Orientadores: Sandro Guedes de Oliveira, Lúcia Duarte Dillon / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Física Gleb Wataghin / Made available in DSpace on 2018-08-17T17:53:23Z (GMT). No. of bitstreams: 1 Abreu_ElitaSelmarade_M.pdf: 3986219 bytes, checksum: 3254aa4fe691af01904819c1fb348ada (MD5) Previous issue date: 2010 / Resumo: O principal propósito dessa dissertação é estudar modelos de meio efetivo de física de rochas que conecte as propriedades petrofísicas e as propriedades elásticas, assim como a sua aplicação na investigação dessas propriedades em rochas carbonáticas. Inicialmente será feita uma introdução a alguns modelos de física de rochas para meio efetivo, conhecidos como modelo de Voigt-Reuss-Hill, modelo de Kuster & Toksöz, modelo Diferencial de Meio Efetivo e relação de Gassmann, com objetivo de estabelecer os parâmetros que serão medidos e utilizados no desenvolver do trabalho. Após essa parte introdutória, baseado no modelo de Xu-Payne, foram realizadas uma série de análises de atributos geométricos, como a distribuição de tipos de poros, obtidas através de lâminas petrográficas com intuito de descrever a correlação entre as propriedades petrofísicas e elásticas e assim poder calibrar o modelo teórico utilizado na predição dessas propriedades. Dessa forma, o modelo calibrado passa a desempenhar um papel mais condizente com o sistema poroso da rocha permitindo uma melhor correlação entre os parâmetros elásticos e petrofísicos. Os resultados obtidos mostram que a utilização da informação de lâminas petrográficas, na parametrização do modelo, torna o método mais robusto na predição e conexão das propriedades elásticas e petrofísicas de rochas carbonáticas, tornando confiável a mudança de escala rocha-perfil, bem como possibilitando a predição qualitativa de propriedade permo-porosas a partir da velocidade da rocha / Abstract: The main purpose of this dissertation is to study rock physics effective models that connect the petrophysics and elastic properties as well as its application on the investigation of these properties on carbonate rocks. Firstly, we make an introduction to some rock physics of effective models as: Voig-Reuss-Hill, Kuster&Toksöz, Differential Effective Medium, Gassmann¿s Relation, aiming at establishing the parameters that will be measured and used latter. After this introductory part and based on the Xu-Payne model, several geometric factors analysis was done like pore types distribution, obtained by thin sections, with the intention of describing the correlation between the petrophysics and elastic properties. In this way, the model becomes more compatible with the rock porous medium, allowing a better correlation between the petrophysics and elastic parameters. Our results show that using the thin section information on the model parametrization, the predictability and connectivity of petrophysics and elastic properties applied to carbonate rocks become more robust, making trustable the upscale rock-well log and also enabling the permo-porosity properties prediction, in a qualitative way, through the velocity measurements / Mestrado / Física / Mestra em Física

Les argilites, sont définies comme étant des roches comportant une large fraction de minéraux argileux. Leurs propriétés physiques et mécaniques présentent un intérêt pour l’étude du comportement hydro-mécanique des failles dans la partie supérieure de la croûte mais aussi pour la compréhension des roches couvertures de réservoirs d’hydrocarbures ou pour l’expertise de la pérennité du stockage de déchets radioactifs.Cette thèse propose deux approches afin de comprendre l’organisation de la déformation dans ce type de matériau, une première purement mécanique sur des échantillons intacts et la seconde s’intéressant aux propriétés physiques de matériaux déformés. Pour cela, nous avons étudié les argilites de Tournemire (Tunnel expérimental de l'IRSN, Aveyron, France).La première partie consiste en une série d’essais triaxiaux. Nous avons déterminé les enveloppes de rupture de trois groupes d’échantillons carottés avec des orientations différentes par rapport au litage (0°, 45°, et 90°). Pour chaque orientation, sept expériences ont été réalisées à différentes pressions de confinement (2.5, 5, 10, 20, 40, 80, 160 MPa). L’influence de la vitesse de déformation a été établie en comparant des expériences réalisées avec des vitesses de déformation différentes (10-7 s-1 et 10-5 s-1). Pendant les expériences, les vitesses d’ondes P et S ont été enregistrées selon différents angles par rapport au litage afin de quantifier l’évolution de l’anisotropie des propriétés élastiques.Cette partie permet de mettre en évidence l’importance de l’orientation du litage par rapport à la contrainte principale sur la résistance mécanique de nos échantillons. De plus, un modèle micromécanique basé sur le « wing crack » permet