Approximation of Obstacle Problems by Continuation Methods

Authors:
  • FCFrancis Conrad
    Francis Conrad
  • RHRaphaele Herbin
    Raphaele Herbin
  • HMHans D. Mittelmann
    Hans D. Mittelmann

Abstract
Nonlinear variational inequalities associated with obstacle problems are considered, which may exhibit multiple solutions and turning points. A predictor-corrector continuation method for following a solution branch is presented. Essentially, it is shown that “regular” or “simple singular” solutions of the nonlinear eigenvalue problem are regular for an augmented problem. The main difficulty here is due to the fact that the equations are not differentiable in the Fréchet sense. The numerical schemes that are used in the implementation of the method are developed. For the predictor step, a deflated block algorithm is used, and for the corrector step, algorithms are set up using projected SOR, projected Newton, and projected preconditioned conjugate gradients. The efficiency of the various algorithms is compared by applying the continuation method to a finite element discretization of a two-dimensional problem. Numerical results are presented.

This publication has 24 references indexed in Scilit: