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

An accurate estimation of wind loads applied on tall buildings is still a challenge for civil engineers. Traditionally performed CFD simulations are computationally expensive enough to be unaffordable for use in the iterative process of design. A framework that exploits a physics-informed deep learning method is hereby proposed that performs fast wind pressure estimations for tall buildings. It will be demonstrated for one geometry through a proof-of-concept study of the CAARC standard tall building. The main principle relies on injecting information on the governing physical laws, in the shape of a physics-informed loss term, into a neural network for training. In this new learning model, the loss is made up by the Navier-Stokes equation itself and not solely on the numerical output of the CFD simulation to ensure physicsconsistent predictions during learning. An investigation scenario using the CAARC standard tall building as a case study is presented, and a physics-informed neural network trained on 11,520 points over several wind directions and with several turbulence models showed that the accuracy of prediction of the coefficient of determination (R) was 0.989, 0.986, 0.942 and 0.918 for pressure and U, V and W velocities, inference times below 0.01 s (4–5 orders of magnitude less than typical CFD simulations) and the result simultaneously incorporated both physically consistent predictions and conservation laws. For the test case presented, a Cd of 0.686 is estimated in line with values for the typical rectangular aerodynamic shape profile. It will be concluded that the proposed strategy would be very promising for estimating wind load on this CAARC building geometry but that future research is warranted for training PINNs that estimate wind load upon varying building geometries.

Authors

Institutions

Publication Details

Journal
Civil and Environmental Engineering
Published
2026-11-18
DOI
https://doi.org/10.2478/cee-2027-0033
Primary Topic
Model Reduction and Neural Networks
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

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

Hameed Shakir Al-Aasm
5 links
Civil and Environmental Engineering
Model Reduction and Neural Networks
article

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

Hameed Shakir Al-Aasm
article en

Abstract

An accurate estimation of wind loads applied on tall buildings is still a challenge for civil engineers. Traditionally performed CFD simulations are computationally expensive enough to be unaffordable for use in the iterative process of design. A framework that exploits a physics-informed deep learning method is hereby proposed that performs fast wind pressure estimations for tall buildings. It will be demonstrated for one geometry through a proof-of-concept study of the CAARC standard tall building. The main principle relies on injecting information on the governing physical laws, in the shape of a physics-informed loss term, into a neural network for training. In this new learning model, the loss is made up by the Navier-Stokes equation itself and not solely on the numerical output of the CFD simulation to ensure physicsconsistent predictions during learning. An investigation scenario using the CAARC standard tall building as a case study is presented, and a physics-informed neural network trained on 11,520 points over several wind directions and with several turbulence models showed that the accuracy of prediction of the coefficient of determination (R) was 0.989, 0.986, 0.942 and 0.918 for pressure and U, V and W velocities, inference times below 0.01 s (4–5 orders of magnitude less than typical CFD simulations) and the result simultaneously incorporated both physically consistent predictions and conservation laws. For the test case presented, a Cd of 0.686 is estimated in line with values for the typical rectangular aerodynamic shape profile. It will be concluded that the proposed strategy would be very promising for estimating wind load on this CAARC building geometry but that future research is warranted for training PINNs that estimate wind load upon varying building geometries.

Civil and Environmental Engineering
Al-Furat Al-Awsat Technical University (IQ)
Climate action
Model Reduction and Neural Networks
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon