ESR5 | Henrik Wusterberg

Today’s engineering heavily relies on simulation software for designing new systems such as cars, aircraft, power plants, just to name a few. Simulations allow the engineer to predict the optimal size, shape and performance of a new system. However the reliability of the software varies due to the complexity of the underlying physics and the finite availability of computational resources.

Within this project, I will work towards more reliable simulations of fluid flow for industrial applications. Examples are the prediction of aerodynamic forces on cars where strong turbulence results from the geometry of mirrors or in the car’s wake. The key aspect of my project is the development of higher-order methods which are capable of resolving the complex physics of turbulent flow. Nevertheless, these high-fidelity simulations require vast amounts of computational resources limiting their application to industrial problems. Therefore, I am developing efficient numerical algorithms for the solution of the incompressible Navier-Stokes equations that describe the physics of low-speed flows.

Numerical algorithms for the Navier-Stokes equations must consider the mathematical, computational and physical aspects of this problem to find an efficient approach. My efforts are especially directed towards the mathematical and computational challenges. In particular, I am studying implicit time-stepping schemes and efficient parallelisation techniques. While implicit time-stepping increases the stability of the algorithm, an efficient parallelisation allows distributing the computational work to thousands of CPUs and, hence, decrease the time-to-solution for a given simulation. In combination, they provide critical improvements for the algorithm and, thereby, enable highly-resolved simulations of challenging industrial applications.