Open-source framework for parametric study of hydrofoil profiles and motivation for using Physics-Informed Neural Networks (PINN)

Grm, Aleksander (author), Vukašinović, Nikola (author)

This research presents a robust, automated, open-source computational framework for generating high throughput hydrodynamic datasets for Physics Informed Neural Network (PINN) applications. Recognising the significant computational overhead of conventional Computational Fluid Dynamics (CFD) in parametric design spaces, this study proposes an integrated approach using GMSH for algorithmic mesh generation and OpenFOAM for high-fidelity fluid flow simulations. The framework is used to conduct an extensive parametric investigation on an HPC system for three distinct hydrofoil geometries: the NACA 0012, NACA 2412, and NACA 4412 profiles. Simulations are performed across a comprehensive operational envelope, covering angles of attack (α) from −15◦ to +15◦ and Reynolds numbers (Re) from 104 to 107. The resulting structured database contains spatial distributions of velocity and pressure fields, providing the empirical foundation for a deep learning architecture developed within the PyTorch ecosystem. To ensure physical consistency, the PINN architecture enforces the governing Navier Stokes equations by incorporating the conservation law residuals directly into the composite loss function. By automating the transition from geometric definition to numerical solution, the framework provides the necessary ground truth for training and validating PINNs. The accuracy of the surrogate model is evaluated by comparing the L2 relative error and pressure coefficient (Cp) distributions with OpenFOAM steady-state results. This approach aims to accelerate the prediction of hydrodynamic performance in foil-assisted vessel analysis while ensuring that the surrogate model strictly adheres to governing physical laws. The proposed methodology offers a scalable and accessible pathway for developing rapid-response surrogate models in maritime engineering.

• Language: English
• Keywords: hydrodynamics, parametric studies, high-performance computing (HPC)
• Work type: Abstract or summary
• Typology: 1.12 - Published Scientific Conference Contribution Abstract
• Organization: FPP - Faculty of Maritime Studies and Transport; FS - Faculty of Mechanical Engineering
• Publication status: Published
• Publication version: Version of Record
• Year: 2026
• Number of pages: Str. 60
• PID: 20.500.12556/RUL-188843
• UDC: 532.5:004.8
• COBISS.SI-ID: 275308547

Copyright

Licenca Creative Commons je navedena v kolofonu publikacije (https://ashpc.eu/event/27/attachments/154/410/ASHPC26_BOOKLET.pdf): “The abstracts in this booklet are licenced under a CC BY 4.0 licence (https://creativecommons.org/licenses/by/4.0/) ”. (Datum opombe: 29. 9. 2026).

• Publication date in RUL: 29.09.2026

Record is a part of a monograph

• Title: ASHPC26 : Austrian-Slovenian HPC Meeting 2026
• Editors: Ivan Vialov
• Place of publishing: Vienna
• Publisher: EuroCC
• ISBN: 978-3-200-10998-8
• COBISS.SI-ID: 275296259

Licences

• License: CC BY 4.0, Creative Commons Attribution 4.0 International
• Link: http://creativecommons.org/licenses/by/4.0/
• Description: This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.

Secondary language

• Language: Slovenian
• Keywords: hidrodinamika, parametrične analize, visokozmogljivostno računalništvo (HPC)