No description available.

La thèse porte sur l’analyse d’erreur a posteriori pour la résolution numérique de l’équation linéaire des ondes , discrétisée en temps par le schéma de Newmark et en espace par la méthode des éléments finis. Nous adoptons un choix particulier de paramètres pour le schéma de Newmark, notamment β = 1/4, γ = 1/2, qui assure que la méthode est conservative en énergie et d’ordre deux en temps. L’estimation d’erreur a posteriori, d’un ordre optimal en temps et en espace, est élaborée à partir de la discrétisation complète. L’erreur est mesurée dans une norme qui découle naturellement de la physique: H1 en espace et Linf en temps. Nous proposons d’abord un estimateur dit «à 3 points» qui fait intervenir la solution discrète en 3 points successifs du temps à chaque pas de temps. Cet estimateur fait appel à une approximation du Laplacien de la solution discrète qui doit être calculée à chaque pas de temps, en résolvant un problème auxiliaire d'éléments finis. Nous proposons ensuite un estimateur d’erreur alternatif qui permet d’éviter ces calculs supplémentaires: l’estimateur dit «à 5 points» puisqu’il met en jeu le schéma des différences finies d’ordre 4, qui fait intervenir la solution discrète en 5 points successifs du temps à chaque pas de temps. Nous démontrons que nos estimateurs en temps sont d’ordre optimal pour des solutions suffisamment lisses, sur des maillages quasi-uniformes en espace et uniformes en temps, en supposant que les conditions initiales soient discrétisées à l’aide de la projection elliptique. La trouvaille la plus intéressante de cette analyse est le rôle capitale de cette discrétisation : des discrétisations standards pour les conditions initiales, telles que l’interpolation nodale, peuvent être néfastes pour les estimateurs d’erreur en détruisant leur ordre de convergence, bien qu’elles fournissent des solutions numériques tout à fait acceptables. Des expériences numériques prouvent que nos estimateurs d’erreur sont d’ordre optimal en temps comme en espace, même dans les situations non couvertes par la théorie. En outre, notre analyse a posteriori s’étend au schéma de Newmark d’ordre deux plus général (γ = 1/2). Nous présentons des comparaisons numériques entre notre estimateur à 3 points généralisé et l’estimateur sur des grilles décalées, proposé par Georgoulis et al. Finalement, nous implémentons un algorithme adaptatif en temps et en espace basé sur notre estimateur d’erreur a posteriori à 3 points. Nous concluons par des expériences numériques qui montrent l’efficacité de l’algorithme adaptatif et révèlent l’importance de l’interpolation appropriée de la solution numérique d’un maillage à un autre, surtout vis à vis de l’optimalité de l’estimation d’erreur en temps. / This thesis focuses on the a posteriori error analysis for the linear second-order wave equation discretized by the second order Newmark scheme in time and the finite element method in space. We adopt the particular choice for the parameters in the Newmark scheme, namely β = 1/4, γ = 1/2, since it provides a conservative method with respect to the energy norm. We derive a posteriori error estimates of optimal order in time and space for the fully discrete wave equation. The error is measured in a physically natural norm: H1 in space, Linf in time. Numerical experiments demonstrate that our error estimators are of optimal order in space and time. The resulting estimator in time is referred to as the 3-point estimator since it contains the discrete solution at 3 points in time. The 3-point time error estimator contains the Laplacian of the discrete solution which should be computed via auxiliary finite element problems at each time step. We propose an alternative time error estimator that avoids these additional computations. The resulting estimator is referred to as the 5-point estimator since it contains the fourth order finite differences in time and thus involves the discrete solution at 5 points in time at each time step. We prove that our time estimators are of optimal order at least on sufficiently smooth solutions, quasi-uniform meshes in space and uniform meshes in time. The most interesting finding of this analysis is the crucial importance of the way in which the initial conditions are discretized: a straightforward discretization, such as the nodal interpolation, may ruin the error estimators while providing quite acceptable numerical solution. We also extend the a posteriori error analysis to the general second order Newmark scheme (γ = 1/2) and present numerical comparasion between the general 3-point time error estimator and the staggered grid error estimator proposed by Georgoulis et al. In addition, using obtained a posteriori error bounds, we implement an efficient adaptive algorithm in space and time. We conclude with numerical experiments that show that the manner of interpolation of the numerical solution from one mesh to another plays an important role for optimal behavior of the time error estimator and thus of the whole adaptive algorithm.

