Contribution to a conference proceedings/Contribution to a book

FZJ-2025-02459

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Predicting Turbulent Boundary Layer Flows Using Transformers Coupled to the Multi-Physics Simulation Tool m-AIA

Sarma, R. (Corresponding author)FZJ* ; Hübenthal, F. ; Orland, F. ; Terboven, C. ; Lintermann, A.FZJ*

2025
Forschungszentrum Jülich GmbH Zentralbibliothek, Verlag Jülich

This record in other databases:

Please use a persistent id in citations: doi:10.34734/FZJ-2025-02459

Abstract: Time-marching of turbulent flow fields is computationally expensive with traditional numerical solvers. In this regard, transformer neural network, which has been largely successful in many other technical and scientific domains, can potentially predict complex flow fields faster compared to physics-based solvers. In this study, a transformer model is trained for a turbulent boundary layer problem, which is then coupled to the multi-physics solver m-AIA to make predictions of velocity fields. The method can potentially contribute to significant reduction in computational effort while maintaining high accuracy.

---

Contributing Institute(s):

1. Jülich Supercomputing Center (JSC)

Research Program(s):

1. 5111 - Domain-Specific Simulation & Data Life Cycle Labs (SDLs) and Research Groups (POF4-511) (POF4-511)
2. RAISE - Research on AI- and Simulation-Based Engineering at Exascale (951733) (951733)

Appears in the scientific report 2025

---

Database coverage:

---