Abstract Summary
We consider methods for dynamic coupled problems, in particular partitioned solvers for fluid structure interaction where different sub-solvers are used for the fluid and solid domain. More specifically we want to create a detailed simulation of a jackdaw's feather to investigate how feathers generate lift and ultimately answer the question of how birds evolved flight. This requires a fast and robust solver. Specifically, we want a solver that uses few coupling iterations, allows for high order in both space and time, is time adaptive and allows for different timesteps in the sub-solvers. Using so called waveform iterations these goals have been achieved for thermal heat transfer problems. The waveform iteration with adaptive time steps consists of nested solvers that each have their termination criteria and tolerance. Despite its success for simulating thermal heat transfer problems, it is still an open question how one should select the tolerances and the termination criteria within the nested solvers. Additionally, there is a problem with convergence of the waveform iteration which does not exist for fixed time steps. For adaptive ones, it can happen that the time grids do not converge and thus neither the waveform iteration. For fluid structure interaction, waveform iterations have previously been combined with interface Quasi-Newton acceleration to achieve an efficient partitioned solver in the case where both sub-solvers use different fixed time steps. In this talk we present the general framework of waveform iterations and their implementation in the open source coupling library PreCICE. We present some insight into how to select the termination criteria and tolerance within the nested solvers in the waveform iteration. We also extend the Quasi-Newton waveform iterations to the time adaptive case, where both of the sub-solvers use an adaptive time stepping scheme. Lastly, we also show that using a time adaptive solver results in faster run times for the simulation of our bird feather.
