SC²S Colloquium - June 01, 2016
|Date:||June 1, 2016|
|Time:||3:00 pm, s.t.|
Alexander Rusch: Extending SU2 to fluid-structure interaction via preCICE
Partitioned fluid-structure interaction (FSI) simulations involve a fluid solver, a solid solver and a tool, which manages the coupling of the former two. In my bachelor’s thesis, the computational fluid dynamics suite Stanford University Unstructured (SU2) is linked with the multiphysics coupling library Precise Code Interaction Coupling Environment (preCICE). Therefore, a C++ adapter is developed, which is integrated into the source code of SU2. The coupling with preCICE allows to flexibly choose any partner (solid mechanics) code, including well-validated, commercial solvers. The coupling approach is successfully tested with two- and three-dimensional, generic scenarios, as well as quantitatively validated with a well-known FSI benchmark problem. Finally, the adapter is used to simulate a single wire of a brush seal under turbulent flow conditions with up to 230 processes on the fluid domain. This exemplifies the suitability of the realized coupling for real-world applications of larger scale.
Jan Sültemeyer: Uncertainty Quantification in Fluid Flows via Polynomial Chaos Methodologies
The focus of this thesis lies on the comparison of different methodologies for uncertainty quantification in computational fluid dynamics. Two methods based on polynomial chaos expansions – namely the pseudo spectral approach and the stochastic Galerkin method – are introduced and employed for modeling the forward propagation of uncertainty. A simulation based on Monte Carlo sampling is performed for the purpose of validation. The chosen flow scenario is a two dimensional lid driven cavity, simulated by solving the Navier Stokes equations via a finite difference scheme. The viscosity of the fluid is assumed to be uncertain, and it is modeled as a random variable with a Gaussian probability distribution. The influence of this uncertainty on the pressure and the velocity of the fluid at different points in the domain is studied. Their statistical properties – i.e. the mean value and the variance – are computed, and their probability density functions are estimated following a kernel density approach. Both methodologies are used for these computations and are then compared with respect to their convergence behavior, their computational cost, and the effort needed for their implementation.