Tagungsband des Ohde-Kolloquiums 2014. Die Fachtagung fand am 26.03.2014 an der TU Dresden statt.

L'état de contrainte initial est l'état de contrainte dans le sous-sol avant toute intervention humaine. Sa connaissance est essentielle pour atteindre des objectifs aussi bien scientifiques (compréhension de la tectonique des plaques) que préventifs (étude et prédiction des séismes) ou industriels (compréhension de la mécanique des réservoirs pétroliers pour leur exploitation). Dans cette thèse, nous présentons une méthode permettant d'estimer l'état de contrainte initial en trois dimensions, à partir de données éparses. Cette méthode repose sur une approche inverse dans laquelle la méthode des éléments finis est utilisée pour résoudre le problème mécanique élastique. Les paramètres de modèles sont les conditions de Neumann décrites par des fonctions linéaires par morceaux. Les données sont des mesures partielles de l'état de contrainte en quelques points du domaine d'étude. Une optimisation d'ensemble est utilisée pour résoudre le problème inverse. La méthode est testée sur un cas synthétique où la solution de référence dans tout le domaine est supposée connue. Sur cet exemple, la méthode présentée est capable de retrouver un état de contrainte en accord avec les mesures, et cohérent avec l'état de contrainte de référence dans tout le domaine. La méthode est ensuite enrichie par des critères mécaniques qui contraignent l'état de contrainte dans les zones où les données sont absentes. La méthode est ensuite appliquée sur un cas réel: le bassin de Neuquèn en Argentine, sur lequel des données de puits sont disponibles. La confrontation de la méthode avec un cas d'étude permet de déterminer quelles en sont les limites. L'impact des failles sur l'état de contrainte et les différentes façons de les modéliser sont discutés. En effet, la prise en compte de ces structures complexes est problématique dans les calculs induits par les méthodes inverses puisqu'elles introduisent des non linéarités rendant le temps de calcul trop important. Nous investiguons alors la possibilité de les modéliser uniquement par des propriétés élastiques. Enfin, nous consacrons un chapitre sur l'environnement logiciel intégratif que nous avons développé pour la réalisation des calculs mécaniques. Cet environnement est composé de RINGMesh, une bibliothèque proposant une structure de données pour les modèles géologiques et de RINGMecha, permettant la réalisation de calculs mécaniques sur ces modèles. RINGMecha interface plusieurs simulateurs, chacun ayant ses spécificités, de manière totalement transparente pour l'utilisateur. RINGMecha a été utilisé pour la réalisation de tous les calculs présentés dans cette thèse et a été pensé pour pouvoir être étendu à d'autres problèmes, avec d'autres simulateurs / Initial stress state is the stress state before any human activity. Its knowledge is essential when dealing with scientific (understanding of plate tectonics), preventive (earthquake prediction) and industrial (understanding reservoirs before their exploitation) purposes. We present a method to estimate the initial stress state in a 3D domain from sparse data. This method relies on an inverse approach which uses the finite elements method to solve the elastic mechanical problem. The model parameters are Neumann conditions, which are defined as piecewise linear functions. The data parameters are stress state observations, such as intensity and orientation at a few points. An ensemble optimization method is used to solve the inverse problem. The method is tested on a synthetic case where the reference solution is known. On this example, the method succeeds in retrieving the stress state at data points as well as in the whole domain. The method is enriched with a mechanical criterion which imposes mechanical constraints in the domain under study. The method is then applied to a real case: the Neuquèn basin in Argentina where borehole stress data is available. This application reveals some of the limits of the presented method. Then, the effect of faults on the stress state is investigated. Different modeling strategies are presented: the objective is to reduce the computing cost, which can be very high when dealing with such complex structures. We propose to model them using only elastic properties. Finally, we present the integrative software which were developed to run mechanical simulations. RINGMesh handles the structural model data structure and RINGMecha runs the mechanical simulations on the model. RINGMecha is interfaced with several simulators. Each of them can be called separately, depending on the problem to be solved. The interface of RINGMecha with third party simulators is done in a user friendly manner. RINGMecha was used for all the computations presented in this thesis. It was built in order to be extended to other problems, with other simulators

