Thesis (PhD (Electrical and Electronic Engineering))--University of Stellenbosch, 2011. / AFRIKAANSE OPSOMMING: Hierdie proefskrif handel oor die natuurlike balanserings meganismes van veelvlakkige en modulêre omsetters wat fase-skuif dragolf puls wydte modulasie gebruik. Die meganismes kan in twee hoof groepe verdeel word: ‘n swak balanserings meganisme wat afhanklik is van die oorvleuling van die skakelfunksies en ‘n sterk meganisme wat voorkom ongeag of die skakelfunksies oorvleul al dan nie. Die sterk meganisme verdeel verder in twee subgroepe, ‘n direkte oordrag van onbalans energie en ‘n meganisme wat afhang van die verliese in die stelsel. Elkeen van die meganismes word aan die hand van ‘n omsetter topologie waarin die spesifieke meganisme oorheers beskryf en ontleed. In die ondersoek word klem geplaas op die daarstelling van uitdrukkings om die tydskonstantes van herbalansering na ’n afwyking vir elk van die omsetter toplologieë te beskryf. / ENGLISH ABSTRACT: This thesis investigates the natural balancing mechanisms in multilevel and modular converters using phase shifted carrier pulse width modulation. Two groups of mechanisms are identified; a weak balancing mechanism that is only present when the switching functions are interleaved and a strong mechanism that occurs irrespective of the interleaving of the switching functions. It is further shown that the strong balancing mechanism can be divided into a balancing mechanism that depends on the direct exchange of unbalance energy and a loss based balancing mechanism. Each of the mechanisms is discussed and analysed using a converter where the specific mechanism dominates as example. Emphasis is placed on the calculation of the rebalancing time constant following a perturbation. Closed form expressions for the rebalancing time constants for each of the analysed converters are presented.

Given two samples from two populations, one could ask how similar the populations are, that is, how close their probability distributions are. For absolutely continuous distributions, one way to measure the proximity of such populations is to use a measure of distance (metric) between the probability density functions (which are unknown given that only samples are observed). In this work, we work with the integrated squared distance as metric. To measure the uncertainty of the squared integrated distance, we first model the uncertainty of each of the probability density functions using a nonparametric Bayesian method. The method consists of estimating the probability density function f (or its logarithm) using Fourier series {f0;f1; :::;fI}. Assigning a prior distribution to f is then equivalent to assigning a prior distribution to the coefficients of this series. We used the prior suggested by Scricciolo (2006) (sieve prior), which not only places a prior on such coefficients, but also on I itself, so that in reality we work with a Bayesian mixture of finite dimensional models. To obtain posterior samples of such mixture, we marginalize out the discrete model index parameter I and use a statistical software called Stan. We conclude that the Bayesian Fourier series method has good performance when compared to kernel density estimation, although both methods often have problems in the estimation of the probability density function near the boundaries. Lastly, we showed how the methodology of Fourier series can be used to access the uncertainty regarding the similarity of two samples. In particular, we applied this method to dataset of patients with Alzheimer. / Dadas duas amostras de duas populações, pode-se questionar o quão parecidas as duas populações são, ou seja, o quão próximas estão suas distribuições de probabilidade. Para distribuições absolutamente contínuas, uma maneira de mensurar a proximidade dessas populações é utilizando uma medida de distância (métrica) entre as funções densidade de probabilidade (as quais são desconhecidas, em virtude de observarmos apenas as amostras). Nesta dissertação, utilizamos a distância quadrática integrada como métrica. Para mensurar a incerteza da distância quadrática integrada, primeiramente modelamos a incerteza sobre cada uma das funções densidade de probabilidade através de uma método bayesiano não paramétrico. O método consiste em estimar a função de densidade de probabilidade f (ou seu logaritmo) usando séries de Fourier {f0;f1; :::;fI}. Atribuir uma distribuição a priori para f é então equivalente a atribuir uma distribuição a priori aos coeficientes dessa serie. Utilizamos a priori sugerida em Scricciolo (2006) (priori de sieve), a qual não coloca uma priori somente nesses coeficientes, mas também no próprio I, de modo que, na realidade, trabalhamos com uma mistura bayesiana de modelos de dimensão finita. Para obter amostras a posteriori dessas misturas, marginalizamos o parâmetro (discreto) de indexação de modelos, I, e usamos um software estatístico chamado Stan. Concluímos que o método bayesiano de séries de Fourier tem boa performance quando comparado ao de estimativa de densidade kernel, apesar de ambos os métodos frequentemente apresentarem problemas na estimação da função de densidade de probabilidade perto das fronteiras. Por fim, mostramos como a metodologia de series de Fourier pode ser utilizada para mensurar a incerteza a cerca da similaridade de duas amostras. Em particular, aplicamos este método a um conjunto de dados de pacientes com doença de Alzheimer.

