Dissertations / Theses on the topic 'High-order approach'

La Màquina de Boltzmann (MB) és una xarxa neuronal estocàstica amb l'habilitat tant d'aprendre com d'extrapolar distribucions de probabilitat. Malgrat això, mai ha arribat a ser tant emprada com d'altres models de xarxa neuronal, com ara el perceptró, degut a la complexitat tan del procés de simulació com d'aprenentatge: les quantitats que es necessiten al llarg del procés d'aprenentatge són normalment estimades mitjançant tècniques Monte Carlo (MC), a través de l'algorisme del Temprat Simulat (SA). Això ha portat a una situació on la MB és més ben aviat considerada o bé com una extensió de la xarxa de Hopfield o bé com una implementació paral·lela del SA. <br/> <br/>Malgrat aquesta relativa manca d'èxit, la comunitat científica de l'àmbit de les xarxes neuronals ha mantingut un cert interès amb el model. Una de les extensions més rellevants a la MB és la Màquina de Boltzmann d'Alt Ordre (HOBM), on els pesos poden connectar més de dues neurones simultàniament. Encara que les capacitats d'aprenentatge d'aquest model han estat analitzades per d'altres autors, no s'ha pogut establir una equivalència formal entre els pesos d'una MB i els pesos d'alt ordre de la HOBM. <br/> <br/>En aquest treball s'analitza l'equivalència entre una MB i una HOBM a través de l'extensió del mètode conegut com a decimació. Decimació és una eina emprada a física estadística que es pot també aplicar a cert tipus de MB, obtenint expressions analítiques per a calcular les correlacions necessàries per a dur a terme el procés d'aprenentatge. Per tant, la decimació evita l'ús del costós algorisme del SA. Malgrat això, en la seva forma original, la decimació podia tan sols ser aplicada a cert tipus de topologies molt poc densament connectades. La extensió que es defineix en aquest treball permet calcular aquests valors independentment de la topologia de la xarxa neuronal; aquest model es basa en afegir prou pesos d'alt ordre a una MB estàndard com per a assegurar que les equacions de la decimació es poden solucionar. <br/> <br/>Després, s'estableix una equivalència directa entre els pesos d'un model d'alt ordre, la distribució de probabilitat que pot aprendre i les matrius de Hadamard: les propietats d'aquestes matrius es poden emprar per a calcular fàcilment els pesos del sistema. Finalment, es defineix una MB estàndard amb una topologia específica que permet entendre millor la equivalència exacta entre unitats ocultes de la MB i els pesos d'alt ordre de la HOBM.<br>La Máquina de Boltzmann (MB) es una red neuronal estocástica con la habilidad de aprender y extrapolar distribuciones de probabilidad. Sin embargo, nunca ha llegado a ser tan popular como otros modelos de redes neuronals como, por ejemplo, el perceptrón. Esto es debido a la complejidad tanto del proceso de simulación como de aprendizaje: las cantidades que se necesitan a lo largo del proceso de aprendizaje se estiman mediante el uso de técnicas Monte Carlo (MC), a través del algoritmo del Temple Simulado (SA). En definitiva, la MB es generalmente considerada o bien una extensión de la red de Hopfield o bien como una implementación paralela del algoritmo del SA. <br/> <br/>Pese a esta relativa falta de éxito, la comunidad científica del ámbito de las redes neuronales ha mantenido un cierto interés en el modelo. Una importante extensión es la Màquina de Boltzmann de Alto Orden (HOBM), en la que los pesos pueden conectar más de dos neuronas a la vez. Pese a que este modelo ha sido analizado en profundidad por otros autores, todavía no se ha descrito una equivalencia formal entre los pesos de una MB i las conexiones de alto orden de una HOBM. <br/> <br/>En este trabajo se ha analizado la equivalencia entre una MB i una HOBM, a través de la extensión del método conocido como decimación. La decimación es una herramienta propia de la física estadística que también puede ser aplicada a ciertos modelos de MB, obteniendo expresiones analíticas para el cálculo de las cantidades necesarias en el algoritmo de aprendizaje. Por lo tanto, la decimación evita el alto coste computacional asociado al al uso del costoso algoritmo del SA. Pese a esto, en su forma original la decimación tan solo podía ser aplicada a ciertas topologías de MB, distinguidas por ser poco densamente conectadas. La extensión definida en este trabajo permite calcular estos valores independientemente de la topología de la red neuronal: este modelo se basa en añadir suficientes pesos de alto orden a una MB estándar como para asegurar que las ecuaciones de decimación pueden solucionarse. <br/> <br/>Más adelante, se establece una equivalencia directa entre los pesos de un modelo de alto orden, la distribución de probabilidad que puede aprender y las matrices tipo Hadamard. Las propiedades de este tipo de matrices se pueden usar para calcular fácilmente los pesos del sistema. Finalmente, se define una BM estándar con una topología específica que permite entender mejor la equivalencia exacta entre neuronas ocultas en la MB y los pesos de alto orden de la HOBM.<br>The Boltzmann Machine (BM) is a stochastic neural network with the ability of both learning and extrapolating probability distributions. However, it has never been as widely used as other neural networks such as the perceptron, due to the complexity of both the learning and recalling algorithms, and to the high computational cost required in the learning process: the quantities that are needed at the learning stage are usually estimated by Monte Carlo (MC) through the Simulated Annealing (SA) algorithm. This has led to a situation where the BM is rather considered as an evolution of the Hopfield Neural Network or as a parallel implementation of the Simulated Annealing algorithm. <br/> <br/>Despite this relative lack of success, the neural network community has continued to progress in the analysis of the dynamics of the model. One remarkable extension is the High Order Boltzmann Machine (HOBM), where weights can connect more than two neurons at a time. Although the learning capabilities of this model have already been discussed by other authors, a formal equivalence between the weights in a standard BM and the high order weights in a HOBM has not yet been established. <br/> <br/>We analyze this latter equivalence between a second order BM and a HOBM by proposing an extension of the method known as decimation. Decimation is a common tool in statistical physics that may be applied to some kind of BMs, that can be used to obtain analytical expressions for the n-unit correlation elements required in the learning process. In this way, decimation avoids using the time consuming Simulated Annealing algorithm. However, as it was first conceived, it could only deal with sparsely connected neural networks. The extension that we define in this thesis allows computing the same quantities irrespective of the topology of the network. This method is based on adding enough high order weights to a standard BM to guarantee that the system can be solved. <br/> <br/>Next, we establish a direct equivalence between the weights of a HOBM model, the probability distribution to be learnt and Hadamard matrices. The properties of these matrices can be used to easily calculate the value of the weights of the system. Finally, we define a standard BM with a very specific topology that helps us better understand the exact equivalence between hidden units in a BM and high order weights in a HOBM.

Development of a novel high-order flux correction method on strand grids is presented. The method uses a combination of flux correction in the unstructured plane and summation-by-parts operators in the strand direction to achieve high-fidelity solutions. Low-order truncation errors are cancelled with accurate flux and solution gradients in the flux correction method, thereby achieving a formal order of accuracy of 3, although higher orders are often obtained, especially for highly viscous flows. In this work, the scheme is extended to high-Reynolds number computations in both two and three dimensions. Turbulence closure is achieved with a robust version of the Spalart-Allmaras turbulence model that accommodates negative values of the turbulence working variable, and the Menter SST turbulence model, which blends the k-ε and k-ω turbulence models for better accuracy. A major advantage of this high-order formulation is the ability to implement traditional finite volume-like limiters to cleanly capture shocked and discontinuous flow. In this work, this approach is explored via a symmetric limited positive (SLIP) limiter. Extensive verification and validation is conducted in two and three dimensions to determine the accuracy and fidelity of the scheme for a number of different cases. Verification studies show that the scheme achieves better than third order accuracy for low and high-Reynolds number flow. Cost studies show that in three-dimensions, the third-order flux correction scheme requires only 30% more walltime than a traditional second-order scheme on strand grids to achieve the same level of convergence. In order to overcome meshing issues at sharp corners and other small-scale features, a unique approach to traditional geometry, coined "asymptotic geometry," is explored. Asymptotic geometry is achieved by filtering out small-scale features in a level set domain through min/max flow. This approach is combined with a curvature based strand shortening strategy in order to qualitatively improve strand grid mesh quality.

The thesis presents the development of an h-multigrid solver for high-order accurate discontinuous Galerkin (DG) discretizations of non-linear systems of conservations laws on unstructured grids. For this purpose, a high-order DG discretization on polyhedral grids is developed first and it is applied to the compressible Navier-Stokes equations. The proposed method employs shape functions, consisting of complete polynomials defined in the real space, which are hierarchic and orthonormal on arbitrarily shaped elements. As regards the discretization of the viscous terms of the Navier-Stokes equations, we use the well-known method introduced in [6] with suitably enlarged values of the stability parameter reported in [1]. The accuracy and the convergence properties of the method have been tested against classical inviscid problems such as the transonic Ringleb flow and the subsonic flow over a Gaussian bump. The Helmholtz problem and the solution of the Navier-Stokes equations around a NACA0012 airfoil have been used as viscous tests. Then we present the development of an h-multigrid method where coarse grid levels are constructed by agglomerating neighbouring elements of the fine grid. First, an elegant yet practical set of transfer operators is derived for general space settings and for the current one. After that, a quasi-implicit multistage h-multigrid iteration strategy for the discontinuous Galerkin discretization of the steady Euler equations is developed and numerically investigated. Results are presented for a subsonic flow over a NACA0012 airfoil at 2o of incidence. These results highlight the main properties of the developed multigrid scheme and its different behaviour with respect to the classic p-multigrid scheme.

High-order methods in Computational Fluid Dynamics (CFD) offer a potential route towards the resolution of hitherto intractable fluid-dynamics problems in industry. The Flux Reconstruction (FR) approach provides a unifying framework for a number of popular high-order methods such as the Discontinuous Galerkin (DG). Its suitability for use on unstructured grids along with its ability to facilitate massively parallelised implementation on architectures such as GPUs provide a means to tackle computationally challenging flows around complex geometries. Such a flow can be found in the rod-aerofoil tandem configuration: Complex, unsteady flow structures generated by and interacting with more than a single solid body are central to a number of applications in the aerospace industry. The current thesis attempts to demonstrate the suitability of the FR approach in successfully simulating the flow around a rod-aerofoil configuration. The in-house CFD solver employed in the research is presented and the FR implementation analysed. Computational grid resolution issues arising from the rod-aerofoil problem are studied and a novel strategy for the stabilisation of the computation is implemented in the form of local entropy stability. The results obtained are analysed and conclusions are drawn on the utility of the FR approach in the absence of a sub-grid scale model (Implicit LES - under-resolved DNS). The present work confirms the utility of local entropy stability for the stabilisation of the rod-aerofoil simulation of aerofoil-chord based Reynolds number of Re=480, 000. It will also demonstrate that the under-resolved DNS setup that resulted in a computational cost of approximately six hours for a single flow pass over the aerofoil chord on 200 Nvidia P100 GPUs resulted in moderate success for a significant portion of the flow dynamics, which not adequately predicted when compared with experiment. The latter led to a series of useful conclusions. The core of the conclusions involved the apparent over-prediction of time-averaged velocity and momentum deficits across wakes and as well as over-prediction of turbulent intensities. An identification of the problematic areas is therefore given and potential alleviation techniques outlined.

a b s t r a c t In this paper, a hybrid method using active control and a High Order Differentiator (HOD) methodology is proposed to synchronize chaotic systems. Compared to some traditional active control methods, this new method can synchronize chaotic systems where only output states of the master system are available, i.e. the system is considered a black box. The HOD is used to estimate the derivative information of the master system followed by an active control methodology relying on HOD information. The Qi hyperchaotic system is used to verify the performance of this hybrid method. The proposed method is also compared to some traditional methods. Experimental results show that the proposed method has high synchronization precision and speed and is robust against uncertainties in the master system. The circus implements of the proposed synchronizing scheme are included in this paper. The simulation results show the feasibility of the proposed scheme.

Magnetic Resonance Imaging (MRI) scanners are becoming increasingly popular with many clinical experts for use in both medical research and clinical imaging of patients, due to their ability to perform high-resolution non-intrusive imaging examinations. Recently, however, there has been an increasing demand for higher resolution scanners that are capable of performing quicker scans with increased patient comfort. With this demand for more advanced MRI systems, there also follows a number of challenges facing designers. Understanding the physical phenomena behind MRI is crucial in the development of scanners that are capable of producing accurate images of the patient with maximum comfort and minimal noise signatures. MRI scanners utilise strong static magnetic fields coupled with rapidly time varying gradient magnetic fields to generate images of the patient. In the presence of these time varying fields, the conducting components of MRI scanners generate eddy currents, which give rise to Lorentz forces and cause the conductors to vibrate. These vibrations cause acoustic waves to form that propagate through the air and result in audible noise which can cause significant discomfort for the patient. They also generate Lorentz currents which feedback into the electromagnetic field and this process results in a fully coupled non-linear acousto-magneto-mechanical system. The determination of the coupling mechanisms involved in such a system is a nontrivial task and so, in order to understand the behaviour of MRI systems during operation, advanced computational tools and techniques are required. Moreover, there exists certain small scale physical phenomena that arise in the coupled system which require high resolutions to obtain accurate results. In this thesis, a new computational framework for the treatment of acoustomagneto-mechanical coupling that arises in low-frequency electro-magneto-mechanical systems, such as MRI scanners, is proposed. The transient Newton-Raphson strategy involves the solution of a monolithic system, obtained from the linearisation of the coupled system of equations and two approaches are considered: (i) the linearised approach and (ii) the non-linear approach. In (i), physically motivated by the excitation from static and time varying current sources of MRI scanners, the fields may be split into a dominant static component and a much smaller dynamic component. The resulting linearised system is obtained by performing the linearisation of the fields about this dominant static component. This approach permits solutions in the frequency domain, for understanding the response of MRI systems under various excitations, and provides a computationally efficient way to solve this challenging problem, as it allows the tangent stiffness matrix to be inverted independently of time or frequency. In (ii), there is no approximation from a physical standpoint and the linearization is performed about the current solution. This approach requires that solutions are obtained in the time domain and thus the focus is then put on transient solutions to the coupled system of equations to address the following two important questions: 1) How good is the agreement between the computationally efficient linearised approach compared with the intensive non-linear approach?; and 2) Over what range of MRI operating conditions can the linearised approach be expected to provide acceptable results for MRI scanner design? Motivated by the need to solve industrial problems rapidly, solutions will be restricted to problems consisting of axisymmetric geometries and current sources. This treatment also discusses, in detail, the computational requirements for the solution of these coupled problems on unbounded domains and the accurate discretisation of the fields using hp-finite elements. A set of academic and industrially relevant examples are studied to benchmark and illustrate both approaches, in a hp- finite element context, as well as performing rigorous comparisons between the approaches.

This work deals with high-order numerical methods for unsteady flows around complex geometries. In order to cope with the low-order industrial Finite Volume Method, the proposed technique consists in computing on structured and unstructured zones with their associated schemes: this is called a hybrid approach. Structured and unstructured meshes are then coupled by a nonconforming grid interface. The latter is analyzed in details with special focus on unsteady flows. It is shown that a dedicated treatment at the interface avoids the reflection of spurious waves. Moreover, this hybrid approach is validated on several academic test cases for both convective and diffusive fluxes. The extension of this hybrid approach to high-order schemes is limited by the efficiency of unstructured high-order schemes in terms of computational time. This is why a new approach is explored: The Spectral Difference Method. A new framework is especially developed to perform the spectral analysis of Spectral Discontinuous Methods. The Spectral Difference Method seems to be a viable alternative in terms of computational time and number of points per wavelength needed for a given application to capture the flow physics.

Les effets d’extrémité jouent un rôle important dans la modélisation et la commande de la Machine Linéaire à Induction (MLI). Ces phénomènes augmentent significativement la non-linéarité du modèle de la machine et génèrent plusieurs difficultés pour contrôler et observer ses états avec de bonnes performances. Cette thèse aborde trois problématiques distinctes : la commande robuste de la MLI, l’estimation de la vitesse et du flux de la MLI et le contrôle robuste à base d’observateur en utilisant la théorie du mode glissant d’ordre supérieur.Dans la première partie de la thèse, trois contrôleurs robustes assurant la poursuite de trajectoire de la vitesse et du flux pour la MIL ont été développés : le Super Twisting (ST), le Super Twisting Adaptatif (STA) et le Twisting Adaptatif (TA). Ces commandes ont été testées en simulations et leurs performances ont été démontrées. Ainsi, le ST assure un contrôle continu avec convergence à temps fini de l’erreur à zéro malgré les perturbations, sous l’hypothèse que les bornes des incertitudes sont connues. Cette hypothèse est relaxée dans le cas du TA et du STA grâce à leurs propriétés adaptatives.Dans la deuxième partie de la thèse, un nouveau modèle du MLI a été proposé et son observabilité a été démontrée. Ensuite un Observateur par Mode Glissant d’Ordre Deux (MGOD) et un Observateur par Mode Glissant d’Ordre Supérieur (MGOS) ont été synthétisé afin d’estimer la vitesse et le flux du MLI, uniquement en utilisant la mesure des tensions et des courants statorique. La stabilité des deux observateurs a été prouvée par une approche de Lyapunov et leurs performances ont été démontrées à travers des simulations.Dans la dernière partie de la thèse, deux commandes par rejet actif des perturbations sont synthétisées. Ainsi et dans un premier temps, le modèle de la MLI est décomposé en deux sous-systèmes du second ordre. Ensuite, deux contrôleurs (le twisting et le super-twsiting) ont été synthétisés afin d’assurer la poursuite du flux et de la vitesse. Le MGOS est utilisé pour estimer les dérivées du flux et de la vitesse, ainsi que pour l’estimation en temps réel de la perturbation. Les contrôleurs quant à eux assurent la compensation des perturbations et la poursuite des trajectoires du flux et de la vitesse. La stabilité et la convergence des deux commandes proposées ont été prouvées et leurs performances démontrées par simulation<br>Dynamic end effects play an important role in the Linear Induction Machine (LIM) control. They increase significantly the nonlinearity of the machine model and generate several difficulties to control and observe states with good performances. This thesis addresses three distinctissues: LIM robust control, LIM speed and flux estimation and observer-based robust control using higher order sliding mode theory.In the first part, to achieve speed and flux tracking,Super Twisting Controller (STC), Adaptive Super Twisting Controller (ASTC), and Adaptive Twisting Controller (ATC) were proposed and implemented into LIM system with great performance, i.e. finite time convergence and robustness properties. Among them, STC ensures continuous control with finite time convergence of the error to zero despite disturbances, under the assumption that their bounds are known. ATC and ASTC can deal with unknown bounded disturbance thanks to their adaptive properties.In the second part, a novel simplified LIM model was proposed and its observability has been proved. Then, Second Order Sliding Mode Observer (SOSMO) and Adaptive High Order Sliding Mode Observer (HOSMO) were proposed to estimate LIM speed, only by using the measured stator voltages and stator currents. SOSMO observer is based on the super twisting algorithm and its stability has been proved with Lyapunov’s theory, which can guarantee finite time convergence with less chattering. Adaptive HOSMO strategy combines speed adaptive algorithm and HOSMO method together to estimate rotor fluxes and speed simultaneously.In the third part, the LIM is viewed as two second order subsystems. Moreover, only the speed and the flux are supposed to be measured. Based on that two differentcontrollers based on HOSMO were presented in order to achieve flux and speed tracking. In both controllers, the idea of active disturbance rejection control is applied. Hence, the HOSMO is used to estimate the derivatives of the flux and the speed, as well as the disturbance. Then, in order to deal with the uncertainty in the measured variables, two different SM controllers are proposed. Firstly, the TC is applied in the LIM. However, the control signal in this case is discontinuous. Then, in order to provide a continuous control signal, the TC is replaced with STC. The stability and convergence of proposed TC-HOSMO and STC-HOSMO approaches were given and simulation validated their performances