A posteriori error estimation is an inseparable component of any reliable numerical method for solving partial differential equations. The aim of the goal-oriented a posteriori error estimates is to control the computational error directly with respect to some quantity of interest, which makes the method very convenient for many engineering applications. The resulting error estimates may be employed for mesh adaptation which enables to find a numerical approximation of the quantity of interest under some given tolerance in a very efficient manner. In this thesis, the goal-oriented error estimates are derived for discontinuous Galerkin discretizations of the linear scalar model problems, as well as of the Euler equations describing inviscid compressible flows. It focuses on several aspects of the goal-oriented error estimation method, in particular, higher order reconstructions, adjoint consistency of the discretizations, control of the algebraic errors arising from iterative solutions of both algebraic systems, and linking the estimates with the hp-anisotropic mesh adaptation. The computational performance is demonstrated by numerical experiments.

This work develops finite element methods with high order stabilization, and robust and efficient adaptive algorithms for Large Eddy Simulation of turbulent compressible flows. The equations are approximated by continuous piecewise linear functions in space, and the time discretization is done in implicit/explicit fashion: the second order Crank-Nicholson method and third/fourth order explicit Runge-Kutta methods. The full residual of the system and the entropy residual, are used in the construction of the stabilization terms. These methods are consistent for the exact solution, conserves all the quantities, such as mass, momentum and energy, is accurate and very simple to implement. We prove convergence of the method for scalar conservation laws in the case of an implicit scheme. The convergence analysis is based on showing that the approximation is uniformly bounded, weakly consistent with all entropy inequalities, and strongly consistent with the initial data. The convergence of the explicit schemes is tested in numerical examples in 1D, 2D and 3D. To resolve the small scales of the flow, such as turbulence fluctuations, shocks, discontinuities and acoustic waves, the simulation needs very fine meshes. In this thesis, a robust adjoint based adaptive algorithm is developed for the time-dependent compressible Euler/Navier-Stokes equations. The adaptation is driven by the minimization of the error in quantities of interest such as stresses, drag and lift forces, or the mean value of some quantity. The implementation and analysis are validated in computational tests, both with respect to the stabilization and the duality based adaptation. / QC 20110627

On considère l'équation de convection--diffusion--réaction instationnaire. On s'intéresse à la dérivation d'estimations d'erreur a posteriori pour la discrétisation de cette équation par la méthode des volumes finis centrés par mailles en espace et un schéma d'Euler implicite en temps. Les estimations, qui sont établies dans la norme d'énergie, bornent l'erreur entre la solution exacte et une solution post-traitée à l'aide de reconstructions $Hdiv$-conformes du flux diffusif et du flux convectif, et d'une reconstruction $H^1_0(Omega)$-conforme du potentiel. On propose un algorithme adaptatif qui permet d'atteindre une précision relative fixée par l'utilisateur en raffinant les maillages adaptativement et en équilibrant les contributions en espace et en temps de l'erreur. On présente également des essais numériques. Enfin, on dérive une estimation d'erreur a posteriori dans la norme d'énergie augmentée d'une norme duale de la dérivée en temps et de la partie antisymétrique de l'opérateur différentiel. Cette nouvelle estimation est robuste dans des régimes dominés par la convection et des bornes inférieures locales en temps et globales en espace sont également obtenues / We consider the time-dependent convection--diffusion--reaction equation. We derive a posteriori error estimates for the discretization of this equation by the cell-centered finite volume scheme in space and a backward Euler scheme in time. The estimates are established in the energy norm and they bound the error between the exact solution and a locally post processed approximate solution, based on $Hdiv$-conforming diffusive and convective flux reconstructions, as well as an $H^1_0(Omega)$-conforming potential reconstruction. We propose an adaptive algorithm which ensures the control of the total error with respect to a user-defined relative precision by refining the meshes adaptively while equilibrating the time and space contributions to the error. We also present numerical experiments. Finally, we derive another a posteriori error estimate in the energy norm augmented by a dual norm of the time derivative and the skew symmetric part of the differential operator. The new estimate is robust in convective-dominated regimes and local-in-time and global-in-space lower bounds are also derived