d’expliquer l’anisotropie mécanique de nos argilites par l’anisotropie de la ténacité KIC.Par ailleurs, on constate que l’évolution de l’anisotropie des propriétés élastiques dépend elle aussi de l’orientation considérée. Lors de la compression, l’orientation 90° présente d’importantes variations pouvant aller jusqu’à une inversion de l’anisotropie, alors que les vitesses n’évoluent que très peu pour l’orientation 0°. Ces variations ont été quantifiées par les paramètres de Thomsen. L’étude des vitesses élastiques et celle des microstructures, permettent de mettre en évidence l’importance des processus plastiques comme la réorientation des minéraux au cours de la déformation.La seconde partie consiste en une étude pétrophysique (vitesses des ondes P, ASM, densité, saturation, porosité) d’échantillons provenant d’un forage traversant une zone de faille. Le but est de quantifier la variation de ces propriétés à l’approche du cœur de faille.Un protocole d’échantillonnage et de mesure a été mis en place. Le protocole comporte une première série de mesures directement sur le terrain afin de s’affranchir des problématiques liées à la préservation des échantillons (notamment pour les mesures de porosité/densité/saturation). Par la suite, des échantillons ont été prélevés pour réaliser des mesures à la fois dans le cadre de cette thèse (vitesses des ondes P et ASM) mais aussi dans d’autres laboratoires (étude de la composition minéralogique, CT-scan).A partir des observations, on caractérise :• Une zone saine caractérisée par des échantillons ne présentant pas ou très peu de fracturation• Une zone endommagée qui présente un grand nombre de fractures calcifiées• Une zone de cœur caractérisée par une déstructuration totale (pas de bedding apparent) et des variations de couleur.Notre étude met en évidence une signature physique propre à chaque zone de cette faille avec notamment une diminution de l’anisotropie des échantillons en zone endommagée fortement marquée. De plus, des mécanismes de rotation de la stratigraphie similaires à ceux observés à l’échelle des microstructures lors de la déformation expérimentale ont été observés. / Shales or clays are defined as rock having a large proportion of clay minerals. Their physical and mechanical properties are of interest for the study of the hydro-mechanical behavior of faults in the uppermost crust but also for the understanding of the cap rocks of hydrocarbon reservoirs or for the expertise of the durability of radioactive waste storage.This thesis proposes two complementary approaches to understand the organization of the deformation in this type of material, a first purely mechanical on undisturbed samples and the second focusing on the physical properties of deformed materials. During this thesis, we studied Tournemire shales (IRSN tunnel, Aveyron, France).The first part consists in triaxial tests. We determined the failure envelopes of three sets of core samples with different orientations with respect to bedding (0 °, 45 ° and 90 °). For each orientation, seven experiments were performed at different confining pressures (2.5, 5, 10, 20, 40, 80, 160 MPa). The influence of the strain rate was determined by comparing experiments with different strain rates extending over two orders of magnitude (between 〖10〗^(-7) s^(-1) and 〖10〗^(-5) s^(-1)). During the experiments, the P and S wave velocities were recorded from different angles with respect to the bedding to quantify the evolution of the anisotropy of the elastic properties according to the imposed stress.This section allows to highlight the importance of the orientation of bedding relatively to the principal stress applied on our samples. It is noted for example that the weakest orientation is 45 ° and 90 ° the strongest orientation. In addition, a micromechanical model based on the "wing crack" theory helps to explain the mechanical anisotropy of our argillites by the anisotropy of the fracture toughness K_Ic.Moreover, it is found that changes in the anisotropy of the elastic properties also depends on the bedding orientation. During compression, the 90 ° orientation has significant variations up to a reversal of the anisotropy, whereas the elastic wave velocities show little changes for 0 ° orientation. These variations were quantified by Thomsen parameters. The study of elastic velocities and of microstructures of our samples highlight the importance of the plastic processes such as reorientation of minerals during deformation.The second part consists of a petrophysical study (P wave velocities, ASM, density, saturation, porosity) of samples from a borehole drilled through a fault zone. The goal here is to quantify the variation of these properties as we approach the fault core.A protocol of sampling and measurement was established to realize a complete study of drillings. The protocol includes a first serie of