Classically, researchers in Computational Physics and specifically in Computational Electromagnetics have sought to find numerical solutions to complex physical problems. Several techniques have been developed to accomplish such tasks, each of which having advantages over their counterparts. Typically, each solution method has been developed separately despite having numerous commonalities with other methods. This fact motivates a unified software tool to house each solution method to avoid duplicating previous efforts. Subsequently, these solution methods can be used alone or in conjunction with one another in a straightforward manner. The aforementioned goals can be accomplished by using an Object Oriented software approach. Thus, the goal of the presented research was to incorporate a specific solution technique, an Integral Equation Nystrm method, into a general, Object Oriented software framework.

The purpose of this work is the development of a numerical tool devoted to the study of the flow field in the components of aerospace propulsion systems. The goal is to obtain a code which can efficiently deal with both steady and unsteady problems, even in the presence of complex geometries. Several physical models have been implemented and tested, starting from Euler equations up to a three equations RANS model. Numerical results have been compared with experimental data for several real life applications in order to understand the range of applicability of the code. Performance optimization has been considered with particular care thanks to the participation to two international Workshops in which the results were compared with other groups from all over the world. As far as the numerical aspect is concerned, state-of-art algorithms have been implemented in order to make the tool competitive with respect to existing softwares. The features of the chosen discretization have been exploited to develop adaptive algorithms (p, h and hp adaptivity) which can automatically refine the discretization. Furthermore, two new algorithms have been developed during the research activity. In particular, a new technique (Feedback filtering [1]) for shock capturing in the framework of Discontinuous Galerkin methods has been introduced. It is based on an adaptive filter and can be efficiently used with explicit time integration schemes. Furthermore, a new method (Enhance Stability Recovery [2]) for the computation of diffusive fluxes in Discontinuous Galerkin discretizations has been developed. It derives from the original recovery approach proposed by van Leer and Nomura [3] in 2005 but it uses a different recovery basis and a different approach for the imposition of Dirichlet boundary conditions. The performed numerical comparisons showed that the ESR method has a larger stability limit in explicit time integration with respect to other existing methods (BR2 [4] and original recovery [3]). In conclusion, several well known test cases were studied in order to evaluate the behavior of the implemented physical models and the performance of the developed numerical schemes.

Mestrado em Econometria Aplicada e Previsão<br>This dissertation proposes a new concept: the usage of Multivariate Markov Chains (MMC) as covariates. Our innovative approach is based on the observation that we can treat possible categorical regressors as a MMC in order to improve the forecast error of a certain dependent variable,provided it is caused, in the Granger sense, by the MMC. We conduct a Monte Carlo simulation study to assess the performance of our model and we archive excellent results in terms of forecast. An empirical illustration, that widely supports the results obtained in the Monte Carlo study, is also provided. Furthermore, the results of our empirical illustration suggest that the sovereign bond markets in peripherical European countries, namely Portugal, are ine cient. The conclusions drawn include implications for policy. We also discuss the ideas behind several methods to estimate MMC, tackling issues with regard to the statistical inference topic. We provide a general framework to allow us to obtain the MMC h-step-ahead forecast closed formulas.<br>Esta dissertação propõe um novo conceito: a utilização de Cadeias de Markov Multivariadas enquanto regressores. A nossa abordagem inovadora baseia-se na observação de que é possível fazer uso de CMM enquanto variáveis explicativas com o intuito de se reduzirem os erros de previsão de uma determinada variável dependente, desde que essa variável dependente seja causada, a la Granger, pela CMM. Com o objectivo de perceber a performance do nosso modelo em termos de previsão operacionalizamos um estudo de simulação de Monte Carlo no qual obtemos excelentes resultados. Também recorremos a uma ilustração empírica que sustenta fortemente os resultados obtidos no estudo de simulação de Monte Carlo. Para além disso, os resultados da ilustração empírica apontam para a circunstância de que os mercados das obrigações das dívidas soberanas dos países da periferia europeia, nomeadamente Portugal, são ine cientes. Podem retirar-se das conclusões obtidas algumas implicações em termos de orientação de política económica. Discutimos ainda algumas ideias subjacentes às diversas metodologias de estimação de CMM, sublinhando as questões relativas ao tópico da inferência estatística. Providenciamos uma utensilagem teórica do seio da qual se obtêm as expressões da previsão a h-passos com CMM.

