Benefits of Structured Cartesian Grids for the Simulation of Fluid-Structure Interactions

Due to the different physical effects involved and the complicated and changing geometries, the simulation of fluid-structure interactions still is a highly challenging task. Thus, simplicity and efficiency are essential for the corresponding software codes. In this paper, we consider the potential of structured Cartesian grids with respect to numerical and hardware efficiency as well as the development of user friendly tools for partitioned simulations of fluid-structure interactions. In contrast to the so-called monolithic approach, the partitioned approach does not establish a new complete solver for the whole scenario but instead embeds two existing solver – for fluid flow and structural mechanics – into a bigger environment handling the combination of both fluid and structure. Our fist task concerns the fluid solver. Here, we exploit the structuredness and spatially recursive properties of structured adaptive Cartesian grids to bring together both numerical efficiency (in the form of multigrid solvers on adaptively refined grids) and hardware efficiency (in particular memory efficiency). The second task is to provide tools for coupling the two solvers involved in an easy, modular, and flexible way. For this purpose, we developed a coupling client FSI*ce with clearly defined interfaces and a central description of the coupling surface between fluid and structure hiding the two solver grids from each other and, thus, making the solver completely independent from each other. In addition, FSI*ce controls the whole coupled simulation which leads to a central implementation of the coupling strategy (instead of an implementation in one of the solvers). Finally, libraries for general tasks such as data interpolation and projection are established. Here, our Cartesian grids come into play again as they are used as a very efficient tool for the administration and modification of the geometric information such as neighborhood relations between nodes of the spatial solver grids and the central surface description (lower dimensional triangulation).