Les processus de localisation de la déformation sont la clé du comportement mécanique des roches et des sols car ils sont, par exemple, souvent précurseurs de la rupture des matériaux. Par conséquent, la compréhension des processus qui conduisent à la localisation de la déformation, et donc à la rupture, dans les géomatériaux est essentielle pour la réussite d'un grand nombre de projets d'ingénierie géotechnique. Ce travail est centré sur l’étude des sols durs, roches tendres et matériaux granulaires cimentés. Dans ce cas, la localisation de la déformation est souvent associée à des phénomènes de fracturation à différentes échelles. Les macrofractures sont généralement entourées de méso et micro fractures et leur propagation est précédée d’une zone de processus; la microfissuration, qui caractérise cette zone, peut être identifiée comme endommagement car cela affecte les propriétés mécaniques du matériau. Les bandes de cisaillement ou de compaction sont souvent associées à des microfissures et elles peuvent également être considérées comme de la localisation d’endommagement. Dans le cadre de la mécanique des roches, l’endommagement peut être causé par des phénomènes de détachement (rupture du ciment) et écrasement des grains. Pour étudier les phénomènes de localisation de la déformation et de l'endommagement, une sorte de mesure de champ non destructive doit être mis en place. Des techniques bien connues dans la géomécanique comprennent la tomographie par rayons X pour étudier la structure des matériels, et la corrélation d'image numérique (DIC) des images 2D ou 3D (volumes), pour étudier le champ de déformation. La DIC est un outil très puissant pour l’étude des phénomènes hétérogènes mais fournit uniquement des données sur la cinématique. Dans cette thèse, nous suggérons l’utilisation d’un autre outil, la tomographie ultrasonore, qui permet la mesure du champ de la vitesse ultrasonique dans un échantillon, ainsi que ses propriétés élastiques, en exploitant la théorie de la propagation des ondes. La tomographie ultrasonore, utilisée comme technique complémentaire de la DIC et d'autres mesures de champ, peut donc fournir de nouvelles informations pour comprendre les processus de déformation. Dans la mécanique des sols et des roches des méthodes acoustiques et ultrasoniques ont été largement utilisées pour mesurer les propriétés élastiques, y compris pendant les essais mécaniques. Cependant, ces mesures ont été limitées à un petit nombre limitant ainsi l'étude des hétérogénéités. La tomographie ultrasonore peut surmonter cette limitation en fournissant une carte des propriétés élastiques. La principale contribution de ce travail est le développement de la tomographie ultrasonore comme technique de mesure de champ pour tester les géomatériaux et son application dans une vaste campagne expérimentale. Des tests ont été effectués sur différents matériaux et dans différentes conditions, y compris pendant le chargement et la tomographie ultrasonore a été comparée avec les résultats obtenus par DIC ainsi que par tomographie à rayons X. Une comparaison entre DIC et tomographie ultrasonore indique que la vitesse de propagation subit des modifications importantes à un niveau de chargement pour lequel la DIC ne présente aucune déformation notable. Puisque la tomographie ultrasonore est sensible aux variations des propriétés élastiques, à savoir l’endommagement, il peut être conclu que le niveau de déformation nécessaire pour causer de l’endommagement dans ce matériau est inférieur à la résolution de la DIC. De plus, la tomographie ultrasonore semble être capable de détecter la zone de processus qui précède la fracture, et donc d’indiquer sa direction de propagation. Les différentes techniques utilisées dans ce travail fournissent des informations différentes et complémentaires. Grâce à leur combinaison, une meilleure compréhension du comportement mécanique des géomatériaux peut être acquise. / The processes of strain localisation are key to the deformation behaviour of rocks and soils, because, for example, localised deformation is often precursor to material failure. Therefore the understanding of the processes leading to localised deformation, and thus to failure, in geomaterials is critical to the success of many geotechnical engineering projects. This work is focused on the analysis of localised deformation in hard soils, soft rocks and, more generally, cemented granular materials. For such materials localized deformation, in the form of localized strain, i.e., shear and compaction bands, is often associated with damage, i.e., inter and intra-granular fractures and cracks, de–bonding and breakage of particles (grain crushing). Furthermore, macrofractures are commonly surrounded by meso and micro cracks and a process zone of microfracturing precedes their propagation. To study localised phenomena such as strain and damage localisation, some kind of non-destructive, full–field measurement has to be used. Well-known techniques in geomechanics include X-ray tomography, to study material structure, and Digital