COUPLED 2023

Recent Developments in Quasi-Newton Techniques for Partitioned Simulation of Coupled Problems

  • Degroote, Joris (Ghent University)
  • Schulte, Miriam (University of Stuttgart)
  • Hosters, Norbert (RWTH Aachen University)

Please login to view abstract download link

Many fluid-structure interaction problems can be simulated in a partitioned way, which means that a flow solver is coupled with a structural solver, and other coupled problems can be simulated in a similar way. If the interaction between the solvers is strong, then usually coupling iterations are used in each time step and quasi-Newton algorithms are frequently applied to stabilize and accelerate the convergence of these iterations. In this contribution, an overview of recent developments in the field of quasi-Newton coupling techniques will be provided. Firstly, different options for the incorporation of data from previous time steps in the approximate Jacobians will be compared and several methods will be classified in the generalized Broyden framework. Second, recent developments to make these methods suitable for parallel and high-performance computing will be reviewed. Third, new combinations of quasi-Newton methods with a Robin-Neumann decomposition will be listed. In addition, we shortly compare quasi-Newton methods developed for partitioned coupling with similar methods used for the acceleration of fixed-point iterations in general as well as for second order optimization and learning. Also several other contributions to the mini-symposium “Quasi-Newton Techniques for Partitioned Simulation of Coupled Problems” will be highlighted. This presentation is meant to increase awareness of all these approaches in order to foster synergies within and across communities.