This thesis systematically derives an innovative approach, called FOMFE, for fingerprint ridge orientation modeling based on 2D Fourier expansions, and explores possible applications of FOMFE to various aspects of a fingerprint identification system. Compared with existing proposals, FOMFE does not require prior knowledge of the landmark singular points (SP) at any stage of the modeling process. This salient feature makes it immune from false SP detections and robust in terms of modeling ridge topology patterns from different typological classes. The thesis provides the motivation of this work, thoroughly reviews the relevant literature, and carefully lays out the theoretical basis of the proposed modeling approach. This is followed by a detailed exposition of how FOMFE can benefit fingerprint feature analysis including ridge orientation estimation, singularity analysis, global feature characterization for a wide variety of fingerprint categories, and partial fin gerprint identification. The proposed methods are based on the insightful use of theory from areas such as Fourier analysis of nonlinear dynamic systems, analytical operators from differential calculus in vector fields, and fluid dynamics. The thesis has conducted extensive experimental evaluation of the proposed methods on benchmark data sets, and drawn conclusions about strengths and limitations of these new techniques in comparison with state-of-the-art approaches. FOMFE and the resulting model-based methods can significantly improve the computational efficiency and reliability of fingerprint identification systems, which is important for indexing and matching fingerprints at a large scale.

Given two samples from two populations, one could ask how similar the populations are, that is, how close their probability distributions are. For absolutely continuous distributions, one way to measure the proximity of such populations is to use a measure of distance (metric) between the probability density functions (which are unknown given that only samples are observed). In this work, we work with the integrated squared distance as metric. To measure the uncertainty of the squared integrated distance, we first model the uncertainty of each of the probability density functions using a nonparametric Bayesian method. The method consists of estimating the probability density function f (or its logarithm) using Fourier series {f0;f1; :::;fI}. Assigning a prior distribution to f is then equivalent to assigning a prior distribution to the coefficients of this series. We used the prior suggested by Scricciolo (2006) (sieve prior), which not only places a prior on such coefficients, but also on I itself, so that in reality we work with a Bayesian mixture of finite dimensional models. To obtain posterior samples of such mixture, we marginalize out the discrete model index parameter I and use a statistical software called Stan. We conclude that the Bayesian Fourier series method has good performance when compared to kernel density estimation, although both methods often have problems in the estimation of the probability density function near the boundaries. Lastly, we showed how the methodology of Fourier series can be used to access the uncertainty regarding the similarity of two samples. In particular, we applied this method to dataset of patients with Alzheimer. / Dadas duas amostras de duas populações, pode-se questionar o quão parecidas as duas populações são, ou seja, o quão próximas estão suas distribuições de probabilidade. Para distribuições absolutamente contínuas, uma maneira de mensurar a proximidade dessas populações é utilizando uma medida de distância (métrica) entre as funções densidade de probabilidade (as quais são desconhecidas, em virtude de observarmos apenas as amostras). Nesta dissertação, utilizamos a distância quadrática integrada como métrica. Para mensurar a incerteza da distância quadrática integrada, primeiramente modelamos a incerteza sobre cada uma das funções densidade de probabilidade através de uma método bayesiano não paramétrico. O método consiste em estimar a função de densidade de probabilidade f (ou seu logaritmo) usando séries de Fourier {f0;f1; :::;fI}. Atribuir uma distribuição a priori para f é então equivalente a atribuir uma distribuição a priori aos coeficientes dessa serie. Utilizamos a priori sugerida em Scricciolo (2006) (priori de sieve), a qual não coloca uma priori somente nesses coeficientes, mas também no próprio I, de modo que, na realidade, trabalhamos com uma mistura bayesiana de modelos de dimensão finita. Para obter amostras a posteriori dessas misturas, marginalizamos o parâmetro (discreto) de indexação de modelos, I, e usamos um software estatístico chamado Stan. Concluímos que o método bayesiano de séries de Fourier tem boa performance quando comparado ao de estimativa de densidade kernel, apesar de ambos os métodos frequentemente apresentarem problemas na estimação da função de densidade de probabilidade perto das fronteiras. Por fim, mostramos como a metodologia de series de Fourier pode ser utilizada para mensurar a incerteza a cerca da similaridade de duas amostras. Em particular, aplicamos este método a um conjunto de dados de pacientes com doença de Alzheimer.

