Thesis (PhD (Statistics and Actuarial Science))—Stellenbosch University, 2008. / Kernel Fisher discriminant analysis (KFDA) is a kernel-based technique that can be used to classify observations of unknown origin into predefined groups. Basically, KFDA can be viewed as a non-linear extension of Fisher’s linear discriminant analysis (FLDA). In this thesis we give a detailed explanation how FLDA is generalized to obtain KFDA. We also discuss two methods that are related to KFDA. Our focus is on binary classification. The influence of atypical cases in discriminant analysis has been investigated by many researchers. In this thesis we investigate the influence of atypical cases on certain aspects of KFDA. One important aspect of interest is the generalization performance of the KFD classifier. Several other aspects are also investigated with the aim of developing criteria that can be used to identify cases that are detrimental to the KFD generalization performance. The investigation is done via a Monte Carlo simulation study. The output of KFDA can also be used to obtain the posterior probabilities of belonging to the two classes. In this thesis we discuss two approaches to estimate posterior probabilities in KFDA. Two new KFD classifiers are also derived which use these probabilities to classify observations, and their performance is compared to that of the original KFD classifier. The main objective of this thesis is to develop criteria which can be used to identify cases that are detrimental to the KFD generalization performance. Nine such criteria are proposed and their merit investigated in a Monte Carlo simulation study as well as on real-world data sets. Evaluating the criteria on a leave-one-out basis poses a computational challenge, especially for large data sets. In this thesis we also propose using the smallest enclosing hypersphere as a filter, to reduce the amount of computations. The effectiveness of the filter is tested in a Monte Carlo simulation study as well as on real-world data sets.

The dissertation introduces Sphere Support Vector Machines (SphereSVM) and Minimal Norm Support Vector Machines (MNSVM) as the new fast classification algorithms that use geometrical properties of the underlying classification problems to efficiently obtain models describing training data. SphereSVM is based on combining minimal enclosing ball approach, state of the art nearest point problem solvers and probabilistic techniques. The blending of the three speeds up the training phase of SVMs significantly and reaches similar (i.e., practically the same) accuracy as the other classification models over several big and large real data sets within the strict validation frame of a double (nested) cross-validation (CV). MNSVM is further simplification of SphereSVM algorithm. Here, relatively complex classification task was converted into one of the simplest geometrical problems -- minimal norm problem. This resulted in additional speedup compared to SphereSVM. The results shown are promoting both SphereSVM and MNSVM as outstanding alternatives for handling large and ultra-large datasets in a reasonable time without switching to various parallelization schemes for SVMs algorithms proposed recently. The variants of both algorithms, which work without explicit bias term, are also presented. In addition, other techniques aiming to improve the time efficiency are discussed (such as over-relaxation and improved support vector selection scheme). Finally, the accuracy and performance of all these modifications are carefully analyzed and results based on nested cross-validation procedure are shown.

De nos jours la quasi-totalité des grandes métropoles du monde comptent avec des réseaux de transport étendus et ont pour la plupart satisfait leurs besoins de construction d’ouvrages souterrains. Cependant le parc des ouvrages existants est vieillissant et une part importante est en service depuis plusieurs décennies, par conséquence, l’entretien et la pérennisation des infrastructures souterraines existantes un enjeu majeur pour la gestion et le développement des villes. En vue d’améliorer leur politique de maintenance, c’est-à-dire pérenniser leur patrimoine et le maintenir dans de bonnes conditions de sécurité et d’exploitation, il est nécessaire que les gestionnaires d’ouvrages puissent mettre à jour leurs techniques de diagnostic (AFTES, 2005). Il est donc impératif de disposer d’outils leur permettant de mieux évaluer l’état réel des ouvrages. Or les méthodes dont disposent à l’heure actuelle les gestionnaires sont le plus souvent soit insuffisantes pour apporter une information quantitative de qualité, soit mal adaptées aux contraintes de ces ouvrages. Dans ce contexte, le projet ANR MéDiTOSS « Méthodologie de Diagnostic des Tunnels et Ouvrages Souterrains en Service » vise à développer une nouvelle méthodologie de diagnostic adaptée à ce type d’ouvrage en caractérisant les différents composants de l’ouvrage (structure du revêtement, terrain encaissant ainsi que leur interface ou contact). L’objectif de cette thèse est de fournir au gestionnaire des indicateurs physiques et mécaniques lui permettant d’orienter sa politique de maintenance en s’appuyant sur une expertise plus certaine. / Maintenance and sustainability of underground structures are the major tasks for infrastructures’ owners. To carry out these tasks, managers need to qualify the actual state of these facilities. There are very few operational methods to describe separately each component of an underground structure (structure lining, contact interface and surrounding soil). Most of them are mainly based on visual inspection. Nevertheless only the intrados of the tunnel is open to visual inspection. To improve their maintenance policy, that is to say sustain their heritage and keep it in good conditions of safety and operational, it is necessary that manager’s can update their diagnosis methods (AFTES, 2005). It is therefore imperative to have tools to better assess the actual condition of structures. However, the methods available at present managers are often either insufficient to provide quantitative information quality or poorly adapted to the constraints of these works. In this context, the ANR project MéDiTOSS "Methodology Diagnostic Tunnels and Underground Structures in Service" aims to develop a new diagnostic methodology adapted to this type of work characterizing the various components of the structure (structure of the lining, enclosing soil and their interface or contact). The objective of this thesis is to provide to the managers of physical and mechanical indicators that will guide its maintenance policy based on a certain expertise.