The thesis consists of four papers on numerical complexityanalysis of weak approximation of ordinary and partialstochastic differential equations, including illustrativenumerical examples. Here by numerical complexity we mean thecomputational work needed by a numerical method to solve aproblem with a given accuracy. This notion offers a way tounderstand the efficiency of different numerical methods. The first paper develops new expansions of the weakcomputational error for Itˆo stochastic differentialequations using Malliavin calculus. These expansions have acomputable leading order term in a posteriori form, and arebased on stochastic flows and discrete dual backward problems.Beside this, these expansions lead to efficient and accuratecomputation of error estimates and give the basis for adaptivealgorithms with either deterministic or stochastic time steps.The second paper proves convergence rates of adaptivealgorithms for Itˆo stochastic differential equations. Twoalgorithms based either on stochastic or deterministic timesteps are studied. The analysis of their numerical complexitycombines the error expansions from the first paper and anextension of the convergence results for adaptive algorithmsapproximating deterministic ordinary differential equations.Both adaptive algorithms are proven to stop with an optimalnumber of time steps up to a problem independent factor definedin the algorithm. The third paper extends the techniques to theframework of Itˆo stochastic differential equations ininfinite dimensional spaces, arising in the Heath Jarrow Mortonterm structure model for financial applications in bondmarkets. Error expansions are derived to identify differenterror contributions arising from time and maturitydiscretization, as well as the classical statistical error dueto finite sampling. The last paper studies the approximation of linear ellipticstochastic partial differential equations, describing andanalyzing two numerical methods. The first method generates iidMonte Carlo approximations of the solution by sampling thecoefficients of the equation and using a standard Galerkinfinite elements variational formulation. The second method isbased on a finite dimensional Karhunen- Lo`eve approximation ofthe stochastic coefficients, turning the original stochasticproblem into a high dimensional deterministic parametricelliptic problem. Then, adeterministic Galerkin finite elementmethod, of either h or p version, approximates the stochasticpartial differential equation. The paper concludes by comparingthe numerical complexity of the Monte Carlo method with theparametric finite element method, suggesting intuitiveconditions for an optimal selection of these methods. 2000Mathematics Subject Classification. Primary 65C05, 60H10,60H35, 65C30, 65C20; Secondary 91B28, 91B70. / QC 20100825

In the first part of this thesis we are concerned with the constitutive the- ory for incompressible fluids characterized by a continuous monotone rela- tion between the velocity gradient and the Cauchy stress. We, in particular, investigate a class of activated fluids that behave as the Euler fluid prior activation, and as the Navier-Stokes or power-law fluid once the activation takes place. We develop a large-data existence analysis for both steady and unsteady three-dimensional flows of such fluids subject either to the no-slip boundary condition or to a range of slip-type boundary conditions, including free-slip, Navier's slip, and stick-slip. In the second part we show that the W−1,q norm is localizable provided that the functional in question vanishes on locally supported functions which constitute a partition of unity. This represents a key tool for establishing local a posteriori efficiency for partial differential equations in divergence form with residuals in W−1,q . In the third part we provide a novel analysis for the pressure convection- diffusion (PCD) preconditioner. We first develop a theory for the precon- ditioner considered as an operator in infinite-dimensional spaces. We then provide a methodology for constructing discrete PCD operators for a broad class of pressure discretizations. The...

