Dissertationen zum Thema „Non-intrusive approach“

PONTIFÍCIA UNIVERSIDADE CATÓLICA DO RIO DE JANEIRO
CONSELHO NACIONAL DE DESENVOLVIMENTO CIENTÍFICO E TECNOLÓGICO
O projeto físico de bancos de dados cumpre um papel primordial para assegurar um desempenho adequado. Atualmente, existe uma grande quantidade de trabalhos e ferramentas na área de seleção automática do projeto físico. Tais ferramentas, contudo, adotam uma abordagem offline na solução do problema e transferem para o DBA, dentre outras tarefas, a decisão de executar ou não as recomendações sugeridas. Todavia, em ambientes dinâmicos, com consultas ad-hoc, torna-se bastante complexo identificar configurações de projeto físico que sejam adequadas. Recentemente, algumas iniciativas apresentaram descrições de protótipos que implementam funcionalidades de sintonia automática. Estes trabalhos, porém, adotam uma abordagem intrusiva e funcionam apenas com um SGBD específico. Neste trabalho, propõe-se uma abordagem não-intrusiva para a manutenção automática e on-the-fly do projeto físico de bancos de dados. A abordagem proposta é completamente desacoplada do código do SGBD, pode ser utilizada com qualquer SGBD e executada sem intervenção humana. A estratégia adotada baseia-se em heurísticas que executam continuamente e, sempre que necessário, modificam o projeto físico corrente, reagindo a alterações na carga de trabalho. Para comprovar a viabilidade das idéias apresentadas, a abordagem proposta foi instanciada para solucionar dois importantes problemas relacionados ao projeto físico: a manutenção automática de índices e de clusters alternativos de dados.
The physical design of a database plays a critical role in performance. There has been considerable work on automated physical design tuning for database systems. Existing solutions require offline invocations of the tuning tool and depend on DBAs identifying representative workloads manually. However, in dynamic environments involving various ad-hoc queries it is difficult to identify potentially useful physical design in advance. Recently, a few initiatives present brief descriptions of prototypes that address some aspects of online physical tuning. Nevertheless, these references work in an intrusive manner and work only with a specific DBMS. In this work, we propose a non intrusive approach to automated and on-the-fly physical design problems, in order to speed up processing of subsequent queries. Specifically, we design algorithms that are always-on and continuously modify the current physical design, reacting to changes in the query workload. To prove the viability of the presented ideas, the proposed approach was instantiated to solve two major problems related to the database physical design: indexing and alternative data clusters automatic maintenance.

PONTIFÍCIA UNIVERSIDADE CATÓLICA DO RIO DE JANEIRO
COORDENAÇÃO DE APERFEIÇOAMENTO DO PESSOAL DE ENSINO SUPERIOR
PROGRAMA DE EXCELENCIA ACADEMICA
Este trabalho apresenta aplicações de métodos espectrais estocásticos para otimização topológica de estruturas na presença de incertezas. Esse procedimento, conhecido como otimização topológica robusta, minimiza uma combinação entre a média e o desvio padrão da função objetivo. Para tanto, uma expansão de caos polinomial não intrusiva é integrada a um algoritmo de otimização topológica para se calcular os dois primeiros momentos estatísticos da resposta do modelo mecânico. A fim de abordar as variabilidades na resposta estrutural, as incertezas são consideradas no carregamento e nas propriedades do material. Na formulação probabilística proposta, as incertezas são representadas como um conjunto de variáveis aleatórias (por exemplo, magnitudes e direções das cargas) ou como campos aleatórios (por exemplo, cargas distribuídas e propriedades do material). Um campo aleatório homogêneo não Gaussiano com uma função de distribuição marginal e covariância especificada é usado para representar as incertezas nas propriedades dos materiais, pois garante a sua admissibilidade física. A transformação não-linear sem memória de um campo Gaussiano homogêneo é usada para obter campos não Gaussianos. A expansão de Karhunen-Loève é empregada para fornecer uma representação do campo Gaussiano em termos de um número finito de variáveis aleatórias independentes. A quadratura de grade esparsa é empregada para reduzir o custo computacional no cálculo dos coeficientes da expansão do caos polinomial. Além disso, é mostrada uma previsão eficiente (isto é, com um baixo custo computacional) da resposta estrutural sob incertezas. A precisão e a aplicabilidade da metodologia proposta são demonstradas por meio de vários exemplos de otimização topológica de estruturas contínuas 2D. Os resultados obtidos estão em excelente concordância com as soluções obtidas pelo método de Monte Carlo. Finalmente, conclusões são apresentadas e possíveis extensões deste trabalho são propostas.
This work presents some applications of stochastic spectral methods for structural topology optimization in the presence of uncertainties. This procedure, known as robust topology optimization, minimizes a combination of the mean and standard deviation of the objective function. For this purpose, a non-intrusive polynomial chaos expansion is integrated into a topology optimization algorithm to calculate the first two statistical moments of the mechanical model response. In order to address variabilities in the structural response, the uncertainties are considered in the loading and the material properties. In this proposed probabilistic formulation, uncertainties are represented as a set of random variables (e.g., magnitudes and directions of the loads) or as random fields (e.g., distributed loads and material properties). A non-Gaussian homogenous random field with a specified marginal distribution and covariance function is used to represent the material uncertainties because it ensures their physical admissibility. Nonlinearm memoryless transformation of a homogeneous Gaussian field is used for obtaining non-Gaussian fields. The Karhunen-Loève expansion is employed to provide a representation of the Gaussian field in terms of countable uncorrelated random variables. The sparse grid quadrature is considered for reducing the computational cost when computing the coefficients of the polynomial chaos expansion. Moreover, an efficient prediction (i.e., with a low computational cost) of the structural response under uncertainties is presented. Accuracy and applicability of the proposed methodology are demonstrated by means of several topology optimization examples of 2D continuum structures. The obtained results are in excellent agreement with the solutions obtained using the Monte Carlo method. Finally, conclusions are inferred and possible extensions of this work are proposed.

