Orientador: Raquel Gonçalves / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Agricola / Made available in DSpace on 2018-08-06T00:21:33Z (GMT). No. of bitstreams: 1 Silva_SergioAugustoMelloda_D.pdf: 3060293 bytes, checksum: 3514c54e017be522e62c07231642a540 (MD5) Previous issue date: 2003 / Resumo: As tecnologias para produção das chapas denominadas ¿Medium Density Fiberboard¿ (MDF), confeccionadas com fibras de madeiras de reflorestamento de baixo custo e resinas fenólicas provenientes de uréia, representam para vários setores industriais uma possibilidade de agregar valores às espécies com pouco interesse comercial, possibilitando para os seguimentos industriais investimento promissores. Um aspecto importante a ser considerado sobre a utilização de resinas fenólicas na confecção de chapas MDF é que produzem emissões tóxicas poluidoras da natureza e nocivas ao ser humano. Dentro deste contexto, de acordo com o ¿Protocolo de Quioto¿ de 1997, os países industrializados precisam diminuir suas emissões combinadas de gases de efeito estufa em pelo menos 5 % até 2012. Considerando-se as exigências do ¿Protocolo de Quioto¿, este trabalho objetivou avaliar o desempenho de chapas MDF confeccionadas com poliuretana (PU) derivada do óleo de mamona, pois este produto apresenta baixos teores de toxidez sendo classificado como não poluente e não tóxico ao ser humano. A caracterização das chapas foi realizada considerando-se a utilização do equipamento de ultra-som da Marca Steinkamp modelo BP7 com transdutoes de 45 kHz e a realização de ensaios físicos e mecânicos propostos pela EuroMDFBoard - EMB. A avaliação desenvolveu-se em três fases distintas: 1. Na fase I foram realizados estudos exploratórios com o objetivo de verificar as características do PU sendo utilizado como adesivo para confecção de chapas de Pinus caribaea. Nesta fase foram confeccionadas chapas com 5 e 10 % de PU3070, que de acordo com a caracterização física e mecânica observou-se que as chapas apresentaram resistências compatíveis com as exigências da EMB, entretanto, a forte exalação de solvente inviabilizou seu manuseio do PU3070. 2. Na fase II foram confeccionadas chapas com fibras de Pinus caribaea e Eucalyptus grandis com uma nova síntese de adesivo, denominada neste trabalho de PU7030. A caracterização física e mecânica determinou valores de resistências compatíveis com as exigências da EMB. Entretanto, verificou-se a necessidade de se ajustar o teor de umidade a temperatura e a pressão de prensagem. 3. Na fase III novas chapas foram confeccionadas com fibras de Pinus caribea e Eucalyptus grandis modificando-se o teor de umidade, a temperatura e a pressão de prensagem. Com o resultado da caracterização física e mecânica, observou-se que as alterações nas umidades, temperatura e pressão de prensagem propiciaram resistências compatíveis com a EMB, com vantagens de utilização de menores teores de PU7030 e diminuição na temperatura de prensagem, representando economia no consumo de energia. Outro aspecto muito importante estudado neste trabalho foi a caracterização das chapas utilizando ensaios não-destrutivos. Neste caso foram realizadas medições, por meio de equipamento de ultra-som, utilizando-se transdutores de faces exponenciais e planas aplicados nas direções x, y e z das chapas. A partir das medições dos tempos de propagação das ondas ultra-sônicas determinaram-se as velocidades das ondas e as constantes dinâmicas das chapas. Estas variáveis foram comparadas com as propriedades de resistência das chapas obtidas nos ensaios estáticos. Em seguida desenvolveu-se análise estatística buscando-se avaliar as correlações entre os resultados de ensaios não-destrutivos e destrutivos tendo sido possível concluir que a utilização dos métodos de ensaios não-destrutivos é viável para inferir sobre as propriedades físicas e mecânicas das chapas. Com relação à utilização da PU7030, concluiu-se que essa poliuretana possibilita a substituição da resina fenólica na confecção das chapas MDF / Abstract: While employing low cost reforestation lumber, the technologies of the fabrication of fiber plates denominated ¿Medium Density Fiberboard¿ (MDF) offer to several industrial sectors the opportunity to aggregate worth to species with little or no commercial interest, opening to them the possibility of promising profitable investments. However, the