No description available.

No description available.

Earthquake ground motions are needed to evaluate the seismic performance of new and existing structures and facilities. In seismically active regions the strong ground motion recordings database is usually sufficiently large to physically constrain the earthquake estimation for seismic risk assessment. However, in areas of low seismicity rate, particularly in the Central and Eastern United States, the estimation of strong ground motions for a specified magnitude, distance, and site conditions represents a significant issue. The only available approach for ground motion estimation in this region is numerical simulation. In this study, earthquake ground motions have been generated for the Upper Mississippi Embayment using a numerical wave propagation formulation. The effects of epistemic and aleatory uncertainties in the earthquake source, path, and site processes, the effect of non-linear soil behavior, and the effects of the geometry of the Embayment have been incorporated. The ground motions are intended to better characterize the seismic hazard in the Upper Mississippi Embayment by representing the amplitude and variability that might be observed in real earthquakes and to provide resources to evaluate the seismic risk in the region.

Assessment of potential ground motions in seismically active regions is essential for purposes of seismic design and analysis. Peak ground motion intensity values and frequency content of seismic excitations are required for reliable seismic design, analysis and retrofitting of structures. In regions of sparse or no strong ground motion records, ground motion simulations provide physics-based synthetic records. These simulations provide not only the earthquake engineering parameters but also give insight into the mechanisms of the earthquakes. This thesis presents strong ground motion simulations in three regions of intense seismic activity. Stochastic finite-fault simulation methodology with a dynamic corner frequency approach is applied to three case studies performed in D&uuml / zce, L&rsquo / Aquila and Erzincan regions. In D&uuml / zce study, regional seismic source, propagation and site parameters are determined through validation of the simulations against the records. In L&rsquo / Aquila case study, in addition to study of the regional parameters, the limitations of the method in terms of simulating the directivity effects are also investigated. In Erzincan case study, where there are very few records, the optimum model parameters are determined using a large set of simulations with an error-minimization scheme. Later, a parametric sensitivity study is performed to observe the variations in simulation results to small perturbations in input parameters. Results of this study confirm that stochastic finite-fault simulation method is an effective technique for generating realistic physics-based synthetic records of large earthquakes in near field regions.

Ces travaux s'articulent autour de la détection et de la prévision des phénomènes turbulents dans la couche limite atmosphérique. Nous proposons tout d'abord une méthode stochastique innovante de reconstruction locale de l'atmosphère. Nous utilisons des systèmes de particules pour modéliser l'écoulement atmosphérique et sa variabilité interne. L'apprentissage des paramètres turbulents et la mise à jour des particules se font à l'aide d'observations mesurées par un lidar Doppler 3D. Nous présentons ensuite une nouvelle méthode de descente d'échelle stochastique pour la prévision de la turbulence sous-maille. A partir du modèle en points de grille Meso-NH, nous forçons un système de particules qui évolue à l'intérieur des mailles. Notre méthode de descente d'échelle permet de modéliser des champs sous-maille cohérents avec le modèle en points de grille. Dans un troisième et dernier temps nous introduisons les problèmes de remontée d'échelle. La reconstruction de l'atmosphère modélise la turbulence dans un volume restreint qui couvre au plus quelques mailles des modèles météorologiques en points de grille. L'objectif de la remontée d'échelle est de construire une méthode d'assimilation de l'atmosphère reconstruite. En utilisant l'algorithme de nudging direct et rétrograde, nous explorons les problèmes liés à la taille du domaine observé. Nous proposons finalement un algorithme de nudging avec apprentissage de paramètre, illustré sur un cas simple. / Our work aims to improve the turbulent phenomena detection and forecast in the atmospheric boundary layer. First, we suggest a new stochastic method to reconstruct locally the turbulent atmosphere. Particle systems are used to model the atmospheric flow and its internal variability. To update particles and lean the turbulent parameters 3D Doppler lidar measurements are used. Then, a new stochastic downscaling technic for sub-grid turbulence forecast is presented. From the grid point model Meso-NH, a sub-grid particle system is forced. Here, the particles evolve freely in the simulated domain. Our downscaling method allows to model sub-grid fields coherent with the grid point model. Next, we introduce the upscaling issue. The atmosphere reconstruction covers at best few cells of meteorological grid point models. The issue is to assimilate the reconstructed atmosphere in such models. Using the back and forth nudging algorithm, we explore the problems induced by the size of the observed domain. Finally we suggest a new way to use the back and forth nudging algorithm for parameter identification.