During my internship at the Upholstery conservation department at the Metropolitan Museum of art in New York, I came across a chair that caught my interest and therefore I chose to dedicate the last two months of my internship, investigating this chair. It was said about the chair that it had the original upholstery and cover, but some questions appeared. These questions I have tried to solve, as I will reveal in the following, by using mostly non-intrusive methods, in attempt to see how much useful information this approach can provide, concerning frame structure, inner-upholstery, showcover and possible campaigns (interventions). I found out, that it was possible to gather a great amount of useful information. In fact I will go so far as to claim that I found out even more than I would have found out, by  isassembling the object manually.

Includes bibliography.
Electrical energy is the life line to every nation’s or continent development and economic progress. Referable to the recent growth in the demand for electricity and shortage in production, it is indispensable to develop strategies for effective energy management and system delivery. Load monitoring such as intrusive load monitoring, non-intrusive load monitoring, and identification of domestic electrical appliances is proposed especially at the residential level since it is the major energy consumer. The intrusive load monitoring provides accurate results and would allow each individual appliance's energy consumption to be transmitted to a central hub. Nevertheless, there are many practical disadvantages to this method that have motivated the introduction of non-intrusive load monitoring system. The fiscal cost of manufacturing and installing enough monitoring devices to match the number of domestic appliances is considered to be a disadvantage. In addition, the installation of one meter per household appliances would lead to congestion in the house and thus cause inconvenience to the occupants of the house, therefore, non-intrusive load monitoring technique was developed to alleviate the aforementioned challenges of intrusive load monitoring. Non-intrusive load monitoring (NILM) is the process of disaggregating a household’s total energy consumption into its contributing appliances. The total household load is monitored via a single monitoring device such as smart meter (SM). NILM provides cost effective and convenient means of load monitoring and identification. Several nonintrusive load monitoring and identification techniques are reviewed. However, the literature lacks a comprehensive system that can identify appliances with small energy consumption, appliances with overlapping energy consumption and a group of appliance ranges at once. This has been the major setback to most of the adopted techniques. In this dissertation, we propose techniques that overcome these setbacks by combining artificial neural networks (ANN) with a developed algorithm to identify appliances ranges that contribute to the energy consumption within a given period of time usually an hour interval.