measures directly in the field in order to overcome the problems linked to the preservation of samples (especially for porosity measurements / density / saturation). Subsequently, samples were taken for measurements both in the context of this thesis (P wave velocities and ASM), but also in other laboratories (study of the mineralogical composition, CT-scan).Three fault zones were identified from field observations:• An intact zone characterized by samples with no or very little fracturing• A damaged zone that includes a large number of calcified fractures• A fault core zone characterized by a complete breakdown (no apparent bedding) and color variations.Although these areas were determined on observation criteria, our study demonstrates an own physical signature for each zone of this fault with an important decrease in the anisotropy of the samples from fault core. In addition, bedding rotation with similar mechanisms to those observed at the microstructural scale during the experimental deformation was observed.

京都大学 / 新制・課程博士 / 博士(工学) / 甲第23175号 / 工博第4819号 / 新制||工||1753(附属図書館) / 京都大学大学院工学研究科都市社会工学専攻 / (主査)教授 小池 克明, 教授 林 為人, 准教授 柏谷 公希 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DGAM

<p>Two longstanding goals in subsurface science are to induce fractures with a desired geometry to adaptively control the interstitial geometry of existing fractures in response to changing subsurface conditions. Many energy and water-related engineering applications that use induced fractures to withdraw and inject fluids from subsurface reservoirs occur in some sedimentary rock.  Sedimentary rock such as shales often exhibit anisotropic mechanical properties because of bedding, layering and mineral texture.  These structural and textural features also affect fracture formation and in turn the resulting fracture geometry. Understanding the interplay between the microscopic mineral fabric and structure and how it effects fracture geometry is important for the prediction of the geometry of induced fractures and to the determination of the most ideal conditions for maximizing energy production and minimizing leaks from sequestration sites in the subsurface. </p> <p><br></p> <p>This Ph.D. thesis research focuses on the formation and geometry of fractures in anisotropic rock and the identification of geophysical signatures of fracture formation using additively manufactured gypsum rock analogs. Specifically, the work is grouped into three topics: (1) material controls on fracture geometry, toughness and roughness in additively manufactured rocks; (2) acoustic emissions (AE) during fracture formation in anisotropic additively manufactured rocks; and (3) determination of the effect of fluid-filled oriented voids in fractures on compressional to shear wave conversions. </p> <p><br></p> <p>For topic (1), unconfined compressive strength (UCS), Brazilian and 3-point bending (3PB) tests under pure and mixed mode mechanical tests were performed on cast and 3D printed gypsum samples that were characterized using 3D Xray microscopy, Xray Diffraction and SEM to examine the micro-structure of the samples. Research on topic 1 discovered microstructural controls on fracture surface roughness and the failure behavior of anisotropic rock and that the failure mode (tensile, mixed mode I and II, mixed mode I and III) affects the fracture propagation path and the surface roughness which is controls to the flow paths through a fracture. The results suggest that detailed mineralogical studies of mineral texture/fabric in laboratory or core samples is important to unravel failure strength, surface roughness, and how fractures propagate in layered geological media. </p> <p><br></p> <p>For topic (2), UCS tests were performed with concurrent measurements of acoustic emissions (AE) on cylindrical specimens: cast gypsum (CG) samples, and 3D printed (3DP) samples with five different orientations of bassanite layer and gypsum texture relative to the loading direction. Mechanical properties and induced fracture surface information were compared with the collected the AE signals to study if there is a way to tell the differences between the induced fracture surfaces with the AE signals patterns together with loading data. Examination of the AE signal amplitude from post-peak loading revealed that more ductile behavior was associated with more AE events that occurred over a longer period of time, and the resultant fracture surfaces were rougher than for narrow time distributions of events. </p> <p><br></p> <p>For topic (3), a detail study of fracture void orientation was performed using ultrasonic compressional, P, and shear, S, waves to determine how energy is partitioned when P-to-S or S-to-P conversions occur for waves normally incident on an air-filled or fluid-filled fracture. In this study, experiments and computer simulations were performed to