An adaptive Newton continuation strategy for the fully implicit finite element immersed boundary method

  • Authors:

  • R. H. W. Hoppe;C. Linsenmann

  • Affiliations:

  • Department of Mathematics, University of Houston, PGH 651, Houston, TX 77204-3008, USA and Institute of Mathematics, University of Augsburg, Universitätsstr. 14, D-86159 Augsburg, Germany;Institute of Mathematics, University of Augsburg, Universitätsstr. 14, D-86159 Augsburg, Germany

  • Venue:
  • Journal of Computational Physics
  • Year:
  • 2012

Quantified Score

Hi-index 31.45

Visualization

Abstract

The immersed boundary method (IB) is known as a powerful technique for the numerical solution of fluid-structure interaction problems as, for instance, the motion and deformation of viscoelastic bodies immersed in an external flow. It is based on the treatment of the flow equations within an Eulerian framework and of the equations of motion of the immersed bodies with respect to a Lagrangian coordinate system including interaction equations providing the transfer between both frames. The classical IB uses finite differences, but the IBM can be set up within a finite element approach in the spatial variables as well (FE-IB). The discretization in time usually relies on the Backward Euler (BE) method for the semidiscretized flow equations and the Forward Euler (FE) method for the equations of motion of the immersed bodies. The BE/FE FE-IB is subject to a CFL-type condition, whereas the fully implicit BE/BE FE-IB is unconditionally stable. The latter one can be solved numerically by Newton-type methods whose convergence properties are dictated by an appropriate choice of the time step size, in particular, if one is faced with sudden changes in the total energy of the system. In this paper, taking advantage of the well developed affine covariant convergence theory for Newton-type methods, we study a predictor-corrector continuation strategy in time with an adaptive choice of the continuation steplength. The feasibility of the approach and its superiority to BE/FE FE-IB is illustrated by two representative numerical examples.