La simulation aux grandes échelles est devenue un outil d’analyse incontournable pour l’étude des écoulements turbulents dans des géométries complexes. Cependant, à cause de l’augmentation constante des ressources de calcul, le traitement des grandes quantités de données générées par les simulations hautement résolues est devenu un véritable défi qu’il n’est plus possible de relever avec des outils traditionnels. En mécanique des fluides numérique, cette problématique émergente soulève les mêmes questions que celles communément rencontrées en informatique avec des données massives. A ce sujet, certaines méthodes ont déjà été développées telles que le partitionnement et l’ordonnancement des données ou bien encore le traitement en parallèle mais restent insuffisantes pour les simulations numériques modernes. Ainsi, l’objectif de cette thèse est de proposer de nouveaux formalismes permettant de contourner le problème de volume de données en vue des futurs calculs exaflopiques que l’informatique devrait atteindre en 2020. A cette fin, une méthode massivement parallèle de co-traitement, adaptée au formalisme non-structuré, a été développée afin d’extraire les grandes structures des écoulements turbulents. Son principe consiste à introduire une série de grilles de plus en plus grossières réduisant ainsi la quantité de données à traiter tout en gardant intactes les structures cohérentes d’intérêt. Les données sont transférées d’une grille à une autre grâce à l’utilisation de filtres et de méthodes d’interpolation d’ordre élevé. L’efficacité de cette méthodologie a pu être démontrée en appliquant des techniques de décomposition modale lors de la simulation 3D d’une pale de turbine turbulente sur une grille de plusieurs milliards d’éléments. En outre, cette capacité à pouvoir gérer plusieurs niveaux de grilles au sein d’une simulation a été utilisée par la suite pour la mise en place de calculs basés sur une stratégie multi-niveaux. L’objectif de cette méthode est d’évaluer au cours du calcul les erreurs numériques et celles liées à la modélisation en simulant simultanément la même configuration pour deux résolutions différentes. Cette estimation de l’erreur est précieuse car elle permet de générer des grilles optimisées à travers la construction d’une mesure objective de la qualité des grilles. Ainsi, cette méthodologie de multi-résolution tente de limiter le coût de calcul de la simulation en minimisant les erreurs de modélisation en sous-maille, et a été appliquée avec succès à la simulation d’un écoulement turbulent autour d’un cylindre<br>Large-Eddy Simulation (LES) has become a major tool for the analysis of highly turbulent flows in complex geometries. However, due to the steadily increase of computational resources, the amount of data generated by well-resolved numerical simulations is such that it has become very challenging to manage them with traditional data processing tools. In Computational Fluid Dynamics (CFD), this emerging problematic leads to the same "Big Data" challenges as in the computer science field. Some techniques have already been developed such as data partitioning and ordering or parallel processing but still remain insufficient for modern numerical simulations. Hence, the objective of this work is to propose new processing formalisms to circumvent the data volume issue for the future 2020 exa-scale computing objectives. To this aim, a massively parallel co-processing method, suited for complex geometries, was developed in order to extract large-scale features in turbulent flows. The principle of the method is to introduce a series of coarser nested grids to reduce the amount of data while keeping the large scales of interest. Data is transferred from one grid level to another using high-order filters and accurate interpolation techniques. This method enabled to apply modal decomposition techniques to a billion-cell LES of a 3D turbulent turbine blade, thus demonstrating its effectiveness. The capability of performing calculations on several embedded grid levels was then used to devise the multi-resolution LES (MR-LES). The aim of the method is to evaluate the modeling and numerical errors during an LES by conducting the same simulation on two different mesh resolutions, simultaneously. This error estimation is highly valuable as it allows to generate optimal grids through the building of an objective grid quality measure. MR-LES intents to limit the computational cost of the simulation while minimizing the sub-grid scale modeling errors. This novel framework was applied successfully to the simulation of a turbulent flow around a 3D cylinder

The transcription in the eukaryotic cells involves epigenetic regulatory mechanisms that control local and higher-order chromatin remodelling. In the skin, keratinocyte-specific genes are organized into distinct loci including Epidermal Differentiation Complex (EDC) and Keratin type I/II loci. This thesis introduces bioinformatics approaches to analyze multi-level regulatory mechanisms that control skin development and keratinocyte-specific differentiation. Firstly, integration of gene expression data with analyses of linear genome organization showed dramatic downregulation of the genes that comprise large genomic domains in the sweat glands including EDC locus, compared to ii hair follicles, suggesting substantial differences in global genome rearrangement during development of these two distinct skin appendages. Secondly, comparative analysis of the genetic programmes regulated in keratinocytes by Lhx2 transcription factor and chromatin remodeler Satb1 revealed that significant number of their target genes is clustered in the genome. Furthermore, it was shown in this study that Satb1 target genes are lineage-specific. Thirdly, analysis of the topological interactomes of Loricrin and Keratin 5 in hair follicle steam cells revealed presence of the cis- and trans-interactions and lineage specific genes (Wnt, TGF-beta/activin, Notch, etc.). Expression levels of the genes that comprise interactomes show correlation with their histone modification status. This study demonstrates the crucial role for integration of transcription factormediated and epigenetic regulatory mechanisms in establishing a proper balance of gene expression in keratinocytes during development and differentiation into distinct cell lineages and provides an integrated bioinformatics platform for further analyses of the changes in global organization of keratinocyte-specific genomic loci in normal and diseased skin.