L'objectif de cette recherche est de construire un modèle réduit de petite dimension pour prévoir les réponses dynamiques dans une bande BF sur les parties rigides d'un véhicule automobile complet. Un tel modèle réduit "léger" est une aide à la phase de conception en "Avant Projet" de ces véhicules qui ont la particularité de présenter de nombreux modes élastiques locaux en BF dues à la présence de nombreuses parties flexibles et d'équipements. Pour la construction du modèle réduit, nous avons introduit une base non usuelle de l'espace admissible des déplacements globaux. La construction de cette base requiert la décomposition en sous-domaines du domaine de la structure qui peut présenter une très grande complexité géométrique et dont les modèles EF font intervenir de très nombreux types d'éléments finis. Cette décomposition en sous-domaines a été réalisée par la Fast Marching Method que nous avons due étendre pour pouvoir traiter la complexité des modèles EF des véhicules automobiles. Puis les équations matricielles du modèle EF sont projetées sur cette base. Afin de prendre en compte les incertitudes sur les paramètres du modèle, les incertitudes de modèle induites par les erreurs de modélisation et enfin les incertitudes liées à la non prise en compte des contributions locales dans le modèle réduit des déplacements globaux, un unique modèle probabiliste non paramétrique de ces trois sources d'incertitude a été implémenté sur le modèle réduit construit avec les vecteurs propres globaux. Les paramètres de dispersion de ce modèle probabiliste ont été identifiés en utilisant le principe du maximum de vraisemblance et des réponses obtenues à l'aide d'un modèle stochastique de référence qui inclut des informations expérimentales résultant de travaux précédents. Le modèle réduit stochastique, pour la prévision des déplacements globaux sur les parties rigides dans la bande BF qui a été développé, a été validé sur un modèle de structure automobile "nue" puis a été appliqué avec succès sur un modèle complet de véhicule automobile / The objective of this research is to construct a reduced-order model to predict the dynamical response, in the LF band, of the stiff parts of a complete automotive vehicle in order to facilitate the draft design. The vehicles under consideration have many elastic modes in LF due to the presence of many flexible parts and equipments. To build such a model, we introduced a non-usual basis of the admissible space of global displacements. The construction of this basis requires the decomposition of the domain of the structure. This subdomain decomposition is performed by using the Fast Marching Method that we have extended to take into account the high complexity of the mesh of an automotive vehicle. Then the matrix equations of the FE model are projected on this basis. To take into account the system parameters uncertainties, the model uncertainties induced by the modeling errors and finally, the uncertainties related to the neglecting of local contributions in the reduced-order model, a nonparametric probabilistic model of the three sources of uncertainties has been implemented on the reduced-order model constructed with the global displacements eigenvectors. The dispersion parameters of the probabilistic model are identified using the maximum likelihood method and the responses obtained from a stochastic reference model which includes experimental data resulting from previous works. This stochastic model which has been designed for the prediction of the global displacements of the rigid parts in the LF band is validated on a simple structure of an automotive model and has been successfully applied on a complete model of automotive vehicle