Un des objectifs prioritaires des industries aéronautiques est la réduction de la masse des structures, tout en permettant l'amélioration de leurs performances. Ceci passe notamment par l'utilisation de matériaux composites et le recours croissant à la simulation numérique, permettant la minimisation du nombre d'essais physiques et l'optimisation des structures.L'enjeu de ces travaux est de pouvoir calculer précisément, sur des matériaux architecturés, l'influence de la microstructure, modélisée par exemple directement par tomographie, sur la tenue de pièces complètes. Pour prendre en compte à la fois l'ensemble de la pièce et les effets de son chargement, une approche global/local multiéchelle semble adaptée tant du point de vue des méthodes de calcul que des modèles matériaux utilisés.Pour répondre à cette problématique, une méthode de couplage entre des modèles qui décrivent une même structure, mais à des échelles différentes, a été développée. Elle repose sur une séparation micro-macro des quantités d’interface, dans la zone de raccord surfacique entre les deux modèles. Pour faciliter son utilisation dans les bureaux d’étude, une technique de résolution itérative non-intrusive est également présentée. Elle permet de mettre en œuvre la méthode de couplage proposée dans un environnement logiciel industriel qui utilise bien souvent des codes éléments finis commerciaux fermés. La méthode est systématiquement comparée à d'autres méthodes de couplage de la littérature et la qualité des solutions est quantifiée par comparaison à une solution de référence obtenue par un calcul direct à l'échelle fine.Les principaux résultats sont encourageants dans la mesure où ils montrent, dans des cas d'étude représentatifs bidimensionnels et tridimensionnels, sous des hypothèses d’élasticité linéaire, des solutions cohérentes avec les théories de l’homogénéisation au premier et second ordre. De fait, les solutions obtenues sont systématiquement de meilleure qualité avec la méthode proposée qu'avec les méthodes de la littérature, non-adaptées à des cas de couplage pour modèles non-compatibles.Finalement, les perspectives sont multiples en raison des différentes alternatives de la méthode qui, dans un contexte industriel, pourrait offrir un véritable outil d'analyse visant à introduire un modèle local décrit à l'échelle fine dans un modèle global macroscopique homogénéisé
One of the priority objectives of the aeronautics industry is to reduce the mass of structures while improving their performances. This involves the use of composite materials and the increasing use of digital simulation to optimize structures.The major challenge of this project is to be able to accurately calculate the local variations of the microstructure - for instance detected by tomography and directly modelled from tomogram - on the behavior of an architectured material part. In order to take into account the whole structure and its load effects, a multi-scale approach seems to be a natural choice. Indeed, the related models to the part and its microstructure might use different formalisms according to each scale.In this context, a coupling formulation was proposed in order to replace, in a non-intrusive way, a part of a homogenized macroscopic finite-element model by a local one described at a microscopic level. It is based on a micro-macro separation of interface quantities in the coupling area between the two models. To simplify its use in design offices, a non-intrusive iterative resolution procedure has also been proposed. It allows the implementation of the proposed coupling method in an industrial software environment that often uses closed commercial finite element codes. Different mechanical problems under linear elasticity assumption are proposed. The proposed method is systematically compared with other coupling methods of the literature and the quality of the solutions is quantified compared to a reference one obtained by direct numerical simulation at a fine scale.The main results are promising as they show, for representatives test cases under linear elasticity assumption in two and three-dimensions, solutions that are consistent with first- and second-order homogenization theories. The solutions obtained with the proposed method are systematically the best approximations of the reference solution whereas the methods of the literature are less accurate and shown to be unsuitable to couple non-compatible models.Finally, there are many perspectives due to the different alternatives of the method which could become, in an industrial context, a real analytic tool that aims to introduce a local model described at a fine scale, into a homogenized macroscopic global one

This study presents a modelling approach for quantifying the Mobile Web Quality of Experience (MWQoE). It builds on current QoE and Web QoE research, and by fusing together data that is available on modern mobile devices, constructs a novel MWQoE model that is user-centered, context-aware and non-intrusive (does not depend on user feedback). This study identifies the factors which affect Web QoE and measures their effect on it in mobile scenarios. Moreover, this study explores scenarios in which Web QoE can be effectively characterized and enhanced, delivering a novel Mobile-to-Cloud system for the continuous evaluation of MWQoE in real-world environments. The significance of defining and evaluating MWQoE is identified. Specifically, MWQoE can be used by online providers to uncover customer insights and illustrate how the experience in using their products is perceived by their customers. In fact, MWQoE can be considered an important key performance indicator showing the technology acceptance or satisfiability of customers for a specific web product or service.

Mänskligt igenkännande med användning av sensorer och maskininlärning är ett fält med många praktiska tillämpningar. Det finns några kommersiella produkter som på ett tillförlitligt sätt kan känna igen människor med hjälp av videokameror. Dock ger videokameror ofta en oro för inkräktning i privatlivet, men genom att läsa det relaterade arbetet kan man hävda att i vissa situationer är en videokamera inte nödvändigtvis mer tillförlitlig än billiga, icke-inkräktande sensorer. Att känna igen antalet människor i ett litet studie / kontorsrum är en sådan situation. Även om det har gjorts många framgångsrika studier för igenkänning av människor med olika sensorer och maskininlärningsalgoritmer, kvarstår en fråga om vilken kombination av sensorer och maskininlärningsalgoritmer som är allmänt bättre. Denna avhandling utgår från att testa fem lovande sensorer i kombination med sex olika maskininlärningsalgoritmer för att bestämma vilken kombination som överträffade resten. För att uppnå detta byggdes en arduino prototyp för att samla in och spara läsningarna från alla fem sensorer i en textfil varje sekund. Arduinon, tillsammans med sensorerna, placerades i ett litet studierum på Malmö universitet för att samla data vid två separata tillfällen medan studenterna använde rummet som vanligt. Den insamlade datan användes sedan för att träna och utvärdera fem maskininlärningsklassificerare för var och en av de möjliga kombinationerna av sensorer och maskininlärningsalgoritmer, för både igenkänningsdetektering och igenkänningsantal. I slutet av experimentet konstaterades det att alla algoritmer kunde uppnå en precision på minst 90% med vanligtvis mer än en kombination av sensorer. Den högsta träffsäkerheten som uppnåddes var 97%.
Human recognition with the use of sensors and machine learning is a field with many practical applications. There exists some commercial products that can reliably recognise humans with the use of video cameras. Video cameras often raises a concern about privacy though, by reading the related work one could argue that in some situations a video camera is not necessarily more reliable than low-cost, non-intrusive, ambient sensors. Human occupancy recognition in a small sized study/office room is one such situation. While there has been a lot of successful studies done on human occupancy recognition with various sensors and machine learning algorithms, a question about which combination of sensors and machine learning algorithms is more viable still remains. This thesis sets out to test five promising sensors in combination with six different machine learning algorithms to determine which combination outperformed the rest. To achieve this, an arduino prototype was built to collect and save the readings from all five sensors into a text file every second. The arduino, along with the sensors, was placed in a small study room at Malmö University to collect data on two separate occasions whilst students used the room as they would usually do. The collected data was then used to train and evaluate five machine learning classifier for each of the possible combinations of sensors and machine learning algorithms, for both occupancy detection and occupancy count. At the end of the experiment it was found that all algorithms could achieve an accuracy of at least 90% with usually more than one combination of sensors. The highest hit-rate achieved was 97%.