demonstrate the link among cross-coupling stiffness, micro-crack orientation and energy partitioning into P, S, and P-S/S-P wave. The cross-coupling stiffness was created by 3D printing samples with linear arrays of micro-cracks oriented at  $0^o$, $\pm15^o$, $\pm30^o$, $\pm45^o$, $\pm60^o$, $\pm75^o$, and $90^o$. For $45^o$ orientation, measurements were made on air-filled and fluid-filled (silicon oil). For the air-filled fractures, the observed energy partitioning matched the simulated behavior obtained from discontinuous Galerkin simulations. Information on local fracture geometry is contained in the far-field waves. When filled with a viscous fluid, the P- and S- waves amplitude exhibited slight increases and decreases, respectively. The P-to-S converted mode amplitude decreased 30\% with an increase in fluid viscosity from 1–300kcSt. This suggests that P-S converted mode provides a potential method to remotely probe changes in fluid viscosity in fractures. </p> <p><br></p> <p>The work from the 3 research topics demonstrated that micro-scale structure impacts macroscale behavior and signals used for monitoring the condition of a rock. Additively manufactured samples enabled the exploration and determination of (1) the impact of mineral fabric orientation in layered media on failure load, fracture propagation path, and fracture surface roughness, (2) the sensitivity of P-to-S conversions to fluid viscosity, and (3) how oriented voids within a fracture effect energy partitioning. These research findings advances our current understanding of role microscopic properties and structure on the generation, propagation and geometry of induced fractures in anisotropic rock, and help to identify the best imaging modalities to use to identify the seismic signatures of the viscosity of fluids in fractures with oriented voids. These contributions will help unravel the complex behavior often observed in natural rock that is structurally and compositionally complex with features and heterogeneity.  </p> <p><br></p>

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第19695号 / 工博第4150号 / 新制||工||1640(附属図書館) / 32731 / 京都大学大学院工学研究科都市社会工学専攻 / (主査)教授 小池 克明, 教授 高岡 昌輝, 准教授 村田 澄彦 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DGAM

Although carbonates hold more than 60 percent of the world's oil reserves, they, nevertheless, exhibit much lower average recovery factor values than terrigenous sandstone reservoirs. Thus, utilization of advanced enhanced oil recovery (EOR) techniques such as high pressure CO2 injection may normally be required to recover oil in place in carbonate reservoirs. This study addresses how different rock types can influence the seismic monitoring of CO2 sequestration in carbonates. This research utilizes an elastic parameter, defined in a rock physics model of poroelasticity and so-­called as the frame flexibility factor, to successfully quantify the carbonate pore types in core samples available from the Great Bahama Bank (GBB). This study shows that for carbonate samples of a given porosity the lower the frame flexibility factors the higher is the sonic wave velocity. Generally, samples with frame flexibility values of <4 are either rocks with visible moldic pores or intraframe porosity; whereas, samples with frame flexibility values of >4 are rocks with intercrystalline and microporosity. Hence, different carbonate pore geometries can be quantitatively predicted using the elastic parameters capable of characterizing the porous media with a representation of their internal structure on the basis of the flexibility of the frame and pore connectivity. In this research, different fluid substitution scenarios of liquid and gaseous CO2 saturations are demonstrated to characterize the variations in velocity for carbonate-specific pore types. The results suggest that the elastic response of CO2 flooded rocks is mostly governed by pore pressure conditions and carbonate rock types. Ultrasonic P-­wave velocities in the liquid-­phase CO2 flooded samples show a marked decrease in the order of 0.6 to 16 percent. On the contrary, samples flooded with gaseous-­phase CO2 constitute an increase in P-­wave velocities for moldic and intraframe porosities, while establishing a significant decrease for samples with intercrystalline and micro-­porosities. Such velocity variations are explained by the stronger effect of density versus compressibility, accounting for the profound effect of pore geometries on the acoustic properties in carbonates. The theoretical results from this research could be a useful guide for interpreting the response of time-­lapse seismic monitoring of carbonate formations following CO2 injection at depth. In particular, an effective rock-­physics model can aid in better discrimination of the profound effects of different pore geometries on seismic monitoring of CO2 sequestration in carbonates.

[pt] O conhecimento do que acontece no reservatório em produção a partir de variações temporais dos atributos sísmicos devido aos processos dinâmicos vem atingindo um valor crescente na indústria do petróleo, especialmente em arenitos. Este processo possui vários desafios, focados em grande parte a desvendar a superposição dos diferentes efeitos provocados pelas mudanças do reservatório nos dados sísmicos. As propriedades sísmicas são afetadas de maneira complexa por vários fatores, sendo a saturação um dos mais importantes, principalmente em rochas porosas como o arenito. Esta propriedade influencia no módulo elástico da rocha e sua resposta sísmica e, ao mesmo tempo, introduz dispersão da velocidade (variação da velocidade com a freqüência). A transição de fluido efetivo (distribuição homogênea e menores velocidades) para fluido com distribuição heterogênea (e maiores velocidades) estabelece um mecanismo de dispersão presente para freqüências sísmicas in situ, especialmente no arenito. O método mais utilizado para aplicar a técnica de substituição de fluidos se baseia na teoria de Gassmann (1951), que considera o meio poroso estático (estado de isostress), onde o fluido não é afetado pela perturbação da onda. No entanto, pesquisas mostram que as velocidades acústicas em rochas saturadas de fluido dependem da freqüência, do tipo de fluido e sua distribuição no meio poroso, viscosidade e outras propriedades que tornam as ondas dispersivas. Neste trabalho são realizadas simulações de fluxo de reservatórios, transformações de física de rochas, upscaling e modelagem sísmica em cenários de injeção de gás com o objetivo de esclarecer a importância de levar em conta a dispersão da velocidade na análise time-lapse. Para isso, são analisados para cada modelo mapas de saturação, velocidade, impedância e sismogramas sintéticos (seções de contraste) calculados com as teorias de substituição Gassmann (1951) e Mavko E Jizba (1991). Os resultados mostram que a resposta sísmica pode ter um incremento de até 15 por cento quando a dispersão devida ao fluxo local é considerada. Porosidade e tortuosidade são parâmetros essenciais que influenciam de maneira diferente na resposta sísmica. / [en] The evaluation of reservoir dynamics during production through time-lapse interpretation has reached a substantial importance in the petroleum industry, mainly in sandstones. This evaluation presents many challenges, mainly concerned to unmask the overlapping of different effects in seismic data due to reservoir changes. Several factors affect seismic properties and saturation is one of the most important. This property influences the rock bulk modulus and seismic response and also causes a velocity dependence on the frequency. This phenomenon is known as velocity dispersion. Furthermore, the transition from effective homogeneous fluid to heterogeneous saturation represents a dispersion mechanism that appears for seismic frequencies in situ in sandstones. The most commonly method used to perform the fluid substitution technique is based in Gassmann theory (1951). This approach considers a static porous media (isostress condition), where fluid is not affected by wave propagation. However, it is well known that acoustic velocities in fluid saturated rocks depends on frequency, according to fluid type and distribution on porous media, viscosity, and others properties that become waves dispersive. In this work reservoir flow-simulation, rock physics transformations, upscaling and seismic modeling were performed in gas injection scenarios. Synthetic seismograms and some contrast sections were generated using Gassmann (1951) and Mavko & Jizba (1991) substitution theories. The goal is to clarify the relevance of considering velocity dispersion on time-lapse seismic analyzing possible differences in the seismic parameters. Results show that seismic response could increase in 15% when squirt flow dispersion is considered. Porosity and tortuosity are essential parameters to analyze seismic response.

