Event Information:

• Thu

  16

  Jul

  2026

  SC Seminar: Muhammed Agca

  10:15Room 32-349

  Muhammed Agca, RPTU Kaiserslautern-Landau, Kaiserslautern

  Title: Sparse Identification of Reduced-Order Dynamics for the Unsteady Flow around a NACA 0012 Airfoil
  

  Abstract:

  High-fidelity CFD simulations of unsteady flows generate large datasets and offer limited direct insight into the underlying dynamics. In this work, we develop a data-driven reduced-order modeling framework for periodic vortex shedding behind an airfoil, using simulation data obtained with SU2. The approach combines Proper Orthogonal Decomposition (POD) with Sparse Identification of Nonlinear Dynamics (SINDy), using a physics-informed Stuart–Landau library with symmetry constraints for the dominant mode pair and algebraic resonant libraries for higher harmonics. The resulting low-dimensional model captures the dominant coherent flow structures and reconstructs the full spatial pressure field with high accuracy. A systematic hyperparameter study suggests that the physics-informed library is essential for obtaining stable models at the considered truncation rank, while generic polynomial libraries fail in the tested cases. Finally, we discuss the influence of transient training data and its role in improving the robustness of the identified parameters.