In the future, the performance of the railway system must be improved to accommodate increasing passenger volumes and service quality demands. Track switches are a vital part of the rail infrastructure, enabling traffic to take different routes. All modern switch designs have evolved from a design first patented in 1832. However, switches present single points of failure, require frequent and costly maintenance interventions, and restrict network capacity. Fault tolerance is the practice of preventing subsystem faults propagating to whole-system failures. Existing switches are not considered fault tolerant. This thesis describes the development and potential performance of fault-tolerant railway track switching solutions. The work first presents a requirements definition and evaluation framework which can be used to select candidate designs from a range of novel switching solutions. A candidate design with the potential to exceed the performance of existing designs is selected. This design is then modelled to ascertain its practical feasibility alongside potential reliability, availability, maintainability and capacity performance. The design and construction of a laboratory scale demonstrator of the design is described. The modelling results show that the performance of the fault tolerant design may exceed that of traditional switches. Reliability and availability performance increases significantly, whilst capacity gains are present but more marginal without the associated relaxation of rules regarding junction control. However, the work also identifies significant areas of future work before such an approach could be adopted in practice.

ABSTRACT One of the most important tasks in the chemical industry is the separation of multicomponent liquid mixtures into one or more high-purity products. Several technologies are feasible for this task, either alone or in combination, such as distillation, extraction, crystallization, ect. Among these, distillation is by far the most widely spread and has a long history in chemical technology. However, until recently, there has been no systematic approach for understanding the separation of complex mixtures where azeotropes and multiple liquid phases may occur. There has been a growing interest in the use of residue curve and column profiles for the preliminary design of distillation columns. Residue curves and column profile are not only used to predict the composition changes in the distillation column but also to determine the feasibility of the proposed separation. Recently, theory underlying column profile maps has been developed by Tapp, Holland and co-workers. However there has been no direct experimental validation of the predictions of the column profile map theory. The main aim of this thesis is to experimentally verify some of the predictions of column profile map theory. A simple experimental batch apparatus has been developed to measure residue curve maps (RCMs) by Tapp and co-workers, the apparatus was modified so that it could be used to measure column profile maps (CPMs) in this thesis. CPM theory has shown that CPMs are linear transforms of the residues curve maps (RCMs). A stable node which was the apex of a mass balance triangle (MBT) was introduced inside the MBT, this was done by transforming the RCMs to CPMs using the appropriate distillate composition xd and reflux ratio R. It was also shown that the saddle point which was on the boundary of the triangle of the RCM can be shifted inside the MBT by transforming the RCM to CPM. This is again in accordance with theoretical predictions of CPM theory. iv Residue curves (RCs) and pinch point curves (PPCs) are used to determine the operation leaves and hence the feasible region for distillation columns operating at a specific distillate and bottoms composition for all fixed reflux ratio. The operating leaves were expanded beyond the pinch point curve by varying the reflux ratio from a higher reflux to a lower reflux ratio. This showed that one can effectively cross the pinch point curve hence expanding the operating leave. Finally the importance of experimentally measuring CPMs is demonstrated. Two thermodynamic models were used to predict the profiles of a complex system. The binary vapor-liquid equilibrium (VLE) diagrams and the residue curves produced from using these two thermodynamic models did not predict the same topology. The composition of the profiles were not the same because there were multiple liquid phases involved in this system, which made it difficult for the researchers to measure the correct profiles. Column profile maps were simulated using the different thermodynamic models, they also showed that there is some discrepancy between the predictions of the two models.