phenolic resins used in the production of MDF plates are well-known pollutants to the environment and therefore harmful to the human beings in such a way that environmentally friendly alternatives have been eagerly sought. According to that, the present work has been aimed to evaluate the physical, chemical and mechanical properties of MDF manufactured with Polyurethane Adhesive (PU) derived from Castor Oil, due to the low toxicity and to the non-polluting characteristics of the resulting adhesive. The characterization of the plates has been accomplished with the help of Steinkamp BP7 ultrasonic equipment using 45 kHz transducers and the physical and mechanical tests recommended by Euro MDF Board - EMB. The work has been developed along three distinct phases. Initially a better understanding of the characteristics of the polyurethane adhesive derived from castor oil has been acquired. It has been learnt that the adhesive exhibits the main characteristic of a monocomponent synthesis, stabilized by air humidity. During that phase, plates with 5% and 10% of polyurethane adhesive have been manufactured, called Adhesive PU3070 in this work, corresponding to 30% of solids and 70% of solvents. The properties of the plates manufactured with this synthesis have been determined and it has been found that although they have met EMB demands, the strong exhalation of solvent has ruled out its usage for MDF manufacture. As for the second phase, fiber plates have been produced with a new synthesis, called Adhesive PU7030, that is, 70% of solids and 30% of solvents. Again mechanical properties have satisfied EMB regulations. However, during the tests, it has been noticed that in order to control the quality of the final product, a fine adjusting to the values of some important Variables in the manufacture process, such as humidity level, temperature and pressing pressure, would be necessary. Accordingly, in the last phase, using Pinus caribea as fiber material, three experimental conditions have been elaborated, one for each controlled Variable, and repeated again for Eucalyptus gandis. Resulting from the adjusting of the variables, it has been observed that during the manufacture process, plates using much less adhesive than the former standards still have satisfied EMB demands fully, with advantages of a significant decrease in the pressing temperature and the corresponding energy savings. The characterization of the plates using non-destructive tests has been an important feature of the present work. Measurements have been performed in the plates by means of ultrasound equipment, using exponential as well as plane face transducers. From the readings, the propagation times of the ultrasonic waves have been evaluated, allowing to the determination of the propagation velocity of the ultrasonic waves through the material and consequently its dynamic parameters, which have been correlated to the mechanical properties of the plates, by means of a suitable statistic model. It has been concluded from the results that non-destructive test methods could as well be employed for the characterization of the physical and mechanical properties of the plates / Doutorado / Construções Rurais / Doutor em Engenharia Agrícola

Abstract Barkhausen noise measurement is an intriguing non-destructive testing method suitable for ferromagnetic materials. It is based on the stochastic movements of magnetic domain walls when the tested sample is placed in an external varying magnetic field. Barkhausen noise is typically utilised so that some features are calculated from the signal and then compared with the studied material properties. Typical features are, for example, the root-mean-square value (RMS), peak height, width and position. Better utilisation of the method, however, requires quantitative predictions of material properties. The aim of this thesis is to study and select a suitable methodology for the quantitative prediction of material properties based on Barkhausen noise measurement. The prediction considered is divided into four steps: feature generation, feature selection, model identification and model validation. In feature generation, a large set of features is calculated with different mathematical procedures. This feature set is explored in the feature selection step to find the most significant