Ces travaux de recherche ont pour but d’évaluer la méthode dite de la "Simulation auxEchelles Adaptées" (SAS pour Scale-Adaptive Simulation). Cette approche coïncide avec uneapproche RANS classique dans les zones pariétales attachées et adapte le niveau de viscositéturbulente dans les zones décollées pour y permettre une résolution partielle des structures turbulentes.Dans une première partie, une analyse théorique du modèle SAS original a été menéeet a permis de développer une correction visant à favoriser l’adaptation du niveau de viscositéturbulente dans les zones sièges d’instabilités de type Kelvin-Helmholtz. Le modèle ainsi corrigéest nommé SAS-αL. Les modèles SAS et SAS-αL ont été implantés dans le code de calculNavier-Stokes elsA de l’ONERA. À l’issue de cette étape, trois cas académiques d’écoulementsturbulents instationnaires, cylindre à grand nombre de Reynolds, marche descendante et cavitétranssonique, ont été simulés grâce aux trois modèles de turbulence SST, SAS et SAS-αL. Outreune comparaison aux bases de données expérimentales disponibles, une attention particulièrea été portée à l’influence de paramètres numériques tels que des schémas numériques d’ordreélevé. Enfin, afin d’étudier la viabilité de l’approche SAS dans un contexte industriel, les troismodèles de turbulence ont été testés sur une configuration issue de l’industrie aéronautique etcorrespondant à la sortie d’air chaud d’un système de dégivrage des nacelles d’avion. La comparaisondes prévisions obtenues avec les modèles SST, SAS et SAS-αL aux données expérimentalesobtenues à l’ONERA a permis de montrer un gain de précision grâce à l’emploi de l’approcheSAS et ce pour un coût de calcul compatible avec un cycle de conception industrielle<br>This research work is meant to assess an upgraded URANS approach, namely the Scale-Adaptive Simulation (SAS). This method is similar to a conventional RANS approach (namelythe SSTmodel) in attached areas and is able to adapt the eddy-viscosity level in detached areas toensure the resolution, at least partially, of the turbulent structures. In a first part of this researchwork, an improvement of the SAS approach is suggestedto allowa better sensitivity of themodelto instabilities such as Kelvin-Helmholtz ones. This "improved" model is referred to as SAS-αLmodel. Both SAS and SAS-αL models were implemented in the ONERA Navier-Stokes solverelsA and both of themaswell as the SSTmodelwere tested on academic test cases : a cylinder in acrossflowat a high Reynolds number, a backward-facing step flowcorresponding to theDriver&amp;Seegmiller experiment and the transonic flow over the M219 cavity experimentally investigatedby de Henshaw. The influence of the numerical parameters was deeply investigated and particularattention was paid to the high-order space-discretization schemes effects. The reliabilityof the SAS approach in an industrial framework was assessed on an aeronautic configurationnamely a nacelle de-icing device. Comparisons between the threemodels (SST, SAS and SAS-αL)and an experimental database available at ONERA - The French Aerospace Lab have shown thebetter accuracy of the SAS approach as well as the high potential of the SAS-αL model

L'analyse des processus non-linéaires de génération et de propagation de la houle a constitué le cadre de cette thèse. Afin d'améliorer la compréhension de ces phénomènes, une méthode numérique dite High-Order Spectral (HOS), résolvant le problème de façon non-linéaire, a été développée. Cette méthode, avec une formulation surfacique et résolue de manière spectrale, associe précision et efficacité. <br /><br />Un traitement original de la génération de houle non-linéaire est proposé. Il permet l'accès à des simulations de champs de vagues tridimensionnels complexes, fortement cambrés, dans un bassin de houle. Diverses comparaisons avec des expériences menées dans le bassin du Laboratoire de Mécanique des Fluides de l'ECN sont présentées.<br /><br />Des simulations océaniques, en milieu ouvert, sont également proposées. Un intérêt particulier est porté à l'étude de l'apparition des vagues scélérates au sein de l'océan. L'importance des effets non-linéaires est pointée ainsi que l'aptitude de la méthode à modéliser de tels phénomènes. Des comparaisons avec les méthodes classiquement employées dans ce genre de problématique indiquent l'intérêt de l'approche utilisée ici.<br /><br />La résolution du problème de tenue à la mer est également envisagée. L'utilisation de la méthode HOS dans les codes couplés, développés au Laboratoire de Mécanique des Fluides (potentiel, RANS, SPH), est envisagée. Elle permettra la description précise de la houle incidente ; le couplage est mis en place et validé sur un certain nombre de cas d'application.

Le bruit de cavité est un phénomène très fréquent dans le domaine des transports aériens.Il survient notamment lors de l'approche à l'atterrissage, où des interactions entre la cellule de l'aéronef et l'écoulement sont à l'origine de fortes émissions tonales. Il devient dès lors une source de pollution acoustique non-négligeable pour les populations résidant à proximité de zones aéroportuaires. Les études numériques et expérimentales décrites jusqu'à présent dans la littérature abordent essentiellement le cas des cavités rectangulaires. Pourtant, les cavités rencontrées en pratique dans l'industrie aéronautique impliquent des géométries souvent plus complexes. Lorsque ces cavités sont soumises à une excitation de nature aérodynamique, leur spécificité géométrique conduit le plus souvent à des réponses acoustiques assez éloignées des estimations issues de modèles académiques construits sur l'observation de cavités rectangulaires. Quelques travaux seulement abordent le cas des cavités cylindriques.Ce travail est consacré à l'étude aéroacoustique des cavités cylindriques, à l'initiative d'Airbus. Il s'inscrit dans le cadre du projet AEROCAV soutenu par la Fondation de Recherche pour l'Aéronautique & l'espace (FRAE). Son objectif est de déterminer les mécanismes impliqués dans les émissions acoustiques intenses et tonales pour les configurations étudiées.Une première partie présente les résultats expérimentaux issus des campagnes de mesures menées dans la soufflerie anéchoïque du Centre Acoustique du LMFA et de l'école Centrale de Lyon. Un modèle semi-empirique, reposant sur l'hypothèse d'une résonance acoustique pilotée par les instabilités présentes dans la couche de cisaillement à l'ouverture de la cavité,est construit à partir du modèle d'Elder (1978). Le modèle permet d'estimer les fréquences susceptibles de dominer l'acoustique rayonnée en champ lointain à partir de la donnée du champ moyen de vitesse longitudinale, que l'on mesure dans le plan de l'écoulement par Vélocimétrie par Imagerie des Particules (PIV).Une seconde partie est destinée au calcul direct du bruit rayonné par un écoulement laminaire ou turbulent affleurant une cavité cylindrique de référence. Il consiste à calculer le champ acoustique directement à partir de la résolution des équations tridimensionnelles de la mécanique des fluides. Le solver Alesia est présenté dans une version modifiée et adaptée à la mise en oeuvre d'une approche multidomaine d'ordre élevé faisant intervenir plusieurs maillages se recouvrant. Des techniques d'interpolation sont spécifiquement développées en vue d'assurer une communication bidirectionnelle entre les différents maillages, malgré des contraintes géométriques fortes. Un modèle d'excitation de l'écoulement est aussi développé afin de disposer de fluctuations dans l'écoulement incident, pour le cas turbulent. Ces deux points font l'originalité des calculs réalisés.Les simulations, menées sur une cavité de rapport d'aspect géométrique égal à 1 et soumise à un écoulement incident à Mach 0.2, montrent que le rayonnement acoustique peut être fidèlement reproduit numériquement. La couche de cisaillement est caractérisée par la présence de deux larges structures tourbillonnaires s'amplifiant lors de leur convection. Leur présence s'accompagne de fortes fluctuations de vitesse à l'origine d'un débit aérodynamique de fluide à l'ouverture qui excite la cavité acoustiquement. Une résonance forcée s'établit dans celle-ci, excitant la couche de mélange au voisinage du point de séparation. Ce couplage auto-entretenu est à l'origine du rayonnement acoustique intense et fortement tonal de la cavité. Il s'établit à une fréquence proche de la fréquence prédite par le modèle semi-empirique développé.

