High-Fidelity and Efficient Direct Aeroacoustic Simulations

Aeroacoustics studies the propagation of sound waves, which are generated by aerodynamic effects. Examples are given by the flow around the wing of an airplane, the flow around buildings or bridges by the wind, or the generation of a sound with an organ. The complex correlations between sound waves (specified by pressure and density variations), flow velocity and temperature are described by the compressible Navier-Stokes equations (NSE). Depending on the problem, an exact solution of the NSE is not given, which is why a simulation is necessary. The goal of the project is to develop a deeper understanding of the causes and effects of aeroacoustic phenomena. For this purpose, a new software package called DREAM (DiRect aEroAcoustic siMulations) is being developed, which will allow an exact approximation of aerodynamically generated sound waves. During the development, we focus on the flexibility of the considered applications, as well as on efficiency and accuracy. As a basis for the simulation, we will make use of the finite element method (FEM). Thereby, the geometry (e.g., a wing of an airplane or a bridge) is divided into small segments which reduces the complexity of the problem and makes a discretization of the NSE possible. A problem that arises in aeroacoustic problems is that the sound waves usually travel very fast and accordingly far away from the (aerodynamic) sound sources. This means that we have to use a large computational domain during the simulation. For example, consider the propagation of the sound of a flute in a large concert hall. The large domain (a few meters) and the subdivision into very small (a few millimeters) segments due to the FEM results in millions to billions of equations that must be solved during a simulation. In order to accomplish this efficiently, DREAM is being developed to run on "scientific computing clusters", i.e. on a network of computers optimized for computations. For this, the unknowns are divided into smaller groups and are computed in parallel on the network, which drastically reduces the simulation time. In addition to the requirement of an efficient computation, we also focus on the accuracy and stability of the FEM considered in this project. In order to provide this, a detailed mathematical analysis is continuously carried out, which also serves as a basis for the further development of DREAM.