In this dissertation, new theoretical results are obtained for bounding convergence and mean-square error in conditional coding. Further new statistical methods for the practical application of conditional coding are developed. Criteria for the uniform convergence are first examined. Conditional coding Markov chains are aperiodic, π-irreducible, and Harris recurrent. By applying the general theories of uniform ergodicity of Markov chains on genera l state space, one can conclude that conditional coding Markov cha ins are uniformly ergodic and further, theoretical convergence rates based on Doeblin's condition can be found. Conditional coding Markov chains can be also viewed as having finite state space. This allows use of techniques to get bounds on the second largest eigenvalue which lead to bounds on convergence rate and the mean-square error of sample averages. The results are applied in two examples showing that these bounds are useful in practice. Next some algorithms for perfect sampling in conditional coding are studied. An application of exact sampling to the independence sampler is shown to be equivalent to standard rejection sampling. In case of single-site updating, traditional perfect sampling is not directly applicable when the state space has large cardinality and is not stochastically ordered, so a new procedure is developed that gives perfect samples at a predetermined confidence interval. In last chapter procedures and possibilities of applying conditional coding to mixture models are explored. Conditional coding can be used for analysis of a finite mixture model. This methodology is general and easy to use.

Diagonally dominant matrices arise in many applications. In this work, we exploit the structure of diagonally dominant matrices to provide sharp entrywise relative perturbation bounds. We first generalize the results of Dopico and Koev to provide relative perturbation bounds for the LDU factorization with a well conditioned L factor. We then establish relative perturbation bounds for the inverse that are entrywise and independent of the condition number. This allows us to also present relative perturbation bounds for the linear system Ax=b that are independent of the condition number. Lastly, we continue the work of Ye to provide relative perturbation bounds for the eigenvalues of symmetric indefinite matrices and non-symmetric matrices.

The thesis treats model reduction for linear time-varying systems. Time-varying models appear in many fields, including power systems, chemical engineering, aeronautics, and computational science. They can also be used for approximation of time-invariant nonlinear models. Model reduction is a topic that deals with simplification of complex models. This is important since it facilitates analysis and synthesis of controllers. The thesis consists of two parts. The first part provides an introduction to the topics of time-varying systems and model reduction. Here, notation, standard results, examples, and some results from the second part of the thesis are presented. The second part of the thesis consists of four papers. In the first paper, we study the balanced truncation method for linear time-varying state-space models. We derive error bounds for the simplified models. These bounds are generalizations of well-known time-invariant results, derived with other methods. In the second paper, we apply balanced truncation to a high-order model of a diesel exhaust catalyst. Furthermore, we discuss practical issues of balanced truncation and approximative discretization. In the third paper, we look at frequency-domain analysis of linear time-periodic impulse-response models. By decomposing the models into Taylor and Fourier series, we can analyze convergence properties of different truncated representations. In the fourth paper, we use the frequency-domain representation developed in the third paper, the harmonic transfer function, to generalize Bode's sensitivity integral. This result quantifies limitations for feedback control of linear time-periodic systems. / QC 20120206

A singularly perturbed reaction diffusion problem is considered. The small diffusion coefficient generically leads to solutions with boundary layers. The problem is discretized by a vertex-centered finite volume method. The anisotropy of the solution is reflected by using \emph{anisotropic meshes} which can improve the accuracy of the discretization considerably. The main focus is on \emph{a posteriori} error estimation. A residual type error estimator is proposed and rigorously analysed. It is shown to be robust with respect to the small perturbation parameter. The estimator is also robust with respect to the mesh anisotropy as long as the anisotropic mesh sufficiently reflects the anisotropy of the solution (which is almost always the case for sensible discretizations). Altogether, reliable and efficient \emph{a posteriori} error estimation is achieved for the finite volume method on anisotropic meshes.

El Método de los Elementos Finitos (MEF) se ha afianzado durante las últimas décadas como una de las técnicas numéricas más utilizadas para resolver una gran variedad de problemas en diferentes áreas de la ingeniería, como por ejemplo, el análisis estructural, análisis térmicos, de fluidos, procesos de fabricación, etc. Una de las aplicaciones donde el método resulta de mayor interés es en el análisis de problemas propios de la Mecánica de la Fractura, facilitando el estudio y evaluación de la integridad estructural de componentes mecánicos, la fiabilidad, y la detección y control de grietas. Recientemente, el desarrollo de nuevas técnicas como el Método Extendido de los Elementos Finitos (XFEM) ha permitido aumentar aún más el potencial del MEF. Dichas técnicas mejoran la descripción de problemas con singularidades, con discontinuidades, etc., mediante la adición de funciones especiales que enriquecen el espacio de la aproximación convencional de elementos finitos. Sin embargo, siempre que se aproxima un problema mediante técnicas numéricas, la solución obtenida presenta discrepancias con respecto al sistema que representa. En las técnicas basadas en la representación discreta del dominio mediante elementos finitos (MEF, XFEM, ...) interesa controlar el denominado error de discretización. En la literatura se pueden encontrar numerosas referencias a técnicas que permiten cuantificar el error en formulaciones convencionales de elementos finitos. No obstante, por ser el XFEM un método relativamente reciente, aún no se han desarrollado suficientemente las técnicas de estimación del error para aproximaciones enriquecidas de elementos finitos. El objetivo de esta Tesis es cuantificar el error de discretización cuando se utilizan aproximaciones enriquecidas del tipo XFEM para representar problemas propios de la Mecánica de la Fractura Elástico Lineal (MFEL), como es el caso del modelado de una grieta. / González Estrada, OA. (2010). Estimación y acotación del error de discretización en el modelado de grietas mediante el método extendido de los elementos finitos [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/7203 / Palancia

A singularly perturbed reaction diffusion problem is considered. The small diffusion coefficient generically leads to solutions with boundary layers. The problem is discretized by a vertex-centered finite volume method. The anisotropy of the solution is reflected by using \emph{anisotropic meshes} which can improve the accuracy of the discretization considerably. The main focus is on \emph{a posteriori} error estimation. A residual type error estimator is proposed and rigorously analysed. It is shown to be robust with respect to the small perturbation parameter. The estimator is also robust with respect to the mesh anisotropy as long as the anisotropic mesh sufficiently reflects the anisotropy of the solution (which is almost always the case for sensible discretizations). Altogether, reliable and efficient \emph{a posteriori} error estimation is achieved for the finite volume method on anisotropic meshes.

