Numerical Analysis of the Stochastic Navier-Stokes equations

DSpace Repositorium (Manakin basiert)


Dateien:

Zitierfähiger Link (URI): http://nbn-resolving.de/urn:nbn:de:bsz:21-opus-63572
http://hdl.handle.net/10900/49705
Dokumentart: Dissertation
Erscheinungsdatum: 2011
Sprache: Englisch
Fakultät: 7 Mathematisch-Naturwissenschaftliche Fakultät
Fachbereich: Mathematik
Gutachter: Prohl, Andreas (Prof. Dr. )
Tag der mündl. Prüfung: 2012-02-21
DDC-Klassifikation: 510 - Mathematik
Schlagworte: Stochastische partielle Differentialgleichung , Navier-Stokes-Gleichung , Finite-Elemente-Methode , Implizites Euler-Verfahren
Freie Schlagwörter: Non-Lipschitz SPDG , Martingallösung , Chorin-Verfahren
Stochastic Navier-Stokes Equations , Finite Element Method , Backward Euler Method , Chorin Method
Lizenz: http://tobias-lib.uni-tuebingen.de/doku/lic_mit_pod.php?la=de http://tobias-lib.uni-tuebingen.de/doku/lic_mit_pod.php?la=en
Gedruckte Kopie bestellen: Print-on-Demand
Zur Langanzeige

Inhaltszusammenfassung:

Die Zielsetzung dieser Arbeit ist die Analyse von Diskretisierungen der stochastischen Navier-Stokes Gleichungen auf beschränkten, zwei- und dreidimensionalen Gebieten. Die Motivation für diese numerische Analysis ist zweifach: Zum einen sind die Gleichungen ein wichtiges Modellproblem, das algebraische Nebenbedingungen mit einer stochastischen Kraft und einer Nichtlinearität, die nicht Lipschitz ist, kombiniert. Die hier entwickelten Methoden können auf andere ähnliche nichtlineare stochastische Partielle Differentialgleichungen angewandt werden. Zum anderen sind diese Methoden für praktische Anwendungen wichtig. Die hier betrachteten Gleichungen dienen Dazu, turbulente Strömungen, Wohlgestelltheit des entsprechenden deterministischen Problems und zufällige Schwankungen in hydrodynamischen Modellen besser zu verstehen. Im ersten Teil der Arbeit wird die auf finiten Elemente basierte Approximation schwacher Martingallösungen in drei Dimensionen betrachtet. Dies sind Systeme, die aus drei Bestandteilen bestehen: einem filtrierten Wahrscheinlichkeitsraum, einem Wienerprozess und einer Lösung der Gleichungen. Die Diskretisierung ist so konstruiert, dass alle Elemente des Lösungssystems unter Anwendung stetiger Störungen der diskreten Iterationen konstruiert werden können. Konvergenz ohne Raten wird hierfür gezeigt. Zusätzlich werden für dieses Schema mittels allgemeiner Zufallsvariablen, die die Zeitinkremente der stochastischen Kraft approximieren, ähnliche Konvergenzeigenschaften gezeigt. Für zweidimensionale Gebiete konvergiert dasselbe Schema gegen die eindeutige starke Lösung, was unter Verwendung lokaler Monotonie gezeigt wird. Im zweiten Teil werden Konvergenzeigenschaften von Projektionsverfahren studiert, die auf die instationären stochastischen Stokes Gleichungen angewandt werden. In diesem vereinfachten Zusammenhang wird untersucht, welchen Einfluss die Irregularität des Lagrange Multiplikator auf die Konvergenz hat. Hierzu wird auch die Konvergenz der finite Elemente Diskretisierung gezeigt. Im dritten Teil der Arbeit werden Schemata für die zweidimensionale Navier-Stokes Gleichungen mit periodischen Randbedingungen untersucht, die auf finiten Elementen und dem impliziten Eulerverfahren basieren, und eine Konvergenzanalyse mit Raten durchgeführt. Da die Nichtlinearität nicht Lipschitz ist, kann Konvergenz im Allgemeinen nur bis auf eine Menge mit beliebig kleiner Wahrscheinlichkeit gezeigt werden. Im Falle additiven Rauschens wird dennoch Konvergenz auf dem ganzen Realisierungsraum für die Zeitdiskretisierung gezeigt. Dazu wird auch die finite Elemente Approximation betrachtet, und die Konvergenzanalyse für verschiedene Fälle durchgeführt. In den jeweils anschliessenden Simulationen wird das Konvergenzverhalten der vorgestellten Algorithmen gezeigt und deren Leistungen verglichen.

Abstract:

The main subject of this thesis is to analyse various discretisations schemes for the stochastic Navier-Stokes equations on bounded two and three dimensional domains. The motivation for this numerical analysis is twofold: First this is an important model problem which combines algebraic constraints, non-Lipschitz nonlinearity, and stochastic forcing. The methods developed for this model problem may be applied to a wide range of nonlinear stochastic partial differential equations driven by a Wiener noise. Second, these methods may be used in applications. Such a system has been introduced to better understand turbulence phenomena, well posedness of the deterministic problem and random fluctuations in hydrodynamic models. It may be used to model relevant physical phenomena, such as turbulence. In the first part of this thesis, we address the finite element based approximation of weak martingale solutions in two and three dimensions, i.e., a system consisting of a filtered probability space, a Wiener process on it, and a solution to the equations. The discretisation is conceived such that all the elements of the system are constructed using continuous perturbations of the discrete iterates, and convergence without rates for subsequences of approximating solutions is proved. Moreover we show the same convergence properties for a scheme which uses general random variables to approximate the time increments of the stochastic forcing. In the two-dimensional case, thanks to a local monotonicity argument, the same scheme with Wiener process increments is shown to produce iterates that converge towards the unique strong solution. In the second part of this thesis, we study the convergence properties of projection based splitting schemes applied to the unsteady stochastic Stokes equations. In this simplified setting, we observe that the Lagrange multiplier affects the convergence behavior of the scheme, due to its irregularity This motivates the introduction of a new discretisation scheme, which is stable under this irregularity. Finite element discretisations are also considered, and their convergence proved. In the third part of the thesis, we consider implicit Euler based approximation schemes for the two-dimensional stochastic Navier-Stokes equations with periodic boundary conditions, and study convergence with rates. Due to the non-Lipschitz character of the nonlinearity, we prove convergence only on a set with probability arbitrarily close to one for the proposed schemes in a general setting. However, for additive noise we show convergence on the whole realisation space for the time discretisation. Finite element approximations for the corresponding time discretisations are considered, and convergence analysed. All the parts are concluded with simulations to illustrate the convergence results, and compare the efficiency of the different discretisations.

Das Dokument erscheint in: