No label defined (Q4933766)

The asymptotic numerical method (ANM) has been widely applied to solve path following problems with nonlinear mechanics. In this paper, a new iterative method is presented based on Padé approximants for numerical analysis of nonlinear problems. This technique is extended to iterative algorithms in order to be able to solve similar problems avoiding the computation of the whole solution path. For instance, the presented algorithms can be used in the correction step of prediction-correction algorithms. The basic idea is, firstly, to introduce an artificially modified problem that depends on a parameter \(\varepsilon\), and secondly to apply the ANM, the path parameter being \(\varepsilon\). This simple trick is well known in a purely analytic framework, but it has seldom been applied in numerical analysis. The ANM associates a perturbation technique with a classical discretization procedure, generally by finite elements. The best results are obtained when the problem can be written in a quadratic framework. Nonlinear elasticity and Navier-Stokes equations can be put easily in such a framework, but also problems of structural analysis involving unilateral contact, plasticity or viscoplasticity. In this paper the presented iterative algorithms based on power series and Padé approximants are tested in the case of an elastic shell problem and compared with classical iterative algorithms.