Image Correlation (DIC) of 2D or 3D (volume) images, to study material kinematics and strain fields. DIC has proven to be a very powerful tool in the study of heterogeneous phenomena, but provides only data on kinematics and strain and not on associated property changes (e.g., elastic properties). In this thesis another tool, ultrasonic tomography, is suggested as a full-field measurement of the elastic property variations in test specimens through mapping of ultrasonic wave propagation velocities. Ultrasonic tomography, as complementary technique to DIC and other full-filed measures, can thus provide new insight into the deformation processes. In rock and soil mechanics, acoustic and ultrasonic methods have long been used to measure the elastic properties of test materials, including during mechanical testing. However, such measures have generally been limited to only a few measurement paths (usually just one) for a whole sample, thus restricting the study of heterogeneity. Ultrasonic tomography can overcome this limitation to provide a full-field measure. The main contributions of this work are the development of ultrasonic tomography analysis for laboratory geomechanics (both in terms of the experimental method and subsequent data analysis) and its application to analyse material deformation and, in particular, material evolution during loading (time-lapse ultrasonic tomography). The developed ultrasonic tomography approach has been applied to investigate geomaterial behaviour in laboratory tests. In this context, experimental campaigns have been carried out on different materials, where the ultrasonic tomography has been complemented by comparisons with displacement and strain fields from 2D and 3D DIC plus structural analysis by X-ray tomography. A comparison between DIC and ultrasonic tomography results reveals that the latter shows important changes inside the sample in a stage of loading where the DIC may not. As ultrasonic tomography is sensitive to damage, it can be concluded that the degree of deformation needed to cause detectable damage is below the resolution of the DIC. Moreover, the ultrasonic tomography seems to be able to detect the damaged zone surrounding fracture tips and thus can indicate where the fractures will propagate. However, when deformation becomes too high or fractures propagate, ultrasonic signals cannot be acquired, so ultrasonic tomography is not possible; in such situations DIC can still provide important information on the deformation mechanisms. The different full-field techniques employed in this work have thus been found to provide different and complementary information. Furthermore, it is shown that better understanding of the mechanical behaviour of geomaterials can be gained through the combination of more than one technique.

Orientador: Denis Jose Schiozer / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecanica e Instituto de Geociencias / Made available in DSpace on 2018-08-12T08:32:24Z (GMT). No. of bitstreams: 1 MontoyaMoreno_JuanManuel_M.pdf: 2098742 bytes, checksum: c03cbaf7c1d95c189a074da09d33d7dd (MD5) Previous issue date: 2007 / Resumo: A injeção de água tem sido um método de recuperação de petróleo muito usado na industria do petróleo. Um dos principais problemas associados a este método e a perda da injetividade. Diferentes procedimentos tem sido propostos para minimizar os efeitos de tal perda, um dos quais e a injeção acima da pressão de fratura, que consiste em injetar água com pressão suficiente para criar canais de alta condutividade, com o objetivo de restaurar a injetividade inicial do poço. Este procedimento tem se mostrado viável, tendo como vantagem à manutenção da injetividade em valores similares aos iniciais, embora apresente dificuldade de modelagem e incertezas em relação à eficiência de varrido, ao comportamento da frente de avanço da fratura e a influencia desta na recirculação da água. Uma forma adequada de modelar esse fenômeno e a integração da geomecanica com a simulação de escoamento, mas este tipo de abordagem envolve maior custo computacional, alem de não ter sido completamente implementado em aplicações comerciais, o que dificulta aplicações praticas. Outras diferentes técnicas de simulação são usadas para modelar o fenômeno, tais como modificadores de transmissibilidade associados a refinamentos locais de malha. Neste trabalho, e proposta a simulação da fratura por meio do uso de um poço horizontal virtual, usando um simulador comercial de escoamento e um software para simulações de processos de faturamento hidráulico acoplados explicitamente. De inicio, modela-se a perda de injetividade num poço injetor de água e, junto com o modelo geomecanica de fratura, usam-se modificadores de transmissibilidade para modelar a fratura como modelo de referencia. Posteriormente, introduz-se um modelo de fratura representado por um poço horizontal virtual mediante programação em simuladores comerciais. Os resultados mostram que o poço