Description
In the aerodynamic design process, many hundreds to thousands of computationally intensive
Computational Fluid Dynamics (CFD) simulations are performed to find an optimal design for the
given boundary conditions. In this work, the focus lies on the variation of the Courant-Friedrichs-
Lewy (CFL) number, which has a significant influence on the running time, but also on the
convergence of the simulation [1, 2].
The objective of this master’s thesis is to design, implement, and evaluate an adaptive CFLnumber
control algorithm integrated into an implicit Reynolds-Averaged Navier-Stokes
Computational Fluid Dynamics (RANS-CFD) solver [3] for steady-state turbomachinery
applications. The primary focus of this research is to leverage Machine Learning (ML) techniques
to adapt the CFL number during the simulation process based on the residual values obtained
from the RANS solver.
The first step will be a comprehensive survey of existing CFL-number control algorithms in CFD,
state-of-the-art ML-based approaches for adaptive controls and previous studies of similar
attempts for CFD solvers.
Afterwards, a provided RANS-CFD setup is used to simulate the reference CFD simulation. Then,
possible ML-based models are developed and systematically tested using this RANS-CFD setup.
Finally, one model is selected based on robustness and convergence speed.
Thematic Focus
- Literature research
- Selection of possible ML approaches
- Run Reference CFD simulation
- Modular Integration of ML model in solver
- Run CFD solver with adaptive control
- Evaluate and Discuss Results
Literatur
[1] H. Ranocha et al., “On Error-Based Step Size Control for Discontinuous Galerkin Methods for
Compressible Fluid Dynamics,” Commun. Appl. Math. Comput., May 2023, doi: 10.1007/s42967-
023-00264-y.
[2] H. Sitaraman, D. Vaidhynathan, R. Grout, T. Hauser, C. M. Hrenya, and J. Musser, “An errorcontrolled
adaptive time-stepping method for particle advancement in coupled CFD-DEM
simulations,” Powder Technology, vol. 379, pp. 203–216, Feb. 2021, doi:
10.1016/j.powtec.2020.10.051.
[3] L. Lapworth, “Hydra-CFD: a framework for collaborative CFG development” in Proceedings of
International Conference on scientific and engineering computation, 2004
Contact