Dans le cadre de ce travail, une technique non-intrusive est proposée pour coupler la méthode Smoothed Particle Hydrodynamics (SPH) à la méthode des Eléments Finis afin de résoudre numériquement des problèmes dynamiques et non-linéaires d’interaction fluide-structure en permettant l’utilisation des pas de temps différents dans les deux domaines de calcul (fluide et solide). Ces développements sont motivés par le besoin de simuler numériquement des phénomènes rapides et très non-linéaires qui prennent en compte des impacts en se servant des intégrateurs temporels explicites dans chaque sous-domaine de calcul (Newmark explicite pour le solide et Runge-Kutta 2 pour le fluide). De ce fait, le pas de temps de stabilité est limité par des caractéristiques intrinsèques au modèle numérique du phénomène étudié et en conséquence, il devient important de pouvoir intégrer chaque sous-domaine numérique avec un pas de temps proche de son pas de temps de stabilité. Pour permettre d’utiliser un pas de temps proche du pas de temps de stabilité pour chaque sous-domaine, des méthodes de décomposition de domaines dual-Schur sont implémentées et validées pour des cas en 1-D, 2-D, et 3-D. Des simulations numériques d’impacts de cailloux sur des aubes des turbines hydrauliques sont aussi effectue´es afin de prédire le dommage que cet évènement peut engendrer
A method to couple smoothed particle hydrodynamics and finite elements methods for nonlinear transient fluid–structure interaction simulations by adopting different time-steps depending on the fluid or solid sub-domains is proposed. These developments were motivated by the need to simulate highly non-linear and sudden phenomena that take into acount solid impacts and hence require the use of explicit time integrators on both sub-domains (explicit Newmark for the solid and Runge–Kutta 2 for the fluid). However, due to critical time-step required for the stability of the explicit time integrators in, it becomes important to be able to integrate each sub-domain with a different time-step while respecting the features that a previously developed mono time-step coupling algorithm offered. For this matter, a dual-Schur decomposition method originally proposed for structural dynamics was considered, allowing to couple time integrators of the Newmark family with different time-steps with the use of Lagrange multipliers

Thesis (Degrees of Naval Engineer and M.S. in system Design and Management)--Massachusetts Institute of Technology, June 2008.
"June 2008." Description based on title screen as viewed on August 26, 2009. DTIC Descriptor(s): Diagnostic Equipment, Electrical Loads, Fault Detection, Electromechanical Devices, Bench Tests, Decomposition, Alternating Current, Theses, Monitoring . DTIC Identifier(s): Electromechanical Loads, Current Spectra, Automated Diagnostic Systems, NILM (Non-Intrusive Load Monitor), Induction Machine Problems, Current Harmonics . Includes bibliographical references (p. 88-90). Also available in print.

Thesis (Nav. E.)--Massachusetts Institute of Technology, Dept. of Mechanical Engineering; and, (S.M.)--Massachusetts Institute of Technology, System Design and Management Program, 2008.
Includes bibliographical references (p. 88-90).
The Non-Intrusive Load Monitor (NILM) is a system that monitors, records and processes voltage and current measurements to establish the operating characteristics of individual loads on a load center from a single aggregate measurement. The NILM can also be used to actively monitor degradation or diagnose specific system failures. Current NILM research conducted at the Massachusetts Institute of Technology's Laboratory for Electromagnetic and Electronic Systems (LEES) is exploring the application and expansion of NILM technology for the use of monitoring a myriad of electromechanical loads. This thesis presents a fundamental guide to understanding NILM operation using laboratory bench testing and demonstrates its potential to detect an array of electric machine failures before they become catastrophic. The NILM's ability to the monitor the current spectrum of electric machines can be used to immediately diagnose multiple common system casualties and detect unusual system operation. Clean current spectrum regions can be exploited by selecting induction machine design characteristics that result in eccentric modulation frequencies occurring in areas free of supply frequency harmonics. Current spectrum analysis was used to demonstrate the NILM's potential to monitor multiple machines from an aggregate source and discuss intersystem impedances. It can be shown that multiple machines with slightly varied physical characteristics, such as induction motor rotor slots, coupled with using clean current spectral regions support automated diagnostic system development. Measurements and experimentation were conducted in the LEES laboratory and the Industrial Support Center electric shop, Boston.
by Ashley E. Fuller.
S.M.
Nav.E.

Cette thèse, une méthode permettant d'enrichir localement un problème éléments finis industriel est présentée. Cette méthode est non-intrusive, c'est-à-dire que ni le modèle ni le logiciel correspondant ne sont modifiés. Elle consiste à définir un modèle local de la zone concernée, contenant les enrichissements voulus, puis à le substituer au sein du problème industriel, à l'aide d'une technique de couplage de codes adaptée des méthodes de décomposition de domaine. L'aspect non-intrusif et la flexibilité de l'approche sont obtenus en utilisant les deux solveurs comme "boîte noires" et en n'échangeant que des chargements nodaux d'un code à l'autre. L'application présentée ici concerne l'introduction de plasticité et de détails géométriques localisés dans un modèle élastique linéaire. Deux sortes de conditions aux limites sur le modèle local sont étudiées (Dirichlet et Robin) et une technique d'accélération quasi-newton non-intrusive est proposée, menant à une convergence très rapide. Les exemples présentés utilisent le logiciel Abaqus et incluent un problème industriel banalisé fourni par Snecma
This work introduces a method for embedding local details, nonlinear phenomena or innovative features into an industrial finite element problem in a non-intrusive way, ie. Without modifying the existing model data or software. It consists in creating a local model of the concerned region, containing the desired features, and then substituting it into the industrial problem, by the means of a solver coupling technique adapted from domain decomposition methods; non-intrusiveness and flexibility are obtained by using both solvers as black boxes and by only exchanging load vectors between the solvers. The method is currently applied to introduce localized plasticity and geometric details into a large linear elastic model. Two kinds of local boundary conditions are studied (Dirichlet and Robin) and a non-intrusive quasi-Newton acceleration technique is introduced, that leads to very fast convergence. Examples are shown using the commercial software package Abaqus, including a large 3D problem provided by our industrial partner Snecma

Dans le contexte industriel actuel, où la simulation numérique joue un rôle majeur, de nombreux outils sont développés afin de rendre les calculs les plus performants et exacts possibles en utilisant les ressources numériques de façon optimale. Parmi ces outils, ceux non-intrusifs, c’est-à-dire ne modifiant pas les codes commerciaux disponibles mais permettant d’utiliser des méthodes de résolution avancées telles que l’analyse isogéométrique ou les couplages multi-échelles, apparaissent parmi les plus attirants pour les industriels. L’objectif de cette thèse est ainsi de coupler l’Analyse IsoGéométrique (AIG) et la Méthode des Éléments Finis (MEF) standard pour l’analyse de détails structuraux par une approche non-intrusive et certifiée. Dans un premier temps, on développe un lien global approché entre les fonctions de Lagrange, classiquement utilisées en éléments finis et les fonctions NURBS bases de l’AIG, ce qui permet d’implémenter des analyses isogéométriques dans un code industriel EF vu comme une boîte noire. Au travers d’exemples linéaires et non-linéaires implémentés dans le code industriel Code_Aster de EDF, nous démontrons l’efficacité de ce pont AIG\MEF et les possibilités d’applications industrielles. Il est aussi démontré que ce lien permet de simplifier l’implémentation du couplage non-intrusif entre un problème global isogéométrique et un problème local éléments finis. Ensuite, le concept de couplage non-intrusif entre les méthodes étant ainsi possible, une stratégie d’adaptation est mise en place afin de certifier ce couplage vis-à-vis d’une quantité d’intérêt. Cette stratégie d’adaptation est basée sur des méthodes d’estimation d’erreur a posteriori. Un estimateur global et des indicateurs d’erreur d’itération, de modèle et de discrétisation permettent de piloter la définition du problème couplé. La méthode des résidus est utilisée pour évaluer ces erreurs dans des cas linéaires, et une extension aux problèmes non-linéaires via le concept d’Erreur en Relation de Comportement (ERC) est proposée
In the current industrial context where the numerical simulation plays a major role, a large amount of tools are developed in order to perform accurate and effective simulations using as less numerical resources as possible. Among all these tools, the non-intrusive ones which do not modify the existing structure of commercial softwares but allowing the use of advanced solving methods, such as isogeometric analysis or multi-scale coupling, are the more attractive to the industry. The goal of these thesis works is thus the coupling of the Isogeometric Analysis (IGA) with the Finite Element Method (FEM) to analyse structural details with a non-intrusive and certified approach. First, we develop an approximate global link between the Lagrange functions, commonly used in the FEM, and the NURBS functions on which the IGA is based. It’s allowed the implementation of isogeometric analysis in an existing finite element industrial software considering as a black-box. Through linear and nonlinear examples implemented in the industrial software Code_Aster of EDF, we show the efficiency of the IGA\FEM bridge and all the industrial applications that can be made. This link is also a key to simplify the non-intrusive coupling between a global isogeometric problem and a local finite element problem. Then, as the non-intrusive coupling between both methods is possible, an adaptive process is introduced in order to certify this coupling regarding a quantity of interest. This adaptive strategy is based on a posteriori error estimation. A global estimator and indicators of iteration, model and discretization error sources are computed to control the definition of the coupled problem. Residual base methods are performed to estimated errors for linear cases, an extension to the concept of constitutive relation errors is also initiated for non-linear problems

