Our lab is pleased to announce that the paper “Task-aware evolution in physics-informed neural networks: Application to Saint-Venant torsion problems” has been published in Engineering Applications of Artificial Intelligence. This study proposes a task-aware PINN framework that combines standard PINN, variable-scaling PINN (VS-PINN), and parametric PINN to accurately and efficiently solve Saint-Venant torsion problems. The approach achieves high accuracy for complex cross-sections, significantly improves stability for stiff geometric transitions, and enables retraining-free, real-time predictions across varying torque conditions—demonstrating strong potential as a mesh-free alternative to FEM for engineering design and optimization.

Our lab is pleased to announce that the paper “Comparative analysis of the flow in a realistic human airway” has been published in Physics of Fluids (2026) [link]. Led by Dr. Mario Rüttgers in collaboration with Inha University, RWTH Aachen University, and the Jülich Supercomputing Centre, this study presents an experimentally validated direct numerical simulation of airflow in a realistic human airway, from the nasal cavity to the sixth bronchial bifurcation. Using a high-fidelity lattice-Boltzmann approach, the work reveals how flow instabilities and pressure losses develop across key anatomical regions under different breathing conditions, providing a valuable benchmark for respiratory CFD, diagnostics, and surgical planning.

Our lab is pleased to announce that our paper, “An efficient multiscale coupling method for simulations of reactor-scale chemical vapor deposition with microstructural features” has been published in Advances in Engineering Software [link], with contributions from collaborators at POSTECH and Samsung Electronics. This work presents a high-fidelity multiscale CVD modeling framework that couples a full 3D reactor-scale model with a parameterized feature-scale model through an effective reaction-rate formulation. The proposed method significantly reduces computational cost while maintaining high accuracy, enabling practical large-scale CVD simulations.

Our lab is pleased to announce that the paper “Towards a widespread usage of computational fluid dynamics simulations for automated virtual nasal surgery planning” has been published in Future Generation Computer Systems [link]. This work was led by Dr. Mario Rüttgers, with contributions from collaborators at RWTH Aachen, University of Latvia, Kobe University, RIKEN, and Forschungszentrum Jülich. he study introduces an efficient hybrid lattice-Boltzmann and level-set–based CFD framework that significantly reduces computational cost for virtual nasal surgery planning. The method enables fast prediction of pressure loss, airflow balance, and tissue-removal implications, making CFD-based surgical planning far more accessible for clinical environments. his publication marks an important advancement toward widespread, automated, and affordable CFD-assisted surgical decision-making.

Professor Sangseung Lee was invited to give an expert talk on AI-CFD at the Korea Society for Computational Fluids Engineering (KSCFE).

Professor Sangseung Lee has been selected as the recipient of the Early Career Researcher Award by the Korean Society for Visualization (KSV), acknowledging excellence in early-stage research achievements.

We developed a neural network-based thermal model for Korea’s first lunar orbiter, Danuri, enabling real-time temperature prediction from the ground. By combining active and transfer learning, our model achieves high accuracy with drastically reduced computation time—serving as a digital twin for spacecraft thermal monitoring. This joint work with POSTECH and KARI pushes the frontier of AI-driven virtual metrology for space missions. Check our work at : [link]

Happy to host Prof. Romit Maulik (Penn State Univ.) for an invited talk and engaging discussion at Inha University! 

We are launching JK-FLOW (Japan-Korea Fluid Mechanics Online Workshop), an online seminar series on a wide range of topics in fluid mechanics. Please check JK-Flow website for participating!

Happy to host Prof. Kai Fukami (Tohoku Univ.) for an invited talk and engaging discussion at Inha University! 

DDFE hosted Prof. Kai Fukami (Tohoku Univ., Generalized Super-Resolution Analysis with Machine Learning of Turbulence), Prof. Heechang Lim (Pusan National Univ., Artificial Intelligence-Driven Patterning and Reconstruction of Turbulent Flow Structures), and Dr. Mario Rüttgers (Planing Urban Wind Turbines by Means of Machine Learning and Computational Fluid Dynamics) for a workshop of AI for fluid dynamics systems (Feb. 13). 

Our group members attended KSME's Scientific Machine Learning Workshop in Seoul, South Korea.

New publication in Physics of Fluids (POF)! We developed a CNN that predicts drag forces on rough surfaces using only topography data through an international collaboration (INHA, KTH (Sweden), and KIT (Germany)). Our model reveals key patterns linked to drag, offering insights into fluid dynamics without needing complex parameters. Check out how we're advancing data-driven drag prediction (Link)!