Enhancing Quasi-Newton Acceleration for Fluid-Structure Interaction

We propose two enhancements of quasi-Newton methods used to accelerate coupling iterations for partitioned fluid-structure interaction. Quasi-Newton have been established as flexible, yet robust, efficient and accurate multi-physics simulations in general. The library preCICE provides several variants, the so-called IQN-ILS method being most commonly used. It uses input output differences coupled solvers collected previous time steps approximate Newton iterations. To make both applicable parallel (where these contain data from different physical fields) provide a robust approach re-using information, combination information filtering scaling fields is typically required. This leads good convergence, but increases cost per iteration. new approaches—pre-scaling weight monitoring new, QR3 filter, substantially improve runtime while not affecting convergence quality. evaluate variety interaction examples. Results show that we achieve drastic speedups pure update steps. In future, intend apply also volume-coupled scenarios, where gains can be decisive feasibility approach.