Introducción. El objetivo del presente estudio es determinar si existe correlación entre los conocimientos sobre las consecuencias de la obesidad y el grado de actividad física de las personas. Métodos. Se realizó un estudio transversal analítico durante los años 2013 y 2014. Participaron 215 alumnos de pregrado seleccionados por conveniencia no relacionados a carreras del campo de la salud en una universidad de Lima, Perú. Se evaluó el grado de actividad física utilizando el International Physical Activity Questionnaire (IPAQ) y el nivel de conocimientos sobre consecuencias de la obesidad utilizando la escala Obesity Risk Knowledge-10 (ORK-10). También, se consignó las fuentes de información de donde obtuvieron el conocimiento para responder dicho cuestionario. Resultados. La mediana de edad fue 20 (rango intercuartílico=4) y 63% eran mujeres. De acuerdo al IPAQ, 53,9% realizaban actividad física alta, 35,4%, moderada y 10,7%, leve. Se encontró una correlación muy baja (rs=0,06) entre el puntaje del ORK-10 y la cantidad de equivalentes metabólicos/minuto consumidos por semana, pero no era significativa (p=0,38). Las personas informadas por medios de comunicación y por personal de salud obtuvieron mayores puntajes en el ORK- 10 que quienes se informaron por otras vías (p<0,05). Conclusiones. La correlación entre los conocimientos sobre consecuencias de la obesidad y el grado de actividad física es muy baja. Es necesario utilizar enfoques multidisciplinarios que incluyan todos los determinantes de la realización de actividad física para poder lograr cambios en la conducta de la población.

碩士<br>朝陽科技大學<br>資訊管理系碩士班<br>101<br>There are many uncertainty problems in the Human society, such as the forecasting of economic growth rate, financial crisis, etc. Since Song and Chissom proposed the concept of fuzzy time series in 1993, many scholars have proposed different models to deal with these problems. However, previous studies usually did not consider the transfer original data to the fuzzy linguistic value by the subjective opinions in fuzzy process, which cannot objectively show the characteristics of the data. Based on above concepts, the purpose of this study is to explore ways of determining the objective lengths of intervals and amount of linguistic in fuzzy time series. This study proposed a high-order weighted fuzzy time series model based on variable length discretization approach (VLDA) and N-th quantile discretization approach (NQDA) to make forecasts. In order to verify the proposed method, the Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) from the Taiwan Stock Exchange Corporation are used in the experiment, and the experiment results are compared with other methods in with this study. The forecasting performance shows that the proposed method having better forecasting ability. An intelligent decision support system (DSS) for stock market will be developed in this study. It is supposed to be a useful decision support tools for the investor to make better trading strategies in the future stock market.

碩士<br>中原大學<br>電機工程研究所<br>92<br>The voltage-mode high-order universal filter structure using weighting signal approach has been studied in this thesis. A fundamental high-order multi-function (low-pass, band-pass, and high-pass) filter structure is realized first. Different order output signals can be obtained from different nodes in the fundamental filter structure. Add the weighting of the voltage input signal (i.e., a transformed current signal) to each node in the fundamental structure and then the proposed voltage-mode high-order universal filter structure is obtained, which employs 2n+2 single-input active element and n grounded capacitors. The second-generation current controlled conveyors (CCCIIs) with electronically adjustable inner resistance are used in the design. Using Nullator-and-Norator equivalent transformation, the CCCII-based filter structure can be transformed into the Operational Transconductance Amplifiers-based filter structure. Note that the transconductance g of an OTA is also tunable by the bias current. H-Spice simulations are used for carrying out the comparison between CCCII-based and OTA-based filter structures in terms of filtering performance, noise, and sensitivity analysis. Simulation results validate the theory predictions.

碩士<br>中原大學<br>電機工程研究所<br>92<br>The <a href="http://www.ntsearch.com/search.php?q=design&v=56">design</a> of active filters by using current conveyors has received much attention in recent several decades. In this study, the voltag-mode high-order universal filter structure using the second generation current controlled conveyors and capacitors (CCCIIs-C) has been obtained by using weighting approach . First, we apply the block building approach to <a href="http://www.ntsearch.com/search.php?q=design&v=56">design</a> a fundamental first-order CCCII-C filter circuit. The extension to nth-order filters leads to realize an nth-order voltage-mode universal filter structure which can simultaneously realize low-pass, band-pass, high-pass, band-reject , and allpass filtering signals from the same configuration. Note that any component matchings are not needed. The proposed nth-order voltage-mode filter structure has the least number, 2n+2, of active elements, only n grounded capacitors and no resistors. Finally, H-Spice simulation results of the proposed 3rd-order filter validate the theory predictions .

