Rosenbrock-Wanner Methods for Scale-Resolving Simulations

Abstract

Scale-resolving simulations in computational fluid dynamics (CFD) have become increasingly relevant in recent years due to ever improving computational capabilities of modern HPC clusters. Such simulations lead to a complex interplay of physical modeling and the spatial and temporal discretization used to compute a numerical solution to the model. In this thesis focus on implicit temporal discretizations, more specifically Rosenbrock-Wanner (ROW) methods and diagonally implicit RungeKutta (DIRK) methods. ROW methods have been added recently to the next generation CFD code jointly developed by ONERA, DLR and Airbus (henceforth referred to as CODA). In this thesis, we would like to offer a general understanding of the advantages and disadvantages of ROW methods. Furthermore, we would like to compare the ROW methods to DIRK methods in terms of accuracy and performance for scale-resolving simulations. We aim to provide a best practice choice of a time-discretization scheme for this application in CODA.