===firstname:
Klaudius
===firstname3:

===affil6:

===lastname3:

===email:
klaudius.scheufele@ipvs.uni-stuttgart.de
===keyword_other2:

===lastname6:

===affil5:

===lastname4:

===lastname7:

===affil7:

===postal:
Institute for Parallel and Distributed Systems,
University of Stuttgart
Universitätsstraße 38
D-70569 Stuttgart
===ABSTRACT:
Multi-secant quasi-Newton methods based on secant information produced with no overhead throughout subsequent solver
iterations have been shown to be particularly suited to solve non-linear fixed-point equations that arise from 
partitioned multi-physics simulations where the exact Jacobian is inaccessible. In all these methods, the multi-secant
equation for the approximate (inverse) Jacobian is enhanced by a norm minimization condition. It is well-known, that
fluid-structure simulations, e.\,g., typically require the use of secant-information from previous time steps. The number 
of these time-steps highly depends on the application, its parameters, the used solvers, and the mesh resolution. Using 
to few leads to a relatively high number of iterations, using too many not only to a computational overhead but also to 
an increase of the number of iterations, as well. Determining the optimal number requires a costly try-and-error process,
which can be avoided using a modified method that considers the difference of the
current (inverse) Jacobian and the one of the previous time step in the norm-minimization. Thus, previous time step
information is taken into account in an implicit and automatized way without magic parameters. We show numerical results
for fluid-structure interactions for both methods proving the robustness and numerical efficiency of the second variant. 
In addition, we propose a new algorithm eliminating the drawback of having to store full interface Jacobian approximations
for the Jacobian difference norm minimization. This results in a highly efficient, parallelizable, and robust iterative
solver applicable for surface coupling in many types of multi-physics simulations.
===affil3:

===title:
Multi-Secant quasi-Newton Variants for Parallel Fluid-Structure Simulations 
-- and Other Multi-Physics Applications
===affil2:
Institute for Parallel and Distributed Systems, University of Stuttgart
===lastname2:
Mehl
===firstname4:

===keyword1:
Nonlinear solution methods, nonlinear least squares
===workshop:
no
===lastname:
Scheufele
===firstname5:

===keyword2:
Coupled multi-physics problems: electromagnetics/fluids
===otherauths:

===affil4:

===competition:
yes
===firstname7:

===firstname6:

===keyword_other1:

===lastname5:

===affilother:

===firstname2:
Miriam