features in terms of predictions. A model with the selected features is identified and some independent data are usually used for model validation. This thesis presents the developed procedures required in feature generation and the results of the studies using different feature selection strategies and modelling techniques. The studied feature selection methods are forward selection, simulated annealing and genetic algorithms. In addition, two-step algorithms are investigated where a pre-selection step is used before the actual selection. The modelling techniques used are multivariable linear regression, partial least squares regression, principal component regression and artificial neural networks. The studies also consider the use and effect of different objective functions. The results of the studies show that the proposed modelling scheme can be used for the prediction task. The models identified mainly include reasonable terms and the prediction accuracy is fairly good considering the challenge. However, the application of Barkhausen noise measurement is very case-dependent and thus conflicts may occur. Furthermore, the changes in unmeasured material properties may lead to the unexpected behaviour of some features. The results show that linear models are adequate for capturing the major interactions between material properties and Barkhausen noise but indicate that the use of neural networks would lead to better model performance. The results also show that genetic algorithms give better selection results but at the expense of the computational cost. / Tiivistelmä Barkhausen-kohina-mittaus on ferromagneettisille materiaaleille soveltuva materiaalia rikkomaton testausmenetelmä. Mittaus perustuu magneettisten alueiden välisten rajapintojen stokastisiin liikkeisiin, kun testattava kappale asetetaan vaihtuvaan magneettikenttään. Tyypillisesti Barkhausen-kohina-mittaussignaalista lasketaan piirteitä, joita sitten verrataan tutkittaviin materiaaliominaisuuksiin. Usein käytettyjä piirteitä ovat signaalin keskineliön neliöjuuri (RMS-arvo) sekä piikin korkeus, leveys ja paikka. Menetelmää voidaan soveltaa paremmin, jos tutkittavia materiaaliominaisuuksia voidaan ennustaa kvantitatiivisesti. Tämän tutkimuksen tavoitteena on tutkia ja valita menetelmiä, jotka soveltuvat materiaaliominaisuuksien kvantitatiiviseen ennustamiseen Barkhausen-kohina-mittauksen perusteella. Ennustusmallit luodaan neljässä vaiheessa: piirteiden laskenta, piirteiden valinta, mallin identifiointi ja mallin validointi. Piirteiden laskennassa yhdistellään erilaisia matemaattisia laskutoimituksia, joista tuloksena saadaan suuri joukko erilaisia piirteitä. Tästä joukosta valitaan ennustukseen soveltuvimmat piirteiden valinta -vaiheessa. Tämän jälkeen ennustusmalli identifioidaan ja viimeisessä vaiheessa sen toimivuus todennetaan riippumattomalla testausaineistolla. Väitöskirjassa esitetään piirteiden laskentaan kehitettyjä algoritmeja sekä mallinnustuloksia käytettäessä erilaisia piirteiden valintamenetelmiä ja mallinnustekniikoita. Tutkitut valintamenetelmät ovat eteenpäin valinta, taaksepäin eliminointi, simuloitu jäähtyminen ja geneettiset algoritmit. Väitöskirjassa esitellään myös kaksivaiheisia valintamenettelyjä, joissa ennen varsinaista piirteiden valintaa suoritetaan esivalinta. Käytetyt mallinnustekniikat ovat monimuuttujaregressio, osittainen pienimmän neliösumman regressio, pääkomponenttiregressio ja neuroverkot. Tarkasteluissa huomioidaan myös erilaisten kustannusfunktioiden vaikutukset. Esitetyt tulokset osoittavat, että käytetyt menetelmät soveltuvat materiaaliominaisuuksien kvantitatiiviseen ennustamiseen. Identifioidut mallit sisältävät pääasiassa perusteltavia termejä ja mallinnustarkkuus on tyydyttävä. Barkhausen-kohina-mittaus on kuitenkin erittäin tapauskohtainen ja täten ristiriitoja kirjallisuuden kanssa voidaan joskus havaita. Näihin ristiriitoihin vaikuttavat myös ei-mitattavat muutokset materiaaliominaisuuksissa. Esitetyt tulokset osoittavat, että lineaariset mallit kykenevät ennustamaan suurimmat vuorovaikutukset materiaaliominaisuuksien ja Barkhausen-kohinan välillä. Tulokset kuitenkin viittaavat siihen, että neuroverkoilla päästäisiin vielä parempiin mallinnustuloksiin. Tulokset osoittavat myös, että geneettiset algoritmit toimivat piirteiden valinnassa paremmin kuin muut tutkitut menetelmät.