As is the case with many other manufacturers of vehicles, Volvo Construction Equipment has a constantly increasing amount of electric and electronic equipment in their articulated haulers. These are of great use in modern machines, bringing functions, and safety that were not possible before, but they also bring more cables to handle. In the case of the articulated haulers a quite thick cable harness of about 15 meters in length reaches from the driver’s cabin in the front to the components in the far back end of the vehicle. This is not only a lot of long cables to handle both during assembly and service, but the nature of signals traveling in cables is that the voltage gets weaker with distance. This phenomenon has to be accounted for by measuring devices dependent on the voltage. It has been suggested that a device referred to as an intelligent node, or ICCS-module, which communicates digitally via CAN could be installed in the back of the articulated haulers. This module would be independent of the mentioned drop in voltage. The ICCS-module will be receiving a small bundle of cables being routed from the driver’s cabin. From it cables would go out to a majority of the components in its vicinity. The components connected to the node would not need to have any other cables. This thesis is focused on the mechanical aspects of installing this ICCS-module. Having electric and electronic equipment on construction vehicles is a challenge when it comes to protecting the device from the harsh environment that is the hauler’s exterior. It will have to withstand being immersed in water for long periods of time, greatly varying temperatures, vibrations and shocks as well as being hit by projectiles. The placement as well as the design of the enclosure should be chosen with respect to both the devices length of life and how well it fulfils its intended role and achieves the expected results. To produce an enclosure and find a placement aiming to satisfy these conditions, a traditional product development process were executed. The articulated haulers as well as relevant literature were researched. Concepts were generated and evaluated by both the author and by employees at Volvo Construction Equipment until a final concept for the enclosure and placement were found. The enclosure were then designed in detail specifying the material, manufacturing techniques, controlled for thermodynamic circumstances, modelled in Catia V5 and controlled for vibrations. It was concluded that the enclosure should be able protect the ICCS-module after some more development and that the placement and cable routing results in a much shorter total cable length.

Современная система нормирования тепловой защиты зданий в России является одной из передовых в мире, потому что СП 50.13330.2012 «Тепловая защита зданий» содержит нормируемые параметры и передовые методы их расчета, учитывающие все известные эффекты теплопереноса в ограждающих конструкциях. Методы нормирования, применяемые в Китае, являются самобытными и отличаются как от российских, так и от европейских. Нормируемые параметры, как и методы их расчета, в России и Китае различаются. Для их сравнения необходимо учитывать их принципиальные особенности. Как для российских, так и для китайских специалистов по тепловой защите зданий, является актуальным изучение и сравнение потребления энергии на отопление зданий, а также нормирование тепловой защиты зданий в КНР и России. Поэтому актуальным является исследование принципов сравнения как тепловой защиты, так и потребления энергии в здания. В работе проанализированы нормирование и расчет тепловой защиты и энергопотребления зданий в России и Китае. Показано, что сравнивать тепловую защиту отдельных ограждающих конструкций можно при помощи сравнения приведенных сопротивлений (или приведенных коэффициентов теплопередачи) ограждающих конструкций. Комплексную характеристику теплозащиты зданий сравнивать нельзя, поскольку она отсутствует в нормах Китая. Потребление энергии зданием (и энергосбережение) можно сравнивать при помощи удельных характеристик, определяющих тепловые потребности здания и климатических характеристик региона строительства. Приведенные в китайских нормах теплотехнические требования к ограждающим конструкциям и зданиям связаны с климатом региона строительства. Каждое из требований требует сопоставления с аналогичными требованиями российских норм. Выполненная, таким образом работа, позволит приблизиться к оптимальным решениям по нормированию тепловой защиты зданий. / The modern system of standardization of thermal protection of buildings in Russia is one of the most advanced in the world, because SP 50.13330.2012 “Thermal protection of buildings” contains standardized parameters and advanced methods for their calculation, considering all known effects of heat transfer in building envelopes. The rationing methods used in China are original and differ from both Russian and European ones. Standardized parameters, as well as methods for calculating them, differ in Russia and China. To compare them, it is necessary to consider their fundamental features. For both Russian and Chinese specialists in thermal protection of buildings, it is relevant to study and compare energy consumption for heating buildings, as well as standardization of thermal protection of buildings in China and Russia. Therefore, it is relevant to study the principles of comparison of both thermal protection and energy consumption in buildings. The work analyzes the standardization and calculation of thermal protection and energy consumption of buildings in Russia and China. It is shown that the thermal protection of individual enclosing structures can be compared by comparing the given resistances (or given heat transfer coefficients) of the enclosing structures. The comprehensive performance of thermal protection of buildings cannot be compared, since it is not included in Chinese standards. A building's energy consumption (and energy savings) can be compared using specific characteristics that determine the building's thermal needs and the climatic characteristics of the construction region. The thermal requirements for enclosing structures and buildings given in Chinese standards are related to the climate of the construction region. The thermal requirements for enclosing structures and buildings given in Chinese standards are related to the climate of the construction region. Each of the requirements requires comparison with similar requirements of Russian standards. The work completed in this way will allow us to get closer to optimal solutions for regulating the thermal protection of buildings.