Cette thèse s’inscrit dans le contexte de la reconnaissance des activités à partir de séquences vidéo qui est une des préoccupations majeures dans le domaine de la vision par ordinateur. Les domaines d'application pour ces systèmes de vision sont nombreux notamment la vidéo surveillance, la recherche et l'indexation automatique de vidéos ou encore l'assistance aux personnes âgées. Cette tâche reste problématique étant donnée les grandes variations dans la manière de réaliser les activités, l'apparence de la personne et les variations des conditions d'acquisition des activités. L'objectif principal de ce travail de thèse est de proposer une méthode de reconnaissance efficace par rapport aux différents facteurs de variabilité. Les représentations basées sur les points d'intérêt ont montré leur efficacité dans les travaux d'art; elles ont été généralement couplées avec des méthodes de classification globales vue que ses primitives sont temporellement et spatialement désordonnées. Les travaux les plus récents atteignent des performances élevées en modélisant le contexte spatio-temporel des points d'intérêts par exemple certains travaux encodent le voisinage des points d'intérêt à plusieurs échelles. Nous proposons une méthode de reconnaissance des activités qui modélise explicitement l'aspect séquentiel des activités tout en exploitant la robustesse des points d'intérêts dans les conditions réelles. Nous commençons par l'extractivité des points d'intérêt dont a montré leur robustesse par rapport à l'identité de la personne par une étude tensorielle. Ces primitives sont ensuite représentées en tant qu'une séquence de sac de mots (BOW) locaux: la séquence vidéo est segmentée temporellement en utilisant la technique de fenêtre glissante et chacun des segments ainsi obtenu est représenté par BOW des points d'intérêt lui appartenant. Le premier niveau de notre système de classification séquentiel hybride consiste à appliquer les séparateurs à vaste marge (SVM) en tant que classifieur de bas niveau afin de convertir les BOWs locaux en des vecteurs de probabilités des classes d'activité. Les séquences de vecteurs de probabilité ainsi obtenues sot utilisées comme l'entrées de classifieur séquentiel conditionnel champ aléatoire caché (HCRF). Ce dernier permet de classifier d'une manière discriminante les séries temporelles tout en modélisant leurs structures internes via les états cachés. Nous avons évalué notre approche sur des bases publiques ayant des caractéristiques diverses. Les résultats atteints semblent être intéressant par rapport à celles des travaux de l'état de l'art. De plus, nous avons montré que l'utilisation de classifieur de bas niveau permet d'améliorer la performance de système de reconnaissance vue que le classifieur séquentiel HCRF traite directement des informations sémantiques des BOWs locaux, à savoir la probabilité de chacune des activités relativement au segment en question. De plus, les vecteurs de probabilités ont une dimension faible ce qui contribue à éviter le problème de sur apprentissage qui peut intervenir si la dimension de vecteur de caractéristique est plus importante que le nombre des données; ce qui le cas lorsqu'on utilise les BOWs qui sont généralement de dimension élevée. L'estimation les paramètres du HCRF dans un espace de dimension réduite permet aussi de réduire le temps d'entrainement / Human activity recognition (HAR) from video sequences is one of the major active research areas of computer vision. There are numerous application HAR systems, including video-surveillance, search and automatic indexing of videos, and the assistance of frail elderly. This task remains a challenge because of the huge variations in the way of performing activities, in the appearance of the person and in the variation of the acquisition conditions. The main objective of this thesis is to develop an efficient HAR method that is robust to different sources of variability. Approaches based on interest points have shown excellent state-of-the-art performance over the past years. They are generally related to global classification methods as these primitives are temporally and spatially disordered. More recent studies have achieved a high performance by modeling the spatial and temporal context of interest points by encoding, for instance, the neighborhood of the interest points over several scales. In this thesis, we propose a method of activity recognition based on a hybrid model Support Vector Machine - Hidden Conditional Random Field (SVM-HCRF) that models the sequential aspect of activities while exploiting the robustness of interest points in real conditions. We first extract the interest points and show their robustness with respect to the person's identity by a multilinear tensor analysis. These primitives are then represented as a sequence of local "Bags of Words" (BOW): The video is temporally fragmented using the sliding window technique and each of the segments thus obtained is represented by the BOW of interest points belonging to it. The first layer of our hybrid sequential classification system is a Support Vector Machine that converts each local BOW extracted from the video sequence into a vector of activity classes’ probabilities. The sequence of probability vectors thus obtained is used as input of the HCRF. The latter permits a discriminative classification of time series while modeling their internal structures via the hidden states. We have evaluated our approach on various human activity datasets. The results achieved are competitive with those of the current state of art. We have demonstrated, in fact, that the use of a low-level classifier (SVM) improves the performance of the recognition system since the sequential classifier HCRF directly exploits the semantic information from local BOWs, namely the probability of each activity relatively to the current local segment, rather than mere raw information from interest points. Furthermore, the probability vectors have a low-dimension which prevents significantly the risk of overfitting that can occur if the feature vector dimension is relatively high with respect to the training data size; this is precisely the case when using BOWs that generally have a very high dimension. The estimation of the HCRF parameters in a low dimension allows also to significantly reduce the duration of the HCRF training phase

