We study the loss of information (measured in terms of the Kullback- Leibler distance) caused by observing "grouped" data (observing only a discretized version of a continuous random variable). We analyse the asymptotical behaviour of the loss of information as the partition becomes finer. In the case of a univariate observation, we compute the optimal rate of convergence and characterize asymptotically optimal partitions (into intervals). In the multivariate case we derive the asymptotically optimal regular sequences of partitions. Forthermore, we compute the asymptotically optimal transformation of the data, when a sequence of partitions is given. Examples demonstrate the efficiency of the suggested discretizing strategy even for few intervals. (author's abstract) / Series: Forschungsberichte / Institut für Statistik

Šiame darbe buvo analizuojama programinė įranga skirta diskretizuoti sritį į baigtinius elementus. Analizuotos MATLAB sistemos ir COMSOL programos diskretzavimo galimybės. Sukurta programa sujungianti COMSOL ir MATLAB, kuri leidžia diskretizuoti sritį į baigtinius elementus. / In this work was done study of software for domain discretization by finite elements. Created program that connects MATLAB and COMSOL, which allows domain discretization by finite elements.

In this thesis, we study uniform controllability properties of semi-discrete approximations for parabolic systems. In a first part, we address the minimization of the Lq-norm (q > 2) of semidiscrete controls for parabolic equation. Our goal is to overcome the limitation of [LT06] about the order 1/2 of unboundedness of the control operator. Namely, we show that the uniform observability property also holds in Lq (q > 2) even in the case of a degree of unboundedness greater than 1/2. Moreover, a minimization procedure to compute the approximation controls is provided. The study of Lq optimality in the first part is in a general context. However, the discrete observability inequalities that are obtained are not so precise than the ones derived then with Carleman estimates. In a second part, in the discrete setting of one-dimensional finite-differences we prove a Carleman estimate for a semi discrete version of the parabolic operator @t − @x(c@x) which allows one to derive observability inequalities that are far more precise. Here we consider in case that the diffusion coefficient has a jump which yields a transmission problem formulation. Consequence of this Carleman estimate, we deduce consistent null-controllability results for classes of linear and semi-linear parabolic equations.

The objective of this study is to analyze theoretically the laminar film condensation of water vapor flowing on a sphere. For this purpose, the problem was handled by including all of the two-phase boundary layer parameters such as gravity, effect of vapor shear, inertia, energy convection and pressure gradient. For this full two-phase boundary layer system, the boundary layer equations, boundary conditions and the interfacial conditions were first analyzed, and then discretized. A computer program in Mathcad, solving these discretized equations, was written to obtain the velocity and temperature profiles within the condensate, the velocity profiles within the vapor, the condensate film thickness and the local Nusselt number. The effects of pressure gradient, gravity, vapor oncoming velocity and sphere radius on these parameters were examined. By alternating the formulation of the problem, the results for the flow over a horizontal cylinder were obtained. These results were then compared with those for the sphere. Finally, the results for the system with Mercury vapor flowing on a sphere were obtained. All of these results were represented as diagrams and tables, and were discussed at the end of the study.

Effervescent atomization has been a topic of considerable investigation in the literature due to its important advantages over other atomization mechanisms. This work contributes to the development of both effervescent atomizers and also laser-based techniques for spray investigation In order to develop non-intrusive measurement techniques for spray applications, a procedure is suggested to characterize the shape of droplets using image-based droplet analyzers. Image discretization which is a major source of error in droplet shape measurement is evaluated using a simulation. The accuracy of StereoPIV system in conducting droplet velocity measurement in a spray field is also investigated. To assist in the design of effervescent atomizers, bubble formation during gas injection from a micro-tube into liquid cross-flow is investigated using a Shadow-PIV/PTV system. The generated spray fields of two effervescent atomizers which operate using a porous and a typical multi-hole air injector are compared using qualitative images and Shadow-PTV measurement.