L’objectif de cette thèse était de concevoir, fabriquer, et tester des transducteurs ultrasonores multi – éléments pour des applications en environnements sévères. Dans ce contexte, des techniques de caractérisation des matériaux ont été développées afin d’adapter la conception des transducteurs en fonction des contraintes de l’environnement, et plus particulièrement la température. Deux types de conceptions ont été étudiées selon les applications visées : La première application consistait à développer un transducteur ultrasonore permettant de réaliser des contrôles non destructifs dans des réacteurs de génération IV (refroidi au sodium liquide) lors d’un arrêt de tranche. Dans ce cas, l’environnement sévère était caractérisé par une température de 200 °C, et la présence de sodium liquide. Des radiations pouvaient également être présentes. Pour cette application, un transducteur (nommé LiNa) sous la forme d’un projecteur ultrasonore linéaire de 20 éléments, fonctionnant à 2.5MHz a été fabriquée, puis testé en eau et en sodium liquide. La seconde application consistait à développer un transducteur ultrasonore permettant de réaliser des contrôles non destructifs en contact de pièces massives en cours de soudage. Une pièce massive est par exemple une cuve de réacteur nucléaire. Ce transducteur permet de détecter des défauts dans une soudure très rapidement après sa réalisation, et ainsi corriger directement cette dernière, sans attendre que les pièces soient refroidies. Pourcette application, un transducteur (nommé MaCo) sous forme de matrice 8 × 8 éléments, fonctionnant à 5MHz a été développé, et caractérisé de la température ambiante, jusqu’à 240 °C. Avec ce capteur, Avec ce capteur, des échos en ondes transversales ont été obtenus jusqu’à la température limite, sans refroidissement. / The goal of this thesis was to design, fabricate and test phased arrays ultrasonic transducers for harshenvironment applications. In this context, characterization methods were developed in order to adapt transducerdesign to the environment. We designed two types of ultrasonic phased array transducers according to theconsidered application:The first application aimed at developping an ultrasonic transducer for non destructive testing in IVth generationnuclear reactors (cooled with liquid sodium) during reactor outages. In this case, the harsh environment wascharacterized by a temperature of 200 °C, and the presence of liquid sodium. Radiations could also be present.For this application, a 20 elements 2.5MHz immersion linear projector (called LiNa) was fabricated then testedunder water and under liquid sodium.The second application aimed at developping an ultrasonic transducer in order to realize non destructive testingin contact with massive pieces during welding. A massive piece for example is a tank for nuclear plant reactor.The transducer will permit to quickly detect defects in a weld after its realization, and then correct it directlyif necessary, without waiting for the pieces to cool down at ambient temperatures. For this application,a 64elements 5MHz matrix transducer (named MaCo) were developed and tested from ambient to 240 °C. Thetransducer worked up to the limit temperature without any cooling system, and some typical defects weredetected.

Les matériaux composites sont largement utilisés dans l'industrie en raison de leur bonne tenue mécanique et de leur faible densité. La diversité des domaines d’application des matériaux composites donne lieu à une grande variété de modes de sollicitation et d’endommagement. De ce fait, l’évaluation de leurs propriétés et le contrôle de leur état présentent un grand intérêt industriel. Dans ce travail, une nouvelle méthode d’évaluation et de contrôle non destructif dite par sonothermographie est explorée. Cette méthode est basée sur l'analyse du champ thermique induit par des ondes ultrasonores de puissance dans les matériaux absorbants tels que les composites. Deux applications complémentaires sont étudiées, d’une part l’évaluation des propriétés thermiques du matériau et d’autre part le contrôle non destructif de structures par thermographie infrarouge. Dans ce cadre, le problème direct de la sonothermographie est résolu numériquement à partir d’un modèle par éléments finis. Ce modèle permet de simuler le champ thermique induit par la propagation d’ondes ultrasonores dans un matériau absorbant dont les propriétés sont connues. Les simulations réalisées permettent de montrer l’applicabilité de la sonothermographie à la détection de défauts. Une nouvelle approche de caractérisation thermique est également développée. Cette approche basée sur la formulation faible de l’équation de conduction de la chaleur permet une estimation robuste de la diffusivité thermique du matériau à partir du champ thermique induit par les ondes ultrasonores de puissance. Des résultats expérimentaux sont présentés pour le cas de plaques minces. / The composite materials are widely used in industry because of their high mechanical resistance and low density. The diversity of composite materials application fields gives rise to a large variety of solicitation and damage conditions. For this reason, the evaluation of their properties and their health monitoring are of great industrial interest. In this work, a new method of evaluation and non-destructive testing named sonothermography is explored. This method is based on the analysis of thermal fields induced by ultrasonic waves in absorbent materials such as composites. Two additional applications are studied: the evaluation of the thermal properties of the material and the non-destructive testing of structures by infrared thermography. In this framework, the direct problem of sonothermography is solved numerically using a model based on the finite element method. This model allows to simulate the thermal field induced by the propagation of ultrasonic waves in absorbent material whose properties are known. The simulations carried out show the applicability of the sonothermography for the detection of defects. An innovative approach for thermal characterization is also developed. This approach based on the weak formulation of the heat conduction equation allows a robust estimate of the thermal diffusivity of the material from the thermal field induced by ultrasonic waves. Experimental results are presented for thin plates.

TRAPIL est une société Française ayant à charge l'exploitation et l'entretien de pipelines d'hydrocarbures. L'entretien de pipelines enterrés nécessite le passage de racleurs équipés de sondes ultrasons réalisant une cartographie de la structure du pipeline, qui est ensuite analysée à la main afin de détecter et d'identifier les différents défauts pouvant apparaître ou évoluer.L'objectif de ce travail de thèse est d'apporter une solution algorithmique permettant d'accélérer et de compléter le travail des analystes à l'aide des méthodes modernes de traitement d'images et du signal.Notre approche suit le mode opératoire des experts et est découpée en trois partie.Tout d'abord nous réalisons une détection des soudures d'aboutage permettant de séparer le pipelines en les différents tubes qui le composent. Les signaux de sondes représentant la circonférence du tube sont regroupés et compressés dans une détection de rupture par comparaison de moyenne à court et long terme, puis les signaux résultants sont fusionnés à l'aide d'une pondération unique permettant une augmentation majeure du contraste entre bruit et soudure, offrant une détection et une localisation presque sans faille.Les tubes subissent ensuite une première segmentation visant à éliminer le plus grand nombre de pixels sains. Usant de modélisation d'histogramme des valeurs d'épaisseur par un algorithme EM initialisé pour notre problématique, l'algorithme suit un principe récursif comparable aux méthodes de type split and merge pour détecter et isoler les zones dangereuses.Enfin, Les zones dangereuses sont identifiées à l'aide d'une foret aléatoire, apprise sur un grand nombre d'exemples de défauts. Cette troisième partie est centrée sur l'étude de différentes méthodes de reconnaissance de forme appliquées à notre nouvelle problématique.Au travers de ces différentes étapes, les solutions que nous avons apportées permettent à TRAPIL un gain de temps significatif sur les tâches les plus fastidieuses du processus d'analyse (par exemple 30% sur la détection de soudures) et leur offre de nouvelles possibilités commerciales, par exemple la possibilité de fournir un pré-rapport à leur clientèle en quelques jours pendant que l'analyse manuelle est réalisée pouvant prendre plus d'un mois. / TRAPIL is a French society who is in charge of exploitation and maintenance of oil pipelines. Maintenance of buried pipeline implies the use of ultrasonic sensor-equipped devices, providing thickness and structural maps of the pipe, which are analysed by experts in order to detect and identify defects that may appear or evolve.The objective of this work is to provide an algoritmic solution allowing to accelerate and aid the experts's work with modern image and signal processing methods.Our approach follows the experts's operating mode and is divided in three sections.First, a weld detection is realized allowing to split the pipe in tubes. The signals of probes representing the circumference of the pipe are regrouped and compressed through an abrupt change detection, using short and long-term average comparison, then the resulting signals are merged using a unique weightening function allowing a massive increase of the contrast between welds and noise, offering near-perfect detection and localization.The tubes then undergoes a first segmentation aiming at eliminating a large amount of sane pixels. Using histogram modelization through an EM algorithm tuned specially for our purpose, the algorithm follows a recursive approach comparable to split and merge methods to detect and isolate dangerous areas.Finally, those dangerous areas are identified with a Random Forest, which has been learnt on a large amount of defect examples. This third part is greatly focused on the study of different pattern recognition methods applied on our new problematic.Through those different steps, the solution we brought allows TRAPIL to save a lot of time on the most tedious tasks of the analysis process (for example 30% of gain in processing time for the weld detection) and offers new commercial possibilities, like for example the possibility to provide their clients a first report in a matter of days, while the manual analysis is completed, which can take more than a month.