<p>U disertaciji se na bazi redefinisanih pojmova kako prostornog (inte-gralnog) izvoda tako i ostatka (rezidijuma) kompleksnih funkcija defini&scaron;e pojam totalne vrednosti nesvojstvenih integrala. Ova redefinisana vrednost integrala iskori&scaron;ćcena je za dokaz teoreme, koja je uop&scaron;tenje Dirihlejeve teoreme o razvoju realnih funkcija u Furijeov trigonometrijski red.</p> / <p>In this thesis we define the notion of the total value of improper integrals based on the redefined notions of the spatial (integral) derivative and the residue of complex functions. This redefined integrals is used to prove the theorem which is generalization of Dirichlet theorem on the expansion of real valued functions into Fourier trigonometric series.</p>

We introduce the notion of the modulus of fractality and consider the problem of approximation of functions with a restriction on the modulus of fractality by partial sums of trigonometric Fourier series (Fourier sums). The upper estimate of the difference between the function and the corresponding Fourier sum in terms of the modulus of continuity and the modulus of fractality is given. Examples of functions from the considered classes with trigonometric Fourier series diverging at some point are constructed. / Вводится понятие модуля фрактальности и рассматривается задача приближения функций с ограничением на модуль фрактальности частичными суммами тригонометрических рядов Фурье (суммами Фурье). Приведена оценка сверху модуля разности функции и соответствующей суммы Фурье, выраженная в терминах модуля непрерывности и модуля фрактальности. Построены примеры функций из рассматриваемых классов с расходящимся в некоторой точке тригонометрическим рядом Фурье.

The vibration of complex mechanical systems that include coupled rotating and stationary bodies motivates this work. A semi-analytical model is developed for high-speed, compliant, rotating bodies. Exploiting the axisymmetry of the rotating body, the developed semi-analytical model only discretizes the two-dimensional radial cross-section; Fourier series are used in the circumferential direction. The corresponding formulation for thin-walled, axisymmetric shells is given. Even though the body is axisymmetric, its deflection as well as external forces, constraints, and supports acting on the body are allowed to be asymmetric. These asymmetric elements can be stationary or rotating. The model includes Coriolis and centripetal effects. The prestress (or geometric) stiffness matrix that arises from external forces and constant centripetal acceleration has additional terms compared to the literature, and these terms can significantly change the natural frequencies. Discrete stiffness-damper elements, elastic foundations, and constraint equations are used to couple the rotating body to other rotating and stationary bodies. The model is developed in a stationary reference frame to avoid time-dependent coefficients in the equations of motion when coupled to stationary components. Surface constraints are developed using equivalent force relations between multiple points on the surface and a reference node. Discrete stiffness-dampers, asymmetric elastic foundation, and asymmetric constraints introduce non-axisymmetry in the system. The speed-dependent natural frequencies and complex-valued vibration modes, presence of multiple Fourier harmonics in each mode, changes to critical speeds, divergence and flutter instability phenomena, and eigenvalue veering are investigated for spinning systems with asymmetric features. The developed semi-analytical model is used for rotationally periodic systems, for example, planetary gears. Rotationally periodic systems consist of multiple vibrating, rotating central components and substructures. The model is developed in a reference frame rotating with the central component that supports the substructures. Structured modal properties of the cyclically symmetric systems and diametrically opposed systems are investigated. The modes of the spinning system are categorized into translational-tilting, rotational-axial, and substructure modes. Time-varying coupling elements act as parametric excitation in the system. Large strain energy in the coupling elements lead to large parametric instability regions. The analytical closed-form expression of the parametric instability bandwidth obtained using a perturbation method compares well with numerical results from Floquet theory. / Doctor of Philosophy / Complex mechanical systems, for example, mechanical transmission, consist of coupled rotating and stationary bodies. The vibrations of rotating bodies are transmitted to the other bodies through coupling elements. To reduce weight of the system, the rotating bodies are made thin-walled resulting in increased flexibility of the body. The existing lumped parameter/rigid body models do not account for the flexibility of these rotating bodies. Conventional three-dimensional finite element models lead to a large number of degrees of freedom in the system, increasing the computational cost. We aim to develop a computationally efficient model to analyze the dynamics and vibration of complex mechanical systems. Most rotating bodies can be approximated as axisymmetric. The axisymmetric property of the rotating body is harnessed to reduce the three-dimensional model of the body to a two-dimensional radial cross-section using Fourier series in the circumferential direction. This reduces the system degrees of freedom. Coriolis, centripetal, and prestress effects are included in the model. Discrete stiffness-dampers, elastic foundations, and constraint equations couple the rotating body to other rotating and stationary bodies. Non-axisymmetric coupling elements and forces introduce asymmetry in the system. The system model for these asymmetric systems are developed in a stationary reference frame to avoid time-dependent coefficient equations of motion. Flexible stationary bodies alter the natural frequencies and vibration modes of the system. Instabilities, critical speeds, effects of asymmetry on the natural frequencies and vibration modes of the system are investigated. The model is extended for rotationally periodic systems, for examples, planetary gears and bearings. This model is developed in the reference frame that rotates with the central component that supports substructures. Structured modal characteristics are observed for the rotationally periodic systems. Changing contact conditions act as a source of parametric excitation in systems. Parametric resonances occur when natural frequencies of vibration with large strain energy in the coupling elements sum to the excitation frequency. Parametric instability regions obtained using an analytical equation compare well with numerical results.

