Viele Resultate über MR- und OR-Verfahren zur Lösung linearer Gleichungssysteme bleiben (in leicht modifizierter Form) gültig, wenn der betrachtete Operator nicht invertierbar ist. Neben dem für reguläre Probleme charakteristischen Abbruchverhalten, kann bei einem singulären Gleichungssystem auch ein so genannter singulärer Zusammenbruch auftreten. Für beide Fälle werden verschiedene Charakterisierungen angegeben. Die Unterrauminverse, eine spezielle verallgemeinerte Inverse, beschreibt die Näherungen eines MR-Unterraumkorrektur-Verfahrens. Für Krylov-Unterräume spielt die Drazin-Inverse eine Schlüsselrolle. Bei Krylov-Unterraum-Verfahren kann a-priori entschieden werden, ob ein regulärer oder ein singulärer Abbruch auftritt. Wir können zeigen, dass ein Krylov-Verfahren genau dann für beliebige Startwerte eine Lösung des linearen Gleichungssystems liefert, wenn der Index der Matrix nicht größer als eins und das Gleichungssystem konsistent ist. Die Berechnung stationärer Zustandsverteilungen zeitstetiger Markov-Ketten mit endlichem Zustandsraum stellt eine praktische Aufgabe dar, welche die Lösung eines singulären linearen Gleichungssystems erfordert. Die Eigenschaften der Übergangs-Halbgruppe folgen aus einfachen Annahmen auf rein analytischem und matrixalgebrischen Wege. Insbesondere ist die erzeugende Matrix eine singuläre M-Matrix mit Index 1. Ist die Markov-Kette irreduzibel, so ist die stationäre Zustandsverteilung eindeutig bestimmt.

The discretization of the stationary diffusion equation with random parameters by the Stochastic Finite Element Method requires the solution of a highly structured but very large linear system of equations. Depending on the stochastic properties of the diffusion coefficient together with the stochastic discretization we consider three solver cases. If the diffusion coefficient is given by a stochastically linear expansion, e.g. a truncated Karhunen-Loeve expansion, and tensor product polynomial stochastic shape functions are employed, the Galerkin matrix can be transformed to a block-diagonal matrix. For the solution of the resulting sequence of linear systems we study Krylov subspace recycling methods whose success depends on the ordering and grouping of the linear systems as well as the preconditioner. If we use complete polynomials for the stochastic discretization instead, we show that decoupling of the Galerkin matrix with respect to the stochastic degrees of freedom is impossible. For a stochastically nonlinear diffusion coefficient, e.g. a lognormal random field, together with complete polynomials serving as stochastic shape functions, we introduce and test the performance of a new Kronecker product preconditioner, which is not exclusively based on the mean value of the diffusion coefficient.

<p>Rational Krylov methods for model order reduction are studied. A dual rational Arnoldi method for model order reduction and a rational Krylov method for model order reduction and eigenvalue computation have been implemented. It is shown how to deflate redundant or unwanted vectors and how to obtain moment matching. Both methods are designed for generalised state space systems---the former for multiple-input-multiple-output (MIMO) systems from finite element discretisations and the latter for single-input-single-output (SISO) systems---and applied to relevant test problems. The dual rational Arnoldi method is designed for generating real reduced order systems using complex shift points and stabilising a system that happens to be unstable. For the rational Krylov method, a forward error in the recursion and an estimate of the error in the approximation of the transfer function are studie.</p><p>A stability analysis of a heat exchanger model is made. The model is a nonlinear partial differential-algebraic equation (PDAE). Its well-posedness and how to prescribe boundary data is investigated through analysis of a linearised PDAE and numerical experiments on a nonlinear DAE. Four methods for generating reduced order models are applied to the nonlinear DAE and compared: a Krylov based moment matching method, balanced truncation, Galerkin projection onto a proper orthogonal decomposition (POD) basis, and a lumping method.</p>

Cette thèse concerne la modélisation et le contrôle actif des instabilités de combustion. L'approche<br />par boite grise est considérée pour l'identification de modèles. La méthode de Krylov-Bogoliubov (K-B) est choisie comme outil principal d'analyse. Deux modèles analytiquement attractifs avec des structures en boucle fermée sont proposés. Le premier modèle est basé sur deux équations de Van der Pol couplées et généralisées. L'analyse a montré que ce modèle ne peut pas décrire le phénomène de<br />coexistence simultanée de deux modes non harmoniques observé en pratique. Le deuxième modèle est établi en complétant le premier modèle par un retard et un filtre passe bas. L'analyse a montré que le modèle est capable de décrire le phénomène de coexistence simultanée de deux modes non harmoniques. Les performances de l'approximation K-B sont largement illustrées par les tests de simulation. Pour l'établissement des conditions d'extinction des oscillations, le contrôle actif par hautes fréquences et par retour de boucle est considéré. Les deux donnent de bons résultats vérifiés par des tests de simulation.