In this thesis, high order accurate discretization schemes for partial differential equations are investigated. In the first paper, the linearized two-dimensional Navier-Stokes equations are considered. A special formulation of the boundary conditions is used and estimates for the solution to the continuous problem in terms of the boundary conditions are derived using a normal mode analysis. Similar estimates are achieved for the discretized equations. For the discretization, a second order finite difference scheme on a staggered mesh is used. In Paper II, the analysis for the second order scheme is used to develop a fourth order scheme for the fully nonlinear Navier-Stokes equations. The fully nonlinear incompressible Navier-Stokes equations in two space dimensions are considered on an orthogonal curvilinear grid. Numerical tests are performed with a fourth order accurate Padé type spatial finite difference scheme and a semi-implicit BDF2 scheme in time. In Papers III-V, a class of high order accurate time-discretization schemes based on the deferred correction principle is investigated. The deferred correction principle is based on iteratively eliminating lower order terms in the local truncation error, using previously calculated solutions, in each iteration obtaining more accurate solutions. It is proven that the schemes are unconditionally stable and stability estimates are given using the energy method. Error estimates and smoothness requirements are derived. Special attention is given to the implementation of the boundary conditions for PDE. The scheme is applied to a series of numerical problems, confirming the theoretical results. In the sixth paper, a time-compact fourth order accurate time discretization for the one- and two-dimensional wave equation is considered. Unconditional stability is established and fourth order accuracy is numerically verified. The scheme is applied to a two-dimensional wave propagation problem with discontinuous coefficients.

The investigation of the dynamics of chemical reactions, both from the theoretical and experimental side, has drawn an increasing interest since Ahmed H. Zewail was awarded the 1999 Nobel Prize in Chemistry for his work on femtochemistry. On the experimental side, new techniques such as femtosecond lasers and attosecond lasers enable laser control of chemical reactions. Numerical simulations serve as a valuable complement to experimental techniques, not only for validation of experimental results, but also for simulation of processes that cannot be investigated through experiments. With increasing computer capacity, more and more physical phenomena fall within the range of what is possible to simulate. Also, the development of new, efficient numerical methods further increases the possibilities. The focus of this thesis is twofold; numerical methods for chemical reactions including dissociative states and methods that can deal with coexistence of spatial regions with very different physical properties. Dissociative chemical reactions are reactions where molecules break up into smaller components. The dissociation can occur spontaneously, e.g. by radioactive decay, or be induced by adding energy to the system, e.g. in terms of a laser field. Quantities of interest can for instance be the reaction probabilities of possible chemical reactions. This thesis discusses a boundary treatment model based on the perfectly matched layer (PML) approach to accurately describe dynamics of chemical reactions including dissociative states. The limitations of the method are investigated and errors introduced by the PML are quantified. The ability of a numerical method to adapt to different scales is important in the study of more complex chemical systems. One application of interest is long-range molecules, where the atoms are affected by chemical attractive forces that lead to fast movement in the region where they are close to each other and exhibits a relative motion of the atoms that is very slow in the long-range region. A numerical method that allows for spatial adaptivity is presented, based on the summation-by-parts-simultaneous approximation term (SBP-SAT) methodology. The accuracy and the robustness of the numerical method are investigated. / eSSENCE

