Dissertationen zum Thema „Geotechnical Problems Of The Witwatersrand“

COORDENAÇÃO DE APERFEIÇOAMENTO DO PESSOAL DE ENSINO SUPERIOR<br>Nos últimos anos, a aplicação da técnica de redes neurais tem sido difundida em diversas áreas do conhecimento, inclusive na engenharia civil. Em meados da década de 90, iniciaram-se no Brasil estudos no sentido de avaliar a eficiência desta técnica numérica na modelagem do comportamento de solos e na análise de problemas envolvendo engenharia geotécnica. Este trabalho é resultado de parte destes estudos, onde algumas das potencialidades do uso das redes neurais em geotecnia podem ser observadas. São apresentadas três aplicações diferentes de redes neurais feedforward em geotecnia, tendo sido treinadas com o algoritmo LM (Levenberg-Marquardt). A primeira aplicação diz respeito à simulação de resultados de provas de carga dinâmica, analisadas pelo método CAPWAP, através de redes neurais, sendo assim viabilizada a realização de uma pré- análise do comportamento da estaca ainda em campo, o que geralmente não acontece quando se trata da análise CAPWAP tradicional. A segunda aplicação relaciona-se com a análise do comportamento mecânico de dois tipos de solo bastante diferentes entre si: a areia de Ipanema e o solo residual gnáissico do Rio de Janeiro. Para tal, foram utilizados resultados de ensaios de cisalhamento direto, submersos e não submersos, e ensaios de compressão triaxial, drenados e não drenados. A terceira aplicação refere-se à simulação das características do subsolo do sítio da Usina Nuclear Angra 2, localizada no litoral do estado do Rio de Janeiro. As informações disponíveis eram advindas de boletins de sondagens do tipo SPT. Foram realizadas simulações envolvendo a disposição das camadas dos diferentes tipos de solo que poderiam existir no local, o nível de água subterrâneo, a resistência à penetração do solo e a topografia do terreno. Em todos os casos foram obtidos resultados bastante satisfatórios. Portanto, conclui-se que a técnica das redes neurais apresenta grande viabilidade na resolução de problemas geotécnicos de diferentes características, muitas vezes se mostrando tanto ou mais eficiente que as técnicas numéricas tradicionais.<br>During the last years, neural networks applications have been disseminated in many knowledge areas, including civil engineering. In the middle 90`s, a research work had been started in Brazil, in order to investigate the efficiency of neural networks in the analysis of soil behavior and problems involving geotechnical engineering. This thesis is the result of part of these studies, where some potentialities of neural networks technique are presented. Three different feedforward NNs applications in geotechnical engineering are presented. Levenberg- Marquardt algorithm was used for training. The first application is the simulation of results of dynamic pile tests, obtained from CAPWAP analysis, showing that it is possible to do a field pre-analysis of the pile behavior, which is still unpracticable when the traditional CAPWAP method is used. The second application is related to the study of two different soils behavior:sand from Ipanema and residual gnaissic soil from Rio de Janeiro. Results of submerged and non submerged direct shear tests and drained and undrained triaxial compression tests were used. The third application involves the simulation of subsoil characteristics of Angra 2 Nuclear Power Plant site. The available information came from SPT bulletins. Simulations involving several types of soil layers spatial distribution, water level position, penetration strength of soils and local topography were performed. The obtained results were very satisfactory. It can be concluded that the neural networks technique presents great applicability in resolution of geotechnical problems with different characteristics, showing an efficiency as good or even better than other traditional numerical techniques.<br>En los últimos anos, la aplicación de técnicas de redes neurales se ha difundido en diversas áreas del conocimento, incluso en la ingeniería civil. A mediados de la década de 90, se iniciaran en Brasil estudios para evaluar la eficiencia de esta técnica numérica em modelos de comportamiento de suelos y en el análisis de problemas de ingeniería geotécnica. Este trabajo es el resultado de parte de estos estudios, donde pueden ser obseravdas algunas de las potencialidades del uso de las redes neurales en geotecnia. Se presentan tres aplicaciones diferentes de redes neurales fedforward en geotecnia, entrenadas con el algoritmo LM (Levenberg Marquardt). La primera aplicación se refiere a la simulación de resultados de pruebas de carga dinámica, analizadas por el método CAPWAP, a través de redes neurales, realizando un pré análisis del comportamiento de la estaca en campo, lo que generalmente no sucede cuando se trata del análisis CAPWAP tradicional. La segunda aplicación se relaciona con el análisis del comportamiento mecánico de dos tipos de suelo bastante diferentes entre sí: la arena de Ipanema y el suelo residual gnáisico de Rio de Janeiro. Para esto, se uilizaron resultados de ensayos de cisallamiento directo, submersos y no submersos, y ensayos de compresión triaxial, drenados y no drenados. La tercera aplicación se refiere a la simulación de las características del subsuelo del sitio de la Planta Nuclear Angra 2, localizada en el litoral del estado del Rio de Janeiro. Las informaciones disponibles provenian de boletines del tipo SPT. Se realizaron simulaciones que involucraban la disposición de los diferentes tipos de suelo que podrían existir en el local, el nível de agua subterránea, la resistencia a la penetración del suelo y la topografia del terreno. En todos los casos fueron obtenidos resultados bastante satisfactorios. Por lo tanto, se concluye que la técnica de redes neurales presenta gran viabilidad en la resolución de problemas geotécnicos de diferentes características, muchas veces mostrándose tanto o más eficiente que las técnicas numéricas tradicionales.