horizontal virtual representa adequadamente o comportamento da fratura em relação ao modelo de modificadores de transmissibilidade, evitando necessidade de refinamentos locais, tornando-se uma ferramenta útil para simular casos de campo em grande escala. Palavras-Chave: simulação de reservatório; injeção de água; perda de injetividade; injeção de água acima da pressão de fratura; geomecanica. / Abstract: Water injection has been the most used method to improve oil recovery. The main problem of this method, related directly with operational efficiency, is the injectivity loss which is the loss of capacity to maintain water injection rates due to formation damage. Different models have been proposed to minimize the injectivity loss effects due to its high economic impact on oil production. One of these methods is water injection above formation parting pressure. It consists in creating high conductivity channels inside the reservoir to restore or to increase well injectivity. This method has advantages associated with injectivity maintenance but it is difficult to model and it presents uncertainties in relation to sweep efficiency, to fracture tip behavior and its influence on the water re-circulation. One way to model the problem is the integration of geomechanical and flow simulation, but it requires a higher computation time and it has not been completely implemented in commercial simulators. Different techniques are used to model the water injection above formation parting pressure, such as transmissibility modifiers associated with local grid refinement. In this work, a methodology is proposed for fracture simulation using both numerical fluid flow and geomechanical simulators, coupled explicitly, using a virtual well to model the fracture. Initially, injectivity loss is modeled and, along with geomechanical fracture model, transmissibility modifiers are used to model the fracture. This model represents the reference fracture model. Then, the fracture is represented by a virtual horizontal well, allowing easy implementation into commercial simulators. The results show that the virtual horizontal well represents adequately the fracture's behavior given by the reference model, avoiding local grid refinement and, allowing full field scale simulations without simulation grid modification. / Mestrado / Reservatórios e Gestão / Mestre em Ciências e Engenharia de Petróleo

Tagungsband des Ohde-Kolloquiums 2014. Die Fachtagung fand am 26.03.2014 an der TU Dresden statt.:Experimentelle Untersuchung der Kapillarität bei Sand unter monotoner und zyklischer Belastung, Marius Milatz Mehrphasen-Modell zur Simulation von Suffosion, Heike Pfletschinger-Pfaff, Jan Kayser, Holger Steeb Experimentelle Ermittlung intergranularer Kräfte unter Nutzung von 2D-DIC, Max Wiebicke, Edward Andò, Denis Caillerie, Gioacchino Viggiani Systeme paralleler Scherbänder - Experimentelle und analytische Untersuchungen, Lars Röchter Rechnerischer Stabilitätsnachweis für verflüssigungsgefährdete Standorte, Nándor Tamáskovics Untersuchung des Einflusses von Gaseinschlüssen unterhalb des Grundwasserspiegels auf Druckausbreitung und Bodenverformungen mittels gekoppelter FE-Berechnungen, Hector Montenegro, Oliver Stelzer Zeitabhängige Setzungen von Sand und FE-Simulationen einer Tagebaukippe, Stefan Vogt, Emanuel Birle, Gero Vinzelberg Über die Berücksichtigung großer Bodendeformationen in numerischen Modellen, Daniel Aubram Die Gefrierkernmethode - Weiterentwicklung des Erkundungsverfahrens zur geohydraulischen Charakterisierung von Sohlsedimenten, Daniel Straßer, Hermann-Josef Lensing, Dominik Richter, Simon Frank, Nico Goldscheider Nutzung von Verfahren der Bildanalyse zur Baugrundbeurteilung, Markus Wacker, Thomas Neumann, Jens Engel, Gunter Gräfe Anwendung von Elektroosmose zur Reduzierung des Herausziehwiderstandes von Spundwänden: Großmaßstäbliche Modellversuche in Ton, Christos Vrettos, Kai Merz Zementfiltration bei der Herstellung von Verpressankern in nichtbindigen Böden, Xenia Stodieck, Thomas Benz Modell- und Elementversuche zur Bodenverflüssigung, Erik Schwiteilo, Ivo Herle Dynamische Probebelastung einer Mikropfahlgründung - Feldversuch und dynamische 3D-FE-Simulation mittels Hypoplastizität, Thomas Meier, Jens Jähnig, Sina Meybodi Numerische und analytische Berechnungen zur Erdbebenbemessung von Böschungen, Hassan AlKayyal

The boundary value problems (BVPs) of geomechanics are challenging due to the complexity in modeling soil behavior and difficulties in modeling large deformations. While traditional numerical schemes have struggled in realistically simulating geomechanical BVPs, new numerical methods –such as the material point method (MPM)–are increasingly being used to tackle these problems. However, algorithms in MPM have not yet been sufficiently developed, scrutinized, and validated. This thesis focuses on the development, verification, and validation of MPM for use in geomechanical BVPs. In particular, the thesis focuses on simulation of cone penetration tests in both controlled environments and in field conditions.<div><br></div><div>To efficiently simulate cone penetration, a silent boundary scheme, known as a cone boundary, was proposed in the generalized interpolation material point method (GIMP), a variant of MPM. The implementation of the cone boundary in GIMP was discussed, and the boundaries were validated by comparison against several benchmark problems. The cone boundaries were shown to be suitable in transmitting energy at the boundary. In addition, the implementation of traction boundaries in GIMP was analyzed. In GIMP, traction boundaries may be implemented either at the centroid of the material point, or at the edge of the material point domain. It was shown that the implementation of traction boundaries at the edge of the domain led to stress oscillations near the boundary, which were minimized when the traction boundaries were implemented at the edge of the domain.<br></div><div><br></div><div>During cone penetration, the soil near the cone-soil interface is pushed to large strains. At large strains, soils reach critical state, a state in which the soil shears at constant volume. Simulation of incompressible materials using low-order shape functions commonly used in GIMP leads to stiffer solutions and stress oscillations. To mitigate the constraints imposed by incompressibility, the non-linear B-bar method was implemented in GIMP. The modifications required for the implementation of the B-bar method in GIMP were discussed, and the efficacy of the method in mitigating incompressibility was demonstrated by analyzing several benchmark problems.<br></div><div><br></div><div>To simulate cone penetration in saturated soil, a coupled formulation was proposed in GIMP.A single material point was used to represent both the soil matrix and water. The governing equations were solved using an explicit scheme with the velocity of the soil matrix and the velocity of water as the primary variables. The formulation was validated through problems for which analytical or numerical solutions are available.<br></div><div><br></div><div>Finally, cone penetration analyses were performed both in dry sand and saturated clays. Two bounding surface models –one for sand and one for clay –were used for accurately capturing the soil response. Cone penetration tests were performed on Ottawa 20-30 sand under a variety of loading conditions at a large calibration chamber. The penetration resistances were measured, and the displacement fields were captured using the digital image correlation technique(DIC). The cone penetration resistances predicted by MPM were within 25% of the measured values, and the displacement fields computed using MPM were similar to those captured using DIC. For saturated clays, cone penetration test results reported in the literature for a Boston Blue Clay (BBC) test site were used. The simulated cone resistance of 650 kPa lied within the CPT resistance range of 580-730 kPa reported in the field. The results demonstrate the capability of MPM in simulating cone penetration in both sands and clays provided that sufficiently accurate algorithms and advanced constitutive models capable of reproducing realistic soil behavior are used in the analyses.<br></div>

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第18933号 / 工博第3975号 / 新制||工||1612(附属図書館) / 31884 / 京都大学大学院工学研究科社会基盤工学専攻 / (主査)教授 木村 亮, 教授 勝見 武, 准教授 木元 小百合 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

<p>Crustal stress measurements play a crucial role in understanding how the subsurface deforms. As one of the most popular methods for stress characterization in deep wellbores, borehole breakout analysis examines the shape of drilling-induced compressive failures to determine stress directions and magnitudes, assuming that the rock formation is both isotropic and linearly elastic. To ensure accurate stress interpretations, the dissertation investigates the validity of underlying presumptions from two perspectives: (1) the effect of rock anisotropy (i.e., elastic anisotropy, and strength anisotropy) on wellbore failure patterns; and (2) the characterization of rock nonlinear elastic mechanical behaviors. </p> <p>The developed computer program, <em><strong>EASAfail</strong></em>, has broad applicability in calculating wellbore failure patterns for a wide range of scenarios. It takes into account factors such as elastic stiffness matrices of the rock, stress tensors in the surrounding environment, and the presence of weak planes. The program's generality allows it to handle various rock types with different degrees of symmetry in their elastic properties, as well as weak planes that are weaker than the intact rock matrix. By analyzing these factors, the program reveals that the patterns of wellbore failure in elastic and strength anisotropic rock formations are highly influenced by the sliding of weak planes. Complications from two modes of borehole failure, either in the intact rock matrix or in the weak planes, can cause the breakout azimuth to deviate from the direction of the minimum horizontal stress. </p> <p>In addition to hypothetical scenarios generated from numerical models, a case study from the field is presented to underscore the impact of foliations on the anomalous rotations of breakout azimuths. The wellbore was located in Northeastern Alberta, Canada, transecting both the sedimentary column and crystalline basement. Breakout rotations identified from caliper and image logs were highly likely caused by the slippage along foliations, supported by the close correlation between breakout azimuths and dip directions of foliations as well as polarization directions analyzed from dipole sonic logs. Stress magnitudes constrained from Monte Carlo simulations further reveal a lower stress field when rock anisotropy is taken into account, compared to what is inferred conventionally. </p> <p>The characterization of rock nonlinear elasticity involves the utilization of the third-order elastic (TOE) model. To measure the TOE moduli in a static manner, test-specific protocols were proposed based on the nonlinear stress-strain behaviors of the rock. By arranging the stress-strain responses obtained from hydrostatic, uniaxial, and triaxial compressive tests into a linear system of equations, it becomes possible to invert the equations for the TOE moduli. These analytical equations were validated through calculations from finite element models. </p> <p>By employing the established protocols, the TOE moduli were derived for four different rock types with varying pore structures when subjected to hydrostatic and uniaxial compressions. The TOE model successfully captured the nonlinear stress-strain responses exhibited by Indiana limestone, Vif-type Fontainebleau sandstone, and Snake River Plain basalt. However, it was found to be inadequate for Franc-type Fontainebleau sandstone, which displayed noticeable hysteresis and experienced significant strains. Future geomechanical applications will undoubtedly gain advantages from utilizing the inverted TOE moduli obtained through static measurements, as they allow for the examination of the impacts of nonlinear elasticity in rocks. </p>

Heavy oil production depends on the understanding of mechanical and flow properties of unconsolidated or weakly consolidated sands under different loading paths and boundary conditions. Reconstituted bitumen-free Athabasca oil-sands samples were used to investigate the geomechanics of a steam injection process such as the Steam Assisted Gravity Drainage (SAGD). Four stress paths have been studied in this work: triaxial compression, radial extension, pore pressure increase and isotropic compression. Absolute permeability, end-point relative permeability to oil & water (kro and krw), initial water saturation and residual oil saturation were measured while the samples deformed. Triaxial compression is a stress path of increasing mean stress while radial extension and pore pressure increase lead to decreasing mean stress. Pore pressure increase experiments were carried out for three initial states: equal axial and confining stresses, axial stress greater than confining stress and confining stress greater than axial stress. Pore pressure was increased under four boundary conditions: 1) constant axial and confining stress; 2) constant axial stress and zero radial strain; 3) zero axial strain and constant confining stress; and 4) zero axial and radial strain. These experiments were designed to mimic geologic conditions where vertical stress was either S1 or S3, the lateral boundary conditions were either zero strain or constant stress, and the vertical boundary conditions were either zero strain or constant stress. Triaxial compression caused a decrease in permeability as the sample compacted, followed by appreciable permeability enhancement during sample dilation. Radial extension led to sample dilation, shear failure and permeability increase from the beginning. The krw and kro increased by 40% and 15% post-compaction respectively for the samples corresponding to lower depths during triaxial compression. For these samples, residual oil saturation decreased by as much as 40%. For radial extension, the permeability enhancement decreased with depth and ranged from 20% to 50% while the residual oil saturation decreased by up to 55%. For both stress paths, more shear-enhanced permeability was observed for samples tested at lower pressures, implying that permeability enhancement is higher for shallower sands. The pore pressure increase experiments showed an increase of only 0-10% in absolute permeability except when the effective stress became close to zero. This could possibly have occurred due to steady state flow not being reached during absolute permeability measurement. The krw curves generally increased as the pore pressure was increased from 0 psi. The increase ranged from 5% to 44% for the different boundary