference(Δp)42.02 Pa
Drag coefficient(Cd)0.686
Table 6:

Comparison of proposed framework with existing studies

StudyModelDataset sizeR2Inference time (seconds)Multi-directional windPhysics- informedNotes on comparability
Wu & Snaiki [1]CNN~5,0000.920~0.10Limited✗Low-rise buildings; geometry differs from present study
Fernández-Cabán et al. [2]ANN~3,2000.890~0.05Limited✗Low-rise; different building geometry
Nav et al. [3]LSTM~8,0000.935~0.02Limited✗CAARC building (comparable)
Meddage et al. [14]Attention-based DNN~10,0000.952~0.015✓✗CAARC-like geometry (comparable)
Zakharov et al. [15]PINN~2,5000.940~0.010Limited✓Single building facade (not full 3D)
Pure DNN (Ablation)DNN11,5200.971~0.010✓✗Same architecture, no physics loss
Present StudyPINN11,5200.989~0.010✓✓CAARC building; comprehensive dataset
Table 7:

Ablation Study - PINN vs. Pure DNN Performance Comparison

Output variablePINN (R2)Pure DNN (R2)ImprovementPINN (MAE)Pure DNN (MAE)Improvement
U-velocity0.9860.965+2.2%0.3270.412+20.6%
V-velocity0.9420.898+4.7%0.2310.298+22.5%
W-velocity0.9180.885+3.6%0.0850.112+24.1%
Pressure0.9890.971+1.8%3.4874.521+22.9%
DOI: https://doi.org/10.2478/cee-2027-0033 | Journal eISSN: 2199-6512 (formerly 1336-5835) | Journal ISSN: 1336-5835
Language: English
Submitted on: Aug 9, 2026
Accepted on: Sep 11, 2026
Published on: Nov 18, 2026
Published by: University of Žilina
In partnership with: Paradigm Publishing Services
Publication frequency: 4 issues per year

© 2026 Hameed Shakir Al-Aasm, published by University of Žilina
This work is licensed under the Creative Commons Attribution 4.0 License.