Under current technical levels and industrial systems, geological carbon sequestration (GCS) is a viable solution to maintain and further reduce carbon dioxide (CO2) concentration and ensure energy security simultaneously. The pre-injection formation characterization and post-injection CO2 monitoring, verification, and accounting (MVA) are two critical and challenging tasks to guarantee the sequestration effect. The tasks can be accomplished using core analyses and well-logging technologies, which complement each other to produce the most accurate and sufficient subsurface information for pore-scale and reservoir-scale studies. In recent years, the unprecedented data sources, increasing computational capability, and the developments of machine learning (ML) and deep learning (DL) algorithms provide novel perspectives for expanding the knowledge from data, which can capture highly complex nonlinear relationships between multivariate inputs and outputs. This work applied ML and DL methods to GCS-related studies at pore and reservoir scales, including digital rock physics (DRP) and the well-logging data interpretation and analysis. DRP provides cost-saving and practical core analysis methods, combining high-resolution imaging techniques, such as the three-dimensional (3D) X-ray computed tomography (CT) scanning, with advanced numerical simulations. Image segmentation is a crucial step of the DRP framework, affecting the accuracy of the following analyses and simulations. We proposed a DL-based workflow for boundary and small target segmentation in digital rock images, which aims to overcome the main challenge in X-ray CT image segmentation, partial volume blurring (PVB). The training data and the model architecture are critical factors affecting the performance of supervised learning models. We employed the entropy-based-masking indicator kriging (IK-EBM) to generate high-quality training data. The performance of IK-EBM on segmentation affected by PVB was compared with some commonly used image segmentation methods on the synthetic data with known ground truth. We then trained and tested the UNet++ model with nested architecture and redesigned skip connections. The evaluation metrics include the pixel-wise (i.e. F1 score, boundary-scaled accuracy, and pixel-by-pixel comparison) and physics-based (porosity, permeability, and CO2 blob curvature distributions) accuracies. We also visualized the feature maps and tested the model generalizations. Contact angle (CA) distribution quantifies the rock surface wettability, which regulates the multiphase behaviors in the porous media. We developed a DL-based CA measurement workflow by integrating an unsupervised learning pipeline for image segmentation and an open-source CA measurement tool. The image segmentation pipeline includes the model training of a CNN-based unsupervised DL model, which is constrained by feature similarity and spatial continuity. In addition, the over-segmentation strategy was adopted for model training, and the post-processing was implemented to cluster the model output to the user-desired target. The performance of the proposed pipeline was evaluated using synthetic data with known ground truth regarding the pixel-wise and physics-based evaluation metrics. The resulting CA measurements with the segmentation results as input data were validated using manual CA measurements. The GCS projects in the Illinois Basin are the first large-scale injection into saline aquifers and employed the latest pulsed neutron tool, the pulsed neutron eXtreme (PNX), to monitor the injected CO2 saturation. The well-logging data provide valuable references for the formation evaluation and CO2 monitoring in GCS in saline aquifers at the reservoir scale. In addition, data-driven models based on supervised ML and DL algorithms provide a novel perspective for well-logging data analysis and interpretation. We applied two commonly used ML and DL algorithms, support vector machine regression (SVR) and artificial neural network (ANN), to the well-logging dataset from GCS projects in the Illinois Basin. The dataset includes the conventional well-logging data for mineralogy and porosity interpretation and PNX data for CO2 saturation estimation. The model performance was evaluated using the root mean square error (RMSE) and R2 score between model-predicted and true values. The results showed that all the ML and DL models achieved excellent accuracies and high efficiency. In addition, we ranked the feature importance of PNX data in the CO2 saturation estimation models using the permutation importance algorithm, and the formation sigma, pressure, and temperature are the three most significant factors in CO2 saturation estimation models. The major challenge for the CO2 storage field projects is the large-scale real-time data processing, including the pore-scale core and