<div> <div> <div> <p>Composite materials have become widely used in engineering applications, in order to reduce the overall weight of structures while retaining their required strength. Due to their light weight, relatively high stiffness properties, and formability into complex shapes, discontinuous fiber composites are advantageous for producing small and medium size components. However, qualifying their mechanical properties can be expensive, and therefore there is a need to improve predictive capabilities to help reduce the overall cost of large scale testing. To address this challenge, a composite material consisting of discontinuous glass fibers in a polypropylene matrix is studied at the microstructural level through coupled experiments and simulations, in order to uncover the mechanisms that cause microvoids to initiate and progress, as well as certain fiber breakage events to occur, during macroscopic tension. Specifically, this work coupled in-situ X-ray micro computed tomography (μ-CT) experiments with a finite element simulation of the exact microstructure to enable a 3D study that tracked damage initiation and propagation, and computed the local stresses and strains in the microstructure. In order to have a comprehensive 3D understanding of the evolution of the microstructure, high fidelity characterization procedures were developed and applied to the μ-CT images in order to understand the exact morphology of the microstructure. To aid in this process, ModLayer - an interactive image processing tool - was created as a MATLAB executable, and the 3D microstructural feature detection techniques were compared to traditional destructive optical microscopy techniques. For damage initiation, this work showed how high hydrostatic stresses in the matrix can be used as a metric to explain and predict the exact locations of microvoid nucleation within the composite’s microstructure. From a damage propagation standpoint, matrix cracking - a mechanism that has been notably difficult to predict because of its apparent stochastic nature - was studied during damage propagation. The analysis revealed the role of shear stress in fiber mediated flat matrix cracking, and the role of hydrostatic stress in fiber-avoidance conoidal matrix cracking. Overall, a sub-fiber simulation and an in-situ experimental analysis provided the microstructural physical phenomena that govern certain damage initiation and progression mechanisms, further enabling the strength and failure predictions of short fiber thermoplastic composites. </p></div></div></div>

The need for reliable and cost-effective testing procedures for Additive Manufacturing (AM) is growing. In this Dissertation, the development of a new computational-experimental method based on the realization of specific testing artifacts to address this need is presented. This research is focused on one of the widely utilized AM technologies, Fused Deposition Modeling (FDM), and can be extended to other AM technologies as well. In this method, testing artifacts are designed with simplified boundary conditions and computational domains that minimize uncertainties in the analyses. Testing artifacts are a combination of thin and thick cantilever structures, which allow measurement of natural frequencies, mode shapes, and dimensions as well as distortions and deformations. We apply Optical Non-Destructive Testing (ONDT) together with computational methods on the testing artifacts to predict their natural frequencies, thermal flow, mechanical properties, and distortions as a function of 3D printing parameters. The complementary application of experiments and simulations on 3D printed testing artifacts allows us to systematically investigate the density, porosity, moduli of elasticity, and Poisson’s ratios for both isotropic and orthotropic material properties to better understand relationships between these characteristics and the selected printing parameters. The method can also be adapted for distortions and residual stresses analyses. We optimally collect data using a design of experiments technique that is based on regression models, which yields statistically significant data with a reduced number of iterations. Analyses of variance of these data highlight the complexity and multifaceted effects of different process parameters and their influences on 3D printed part performance. We learned that the layer thickness is the most significant parameter that drives both density and elastic moduli. We also observed and defined the interactions among density, elastic moduli, and Poisson’s ratios with printing speed, extruder temperature, fan speed, bed temperature, and layer thickness quantitatively. This Dissertation also shows that by effectively combining ONDT and computational methods, it is possible to achieve greater understanding of the multiphysics that governs FDM. Such understanding can be used to estimate the physical and mechanical properties of 3D printed components, deliver part with improved quality, and minimize distortions and/or residual stresses to help realize functional components.