Dans les maisons et bâtiments intelligents, il devient nécessaire de limiter l'intervention humaine sur le système énergétique, afin de fluctuer automatiquement l'énergie consommée par les appareils consommateurs. Pour cela, un système de mesure de la consommation électrique d'équipements est aussi nécessaire et peut être déployé de deux façons : intrusive ou non-intrusive. La première solution consiste à relever la consommation de chaque appareil, ce qui est inenvisageable à une grande échelle pour des raisons pratiques liées à l'entretien et aux coûts. Donc, la solution non-intrusive (NILM pour Non-Intrusive Load Monitoring), qui est capable d'identifier les différents appareils en se basant sur les signatures extraites d'une consommation globale, est plus prometteuse. Le problème le plus difficile des algorithmes NILM est comment discriminer les appareils qui ont la même caractéristique énergétique. Pour surmonter ce problème, dans cette thèse, nous proposons d'utiliser une information externe pour améliorer la performance des algorithmes existants. Les premières informations additionnelles proposées considèrent l'état précédent de chaque appareil comme la probabilité de transition d'état ou la distance de Hamming entre l'état courant et l'état précédent. Ces informations sont utilisées pour sélectionner l'ensemble le plus approprié des dispositifs actifs parmi toutes les combinaisons possibles. Nous résolvons ce problème de minimisation en norme l1 par un algorithme d'exploration exhaustive. Nous proposons également d'utiliser une autre information externe qui est la probabilité de fonctionnement de chaque appareil fournie par un réseau de capteurs sans fil (WSN pour Wireless Sensor Network) déployé dans le bâtiment. Ce système baptisé SmartSense, est différent de la solution intrusive car seul un sous-ensemble de tous les dispositifs est surveillé par les capteurs, ce qui rend le système moins intrusif. Trois approches sont appliquées dans le système SmartSense. La première approche applique une détection de changements de niveau sur le signal global de puissance consommé et les compare avec ceux existants pour identifier les dispositifs correspondants. La deuxième approche vise à résoudre le problème de minimisation en norme l1 avec les algorithmes heuristiques de composition Paréto-algébrique et de programmation dynamique. Les résultats de simulation montrent que la performance des algorithmes proposés augmente significativement avec la probabilité d'opération des dispositifs surveillés par le WSN. Comme il n'y a qu'un sous-ensemble de tous les appareils qui sont surveillés par les capteurs, ceux qui sont sélectionnés doivent satisfaire quelques critères tels qu'un taux d'utilisation élevé ou des confusions dans les signatures sélectionnées avec celles des autres
In smart homes, human intervention in the energy system needs to be eliminated as much as possible and an energy management system is required to automatically fluctuate the power consumption of the electrical devices. To design such system, a load monitoring system is necessary to be deployed in two ways: intrusive or non-intrusive. The intrusive approach requires a high deployment cost and too much technical intervention in the power supply. Therefore, the Non-Intrusive Load Monitoring (NILM) approach, in which the operation of a device can be detected based on the features extracted from the aggregate power consumption, is more promising. The difficulty of any NILM algorithm is the ambiguity among the devices with the same power characteristics. To overcome this challenge, in this thesis, we propose to use an external information to improve the performance of the existing NILM algorithms. The first proposed additional features relate to the previous state of each device such as state transition probability or the Hamming distance between the current state and the previous state. They are used to select the most suitable set of operating devices among all possible combinations when solving the l1-norm minimization problem of NILM by a brute force algorithm. Besides, we also propose to use another external feature that is the operating probability of each device provided by an additional Wireless Sensor Network (WSN). Different from the intrusive load monitoring, in this so-called SmartSense system, only a subset of all devices is monitored by the sensors, which makes the system quite less intrusive. Two approaches are applied in the SmartSense system. The first approach applies an edge detector to detect the step-changes on the power signal and then compare with the existing library to identify the corresponding devices. Meanwhile, the second approach tries to solve the l1-norm minimization problem in NILM with a compositional Pareto-algebraic heuristic and dynamic programming algorithms. The simulation results show that the performance of the proposed algorithms is significantly improved with the operating probability of the monitored devices provided by the WSN. Because only part of the devices are monitored, the selected ones must satisfy some criteria including high using rate and more confusions on the selected patterns with the others