conditions and differential stresses. The kro curves also showed an increasing trend for most of the cases. The residual oil saturation decreased by 40-60% for samples corresponding to shallow depths while it increased by 0-10% for samples corresponding to greater depths. The reservoirs with high differential stress are more conducive to favorable changes in permeability and residual oil saturation. These results suggested that a decreasing mean stress path is more beneficial for production increase than an increasing mean stress path. The unconsolidated sands are over-consolidated because of previous ice loading which makes the sand matrix stiffer. In this work, it was found that over-consolidation, as expected, decreased the porosity and permeability (40-50%) and increased the Young’s and bulk moduli of the sand. The result is sand which failed at higher than expected stress during triaxial compression. Overall, results show that lab experiments support increased permeability due to steam injection operations in heavy oil, and more importantly, the observed reduction in residual oil saturation implies SAGD induced deformation should improve recovery factors. / text

L’objectif de ce travail est double : d’une part, il s’agit de développer une méthodologie intégrée pour la construction d’un modèle géomécanique ainsi que la représentation des incertitudes associées aux propriétés poro-élastiques des roches constitutives, en exploitant l’ensemble des données disponibles et en s’appuyant de façon cohérente sur les modèles de réservoir statique et dynamique classiquement utilisés par les géologues et les ingénieurs réservoir ; d’autre part, il s’agit d’analyser quel est l’impact des hétérogénéités géologiques, souvent négligées, dans la réponse mécanique du réservoir sollicité par son exploitation, et d’aboutir à des incertitudes sur les champs de contraintes et de déplacements, issues des incertitudes sur ces hétérogénéités et leurs paramètres mécaniques. Pour ce faire, une méthodologie intégrée s’appuyant sur des simulations géostatistiques a été développée. Après la construction du cadre géométrique 3D, le remplissage des propriétés au sein du réservoir suit une démarche de simulations géostatistiques 3D emboîtées, dans laquelle la représentation des hétérogénéités lithologiques conditionne la génération des propriétés poro-élastiques. La démarche consiste ensuite à représenter les incertitudes sur le modèle géomécanique par des ensembles de réalisations géostatistiques dont la réponse mécanique est alors calculée avec un simulateur mécanique aux éléments finis. Les incertitudes sur les champs de contraintes et de déformations sont déduites ensuite des différentes réponses mécaniques obtenues. La démarche a été mise en œuvre sur un réservoir réel, dans un environnement fluvio-deltaïque, produisant en Mer du Nord. Dans ce cadre, il a été démontré que les hétérogénéités du réservoir et leurs incertitudes influencent significativement les calculs des champs de contraintes et de déformations, ainsi que les risques mécaniques de rupture. Des incertitudes sur les quantités mécaniques analysées (premier invariant du tenseur des contraintes et subsidence) ont été aussi estimées / This work has two main objectives. The first one is to develop an integrated methodology allowing to build a 3D geomechanical model and also to image the uncertainties attached to the poro-mechanical properties of the constitutive rocks. This geomechanical model should be based on all related available data and should be consistent with the static and dynamic models, currently built by reservoir geologists and engineers. The second objective is to analyse the impact of geological heterogeneities, which are often neglected, in the mechanical response of the reservoir induced by its exploitation, and furthermore to derive uncertainties on the stress and deformation fields related to the uncertainties on the input properties of the geomechanical model. An integrated methodology based on geostatistical simulations is developed. First, the geometric frame is built; then an approach of embedded stochastic simulations is carried out to infill the different reservoir properties, the lithological description constraining the petrophysical and poro-elastic descriptions. The next step is to generate the mechanical responses of the stochastic realisations, using a finite-element mechanical simulator. The uncertainties on the resulting stress and displacement fields are then deduced from the multiple mechanical responses which are computed. This approach is demonstrated on a real field case, a fluvio-deltaic reservoir in North Sea. It is shown on this example that the reservoir heterogeneities and their uncertainties significantly influence the calculations of stress and strain fields, and also the risks of mechanical failure. Uncertainties on the mechanical quantities under analysis (first invariant of the stress tensor and subsidence) are also derived