O modelo semiparamétrico de Cox é frequentemente utilizado na modelagem de dados de sobrevivência, pois é um modelo muito flexível e permite avaliar o efeito das covariáveis sobre a taxa de falha. Uma das principais vantagens é a fácil interpretação, de modo que a razão de riscos de dois indivíduos não varia ao longo do tempo. No entanto, em algumas situações a proporcionalidade dos riscos para uma dada covariável pode não ser válida e, este caso, uma abordagem que não dependa de tal suposição é necessária. Nesta tese, propomos um modelo tipo Cox em que o efeito da covariável e a função de risco basal são representadas via bases de Fourier e ondaletas de Haar clássicas e deformadas. Propomos também um procedimento de predição da função de sobrevivência para um paciente específico. Estudos de simulações e aplicações a dados reais sugerem que nosso método pode ser uma ferramenta valiosa em situações práticas em que o efeito da covariável é dependente do tempo. Por meio destes estudos, fazemos comparações entre as duas abordagens propostas, e comparações com outra já conhecida na literatura, onde verificamos resultados satisfatórios. / The semiparametric Cox model is often considered when modeling survival data. It is very flexible, allowing for the evaluation of covariates effects. One of its main advantages is the easy of interpretation, as long as the rate of the hazards for two individuals does not vary over time. However, this proportionality of the hazards may not be true in some practical situations and, in this case, an approach not relying on such assumption is needed. In this thesis we propose a Cox-type model that allows for time-varying covariate effects, for which the baseline hazard is based on Fourier series and wavelets on a time-frequency representation. We derive a prediction method for the survival of future patients with any specific set of covariates. Simulations and an application to a real data set suggest that our method may be a valuable tool to model data in practical situations where covariate effects vary over time. Through these studies, we make comparisons between the two approaches proposed here and comparisons with other already known in the literature, where we verify satisfactory results.

O modelo semiparamétrico de Cox é frequentemente utilizado na modelagem de dados de sobrevivência, pois é um modelo muito flexível e permite avaliar o efeito das covariáveis sobre a taxa de falha. Uma das principais vantagens é a fácil interpretação, de modo que a razão de riscos de dois indivíduos não varia ao longo do tempo. No entanto, em algumas situações a proporcionalidade dos riscos para uma dada covariável pode não ser válida e, este caso, uma abordagem que não dependa de tal suposição é necessária. Nesta tese, propomos um modelo tipo Cox em que o efeito da covariável e a função de risco basal são representadas via bases de Fourier e ondaletas de Haar clássicas e deformadas. Propomos também um procedimento de predição da função de sobrevivência para um paciente específico. Estudos de simulações e aplicações a dados reais sugerem que nosso método pode ser uma ferramenta valiosa em situações práticas em que o efeito da covariável é dependente do tempo. Por meio destes estudos, fazemos comparações entre as duas abordagens propostas, e comparações com outra já conhecida na literatura, onde verificamos resultados satisfatórios. / The semiparametric Cox model is often considered when modeling survival data. It is very flexible, allowing for the evaluation of covariates effects. One of its main advantages is the easy of interpretation, as long as the rate of the hazards for two individuals does not vary over time. However, this proportionality of the hazards may not be true in some practical situations and, in this case, an approach not relying on such assumption is needed. In this thesis we propose a Cox-type model that allows for time-varying covariate effects, for which the baseline hazard is based on Fourier series and wavelets on a time-frequency representation. We derive a prediction method for the survival of future patients with any specific set of covariates. Simulations and an application to a real data set suggest that our method may be a valuable tool to model data in practical situations where covariate effects vary over time. Through these studies, we make comparisons between the two approaches proposed here and comparisons with other already known in the literature, where we verify satisfactory results.