Les méthodes d’optimisation de forme basées sur le gradient reposent sur le calcul de la dérivée de forme. Dans beaucoup d’applications, la fonctionnelle coût dépend de la solution d’une EDP. Il s’en suit qu’elle ne peut être résolue exactement et que seule une approximation de celle-ci peut être calculée, par exemple par la méthode des éléments finis. Il en est de même pour la dérivée de forme. Ainsi, les méthodes de gradient en optimisation de forme - basées sur des approximations du gradient - ne garantissent pas a priori que la direction calculée à chaque itération soit effectivement une direction de descente pour la fonctionnelle coût. Cette thèse est consacrée à la construction d’une procédure de certification de la direction de descente dans des algorithmes de gradient en optimisation de forme grâce à des estimations a posteriori de l’erreur introduite par l’approximation de la dérivée de forme par la méthode des éléments finis. On présente une procédure pour estimer l’erreur dans une Quantité d’Intérêt et on obtient une borne supérieure certifiée et explicitement calculable. L’Algorithme de Descente Certifiée (CDA) pour l’optimisation de forme identifie une véritable direction de descente à chaque itération et permet d’établir un critère d’arrêt fiable basé sur la norme de la dérivée de forme. Deux applications principales sont abordées dans la thèse. Premièrement, on considère le problème scalaire d’identification de forme en tomographie d’impédance électrique et on étudie différentes estimations d’erreur. Une première approche est basée sur le principe de l’énergie complémentaire et nécessite la résolution de problèmes globaux additionnels. Afin de réduire le coût de calcul de la procédure de certification, une estimation qui dépend seulement de quantités locales est dérivée par la reconstruction des flux équilibrés. Après avoir validé les estimations de l’erreur pour un cas bidimensionnel, des résultats numériques sont présentés pour tester les méthodes discutées. Une deuxième application est centrée sur le problème vectoriel de la conception optimale des structures élastiques. Dans ce cadre figure, on calcule l’expression volumique de la dérivée de forme de la compliance à partir de la formulation primale en déplacements et de la formulation duale mixte pour l’équation de l’élasticité linéaire. Quelques résultats numériques préliminaires pour la minimisation de la compliance sous une contrainte de volume en 2D sont obtenus à l’aide de l’Algorithme de Variation de Frontière et une estimation a posteriori de l’erreur de la dérivée de forme basée sur le principe de l’énergie complémentaire est calculée. / Gradient-based shape optimization strategies rely on the computation of the so-called shape gradient. In many applications, the objective functional depends both on the shape of the domain and on the solution of a PDE which can only be solved approximately (e.g. via the Finite Element Method). Hence, the direction computed using the discretized shape gradient may not be a genuine descent direction for the objective functional. This Ph.D. thesis is devoted to the construction of a certification procedure to validate the descent direction in gradient-based shape optimization methods using a posteriori estimators of the error due to the Finite Element approximation of the shape gradient.By means of a goal-oriented procedure, we derive a fully computable certified upper bound of the aforementioned error. The resulting Certified Descent Algorithm (CDA) for shape optimization is able to identify a genuine descent direction at each iteration and features a reliable stopping criterion basedon the norm of the shape gradient.Two main applications are tackled in the thesis. First, we consider the scalar inverse identification problem of Electrical Impedance Tomography and we investigate several a posteriori estimators. A first procedure is inspired by the complementary energy principle and involves the solution of additionalglobal problems. In order to reduce the computational cost of the certification step, an estimator which depends solely on local quantities is derived via an equilibrated fluxes approach. The estimators are validated for a two-dimensional case and some numerical simulations are presented to test the discussed methods. A second application focuses on the vectorial problem of optimal design of elastic structures. Within this framework, we derive the volumetric expression of the shape gradient of the compliance using both H 1 -based and dual mixed variational formulations of the linear elasticity equation. Some preliminary numerical tests are performed to minimize the compliance under a volume constraint in 2D using the Boundary Variation Algorithm and an a posteriori estimator of the error in the shape gradient is obtained via the complementary energy principle.

This thesis is concerned with several issues of a posteriori error estimates for linear problems. In its first part error estimates for the heat conduction equation discretized by the backward Euler method in time and discontinuous Galerkin method in space are derived. In the second part guaranteed and locally efficient error estimates involving algebraic error for Poisson equation discretized by the discontinuous Galerkin method are derived. The technique is based on the flux reconstruction where meshes with hanging nodes and variable polynomial degree are allowed. An adaptive strategy combining both adaptive mesh refinement and stopping criteria for iterative algebraic solvers is proposed. In the last part a numerical method for computing guaranteed lower and upper bounds of principal eigenvalues of symmetric linear elliptic differential operators is presented. 1