Magdeburg, Universiẗat, Diss., 2005.

Berlin, Freie Universiẗat, Diss., 2002. / Dateiformat: zip, Dateien im PDF-Format.

Techn. University, Diss., 2005--Dresden.

The topic of this thesis is the theoretical and numerical research of optimal control problems for uncertain nonlinear systems, described by semilinear parabolic differential equations with additive noise, where the state is not completely available. Based on a paper by Kazufumi Ito and Karl Kunisch, which was published in 2006 with the title "Receding Horizon Control with Incomplete Observations", we analyze a Model Predictive Control (MPC) approach where the resulting linear problems on small intervals are solved with a Linear Quadratic Gaussian (LQG) design. Further we define a performance index for the MPC/LQG approach, find estimates for it and present bounds for the solutions of the underlying Riccati equations. Another large part of the thesis is devoted to extensive numerical studies for an 1+1- and 3+1-dimensional problem to show the robustness of the MPC/LQG strategy. The last part is a generalization of the MPC/LQG approach to infinite-dimensional problems.

Techn. University, Diss., 2003--Berlin.

Techn. University, Diss., 2003--Berlin.

Singularly perturbed reaction-diffusion problems exhibit in general solutions with anisotropic features, e.g. strong boundary and/or interior layers. This anisotropy is reflected in the discretization by using meshes with anisotropic elements. The quality of the numerical solution rests on the robustness of the a posteriori error estimator with respect to both the perturbation parameters of the problem and the anisotropy of the mesh. An estimator that has shown to be one of the most reliable for reaction-diffusion problem is the <i>equilibrated residual method</i> and its modification done by Ainsworth and Babuška for singularly perturbed problem. However, even the modified method is not robust in the case of anisotropic meshes. The present work modifies the equilibrated residual method for anisotropic meshes. The resulting error estimator is equivalent to the equilibrated residual method in the case of isotropic meshes and is proved to be robust on anisotropic meshes as well. A numerical example confirms the theory.

Singularly perturbed reaction-diffusion problems exhibit in general solutions with anisotropic features, e.g. strong boundary and/or interior layers. This anisotropy is reflected in the discretization by using meshes with anisotropic elements. The quality of the numerical solution rests on the robustness of the a posteriori error estimator with respect to both the perturbation parameters of the problem and the anisotropy of the mesh. An estimator that has shown to be one of the most reliable for reaction-diffusion problem is the <i>equilibrated residual method</i> and its modification done by Ainsworth and Babuška for singularly perturbed problem. However, even the modified method is not robust in the case of anisotropic meshes. The present work modifies the equilibrated residual method for anisotropic meshes. The resulting error estimator is equivalent to the equilibrated residual method in the case of isotropic meshes and is proved to be robust on anisotropic meshes as well. A numerical example confirms the theory.

The topic of this thesis is the theoretical and numerical research of optimal control problems for uncertain nonlinear systems, described by semilinear parabolic differential equations with additive noise, where the state is not completely available. Based on a paper by Kazufumi Ito and Karl Kunisch, which was published in 2006 with the title "Receding Horizon Control with Incomplete Observations", we analyze a Model Predictive Control (MPC) approach where the resulting linear problems on small intervals are solved with a Linear Quadratic Gaussian (LQG) design. Further we define a performance index for the MPC/LQG approach, find estimates for it and present bounds for the solutions of the underlying Riccati equations. Another large part of the thesis is devoted to extensive numerical studies for an 1+1- and 3+1-dimensional problem to show the robustness of the MPC/LQG strategy. The last part is a generalization of the MPC/LQG approach to infinite-dimensional problems.

Viele technische Prozesse führen auf Randwertprobleme mit partiellen Differentialgleichungen, die mit Finite-Elemente-Methoden näherungsweise gelöst werden können. Spezielle Varianten dieser Methoden sind Finite-Elemente-Mortar-Methoden. Sie erlauben das Arbeiten mit an Teilgebietsschnitträndern nichtzusammenpassenden Netzen, was für Probleme mit komplizierten Geometrien, Randschichten, springenden Koeffizienten sowie für zeitabhängige Probleme von Vorteil sein kann. Ebenso können unterschiedliche Diskretisierungsmethoden in den einzelnen Teilgebieten miteinander gekoppelt werden. In dieser Arbeit wird das Finite-Elemente-Mortaring nach einer Methode von Nitsche für elliptische Randwertprobleme auf zweidimensionalen polygonalen Gebieten untersucht. Von besonderem Interesse sind dabei nichtreguläre Lösungen (u \in H^{1+\delta}(\Omega), \delta>0) mit Eckensingularitäten für die Poissongleichung sowie die Lamé-Gleichung mit gemischten Randbedingungen. Weiterhin werden singulär gestörte Reaktions-Diffusions-Probleme betrachtet, deren Lösungen zusätzlich zu Eckensingularitäten noch anisotropes Verhalten in Randschichten aufweisen. Für jede dieser drei Problemklassen wird das Nitsche-Mortaring dargelegt. Es werden einige Eigenschaften der Mortar-Diskretisierung angegeben und a-priori-Fehlerabschätzungen in einer H^1-artigen sowie der L_2-Norm durchgeführt. Auf lokal verfeinerten Dreiecksnetzen können auch für Lösungen mit Eckensingularitäten optimale Konvergenzordnungen nach gewiesen werden. Bei den Lösungen mit anisotropen Verhalten werden zusätzlich anisotrope Dreiecksnetze verwendet. Es werden auch hier Konvergenzordnungen wie bei klassischen Finite-Elemente-Methoden ohne Mortaring erreicht. Numerische Experimente illustrieren die Methode und die Aussagen zur Konvergenz.