De nombreux systèmes utilisent des convertisseurs électroniques de puissance fonctionnant en modulation de largeur d'impulsions (MLI) comme organe de réglage ou de commande. On peut en particulier citer les entraînements à vitesse variable, les alimentations à découpage ou les filtres actifs. L'objectif du travail est de contribuer à l'étude en temps discret du fonctionnement de ces systèmes lorsqu'ils sont munis d'une commande numérique dont la période d'échantillonnage est synchronisée sur la période de modulation de largeur d'impulsions. Pour cela nous avons développé une méthode originale qui consiste à écrire les équations différentielles qui décrivent la dynamique du système en remplaçant sur chaque période de modulation les fonctions binaires qui représentent les états successifs (ON ou OFF) des semi-conducteurs de puissance par un développement limité en série de Fourier. En introduisant de manière judicieuse dans les équations différentielles du système, des termes harmoniques des fonctions binaires qui décrivent les états des interrupteurs, cette manière de procéder nous a permis de décomposer l'étude en deux étapes : · dans une première étape en limitant le développement en série de Fourier des fonctions binaires décrivant les états des interrupteurs à leur premier terme (c'est-à-dire à leurs valeurs moyennes) nous avons construit un modèle discret d'ordre zéro qui rend compte de l'effet moyen de la découpe MLI, · dans une deuxième étape nous avons construit un modèle dynamique approché de l'écart existant entre le modèle d'ordre zéro et le modèle détaillé du système. Ce modèle fournit une bonne estimation des ondulations induites par la découpe MLI dans les grandeurs du système. Pour le cas des entraînements par moteurs synchrones à aimants permanents alimentés par onduleurs de tension nous avons ainsi pu mettre en évidence que l'étude de la stabilité des boucles des courants du moteur à partir du modèle d'ordre zéro reste valable en présence des ondulations dues à la MLI pour autant qu'on utilise une MLI symétrique avec rafraîchissement des références une fois par période. / Nowadays, a large number of systems are using Pulse Width Modulation (PWM) Power Electronic Converters as control part. These systems can, for example, be variable-speed drives, switching power supplies or active filters. The goal of this work is to contribute to the study of operation of this type of system when they are equipped with a digital control part of which the sampling period is synchronized on the modulation one. After having shown how the equations of these systems can be written down and indicated how their evolution in closed loop operation can be described by using iterative maps, we have tried to develop an approach which allows simplifying this study. The iterative map method has been used in order to take into account the “sequential “character of these systems, i.e. the fact that the control signals are sequences of events which correspond to the changes of the state of the converter semiconductor switches. Therefore we have developed an original method which consists in replacing on each modulation period, in the differential equations describing the system dynamics, the binary functions representing the successive ON or OFF states of the power semiconductors, by a limited Fourier series development of these functions. This method has allowed splitting the study of these systems into two steps: • in the first step a “zero order” model was built. This first model, which provides the average effect of the PWM pulse pattern, has been obtained by limiting the Fourier series development of the binary functions describing the states of the converter switches, to their first term, the one corresponding to their average values on the PWM period, • in the second step we have introduced an approached dynamic model characterising the error between the zero order model and the exact model of the system. This model was built by considering a well chosen number of harmonic terms of the binary functions describing the ON-OFF states of the switches in the differential equations of the system. By combining the error model and the zero order model we have been able to estimate in which measure the ripples induced by the PWM modulation affect the results of the system stability study made by using the zero order model. For the case of Permanent Magnets Synchronous Motors fed by Voltage Source Inverters we have shown that the study of the stability of the motor currents loops made with the zero order model remains valid in the presence of ripples induced by the PWM, as long as a symmetrical modulation is used, with references which are refreshed one time on each switching period.