Partioned Solution and Convergence Acceleration
Highlights
- Partitioned coupling is contrasted with monolithic approaches, offering the flexibility to reuse specialized black-box solvers via coupling software like comana or preCICE at the cost of coupling iterations.
- The coupled interface problem is formulated as a fixed-point iteration, using minimal invasive solver modifications to exchange boundary conditions and states within each time step.
- Convergence acceleration is required to stabilize iterations and reduce computational effort, with relaxation, line search, and Quasi-Newton methods presented as the main families.
- Aitken (Irons-Tuck) relaxation dynamically updates the relaxation factor from residuals, achieving 17% fewer iterations than constant relaxation without parameter tuning, while Quasi-Newton Least Squares reduces iterations by threefold.
- Temporal discretization schemes compare implicit and explicit coupling, highlighting an implicit-explicit setup with subcycling to pair different solver time step sizes.
Updated on Nov 3, 2026