JK-FLOW Japan-Korea Fluid Mechanics Online Workshop

Speaker: Mr. Shilaj Baral (Ph.D. student, POSTECH) [GS]

Author list: Shilaj Baral [1], Youngkyu Lee [2], Sangam Khanal [1], and Joongoo Jeon [3]

1. Graduate School of Integrated Energy-AI, Jeonbuk National University, Republic of Korea 

2. Division of Applied Mathematics, Brown University, United States

3. Division of Advanced Nuclear Engineering, Pohang University of Science and Technology, South Korea

Abstract: The practical utility of hybrid methods for accelerating computational fluid dynamics (CFD) simulations hinges on their ability to generalize across diverse conditions and scale to complex, three-dimensional problems. This study investigates these critical properties using XRePIT, a novel, fully automated framework designed for this purpose. The framework couples OpenFOAM with machine learning surrogates, using a residual-guided loop to ensure long-term stability and physical accuracy. To assess generalization, we tested the hybrid method against multiple boundary conditions and demonstrated its architectural extensibility by seamlessly swapping between a finite-volume method network (FVMN) and a Fourier neural operator variant (FVFNO). To prove scalability, we extended the application from 2D benchmarks to a full 3D simulation of buoyancy-driven flow. Our results validate the approach, achieving stable accelerations of up to 3.68× in 2D and a significant 4.98× in 3D. Across all configurations, the method maintained long-term stability for over 10,000 timesteps with less than 1% error. This work provides critical evidence that the residual-guided hybrid strategy is not only a viable concept but a scalable and generalizable solution, marking a practical step towards applying ML-accelerated CFD to real-world, 3D engineering challenges.

Speaker: Dr. Yuting Guo (Assistant Professor, Kyoto University) [GS]

Abstract: To clarify the reaction behavior of ammonia in the presence of water, this study combines experimental evaluations, molecular dynamics (MD) simulations, and density functional theory (DFT) calculations to investigate the adsorption, diffusion, and dissociation of ammonia and water at the Ni-YSZ interface. Experimental results indicate that water inhibits the ammonia decomposition, especially at lower temperature. MD simulations reveal that on the YSZ surface, water forms multilayer adsorption structures, while ammonia exhibits monolayer adsorption with higher surface mobility and temperature dependence. Competitive adsorption and diffusion between water and ammonia alter gas composition near the Ni surface and modulate reaction pathways. DFT calculations further show that the Ni–YSZ interface provides stronger adsorption sites and lower dissociation barriers for ammonia and water molecules than individual Ni or YSZ surfaces, serving as a favorable site for molecular accumulation and activation. The comparable dissociation activation energies of water and ammonia enable competition at the initial dissociation step, thereby suppressing the ammonia decomposition.

Speaker: Mr. Ryo Koshikawa (Undergraduate student, Tohoku University)

Author list: Ryo Koshikawa [1], Ryo Araki [2], Qiong Liu [3], and Kai Fukami [1]

1. Department of Aerospace Engineering, Tohoku University 

2. Department of Mechanical and Aerospace Engineering, Tokyo University of Science 

3. Department of Mechanical and Aerospace Engineering, New Mexico State University

Abstract: Aerodynamicists have aimed to capture dominating features of flow fields to understand complex unsteady flow and aerodynamic phenomena. Indeed, our eyeballs have persistently sought the causal relationship between vortical structures and aerodynamic forces. To grasp such an aerodynamic causal relationship in an interpretable manner, this talk discusses information-theoretic manifold learning that provides a time-varying informative modal structure with respect to variables of interest in the future. The current convolutional neural network-based approach produces spatially continuous modes, which has been challenging to achieve with conventional techniques. Furthermore, a low-order submanifold representing informative vortical structures and the aerodynamic coefficient is identified. The proposed technique is applied to a range of aerodynamic flow examples, including extreme vortex-gust airfoil interaction, experimental measurements of transverse jet-wing interaction, and a turbulent separated wake. We reveal the causal relationship between vortical structures and the lift dynamics relying only on flow data and information metrics. This work will contribute not only to a deeper understanding of unsteady aerodynamics but also to obtaining insights into modeling and controlling fluid flows.

Speaker: Mr. Jaewon Jang (Graduate student, Inha University)

Abstract: Generative models such as GANs and diffusion models have been widely adopted for surrogate modeling and super-resolution in fluid dynamics. However, these approaches, originally developed for computer vision tasks, prioritize generation over prediction and introduce stochastic artifacts through Gaussian noise-based processes, compromising physical consistency in data-scarce regimes. We present the Energy Dissipation Model (EDM), a deterministic framework fundamentally designed for flow field prediction rather than generation. Inspired by the Kolmogorov energy cascade, EDM employs frequency-based progressive refinement, sequentially reconstructing flow structures from large-scale (low-frequency) to small-scale (high-frequency) components. This physics-oriented hierarchy enables accurate predictions with substantially reduced training data while minimizing both global and local reconstruction errors. Unlike existing methods that adapt computer vision techniques to fluid dynamics, EDM establishes an independent framework grounded in turbulence physics, offering a principled approach to data-driven flow prediction that respects the multi-scale nature of turbulent flows.

Speaker: Ms. Morie Koseki (Ph.D. student, Okinawa Institute of Science and Technology)

Abstract: When fluids flow over deformable walls, the flow field is modified by the fluid-wall interactions. The complex coupling can be described by a combination of many undistinguished effects, e.g., roughness effects (due to the wall deformation by the hydrodynamic force), non-zero wall-normal fluctuations (coming from the wall movement), and wall motion (owing to the wave propagation on the surface and inside materials). This study aims to disentangle the effects of fluid-structure interaction and wall shape/undulations individually in a turbulent flow. We conduct direct numerical simulations of turbulent channel flows over deformable and statistically equivalent rough walls.  Turbulent flows over relatively rigid compliant walls share similar features to those over rough walls; however, as wall flexibility increases, distinct effects are observed that are specific to the mutual fluid-structure interaction.

Speaker: Dr. Seongsu Cho (Postdoctoral Researcher, University of Pennsylvania) [GS]

Abstract: Droplet microfluidics offers great versatility for applications ranging from functional particle fabrication to biological assays. Despite its potential, manually refining experimental conditions to achieve desired droplet properties remains a significant bottleneck, hindering widespread adoption. To address this, we developed an AI-driven control system for the automated optimization of droplet generation. Integrating convolutional neural networks (CNNs) for real-time analysis and Bayesian optimization (BO) for refining experimental conditions, the system works with minimal human intervention. Due to efficient utilization of datasets in BO, the amount of training datasets was reduced, identifying optimal conditions within 15 iterations on average. We demonstrated robustness of the system across various working fluids, channel geometries, and droplet morphologies. This system is expected to accelerate research using droplet microfluidic system.