» Back to overview
Proposing Institution

Simulationstechnik und Wissenschaftliches Rechnen, Uni Siegen
Project Manager

Prof. Dr.-Ing. Sabine Roller
Adolf-Reichwein-Straße 2
57068 Siegen
The increasing compute power on supercomputers makes it possible to simulate more physical effects as well as scales in single simulations. In the DFG funded SPPEXA project ExaFSA 2 we consider the simulation of sound generation in turbulent flow, the interaction with structure and the propagation of sound waves across large distances. The transition of turbulent flow to acoustic wave propagation is one of the major parts in this project and the most demanding part in terms of computational resources.This part we want to investigate especially in detail with this proposal.Simulating this problem with a monolithic approach is still too expensive and not feasible, when considering the numerical effort. Since these different effects appear in spatially separable areas, the whole domain can be subdivided into smaller domains and solved according with the best suited numerical approximation. For this purpose a coupling approach is necessary to allow the communication between the subdomains.Through the coupling approach we are able to solve linearized equations for the acoustic wave propagation, covering the large acoustic domain in a high-order method, which yields a low dissipation and dispersion error for the propagation of waves. The area, where nonlinearities have to be considered can than be restricted to a small area and resolved with a finer grid. To solve the necessary equations, the Discontinuous Galerkin Method (DGM) with a high-order scheme is used for both the linear equations as well as for the nonlinear equations. The implementation of the Discontinuous Galerkin scheme for hyperbolic conservation laws in our solver Ateles has already been proven to be highly scalable on SuperMUC. Our solver provides arbitrary high-order discretization and can be used for various equation systems. With the modal Discontinuous Galerkin scheme the complexity of the computation grows linearly with increasing number of degrees of freedoms per element. Thus, it allows high scheme orders beyond order 100, even for 3D testcases.To enable the interaction of the flow and the acoustic domain, two coupling approaches APESmate and preCICE are available. The coupling approach preCICE is based on a black box coupling, where just the point values at the surface of the coupling domains are known. In contrast AP E S mate has knowledge about the numerical schemes within the domain. Thus, preCICE needs to interpolate values, while AP ESmate can eval- uate the high order polynomials of the underlying Discontinous Galerkin scheme. Hence, the preCICE approach is more generally applicable, while the AP E S mate approach is more efficient, especially in the context of high order schemes. preCICE is developed by the chairs of Scientific Computing in Munich and Simulation of Large Systems in Stuttgart. APESmate is part of our APES framework, which provides pre- and post-processing tools. In this proposal we mainly concentrate on the coupling with the coupling approach APESmate, while in the previous work mainly the coupling approach preCICE was used.

Impressum, Conny Wendler