Description of Ultrasonic Cleaning technology. Description of Non-destructive testing technology process. Design of Ultrasonic Cleaning equipment. Design of equipment for the ultrasonic cleaning machine. Design of non-destructive testing equipment (Liquid Penetrant Inspection and Magnetic Testing). Feasibility study of designed technologies and supporting processes.

The diploma thesis deals with the problematics of weld defects of cranes localisation and check. The aim is to analyse non-destructive methods, monitor the occurence of defects and their form and also to suggest the procedure of correction. The work is divided into a theoretical and practical part. In the theoretical part, there is an analysis of individual methods of non-destructive testing that are used for weld check on a particular girder. There is also an analysis of welding technologies used when making a crane girder. The practical part deals with evaluation of weld defects using two methods of non-destructive testing. Furthermore, statistics of defects is realised. In conclusion, the statistics outcome is resumed and evaluated.

La tomographie en 2D et 3D sont largement utilisée dans l’imagerie médicale ainsi que dans le Contrôle Non Destructif (CND) pour l’industrie. Dans toutes les deux applications, il est nécessaire de réduire le nombre de projections. Dans certains cas, la reconstruction doit être faite avec un nombre d’angle de projections limité. Les données mesurées sont toujours avec des erreurs (erreurs de mesure et de modélisation). Nous sommes donc presque toujours dans la situation de problèmes inversés mal posés. Le rôle des méthodes probabilistes et de la modélisation a priori devient crucial. Pour la modélisation a priori, en particulier dans les applications NDT, l’objet à l’examen est composé de plusieurs matériaux homogènes, avec plusieurs blocs continus séparés par des discontinuités et des contours. Ce type d’objet est dit continu par morceaux. L’objet de cette thèse est sur la reconstruction des objets continu ou constante par morceaux, ou plus généralement homogène par morceaux. En résumé, deux méthodes principales sont proposées dans le contexte de l’inférence bayésienne. La première méthode consiste à reconstruire l’objet en imposant que sa transformée de Haar soit parcimonieuse. Un modèle bayésien hiérarchique est proposé. Dans cette méthode, les variables et les paramètres sont estimés et les hyper-paramètres sont initialisés selon la définition des modèles antérieurs. La deuxième méthode reconstruit les objets en estimant simultanément les contours. L’objet continu par morceaux est modélisé par un modèle markovien non-homogène, qui dépend du gradient de l’objet, et le gradient dépend aussi de l’estimation de l’objet. Cette méthode est également semi-supervisé, avec les paramètres estimés automatiquement. Ces méthodes sont adaptées aux reconstructions de grande taille de données 3D, dans lesquelles le processeur GPU est utilisé pour accélérer les calculs. Les méthodes sont validées avec des données simulées et des données réelles, et sont comparées avec plusieurs méthodes classiques. / 2D and 3D X-ray Computed Tomography (CT) is widely used in medical imaging as well as in Non Destructive Testing (NDT) for industrial applications. In both domains, there is a need to reduce the number of projections. In some cases we may also be limited in angles. The measured data are always with errors (measurement and modelling errors). We are consequently almost always in the situation of ill-posed inverse problems. The role of the probabilistic methods and the prior modelling become crucial. For prior modelling, in particular in NDT applications, the object under examination is composed with several homogeneous materials, with several continuous blocs separated by some discontinuities and contours. This type of object is called the piecewise-continuous object. The focus of this thesis on the reconstruction of the picewise continuous or constant, or more generally piecewise homogeneous objects. In summary two main methods are proposed in the context of the Bayesian inference. The first method consists in reconstructing the object while enforcing the sparsity of the discrete Haar transformation coefficients of the object. A hierarchical Bayesian model is proposed. In this method, the unknown variables and parameters are estimated and the hyper-parameters are initialized according to the definition of prior models. The second method reconstruct objects while the contours are estimated simultaneously. The piecewise continuous object is modeled by a non-homogeneous Markovian model, which depends on the gradient of the object, while the gradient also depends on the estimation of the object. In this methods, the semi-supervised system model is also achieved, with the parameters estimated automatically. Both methods are adapted to the 3D big data size reconstructions, in which the GPU processor is used to accelerate the computation. The methods are validated with both simulated and real data, and are compared with several conventional state-of-the-art methods.