Metamodeling is regarded as a powerful analysis tool to learn the input-output relationship of a system based on a limited amount of data collected when experiments with real systems are costly or impractical. As a popular metamodeling method, Gaussian process regression (GPR), has been successfully applied to analyses of various engineering systems. However, GPR-based metamodeling for time-dependent systems (TDSs) is especially challenging due to three reasons. First, TDSs require an appropriate account for temporal effects, however, standard GPR cannot address temporal effects easily and satisfactorily. Second, TDSs typically require analytics tools with a sufficiently high computational efficiency to support online decision making, but standard GPR may not be adequate for real-time implementation. Lastly, reliable uncertainty quantification is a key to success for operational planning of TDSs in real world, however, research on how to construct adequate error bounds for GPR-based metamodeling is sparse. Inspired by the challenges encountered in GPR-based analyses of two representative stochastic TDSs, i.e., load forecasting in a power system and trajectory prediction for unmanned aerial vehicles (UAVs), this dissertation aims to develop novel modeling, sampling, and statistical analysis techniques for enhancing the computational and statistical efficiencies of GPR-based metamodeling to meet the requirements of practical implementations. Furthermore, an in-depth investigation on building uniform error bounds for stochastic kriging is conducted, which sets up a foundation for developing robust GPR-based metamodeling techniques for analyses of TDSs under the impact of strong heteroscedasticity. / Ph.D. / Metamodeling has been regarded as a powerful analysis tool to learn the input-output relationship of an engineering system with a limited amount of experimental data available. As a popular metamodeling method, Gaussian process regression (GPR) has been widely applied to analyses of various engineering systems whose input-output relationships do not depend on time. However, GPR-based metamodeling for time-dependent systems (TDSs), whose input-output relationships depend on time, is especially challenging due to three reasons. First, standard GPR cannot properly address temporal effects for TDSs. Second, standard GPR is typically not computationally efficient enough for real-time implementations in TDSs. Lastly, research on how to adequately quantify the uncertainty associated with the performance of GPR-based metamodeling is sparse. To fill this knowledge gap, this dissertation aims to develop novel modeling, sampling, and statistical analysis techniques for enhancing standard GPR to meet the requirements of practical implementations for TDSs. Effective solutions are provided to address the challenges encountered in GPR-based analyses of two representative stochastic TDSs, i.e., load forecasting in a power system and trajectory prediction for unmanned aerial vehicles (UAVs). Furthermore, an in-depth investigation on quantifying the uncertainty associated with the performance of stochastic kriging (a variant of standard GPR) is conducted, which sets up a foundation for developing robust GPR-based metamodeling techniques for analyses of more complex TDSs.

A stochastic shortest path (SSP) problem is an undiscounted Markov decision process with an absorbing and zero-cost target state, where the objective is to reach the target state with minimum expected cost. This problem provides a foundation for algorithms for decision-theoretic planning and probabilistic model checking, among other applications. This thesis describes an implementation and evaluation of recently developed error bounds for SSP problems. The bounds can be used in a test for convergence of iterative dynamic programming algorithms for solving SSP problems, as well as in action elimination procedures that can accelerate convergence by excluding provably suboptimal actions that do not need to be re-evaluated each iteration. The techniques are shown to be effective for both decision-theoretic planning and probabilistic model checking.

The solution of a single optimization problem often requires computationally-demanding evaluations; this is especially true in optimal design of engineering components and systems described by partial differential equations. We present a technique for the rapid and reliable optimization of systems characterized by linear-functional outputs of partial differential equations with affine parameter dependence. The critical ingredients of the method are: (i) reduced-basis techniques for dimension reduction in computational requirements; (ii) an "off-line/on-line" computational decomposition for the rapid calculation of outputs of interest and respective sensitivities in the limit of many queries; (iii) a posteriori error bounds for rigorous uncertainty and feasibility control; (iv) Interior Point Methods (IPMs) for efficient solution of the optimization problem; and (v) a trust-region Sequential Quadratic Programming (SQP) interpretation of IPMs for treatment of possibly non-convex costs and constraints. / Singapore-MIT Alliance (SMA)

Linear time-invariant models are widely used in the control community. They often serve as approximations of nonlinear systems. For control purposes linear approximations are often good enough since feedback control systems are inherently robust to model errors. In this thesis some of the possibilities for linear time-varying modeling are studied. In the thesis it is shown that the balanced truncation procedure can be applied to reduce the order of linear time-varying systems. Many of the attractive properties of balanced truncation for time-invariant systems can be generalized into the time-varying framework. For example, it is shown that a truncated input-output stable system will be input-output stable, and computable simple worst-case error bounds are derived. The method is illustrated with model reduction of a nonlinear diesel exhaust catalyst model. It is also shown that linear time-periodic models can be used for analysis of systems with power converters. Power converters produce harmonics in the power grids and give frequency coupling that cannot be modeled with standard time-invariant linear models. With time-periodic models we can visualize the coupling and also use all the available tools for linear time-varying systems, such as balanced truncation. The method is illustrated on inverter locomotives. / QC 20120208