This dissertation explores the problem of energy estimation in supervised Non-Intrusive Load Monitoring (NILM). NILM refers to a set of techniques used to estimate the electricity consumed by individual loads in a building from measurements of the total electrical consumption. Most commonly, NILM works by first attributing any significant change in the total power consumption (also known as an event) to a specific load and subsequently using these attributions (i.e. the labels for the events) to estimate energy for each load. For this last step, most proposed solutions in the field impart simplifying assumptions to make the problem more tractable. This has severely limited the practicality of the proposed solutions. To address this knowledge gap, we present a framework for creating appliance models based on classification labels and aggregate power measurements that can help relax many of these assumptions. Within the framework, we model the problem of utilizing a sequence of event labels to generate energy estimates as a broader class of problems that has two major components (i) With the understanding that the labels arise from a process with distinct states and state transitions, we estimate the underlying Finite State Machine (FSM) model that most likely generated the observed sequence (ii) We allow for the observed sequence to have errors, and present an error correction algorithm to detect and correct them. We test the framework on data from 43 appliances collected from 19 houses and find that it improves errors in energy estimates when compared to the case with no correction in 19 appliances by a factor of 50, leaves 17 appliances unchanged, and negatively impacts 6 appliances by a factor of 1.4. This approach of utilizing event sequences to estimate energy has implications in virtual metering of appliances as well. In a case study, we utilize this framework in order to substitute the need of plug-level sensors with cheap and easily deployable contacless sensors, and find that on the 6 appliances virtually metered using magnetic field sensors, the inferred energy values have an average error of 10:9%.

Le stockage des déchets radioactifs dans des formations géologiques profondes nécessite d'excaver de la roche en grande profondeur pour accueillir les différentes installations. L'argilite du Callovo-Oxfordian a été choisie comme une potentielle formation hôte grâce à sa capacité de rétention et ses propriétés hydromécaniques. L'objectif de la thèse est d'étudier numériquement la fissuration induite lors du creusement des galeries souterraines à -490 m de profondeur. Un modèle Éléments Finis 3D (méthode E-FEM) est développé pour représenter la fissuration. Plusieurs critères sont proposés pour caractériser les fissures avec une ouverture en mode I ou un glissement en mode II. L'influence de l'anisotropie des propriétés de l'argilite et du champ de contrainte in situ sur les réseaux de fractures et la convergence des galeries est principalement étudiée. L'origine géologique de l'argilite, la complexité de sa microstructure et les grandes dimensions des galeries amènent à des incertitudes sur les propriétés hydromécaniques de ce matériau. La prise en compte des variabilités spatiales des paramètres mécaniques de la roche se fait au travers de champs aléatoires corrélés. Des Méthodes Éléments Finis probabilistes avec des formulations de Galerkin (méthodes d’intégration stochastique indirecte) sont ensuite développées avec une approche non intrusive pour propager ces incertitudes pour des systèmes linéaires et non linéaires (avec et sans fissuration). Ces méthodes sont ensuite appliquées à des problèmes d’excavation pour propager les incertitudes paramétriques associées au comportement de l'argilite. Les résultats sont comparés par rapport à ceux obtenus par des méthodes d’intégration stochastique directe (famille des méthodes dites de Monte-Carlo)
A 3D numerical modelling using the Enhanced Finite Element Method (E-FEM) is developped to address induced fracture networks around drifts after an excavation in Callovo-Oxfordian claystone (COx). A transversely isotropic behaviour is considered for the host rock and two failure criteria are proposed and implemented to characterize shear and tensile fractures: an anisotropic/isotropic Mohr Coulomb criterion with sliding (mode II) and an anisotropic principal strain criterion with mode I opening. Numerical simulations of underground excavations are presented and the results are discussed compared to in situ observations. The influence of the anisotropy of rock properties and in situ stress field on the induced fractures and the convergence of drifts are also studied. The geological origin of the COx as well as the large size of drifts lead to a large uncertainty related to hydro-mechanical properties of this material. Stochastic problems for linear and non linear systems are more and more of interest because it is an important issue to quantify uncertainties when parameters (loading, material properties) are modelled by correlated random fields. Despite a slow convergence, Monte Carlo methods are the most often used to solve these problems thanks to its easy implementation (non intrusive computation). Probabilistic Finite Element methods like Galerkin formulations which produce a coupled system are often viewed as intrusive. Here, we want to show that is possible to compute them in a non intrusive way and with the same accuracy as Monte Carlo methods which are considered as a reference. These methods are applied to take into account and to propagate parametric uncertainties related to claystone's behavior. 3D Numerical results are presented and discussed concerning the induced fracture networks around drifts and the anisotropic convergence of walls