碩士<br>中原大學<br>電機工程研究所<br>92<br>Active filter design by using current conveyor has received much attention in recent several decades. In this study, the high order voltage-mode universal filter structure using Second generation Current Controlled Conveyors (CCCIIs) is presented. The advantage of the CCCII with electronically tunable intrinsic resistance at terminal X leads to a synthesized network without any resistors which can simultaneously. In the realize lowpass, bandpass, highpass, bandreject, and allpass signals from the same configuration. Only 2n+4 single-input CCCIIs and n grounded capacitors are needed to construct such a high-order voltage-mode universal filter. First, we synthesize a voltage-mode high-order lowpass, bandpass and highpass filter structure by using n+2 single-input active elements and n grounded capacitors. As a matter of fact, n+1 different order output signals can be taken out from n+1 different nodes in the synthesized network. Therefore, the linear combination of all the n+1 different order output signals leads to an arbitrary filter structure which can realize any kinds of filtering signals. Finally, HSPICE simulations with UMC05 level 49 parameters agree with theory very well.

碩士<br>中原大學<br>電機工程研究所<br>92<br>Abstract The voltage-mode high-order universal filter structure using linear combination approach has been studied ing this thesis [7-20]. A fundamental high-order multi-function ( lowpass , bandpass, and highpass ) filter structure is realized first. Different order output signals can be obtained from different nodes in the fundarnental filter structure. Therefore, the linear combination of the output signals of the fundamental filter structure maker any kinds of filtering functions. The synthesize voltage-mode high-order universal filter structure employs 2n+4 single-input active elements and n grounded capacitors. The second-generation current controlled conveyors (CCCII) with electronically tunable inner resistance and used in the <a href="http://www.ntsearch.com/search.php?q=design&v=56">design</a>. Using nullator-and norator equivalent transformationk, the CCCII-based filter structure can be transformed into the operational Transconductance Amplifier based filter structure. Note that the transconductance of an OTA is also adjustable by the bias current. H-Spice simulations are used for carrying out the comparison between CCCII-absed and OTA-based filter structures in terms of filtering performance, noise, and sensitivity analysis. Simulation results validate the theory predictons

In this thesis, we first propose a new scalar auxiliary variable (SAV) approach for general dissipative nonlinear systems. This new approach is half computational cost of the original SAV approach, can be extended to high order unconditionally energy stable backward differentiation formula (BDF) schemes and not restricted to the gradient flow structure. Rigorous error estimates for this new SAV approach are conducted for the Allen-Cahn and Cahn-Hilliard type equations from the BDF1 to the BDF5 schemes in a unified form. As an application of this new approach, we construct high order unconditionally stable, fully discrete schemes for the incompressible Navier-Stokes equation with periodic boundary condition. The corresponding error estimates for the fully discrete schemes are also reported. Secondly, by combining the new SAV approach with functional transformation, we propose a new method to construct high-order, linear, positivity/bound preserving and unconditionally energy stable schemes for general dissipative systems whose solutions are positivity/bound preserving. We apply this new method to second order equations: the Allen-Cahn equation with logarithm potential, the Poisson-Nernst-Planck equation and the Keller-Segel equations and fourth order equations: the thin film equation and the Cahn-Hilliard equation with logarithm potential. Ample numerical examples are provided to demonstrate the improved efficiency and accuracy of the proposed method.

We present an Embedded Boundary with Adaptive Mesh Refinement technique for solving the compressible Navier Stokes equations in arbitrary complex domains; followed by a numerical studies for the effect of circular cylinders on the transient dynamics of the Richtmyer-Meshkov Instability. A PDE multidimensional extrapolation approach is used to reconstruct the solution in the ghost-fluid regions and imposing boundary conditions on the fluid-solid interface, coupled with a multi-dimensional algebraic interpolation for freshly cleared cells. The Navier Stokes equations are numerically solved by the second order multidimensional upwind method. Block-structured AMR, implemented with the Chombo framework, is utilized to reduce the computational cost while keeping high resolution mesh around the Embedded Boundary and regions of high gradient solutions. The versatility of the method is demonstrated via several numerical examples, in both static and moving geometry, ranging from low Mach number nearly incompressible to supersonic flows. Our simulation results are extensively verified against other numerical results and validated against available experimental results where applicable. The effects on the transient dynamics of the Richtmyer-Meshkov instability due to small scale perturbations introduced on the shock-wave or the material interface by a single or set of solid circular cylinders were computationally investigated using the developed technique. First, we discuss the RMI initiated on a flat interface by a rippled shock-wave that is disturbed by a single circular cylinder. Then, we study the effect of introducing a number of circular cylinders on the interface. The arrangement of the cylinders set mimic (in a two dimensional domain) the presence of the solid supporting grid wires used in the formation of the material interface in the experimental setup. We analyzed their effects on the mixing layer growth and the mixedness level, and qualitatively demonstrate the cylinders' perturbation effects on the mixing layer structure. We modeled the cylinders' influence based on their diameters; and showed the model ability to predict the variation of the mixing layer growth for different flow parameters.

碩士<br>中原大學<br>電機工程研究所<br>92<br>Many Operational Transconductance Amplifier (OTA)-based nth-order voltage-mode universal filter structures were proposed. Very recently, both linear combination and singnal weighting approaches archieve two new nth-order voltage-mode universal filter structures. The former employs 2n+4 single-input OTAs and n grounded capacitors. The latter employs 2n+2 single-input OTAs and n grounded capacitors. In this study, we have done the comparison between linear combination and weighting signal approaches in terms of filtering performance, noise, sensitivity, component spread and cascadability. H-spice simulations with UMC05 Level 49 parameters are used to do the comparison. When non-ideal active devices are considered, differences in frequency responses of the linear combination and weighting signal approaches will be observed. Clearly these differences can be attributed to non-ideal active devices. In the research of OTA-C filters, weighting signal approach is better than linear combination method in terms of filtering performance, noise and sensitivity because the signal weighting method uses two fewer single-ended-input OTAs than the linear combination approach. As we know, for the universal nth-order, the number of capacitors needed should be at least n and to achieve the independent tunability of each coefficient at least 2n+2 OTAs are required. In this sense our technique needs the fewest element, which should be very important for integration. In terms of component spread and cascadability, linear combination method is better than weighting signal approach. This is because the signal weighting method have more input path than linear combination approach. This paper offers the signal weighting and linear combination approaches differences in terms of filtering performance, noise, sensitivity, component spread and cascadability. We have to make the most of this comparsion’s conclusions, which should be very useful for filters implemented using linear combination and singnal weighting approaches archieve nth-order voltage-mode universal filter structures.