reservoir-scale well-logging data. Compared with the traditional data processing methods, ML and DL methods achieved accuracy and efficiency simultaneously. This work developed ML and DL-based workflows and models for X-ray CT image segmentation and well-logging data interpretations based on the available datasets. The performance of data-driven surrogate models has been validated regarding comprehensive evaluation metrics. The findings fill the knowledge gap regarding formation evaluation and fluid behavior simulation across multiple scales, ensuring sequestration security and effect. In addition, the developed ML and DL workflows and models provide efficient and reliable tools for massive GCS-related data processing, which can be widely used in future GCS projects. / Doctor of Philosophy / Geological carbon sequestration (GCS) is the solution to ease the tension between the increasing carbon dioxide (CO2) concentrations in the atmosphere and the high dependence of human society on fossil energy. The sequestration requires the injection formation to have adequate storage capability, injectivity, and impermeable caprock overlain. Also, the injected CO2 plumes should be monitored in real-time to prevent any migration of CO2 to the surface. Therefore, pre-injection formation characterization and post-injection CO2 saturation monitoring are two critical and challenging tasks to guarantee the sequestration effect and security, which can be accomplished using the combination of pore-scale core analyses and reservoir-scale well-logging technologies. This work applied machine learning (ML) and deep learning (DL) methods to GCS-related studies across multiple spatial scales. We developed supervised and unsupervised DL-based workflows to segment the X-ray computed-tomography (CT) image of digital rocks for the pore-scale studies. Image segmentation is a crucial step in the digital rock physics (DRP) framework, and the following analyses and simulations are conducted on the segmented images. We also developed ML and DL models for well-logging data interpretation to analyze the mineralogy and estimate CO2 saturation. Compared with the traditional well-logging analysis methods, which are usually time-consuming and prior knowledge-dependent, the ML and DL methods achieved comparable accuracy and much shorter processing time. The performance of developed workflows and models was validated regarding comprehensive evaluation metrics, achieving excellent accuracies and high efficiency simultaneously. We are at the early stage of CO2 sequestration, and relevant knowledge and tools are inadequate. In addition, the main challenge of CO2 sequestration field projects is the large-scale and real-time data processing for fast decision-making. The findings of this dissertation fill the knowledge gap in GCS-related formation evaluation and fluid behavior simulations across multiple spatial scales. The developed ML and DL workflows provide efficient and reliable tools for massive data processing, which can be widely used in future GCS projects.

[pt] Desde a década de 80, abordagens data-driven têm sido utilizadas para identificação de fluidos e caracterização de reservatórios carbonáticos e siliciclásticos principalmente em relação à análise das amplitudes sísmicas. No entanto, técnicas aplicadas com sucesso para rochas siliciclásticas, como por exemplo: Análise AVO, inversões sísmicas e IDH (Indicadores Diretos de Hidrocarbonetos) revelaram não obter o mesmo êxito para reservatórios carbonáticos heterogêneos. Em contrapartida, diversos artigos mostram que fluxos de caracterização de reservatórios com modelos de física de rochas incorporados têm alcançado grande sucesso para obtenção de propriedades petrofísicas e atributos elásticos de ambas as rochas, utilizando sísmicas e well logs, em uma abordagem model-driven, focada nas características microestruturais do reservatório. Dessa forma, levando em consideração a importância de se utilizar modelos de física de rochas no escopo da caracterização de reservatórios, dois modelos de física de rochas - Xu e Payne e T-Matrix - foram aplicados, comparados e seus parâmetros foram estocasticamente avaliados e otimizados em um arcabouço Bayesiano. Através dessa abordagem, foi possível estimar, de uma forma confiável, os atributos elásticos de um reservatório carbonático (coquinas) levando em consideração diversos tipos de incertezas. Além disso, após a calibração e validação de ambos os modelos de física de rochas para diferentes poços, análises de sensibilidade foram realizadas para compreensão de forma quantitativa do comportamento dos atributos elásticos das coquinas em relação às alterações do conteúdo mineralógico, tipos de poro e fluidos desse reservatório. / [en] Since the 80 s, data-driven approaches have been used for fluids identification and reservoir characterization of