Cette thèse porte sur des méthode de résolution d'équations matricielles appliquées à la résolution numérique d'équations aux dérivées partielles ou des problèmes de contrôle linéaire. On s'intéressen en premier lieu à des équations matricielles linéaires. Après avoir donné un aperçu des méthodes classiques employées pour les équations de Sylvester et de Lyapunov, on s'intéresse au cas d'équations linéaires générales de la forme M(X)=C, où M est un opérateur linéaire matriciel. On expose la méthode de GMRES globale qui s'avère particulièrement utile dans le cas où M(X) ne peut s'exprimer comme un polynôme du premier degré en X à coefficients matriciels, ce qui est le cas dans certains problèmes de résolution numérique d'équations aux dérivées partielles. Nous proposons une approche, noté LR-BA-ADI consistant à utiliser un préconditionnement de type ADI qui transforme l'équation de Sylvester en une équation de Stein que nous résolvons par une méthode de Krylox par blocs. Enfin, nous proposons une méthode de type Newton-Krylov par blocs avec préconditionnement ADI pour les équations de Riccati issues de problèmes de contrôle linéaire quadratique. Cette méthode est dérivée de la méthode LR-BA-ADI. Des résultats de convergence et de majoration de l'erreur sont donnés. Dans la seconde partie de ce travail, nous appliquons les méthodes exposées dans la première partie de ce travail à des problèmes d'équations aux dérivées partielles. Nous nous intéressons d'abord à la résolution numérique d'équations couplées de type Burgers évolutives en dimension 2. Ensuite, nous nous intéressons au cas où le domaine borné est choisi quelconque. Nous établissons des résultats théoriques de l'existence de tels interpolants faisant appel à des techniques d'algèbre linéaire.

La simulation numérique de systèmes d'équations différentielles raides ordinaires ou algébriques est devenue partie intégrante dans le processus de conception des systèmes mécaniques à dynamiques complexes. L'objet de ce travail est de développer des méthodes numériques pour réduire les temps de calcul par le parallélisme en suivant deux axes : interne à l'intégrateur numérique, et au niveau de la décomposition de l'intervalle de temps. Nous montrons l'efficacité limitée au nombre d'étapes de la parallélisation à travers les méthodes de Runge-Kutta et DIMSIM. Nous développons alors une méthodologie pour appliquer le complément de Schur sur le système linéarisé intervenant dans les intégrateurs par l'introduction d'un masque de dépendance construit automatiquement lors de la mise en équations du modèle. Finalement, nous étendons le complément de Schur aux méthodes de type "Krylov Matrix Free". La décomposition en temps est d'abord vue par la résolution globale des pas de temps dont nous traitons la parallélisation du solveur non-linéaire (point fixe, Newton-Krylov et accélération de Steffensen). Nous introduisons les méthodes de tirs à deux niveaux, comme Parareal et Pita dont nous redéfinissons les finesses de grilles pour résoudre les problèmes raides pour lesquels leur efficacité parallèle est limitée. Les estimateurs de l'erreur globale, nous permettent de construire une extension parallèle de l'extrapolation de Richardson pour remplacer le premier niveau de calcul. Et nous proposons une parallélisation de la méthode de correction du résidu.

Advances in numerical optimization have raised the possibility that efficient and novel aircraft configurations may be ``discovered'' by an algorithm. To begin exploring this possibility, a fast and robust set of tools for aerodynamic shape optimization is developed. Parameterization and mesh-movement are integrated to accommodate large changes in the geometry. This integrated approach uses a coarse B-spline control grid to represent the geometry and move the computational mesh; consequently, the mesh-movement algorithm is two to three orders faster than a node-based linear elasticity approach, without compromising mesh quality. Aerodynamic analysis is performed using a flow solver for the Euler equations. The governing equations are discretized using summation-by-parts finite-difference operators and simultaneous approximation terms, which permit nonsmooth mesh continuity at block interfaces. The discretization results in a set of nonlinear algebraic equations, which are solved using an efficient parallel Newton-Krylov-Schur strategy. A gradient-based optimization algorithm is adopted. The gradient is evaluated using adjoint variables for the flow and mesh equations in a sequential approach. The flow adjoint equations are solved using a novel variant of the Krylov solver GCROT. This variant of GCROT is flexible to take advantage of non-stationary preconditioners and is shown to outperform restarted flexible GMRES. The aerodynamic optimizer is applied to several studies of induced-drag minimization. An elliptical lift distribution is recovered by varying spanwise twist, thereby validating the algorithm. Planform optimization based on the Euler equations produces a nonelliptical lift distribution, in contrast with the predictions of lifting-line theory. A study of spanwise vertical shape optimization confirms that a winglet-up configuration is more efficient than a winglet-down configuration. A split-tip geometry is used to explore nonlinear wake-wing interactions: the optimized split-tip demonstrates a significant reduction in induced drag relative to a single-tip wing. Finally, the optimal spanwise loading for a box-wing configuration is investigated.