Dans cette thèse, nous étudions les propriétés de contrôlabilité uniforme des semidiscrets approximations de systèmes paraboliques. Dans une première partie, nous nous intéressons à la minimisation de Lq-norme (q > 2) des contrôles semidiscrete pour l'équation parabolique. Notre objectif est de dépasser la limitation de [LT06] à propos de l'ordre ½ de l'absence de limites d'opérateur de contrôle. Plus précisément, nous montrons que la propriété d'observabilité uniforme est également titulaire dans Lq (q > 2), même dans le cas d'un degré d'absence de limites supérieure à 1/2. En outre, une procédure de minimisation pour calculer les commandes d'approximation est fournie. L'étude de l'optimalité Lq dans lapremière partie est dans un contexte général. Cependant, les inégalités d'observabilité discrets qui sont obtenus ne sont pas aussi précises que celles dérivées puis avec des estimations de Carleman. Dans une seconde partie, dans le contexte particulier de unidimensionnels-finis différences nous démontrons une inégalité de Carleman pour une version semi-discret de l'opérateur parabole @t − @x(c@x) qui permet pour dériver les inégalités d'observabilité qui sont beaucoup plus précis. On considère ici que dans le cas où le coefficient de diffusion a un saut qui donne une formulation du problème de transmission. Conséquence de cette inégalité de Carleman, on en déduit cohérentes nul contrôlabilité des résultats pour les classes de linéaires et semi-linéaire des équations paraboliques. / In this thesis, we study uniform controllability properties of semi-discrete approximations for parabolic systems. In a first part, we address the minimization of the Lq-norm (q > 2) of semidiscrete controls for parabolic equation. Our goal is to overcome the limitation of [LT06] about the order 1/2 of unboundedness of the control operator. Namely, we show that the uniform observability property also holds in Lq (q > 2) even in the case of a degree of unboundedness greater than 1/2. Moreover, a minimization procedure to compute the approximation controls is provided. The study of Lq optimality in the first part is in a general context. However, the discrete observability inequalities that are obtained are not so precise than the ones derived then with Carleman estimates. In a second part, in the discrete setting of one-dimensional finite-differences we prove a Carleman estimate for a semi discrete version of the parabolic operator @t − @x(c@x) which allows one to derive observability inequalities that are far more precise. Here we consider in case that the diffusion coefficient has a jump which yields a transmission problem formulation. Consequence of this Carleman estimate, we deduce consistent null-controllability results for classes of linear and semi-linear parabolic equations.

Este trabalho tem por objetivo apresentar uma nova forma de analisar as redes de sincronismo de fase mutuamente conectadas. Estas redes são formadas por Phase-Locked Loops digitais ou DPLLs. O sinal gerado por cada DPLL é enviado a todos os demais dispositivos, formando a rede mutuamente conectada. Parte-se do pressuposto de que as ligações entre os dispositivos são dotadas de atrasos, o que dificulta o tratamento do problema. No entanto, é apresentado aqui um método para análise das malhas de sincronismo via discretização do modelo de tempo contínuo, objetivando dirimir essa dificuldade, já que atrasos são facilmente representados em modelos de tempo discreto. Para tanto, o modelo da rede no espaço de estados é equacionado a partir da rede. Esse modelo no espaço de estados é, então, discretizado e, enfim, pode-se determinar o estado síncrono da rede incluindo a freqüência de sincronismo e analisar sua estabilidade. Como se poderá constatar, escolhendo um período de amostragem adequado, pode-se representar o comportamento das redes de sincronismo com modelos discretos, obtendo elevado grau de precisão. / This work introduces a new method for studying a mutually-delayed-connected network of Digital Phase-Locked Loops DPLLs. The signal generated by a DPLL in the network is sent to all other devices in this same network. Because of delayed signals, it is difficult to treat this problem. So, its shown here a method for analyzing the networks via discretization of continuous time delay model in order to deal with this issue easily, considering that delays are naturally represented in discrete time models. First of all, a continuous state space model is obtained from mutually-connected network. Then, this model is discretized and, finally, the synchronous state can be determined and the stability can be analyzed. As shown below, choosing a proper time sample, the behavior of mutually-delayed-connected networks can be approximately represented by a discrete time model.

Probabilistic Spacetime is a simple generalization of the classical model of spa- cetime in General Relativity, such that it allows to consider multiple metric field realizations endowed with probabilities. The motivation for such a generalization is a possible application in the context of some quantum gravity approaches, na- mely those using the path integral. It is argued that this model might be used to restrict the precision of the geometry on small scales without postulating discrete structure; or it may be used as an effective description of a probabilistic geometry resulting from a full-fledged quantum gravity computation.