siliciclastic and carbonate rocks mainly regarding seismic amplitudes analyses. However, techniques successfully applied for siliciclastic rocks, such as: AVO analysis, seismic inversions and DHI (Direct Hydrocarbon Indicators) ranking revealed not have achieved the same outstanding and reliable results for heterogeneous carbonate rocks. On the other hand, several articles demonstrate that reservoir characterization workflows with rock physics models embedded have been reaching a robust success in order to obtain petrophysical properties and elastic attributes of both rocks, from the seismic and well logs, in a model-driven approach focused on the reservoirs microstructural information. In this way, taking into account the importance of applying rock physics models in the scope of reservoir characterization, two rock physics models - Xu and Payne and T-Matrix - were applied, compared and their parameters were stochastically evaluated and optimized in a Bayesian framework. Through this approach, it was possible to estimate, in a reliable manner, the elastic attributes of a carbonate reservoir (coquinas) taking into consideration different kinds of uncertainties. Furthermore, after the calibration in the well location and validation of both rock physics models for other wells, sensitivity analyses were conducted in order to quantitatively understand how the coquinas elastic attributes behave regarding the variations in the reservoir mineralogical content, pore shapes and fluids.

Worldwide interest in gas production from shale formations has rapidly increased in recent years, mostly by the successful development of gas shales in North America. The Haynesville Shale is a productive gas shale resource play located in Texas and Louisiana. It produces primarily through enhanced exposure to the reservoir and improved permeability resulting from horizontal drilling and hydraulic fracturing. Accordingly, it is important to estimate the reservoir properties that influence the elastic and geomechanical properties from seismic data. This thesis estimates pore shapes, which affect the transport, elastic, and geomechancial properties, from wellbore seismic velocity in the Haynesville Shale. The approach for this work is to compare computed velocities from an appropriate rock physics model to measured velocities from well log data. In particular, the self-consistent approximation was used to calculate the model-based velocities. The Backus average was used to upscale the high-frequency well log data to the low-frequency seismic scale. Comparisons of calculated velocities from the self-consistent model to upscaled Backus-averaged velocities (at 20 Hz and 50 Hz) with a convergence of 0.5% made it possible to estimate pore aspect ratios as a function of depth. The first of two primary foci of this approach was to estimate pore shapes when a single fluid was emplaced in all the pores. This allowed for understanding pore shapes while minimizing the effects of pore fluids. Secondly, the effects of pore fluid properties were studied by comparing velocities for both patchy and uniform fluid saturation. These correspond to heterogeneous and homogeneous fluid mixing, respectively. Implementation of these fluid mixtures was to model them directly within the self-consistent approximation and by modeling dry-rock velocities, followed by standard Gassmann fluid substitution. P-wave velocities calculated by the self-consistent model for patchy saturation cases had larger values than those from Gassmann fluid substitution, but S-wave velocities were very similar. Pore aspect ratios for variable fluid properties were also calculated by both the self-consistent model and Gassmann fluid substitution. Pore aspect ratios determined for the patchy saturation cases were the smallest, and those for the uniform saturation cases were the largest. Pore aspect ratios calculated by Gassmann fluid substitution were larger because the velocity is inversely related to the aspect ratio in this particular modeling procedure. Estimates of pore aspect ratios for uniform saturation were 0.051 to 0.319 with the average of 0.171 from the velocity modeling using the self-consistent model. For patchy saturation, the aspect ratios were 0.035 to 0.296 with a mean of 0.145. These estimated pore aspect ratios from the patchy saturation case within the self-consistent model are considered the most reasonable set of values I determined. This is because the most likely in-situ fluid distribution is heterogeneous due to the extremely low permeability of the Haynesville Shale. Estimated pore aspect ratios using this modeling help us to understand elastic properties of the Haynesville Shale. In addition, this may help to find zones that correspond to optimal locations for fracturing the shale while considering brittleness and in-situ stress of the formation. / text