Exclusive breastfeeding for at least six months and continued breastfeeding for the first year or two is promoted as the preferred method for feeding infants. In the United States, about 79% of infants are ever breastfed and less than 27% are breastfed through the first year (Centers for Disease Control and Prevention, 2014b). The purpose of this study was to understand the role of breastfeeding experiences and underlying factors that may facilitate the maintenance of breastfeeding one’s first child as well as the re-engagement in breastfeeding with one's second child. This dissertation includes two quantitative studies using secondary data and one qualitative study involving data specifically collected for this study. The quantitative studies used data from the Infant Feeding Practices Study II and the Year 6 Follow Up surveys. Findings indicate barriers in the early breastfeeding experience (pain, trouble with milk supply, and trouble with the baby’s suck or latch) are associated with less time spent in breastfeeding maintenance among first-time mothers and experiencing trouble with the first baby’s suck or latch is associated with non-initiation with one's second child. Conversely, as levels of perceived breastfeeding self-efficacy, opinion about the best way to feed a baby, and belief about breast milk increased, or became more pro-breastfeeding over time, time spent in breastfeeding maintenance among first-time mothers also increased. Those with a high level of response efficacy were also more likely to initiate breastfeeding with a second child. In the qualitative study, 28 first-time mothers who recently stopped breastfeeding and live in Iowa participated in semi-structured interviews. The objectives of this third study were to understand and describe first-time mothers’ experiences of infant feeding throughout the entire breastfeeding maintenance period, and explore how contextual factors may act as facilitators or barriers to breastfeeding maintenance by identifying common turning points. Turning points were defined as events or periods in time perceived as personally significant, that promoted a change, maintenance, or where a new meaning associated with the experience was acquired. Findings include seven turning points (breastfeeding problems, overcoming the unknown, recurring stressors, learning how to live life with the new baby, re-entering social roles, special occasions, and letting go) that were usually associated with stressful situations. The availability of intrapersonal (e.g., individual determination) and interpersonal (e.g., social support from others) coping resources at these turning points facilitated breastfeeding maintenance. The first study is significant in the longitudinal consideration of both experiences and changes in intrapersonal and interpersonal factors and how these changes relate to breastfeeding maintenance. The second study is significant with the focus on multiparous mothers and the effect of previous breastfeeding thoughts and experiences on breastfeeding initiation with a second child. The third study is unique in the identification of breastfeeding turning points that may lead to identifying opportune times and strategies for maintenance interventions. Findings from these studies have implications for intervention planning and future research. The synthesis of findings across the three studies suggests the early breastfeeding period is paramount to breastfeeding maintenance, along with the role of one’s social environment, particularly the perceived availability of social support to meet mothers' needs. Finally, findings from this dissertation can be used to improve social support provision and multi-level interventions targeting mothers’ most salient needs at the most opportune times.

This dissertation research addresses how to detect structural changes in stochastic linear models. By introducing a special structure to the design matrix, we convert the structural change detection problem to a variable selection problem. There are many existing variable selection strategies, however, they do not fully cope with structural change detection. We design two penalized regression algorithms specifically for the structural change detection purpose. We also propose two methods involving these two algorithms to accomplish a bi-level structural change detection: they locate the change points and also recognize which predictors contribute to the variation of the model structure. Extensive simulation studies are shown to demonstrate the effectiveness of the proposed methods in a variety of settings. Furthermore, we establish asymptotic theoretical properties to justify the bi-level detection consistency for one of the proposed methods. In addition, we write an R package with computationally efficient algorithms for detecting structural changes. Comparing to traditional methods, the proposed algorithms showcase enhanced detection power and more estimation precision, with added capacity of specifying the model structures at all regimes.

Subnivean prey appeared to be the primary reason for subnivean access point use by martens. A logistic regression was used to create a predictive model for differential access point use. Prey biomass in grams and percent ground cover of coarse woody debris (CWD) were used as variables in the model. Goodness of fit of the multivariate model was 0.216; biomass was significant al p = 0.0003, CWD was significant at p = 0.0718. Mean values for prey biomass at used and unused access points were 174.2 g and 81 g, respectively, while mean values of CWD were 24.7% and 18.5%, respectively. Both CWD and prey can be used to predict access point use by martens. CWD provides access to the subnivean zone. Martens appear lo key in on access points with higher levels of prey. Red squirrel middens were found at 33% of used and 16% of unused access points (p = 0.015, n = 90). There was no significant difference in prey biomass or CWD based upon the presence or absence of a squirrel midden. Prey biomass was significantly related to access point use (p = 0.0022) and the relationship was strengthened when squirrels were included in the biomass estimates (p = 0.0001). It is likely that red squirrel middens were used as access points by martens because of the opportunity to prey on red squirrels as a prey item. Seed boluses were used to estimate the relative prey abundance at subnivean access points. Use of seed boluses in winter was correlated with prey abundance values obtained by snap trapping after the snow melt (J2 = 0.0134, ¢2 = 0.435). A program for direct entry of raw telemetry data in a Geographic Information System (GIS) data base was developed. With this method telemetry data can be interpreted directly al a number of scales to determine habitat patterns using area rather than point data.

