Integration of Hybrid Deep Learning with CFD Simulations for Flow and Temperature Field Prediction
International Journal of Advanced Research in Science Communication and Technology · 10.04.2026 · 244
DOI: 10.48175/ijarsct-32737 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Advanced Research in Science Communication and Technology, online erschienen am 10.04.2026. Verfasst von Mr. S. Manikatnta und 4 weiteren Autor:innen. Die Arbeit nennt 10 Literaturangaben.
Abstract (Originalsprache)
Accurate prediction of indoor airflow and temperature distribution is essential for improving thermal comfort, indoor air quality, and energy efficiency in buildings. Computational Fluid Dynamics (CFD) is widely used to simulate airflow and heat transfer in indoor environments; however, high computational cost and long simulation time limit its application in real-time analysis and rapid design optimization. To address this challenge, this study proposes a hybrid deep learning surrogate modeling framework that integrates CFD simulations with advanced deep learning techniques for fast and accurate prediction of airflow velocity and temperature fields. High-fidelity CFD simulations are first conducted to generate datasets under different inlet configurations and boundary conditions within a two- dimensional indoor environment. The generated datasets are then used to train a hybrid Bidirectional Long Short-Term Memory (BiLSTM) and Transformer model. The BiLSTM component captures sequential dependencies in the data, while the Transformer attention mechanism effectively learns complex feature interactions within the flow and thermal fields. The trained model is capable of rapidly predicting airflow velocity and temperature distributions with high accuracy. Model performance is evaluated using statistical metrics including Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). Experimental results demonstrate strong agreement between the predicted results and CFD simulations, achieving R² values greater than 0.98 for both velocity and temperature predictions. Furthermore, the model shows strong generalization capability when tested on unseen inlet configurations. The proposed framework significantly reduces computational time compared to traditional CFD simulations while maintaining reliable prediction accuracy. This approach provides an efficient and practical solution for rapid airflow analysis, smart ventilation design, and energy-efficient building management.
Zitieren
APA
Manikatnta, M. S., Suresh, M. G., Rahul, J., Kumar, V., & kumar, P. A. (2026). Integration of Hybrid Deep Learning with CFD Simulations for Flow and Temperature Field Prediction. International Journal of Advanced Research in Science Communication and Technology, 244. https://doi.org/10.48175/ijarsct-32737
Vancouver
Manikatnta MS, Suresh MG, Rahul J, Kumar V, kumar PA. Integration of Hybrid Deep Learning with CFD Simulations for Flow and Temperature Field Prediction. International Journal of Advanced Research in Science Communication and Technology. 2026:244. doi:10.48175/ijarsct-32737
BibTeX
@article{Manikatnta2026,
author = {Manikatnta, Mr. S. and Suresh, Mr. G. and Rahul, J. and Kumar, Vinay and kumar, P. Ajay},
title = {Integration of Hybrid Deep Learning with CFD Simulations for Flow and Temperature Field Prediction},
journal = {International Journal of Advanced Research in Science Communication and Technology},
year = {2026},
pages = {244},
doi = {10.48175/ijarsct-32737}
}
EUVIMED