Computational & Technology Resources
an online resource for computational,
engineering & technology publications
PROCEEDINGS OF THE SIXTH INTERNATIONAL CONFERENCE ON ENGINEERING COMPUTATIONAL TECHNOLOGY
Edited by: M. Papadrakakis and B.H.V. Topping
A Modular and Efficient Simulation Environment for Fluid-Structure Interactions with Large Domain Deformation
M. Mehl, M. Brenk, I.L. Muntean, T. Neckel and T. Weinzierl
Department of Computer Science, Technical University Munich, Germany
M. Mehl, M. Brenk, I.L. Muntean, T. Neckel, T. Weinzierl, "A Modular and Efficient Simulation Environment for Fluid-Structure Interactions with Large Domain Deformation", in M. Papadrakakis, B.H.V. Topping, (Editors), "Proceedings of the Sixth International Conference on Engineering Computational Technology", Civil-Comp Press, Stirlingshire, UK, Paper 9, 2008. doi:10.4203/ccp.89.9
Keywords: fluid-structure interactions, partitioned approach, modularity, large deformations, coupling environment, Cartesian grids, hardware-efficiency.
Modularity and flexibility are properties which seem to be inherent to partitioned fluid-structure interaction simulations. This is surely true from the conceptual point of view. However, it is not trivial not to loose the independence of the involved components in an actual implementation. Many approaches introduce dependencies between the two solver codes (fluid and structure) by the direct mapping of data between the two (non-matching) solver grids at the interface between fluid and structure. In other cases, the coupling strategy (for example explicit-implicit coupling) is implemented in one of the solvers which unneccessarily complicates the exchange of the respective solver. Our newly developed coupling environment FSI*ce avoids these drawbacks by the introduction of a third, independent mesh describing the fluid-structure interface and by a client-server approach with the two solvers acting as servers and a client controlling the whole coupled simulation and, therewith, of course also defining the coupling strategy regardless of the actual solvers. This gives us maximal flexibility both in the choice of the two solvers but also in the choice of coupling strategies including sophisticated methods such as multilevel algorithms or reduced order models which are very hard or even impossible to implement without a strict independence of solvers and coupling strategy.
In addition to these requirements, a widely usable software environment for fluid-structure interactions has to be able to efficiently handle large deformations of the computational domain. In particular this holds for the fluid domain. Thinking of particles moving through the fluid, e.g. to handle this requirement, we present a fluid solver working on (adaptive) Cartesian grids in an Eulerian framework that is with a fixed grid. We show methods for the fast update of the grid after a change of the geometry, to efficiently map data from these Cartesian grids to the central interface mesh of FSI*ce as well as robust interpolation schemes tailored to Cartesian grids and preventing instabilities.
Finally, we will present results achieved with our Cartesian flow solvers and FSI*ce for benchmark and application scenarios in various configurations.
purchase the full-text of this paper (price £20)