Despite a paucity of experimental evidence, clinical opinion remains that though LTrPs allow pain-free movement, they are primarily associated with deleterious motor effects and occur commonly in 'healthy' muscles. The primary aim of this study was to investigate the effects of LTrPs on the muscle activation patterns (MAPs) of key shoulder girdle muscles during scapular plane elevation of the arm in the unloaded, loaded and fatigued states. In connection with the main aim, a preliminary study was carried out to examine the frequency with which LTrPs occur in the scapular positioning muscles in a group of normal subjects. After establishing intra-examiner reliability for the clinical examination process, 154 healthy subjects volunteered to be screened for normal shoulder girdle function, then undergo a physical examination for LTrPs in the trapezius, rhomboids, levator scapulae, serratus anterior and the pectoralis minor muscles bilaterally. Of these 'healthy' subjects, 89.8% had at least one LTrP in the scapular positioning muscles (mean=10.65 ± 6.8, range=1-27), with serratus anterior and upper trapezius harbouring the most LTrPs on average (2.46 ± 1.8 and 2.36 ± 1.3 respectively). Consistent with clinical opinion, this study found that LTrPs occur commonly in the scapular positioning muscles. To investigate the motor effects of LTrPs, surface electromyography (sEMG) was used to measure the timing of muscle activation of the upper and lower trapezius and serratus anterior (upward scapular rotators), the infraspinatus (rotator cuff) and middle deltoid (arm abductor). These studies found that LTrPs housed in the scapular upward rotator muscles affected the timing of activation and increased the variability of the activation times of this muscle group and were also associated with altered timing of activation in the functionally related but LTrP-free infraspinatus and middle deltoid. Compared with the control group (all muscles LTrP-free), the MAPs of the LTrP group appeared to be sub-optimal, particularly in relation to preserving the subacromial space and the loading of the rotator cuff muscles. After the initial sEMG evaluations, the LTrP subjects were randomly assigned to one of two interventions: superficial dry needling (SDN) followed by post-isometric relaxation (PIR) stretching to remove LTrP s or sham ultrasound, to act as a placebo treatment where LTrPs remained. Where LTrPs were removed, a subsequent sEMG evaluation found MAPs to be similar to the control group in most of the experimental conditions investigated. Of particular note, when LTrPs had been treated and the subjects repeated the fatiguing protocol, the resultant MAP showed no significant difference with that of the control group in the rested state, suggesting treating LTrPs was associated with an improved response to fatigue induced by repetitive overhead movements. In conclusion, the findings indicate that LTrPs commonly occur in scapular positioning muscles and have deleterious effects on MAPs employed to perform scapular plane elevation and thus affect motor control mechanisms. Treating LTrPs with SDN and PIR stretching increases pressure-pain thresholds, removes associated taut bands and at least transiently optimises the MAP during scapular plane elevation. Discussion includes possible neuromuscular pathophysiology that might explain these results.

Cette these a pour theme la geometrie des courbes algebriques et de leur jacobienne (en caracteristique zero). Elle a, en particulier, pour objet l'etude du groupe engendre dans la jacobienne par les points de Weierstrass pour certaines courbes planes lisses de genre trois. Nous determinons ce groupe pour certaines familles de courbes de genre trois. Pour ce faire, nous procedons en deux etapes. Nous utilisons tout d'abord la geometrie de la courbe et de sa jacobienne pour restreindre le groupe cherche. Les restrictions obtenues par ces arguments geometriques s'avereront etre optimales. Pour demontrer cela, nous utilisons differentes techniques: dans la deuxieme partie, nous appliquons une descente explicite via une isogenie; dans la troisieme partie, nous utilisons des arguments de reduction modulo un nombre premier. Lorsque nous nous interessons a des familles, ces restrictions ``d'ordre geometrique'' s'obtiennent pour toute la famille. Par contre, les techniques mises en oeuvre lors de la seconde etape ne nous donnent le resultat que pour une courbe particuliere. Dans chaque cas, un argument de specialisation nous permet de conclure. De plus, nous determinons ce groupe pour la seule quartique, autre que le quartique de Fermat, possedant le nombre minimal de points de Weierstrass, a savoir douze; la encore, la geometrie de la jacobienne intervient dans la determination de ce groupe. Ces calculs nous permettent de donner des estimations sur le rang de ce groupe et sur la partie de torsion dans le cas d'une quartique generique, selon le nombre de points d'hyper-inflexion (c'est-a-dire de points de la courbe ou la tangente a multiplicite d'intersection quatre avec la courbe).