Thesis (MScEng)--Stellenbosch University, 2002. / ENGLISH ABSTRACT: The field of topic spotting in conversational speech deals with the problem of identifying "interesting" conversations or speech extracts contained within large volumes of speech data. Typical applications where the technology can be found include the surveillance and screening of messages before referring to human operators. Closely related methods can also be used for data-mining of multimedia databases, literature searches, language identification, call routing and message prioritisation. The first topic spotting systems used words as the most basic units. However, because of the poor performance of speech recognisers, a large amount of topic-specific hand-transcribed training data is needed. It is for this reason that researchers started concentrating on methods using phonemes instead, because the errors then occur on smaller, and therefore less important, units. Phoneme-based methods consequently make it feasible to use computer generated transcriptions as training data. Building on word-based methods, a number of phoneme-based systems have emerged. The two most promising ones are the Euclidean Nearest Wrong Neighbours (ENWN) algorithm and the newly developed Stochastic Method for the Automatic Recognition of Topics (SMART). Previous experiments on the Oregon Graduate Institute of Science and Technology's Multi-Language Telephone Speech Corpus suggested that SMART yields a large improvement over ENWN which outperformed competing phoneme-based systems in evaluations. However, the small amount of data available for these experiments meant that more rigorous testing was required. In this research, the algorithms were therefore re-implemented to run on the much larger Switchboard Corpus. Subsequently, a substantial improvement of SMART over ENWN was observed, confirming the result that was previously obtained. In addition to this, an investigation was conducted into the improvement of SMART. This resulted in a new counting strategy with a corresponding improvement in performance. / AFRIKAANSE OPSOMMING: Die veld van onderwerp-herkenning in spraak het te doen met die probleem om "interessante" gesprekke of spraaksegmente te identifiseer tussen groot hoeveelhede spraakdata. Die tegnologie word tipies gebruik om gesprekke te verwerk voor dit verwys word na menslike operateurs. Verwante metodes kan ook gebruik word vir die ontginning van data in multimedia databasisse, literatuur-soektogte, taal-herkenning, oproep-kanalisering en boodskap-prioritisering. Die eerste onderwerp-herkenners was woordgebaseerd, maar as gevolg van die swak resultate wat behaal word met spraak-herkenners, is groot hoeveelhede hand-getranskribeerde data nodig om sulke stelsels af te rig. Dit is om hierdie rede dat navorsers tans foneemgebaseerde benaderings verkies, aangesien die foute op kleiner, en dus minder belangrike, eenhede voorkom. Foneemgebaseerde metodes maak dit dus moontlik om rekenaargegenereerde transkripsies as afrigdata te gebruik. Verskeie foneemgebaseerde stelsels het verskyn deur voort te bou op woordgebaseerde metodes. Die twee belowendste stelsels is die "Euclidean Nearest Wrong Neighbours" (ENWN) algoritme en die nuwe "Stochastic Method for the Automatic Recognition of Topics" (SMART). Vorige eksperimente op die "Oregon Graduate Institute of Science and Technology's Multi-Language Telephone Speech Corpus" het daarop gedui dat die SMART algoritme beter vaar as die ENWN-stelsel wat ander foneemgebaseerde algoritmes geklop het. Die feit dat daar te min data beskikbaar was tydens die eksperimente het daarop gedui dat strenger toetse nodig was. Gedurende hierdie navorsing is die algoritmes dus herimplementeer sodat eksperimente op die "Switchboard Corpus" uitgevoer kon word. Daar is vervolgens waargeneem dat SMART aansienlik beter resultate lewer as ENWN en dit het dus die geldigheid van die vorige resultate bevestig. Ter aanvulling hiervan, is 'n ondersoek geloods om SMART te probeer verbeter. Dit het tot 'n nuwe telling-strategie gelei met 'n meegaande verbetering in resultate.

No / This paper presents a new partitioning strategy with the objective of increasing scalability by reducing computational effort of routing in networks. The original network is partitioned into blocks (subnetworks) so that there is a bi-directional link between any two blocks. When there is a connection request between a pair of nodes, if the nodes are in the same block, we only use the small single block to derive routings. Otherwise we combine the two blocks where the two nodes locate and in this way the whole network will never be used. The strategy is generic in that it can be used in any underlying routing algorithms in the network layer and can be applied to any networks with fixed topology such as fixed wired subnetworks of the Internet. The performance of this strategy has been investigated by building a simulator in Java and a comparison with existing stochastic partitioning techniques is shown to give superior performance in terms of trade-off in blocking probability (the probability of failure to find a path between source and destination satisfying QoS constraints) and reduction of computational effort.