Cette thèse est une contribution à un thème de recherche sur les applications de la réduction de modèle à l’ingénierie RENAULT. Elle fait suite aux travaux du projet IRT SystemX ROM (Réduction de modèles et Optimisation Multidisciplinaire) et aux thèses CIFRE RENAULT [Vuong], [Charrier]. L’application industrielle principale du thème est la mise au point de la structure d’un véhicule sollicité en crash; des travaux sont également en cours sur la combustion, l’acoustique et l’aérodynamique. Les travaux de cette thèse sont à la fois un apport à la méthode générique ReCUR et son adaptation à l’optimisation de la caisse d’un véhicule pour le crash. RENAULT utilise largement l’optimisation pour la mise au point de la caisse en crash, avec un outil d’optimisation numérique basé sur la méthode des plans d’expériences. Cette méthode nécessite beaucoup de calculs crash car cette simulation est considérée comme une boite noire, en utilisant uniquement les entrées et sorties. La méthode ReCUR prend le contre-pied en exploitant un maximum d’informations de chaque simulation crash, dans le but de réduire fortement leur nombre. Les travaux de cette thèse ont permis de remplir cet objectif pour les applications de mise au point au nominal et pour l’optimisation robuste dans des cas complexes comme le choc frontal et arrière
This thesis is a part of a global research work dedicated to reduced-order modelling applications in the Renault engineering direction. It's research topic has been improved in the IRT System)('s project on Reduced Order Model and Multi-disciplinary Optimization. Some previous thesis can help understand the context. ([Vuong], [Charrier]). The main industrial application of the research theme is the focus on a body structure, in a crash loading. Some research works on acoustic, combustion and aerodynamic are currently ongoing. This thesis is both a contribution to the generic ReCUR method, and its application to a car body structure optimization for crash loadings. Engineering teams at Renault uses optimization to obtain the best crash simulation, with a numerical optimization software, based on designs of experiments. It requires a lot of crash simulation because each simulation is considered as unique, with only one response for each parameter. Only Inputs and Outputs are known. The ReCUR method consider that each simulation is a huge mine that needs our attention. We hope that we can decrease the number of crash simulation required to compute a model, by using much more data for each simulation

碩士
國立中央大學
資訊工程學系
101
With the progress of the smartphone, many applications appear fast. Some applications have something to do with the user's private information and individual finance, hence, if the information is stolen, it will be a great loss. Traditional authentication only authenticates user's password once, but if the password is stolen, the information of the smartphone will not be protected. Therefore, in this paper, we propose a method that we directly use the smartphone orientation and touch screen to catch the information of user's behavior, then according to this information, the smartphone will non-intrusively and continuously authenticate the smartphone user. In this study, we use two applications which include user up/down flick and left/right flick to simulate user's situation of using smartphone, and collect the information of user's behavior. The experimental result indicates that the EER(Equal Error Rate) of authentication mechanism less than 7%, it confirms that the combination of orientation sensor and touch screen is useful in user verification.

碩士
國立中央大學
軟體工程研究所
102
With the advances in smartphone applications, data on the phone to save a lot of important personal privacy, which makes mobile phone security, is increasingly valued. In the traditional phone lock, are one-time validation mechanisms, so I hope by reading the phone sensor without interrupting the user's behavior for authentication the user. This study will be used in mobile applications can continue to collect user operation information to touch the screen and orientation sensors to identify the user as a characteristic information, and in accordance with the existing literature encountered issues raised to improve and strengthen their identification effect. Finally, this study will propose a combination of the ensemble learning and authentication mechanism by multiple models, according to the experimental results of this verification mechanism to identify EER (Equal Error Rate) of 1.5% or less, to enhance the existing authentication mechanisms.

碩士
國立中央大學
資訊工程學系
102
Smartphone sales obviously grew in this years, so the associated security issues about smartphone has become more important. In order to protect the data within the smartphone, intrusive and non-intrusive user authentication mechanisms were developed. Traditional authentication mechanisms like number lock and pattern lock are intrusive user authentication mechanism. Non-intrusive user authentication mechanism doesn’t require any user interface, but collect user’s behavior in the background and authenticate it. Several non-intrusive authentication mechanisms were proposed, but all of them collected user behavior in one fixed posture. These mechanisms didn’t take posture’s effect into consideration. First, for this study, we analyze user’s behavior data in different postures and confirm that user’s behavior data has significant differences in different postures. Second, from the view of the application, if the accuracy that use mixed posture behavior data’s model to predict isn’t significantly lower than the accuracy that separately use single posture behavior data’s model to predict, we can directly neglect posture’s effect and use mixed posture behavior data’s model to predict. This problem will be discussed by doing the experiment in dynamics-based approach and then inform the future researchers that they can use mixed posture behavior data’s model to predict according to the experiment result. Finally, we recommend the best classifier that can avoid the posture’s effect and have the best prediction accuracy.