Analyse théorique et numérique des conditions de glissement pour les fluides et les solides par la méthode de pénalisation

Authors: Dione, Ibrahima
Advisor: Urquiza, José ManuelFortin, André
Abstract: We are interested in the classical stationary Stokes and linear elasticity equations posed in a bounded domain [symbol] with a curved and smooth boundary [symbol], associated with slip and ideal contact boundary conditions, respectively. The finite element approximation of such problems can present difficulties because of a Babuška-Sapondžyan’s like paradox: solutions in polygonal domains approaching the smooth domain do not converge to the solution in the limit domain. The objective of this thesis is to explore the application of the penalty method to these slip boundary conditions, in particular in order to overcome this paradox. The penalty method is a classic method widely used in practice because it allows to work in functional spaces without constraints and avoids adding new unknowns like with the Lagrange multiplier method. The first part of this thesis is devoted to the 2D numerical study of different finite elements choices and, most importantly, of different choices of the approximation of the normal vector to the boundary of the domain. With the (discontinuous) normal vector to polygonal domains [symbol] generated with the meshing of [symbol], the finite element solutions do not seem to converge to the exact solution. However, if we use a (continuous) regularization of the normal, isoparametric finite elements of degree 2 for the velocity (or the displacement for elasticity) or a reduced integration of the penalty term, convergence is obtained, with optimal rates in some cases. In a second part, we make a theoretical analysis (in dimensions 2 and 3) of the convergence. The a priori estimates obtained allow to say that even with the (discontinuous) normal vector to polygonal domains, the finite element approximation converges to the exact solution when the penalty parameter is selected appropriately in terms of the size of the elements, showing that the paradox can be circumvented with the penalty method.
Document Type: Thèse de doctorat
Issue Date: 2013
Open Access Date: 19 April 2018
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
Description SizeFormat 
30379.pdfTexte4.87 MBAdobe PDFThumbnail
All documents in CorpusUL are protected by Copyright Act of Canada.