A partitioned quasi-Newton solution technique for fluid-structure interaction problems using a coarsened grid to accelerate the convergence of the coupling iterations

Previous stability analyses on Gauss-Seidel coupling iterations in partitioned fluid-structure interaction simulations have demonstrated that Fourier modes with a low wave-number in the difference between the current and correct interface displacement are unstable. To stabilize these modes, the IQN-ILS technique automatically constructs a least-squares model of the flow solver and structural solver. In this work, the multi-level IQN-ILS technique (ML-IQN-ILS) is presented, which uses a coarsened grid of the fluid and structure subdomains to initialize this least-squares model. As the modes that need to be present in this least-squares model have a low wave-number, they can be resolved on a coarsened grid. Therefore, in each time step, a number of cheap coupling iterations is first performed on the coarsened grid to construct the model, followed by a smaller number of coupling iterations on the fine grid. As the iterations on the coarse grid are fast and fewer iterations are performed on the fine grid, the total duration of the simulation decreases compared to a simulation on the fine grid only