Thursday 10 Mar 2016: Partitioned Approaches for Multi-Physics Simulations -- From First Ideas to Today's Robust Approaches
Miriam Mehl - University of Stuttgart
Partitioned approaches for multi-phsics simulations have been established for fluid-structure interactions already in the late 90s (Park, Felippa et al.). The idea was to set up this new class of simulation without the time-consuming process of re-implementing a whole simulation code from scratch. Instead, sophisticated and trusted simulation software for fluid flow and structural dynamics were re-used and coupled in an as tight as neccessary fashion. These first approaches allowed for an easy and fast realization of fluid-structure simulation environments but still had issues in terms of stability and convergence speed of coupling iterations. In addition, they were not designed to efficiently make use of parallel compute architectures. Due to these issues and with this type of simulation becoming an important and widely used application, more and more monolithic codes setting up and solving the coupled system as a whole were developed. Obviously, they allow for highly optimized solver implementations for the underlying ill-conditioned systems of equations. At the same time also the partitioned coupling approach has developed into a robust class of iterative solvers for multi-physics problems and still has the great advantage of proving full flexibility in exchanging involved models, solvers, coupling methods, or adding further physical fields in a very short development time. The presentation gives an overview of the past and current developments in partitioned multi-physics simulations including inter-code communication, data mapping between non-matching meshes, iterative equation coupling, and software tools.