Reconstruction of Airflows for an Agricultural Building Assisted by a Physics-Informed Diffusion Model

Computational Fluid Dynamics (CFD) provides robust airflow modeling but requires high-performance computing. Alternatively, this study presents a physics-informed diffusion model (PIDM) that reconstructs velocity fields from limited information while enforcing mass conservation. The model was trained on two-dimensional CFD simulations of a naturally ventilated building (NVB), which was generated by solving the Reynolds-averaged Navier–Stokes (RANS) equations with the realizable k-ϵ turbulence model. The dataset comprised 1144 numerical solutions across air inlet velocities of 1–12 m/s. Velocity fields were converted to images with 87.5% used for training and 12.5% reserved for testing. Physics-guided images, defined by the mean and standard deviation of the velocity field, were used to test the PIDM. We found that the PIDM improved reconstruction accuracy by 47% based on the Root Mean Square Error (RMSE) of the testing dataset. These results demonstrated that the PIDM reconstructed complex velocity fields from CFD-derived sparse input images, highlighting its potential to fill data gaps in agricultural buildings. Realizing this potential will further motivate studies on the synergistic integration of CFD, field sensing and PIDM.

Authors

Institutions

Publication Details

Journal
Buildings
Published
2026-11-16
DOI
https://doi.org/10.3390/buildings16224573
Primary Topic
Model Reduction and Neural Networks
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Reconstruction of Airflows for an Agricultural Building Assisted by a Physics-Informed Diffusion Model

Fernando Rojano, Alba Closa-Tarres, Sisi Guevara-Garcia, Pierre-Emmanuel Bournet
3 links
Buildings
Model Reduction and Neural Networks
article

Reconstruction of Airflows for an Agricultural Building Assisted by a Physics-Informed Diffusion Model

Fernando Rojano, Alba Closa-Tarres, Sisi Guevara-Garcia, Pierre-Emmanuel Bournet
article en

Abstract

Computational Fluid Dynamics (CFD) provides robust airflow modeling but requires high-performance computing. Alternatively, this study presents a physics-informed diffusion model (PIDM) that reconstructs velocity fields from limited information while enforcing mass conservation. The model was trained on two-dimensional CFD simulations of a naturally ventilated building (NVB), which was generated by solving the Reynolds-averaged Navier–Stokes (RANS) equations with the realizable k-ϵ turbulence model. The dataset comprised 1144 numerical solutions across air inlet velocities of 1–12 m/s. Velocity fields were converted to images with 87.5% used for training and 12.5% reserved for testing. Physics-guided images, defined by the mean and standard deviation of the velocity field, were used to test the PIDM. We found that the PIDM improved reconstruction accuracy by 47% based on the Root Mean Square Error (RMSE) of the testing dataset. These results demonstrated that the PIDM reconstructed complex velocity fields from CFD-derived sparse input images, highlighting its potential to fill data gaps in agricultural buildings. Realizing this potential will further motivate studies on the synergistic integration of CFD, field sensing and PIDM.

BuildingsVol. 16(22)
Institut Agro Rennes-Angers (FR), L'Institut Agro (FR), Université d'Angers (FR), West Virginia State University (US)
National Institute of Food and Agriculture
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