A journal article, “Enhancing the accuracy of physics-informed neural networks for indoor airflow simulation with experimental data and Reynolds-averaged Navier–Stokes turbulence model”, authored by a team of PolyU research students and researchers, was featured in Volume 36, Issue 6 of Physics of Fluids. The paper pertains to the physics-informed neural network (PINN). The lack of enough validations on more complex flow problems has curbed further development and application of PINN.

Supervised by Prof. WEN Chih-Yung, Head and Chair Professor from the Department of Aeronautical and Aviation Engineering and Associate Director of the Research Institute for Sports Science and Technology, the team included Ms JIA Yuan, PhD student; Dr LI Zhengtong, Research Assistant Professor from the Department of Aeronautical and Aviation Engineering and Corresponding Author; as well as Mr ZHANG Chi, PhD student; and Dr CHEN Zheng-wei, Research Assistant Professor from the Department of Civil and Environmental Engineering.

The research finds that the PINN prediction accuracy can be significantly improved by exploiting its ability to assimilate high-fidelity data during training, despite the challenge experienced by the PINN to reach an ideal accuracy for the problem through a single purely physics- driven training. Meanwhile, the influence of the number of data points is also examined, suggesting a balance between prediction accuracy and data acquisition cost can be reached. Furthermore, applying Reynolds-averaged Navier–Stokes (RANS) equations and turbulence model has also been proved to refine prediction accuracy remarkably.

Zhang, C., Wen, C.-Y., Jia, Y., Juan, Y.-H., Lee, Y.-T., Chen, Z., Yang, A.-S., & Li, Z. (2024). Enhancing the accuracy of physics-informed neural networks for indoor airflow simulation with experimental data and Reynolds- averaged Navier–Stokes turbulence model. Physics of Fluids (1994), 36(6). https://doi.org/10.1063/5.0216394