La détermination de la tenue mécanique de points soudés est un enjeu industriel important, eten particulier dans le cas d'aciers à Très Haute Résistance. Actuellement, l'essai de tractionen croix est l'essai mécanique le plus répandu caractérisant la tenue des points soudés.Toutefois, l'étude de l'influence des différentes zones constituant le point soudé sur la tenueest difficile en se basant sur cet essai.Un essai d'enfoncement de coin a été développé afin de caractériser les ruptures interfacialesdes points soudés. Une section transverse d'un point soudé est observée tandis qu'un coin estinséré entre les deux tôles soudés. Une caméra CCD enregistre l'observation de lapropagation de la fissure.Le pliage limité des tôles durant la fissuration rend efficace la classification des différentspoints soudés par l'énergie totale dissipée par unité de surface rompue. De plus, la mesure insitu de l'angle d'ouverture de fissure caractérise la fissuration stable de la zone fondue. Dessimulations par éléments finis de l'essai sont conduites afin d'estimer un lien entre cesmesures et la rupture du matériau, modélisée par des zones cohésives.La rupture interfaciale de points soudés d'aciers DP et TRIP a été étudiée. Les mesuresexpérimentales permettent d'estimer les paramètres de modèles cohésifs représentatifs de larupture de la zone fondue, constituant des données fiables susceptibles d'être utilisées danstoute simulation numérique du comportement du point soudé.

Durant ces dernières années, un grand intérêt a été accordé aux systèmes à antennes multiples à cause de leur capacité à augmenter les débits. Une multitude de codes Espace-Temps existent dans la littérature. Les codes Espace-Temps optimaux sont ceux qui satisfont lespropriétés suivantes: rendement plein, ordre de diversité maximal, gain de codage optimal. Malheureusement, les meilleurs codes existants souffrent de déterminants minimaux s'évanouissant lorsque l'efficacité spectrale augmente. Nous proposons deux nouvelles constructions de codes Espace-Temps ayant un rendement plein, une diversité pleine et des déterminantsminimaux ne s'évanouissant pas lorsque l'efficacité spectrale augmente.Nous nous basons dans nos constructions sur les algèbres cycliques de division de centre Q(i) et Q(j). Les premiers codes construits sont les "codes Quaternioniques". Il s'est avéré que la répartition non uniforme de lénergie dans la matrice mot de code pénalise leurs performances lorsque le nombre d'antennes à l'émission augmente. Pour pallier ce problème énergétique, nous avons construit une nouvelle famille de codes Espace-Temps, appelée "codes Parfaits". Ces derniers ont une efficacité énergétique qui se traduit par une distribution énergétique uniforme au sein du mot de code et des constellations transmises ne présentant aucune perte de forme par rapport aux constellations émises. Les codes Quaternioniques et les codes Parfaits atteignent le compromis gain de multiplexage-diversité optimal. La représentation en réseaux de points des codes Quaternioniques et des codes Parfaits permet leur décodage par les décodeurs de réseaux de points. Les décodeurs les plus connus dans la littérature sont le décodeur par sphères et le Schnorr-Euchner. Ces derniers sont utilisés pour décoder des réseaux de points infinis. Étant donné que nous considérons des constellations finies, des versions modifiées des deux algorithmes ont été proposées. La comparaison des complexités correspondants aux deux versions modifiées de ces décodeurs nous a permis de choisir le meilleur, à savoir, le Schnorr-Euchner. Le décodage des réseaux de points peut être considérablement accéléré en utilisant une réduction de réseaux de points. A ce jour, la réduction n'est appliquée qu'aux réseaux de points infinis . L'utilisation du schéma de codage/décodage en mod-Lamda rend l'application de la réduction possible en considérant des constellations finies. Nos nouvelles constructions de codes Espace-Temps se basent sur des réseaux de points algébriques. Nous proposons dans ce sens une nouvelle réduction algébrique adaptée aux réseaux de points algébriques pour les systèmes mono-antenne sur canal à évanouissements rapides. Cette méthode sera étendue au cas des systèmes à antennes multiples dans un proche avenir.