The objective of this PhD research programme is to investigate the effectiveness of a finite volume method incorporating radial basis functions for simulating nonlinear transport processes. The finite volume method is the favoured numerical technique for solving the advection-diffusion equations that arise in transport simulation. The method transforms the original problem into a system of nonlinear, algebraic equations through the process of discretisation. The accuracy of this discretisation determines to a large extent the accuracy of the final solution. A new method of discretisation is presented that employs radial basis functions (rbfs) as a means of local interpolation. When combined with Gaussian quadrature integration methods, the resulting finite volume discretisation leads to accurate numerical solutions without the need for very fine meshes, and the additional overheads they entail. The resulting nonlinear, algebraic system is solved efficiently using a Jacobian-free Newton-Krylov method. By employing the new method as an extension of existing shape function-based approaches, the number of nonlinear iterations required to obtain convergence can be reduced. Furthermore, information obtained from these iterations can be used to increase the efficiency of subsequent rbf-based iterations, as well as to construct an effective parallel reconditioner to further reduce the number of nonlinear iterations required. Results are presented that demonstrate the improved accuracy offered by the new method when applied to several test problems. By successively refining the meshes, it is also possible to demonstrate the increased order of the new method, when compared to a traditional shape function basedmethod. Comparing the resources required for both methods reveals that the new approach can be many times more efficient at producing a solution of a given accuracy.

Sea ice plays a key role in the global climate system. Indeed, through the albedo effect it reflects significant solar radiation away from the oceans, while it also plays a key role in the momentum and heat transfer between the atmosphere and ocean by acting as an insulating layer between the two. Furthermore, as more sea ice melts due to climate change, additional fresh water is released into the upper oceans, affecting the global circulation of the ocean as a whole. While there has been significant effort in recent decades, the ability to simulate sea ice has lagged behind other components of the climate system and most Earth System Models fail to capture the observed losses of Arctic sea ice, which is largely attributed to our inability to resolve sea ice dynamics. The most widely accepted model for sea ice dynamics is the Viscous- Plastic (VP) rheology, which leads to a very non-linear set of partial differential equations that are known to be intrinsically difficult to solve numerically. This work builds on recent advances in solving these equations with a Jacobian-Free Newton- Krylov (JFNK) solver. We present an improved JFNK solver, where a fully second order discretization is achieved via the Crank Nicolson scheme and consistency is improved via a novel approach to the rheology term. More importantly, we present a significant improvement to the Jacobian approximation used in the Newton iterations, and partially form the action of the matrix by expressing the linear and nearly linear terms in closed form and approximating the remaining highly non-linear term with a second order approximation of its Gateaux derivative. This is in contrast with the previous approach which used a first order approximation for the Gateaux derivative of the whole functional. Numerical tests on synthetic equations confirm the theoretical convergence rate and demonstrate the drastic improvements seen by using a second order approximation in the Gateaux derivative. To produce a fast and efficient solver for VP sea ice dynamics, the improved JFNK solver is then coupled with a non- oscillatory, central differencing scheme for transporting sea ice as well as a novel method for tracking the ice domain using a level set method. Two idealized test cases are then presented and simulation results discussed, demonstrating the solver’s ability to efficiently produce Viscous-Plastic, physically motivated solutions. / Graduate

Miniaturisation of electronic chips poses challenges at the design stage. The progressively decreasing circuit dimensions result in complex electrical behaviour that necessitates complex models. Simulation of complex circuit models involves extraordinarily large compu- tational complexity. Such complexity is better managed through Model Order Reduction. Model order reduction has been successful in large reductions in system order for most types of circuits, at high levels of accuracy. However, multiport circuits with large number of inputs/outputs, pose an additional computational challenge. A strategy based on exible clustering of interconnects results in more e cient reduction of multiport circuits. Clustering methods traditionally use Krylov-subspace methods such as PRIMA for the actual model reduction step. These clustering methods are unable to reduce the model order to the optimum extent. SVD-based methods like Truncated Balanced Realization have shown higher reduction potential than Krylov-subspace methods. In this thesis, the di erences in reduction potential and computational cost thereof between SVD-based methods and Krylov-subspace methods are identi ed, analyzed and quanti ed. A novel algorithm has been developed, utilizing a particular combination of both these methods to achieve better results. It enhances the clustering method for model reduction using Truncated Balanced Realization as a second-level reduction technique. The algorithm is tested and signi cant gains are illustrated. The proposed novel algorithm preserves the other advantages of the current clustering algorithm.