Oxidation of pyrite can significantly affect properties and the behavior of soil and rock in civil construction. Problems with pyritic rock and soil extend globally and across many disciplines. Consequences of pyrite oxidation include heave, concrete degradation, steel corrosion, environmental damage, acid mine drainage, and accelerated weathering of rock with concomitant effects on strength and stability. Affected disciplines include soil science, mining, engineering geology, geochemistry, environmental engineering, and geotechnical engineering. <p> While pyrite problems may be well known in their respective disciplines, there has been to date relatively little cross-disciplinary communication regarding problems with pyritic geomaterials. Thus, there is a need to establish an inter-disciplinary and inter-regional awareness regarding the effects of pyrite oxidation and their prevention or mitigation. <p> This engineering research is a compilation of information about geotechnical problems and engineering behavior of pyritic rock and soil, the underlying physicochemical processes, site investigation strategies, and known problematic formations. Several case histories documenting consequences of pyrite oxidation are provided. The results of chemical analyses performed on pyritic shale samples from a formation with acknowledged heave problems are presented. Digital data and ESRIâ s ArcGIS digital mapping program were used to create maps showing results of sampling and testing performed during this study. Appendices include mitigation options, results of a practitioner survey, chemical test procedures, a glossary, a visual identification key for sulfidic geomaterials, and a summary table of the literature review for this research.<br>Master of Science

The research carried out in this thesis has concentrated on the application of numerical solutions to geotechnical penetration problems in offshore engineering. Several important issues closely relevant to deep-water oil and gas developments were investigated, covering installation of suction caisson foundations, interpretation of fullflow penetrometers and shallow penetration of a cylindrical object (submarine pipeline or T-bar), all in clayey sediments such as are often encountered in deep-water sites. These problems are commonly characterised by large vertical movements of structural elements relative to the seabed. A large deformation finite element method was adopted and further developed to simulate these challenging problems, referred to as Remeshing and Interpolation Technique with Small Strain. In this approach, a sequence of small strain Lagrangian increments, remeshing and interpolation of stresses and material properties are repeated until the required displacement has been reached. This technique is able to model relative motion between the penetrating objects and the soil, which is critical for evaluating soil heave inside the caissons, the effect of penetration-induced remoulding on the resistance of full-flow penetrometers, and influence of soil surface heave on the embedment of pipelines. '...' Simple expressions were presented allowing the resistance factors for the T-bar and ball penetrometers to be expressed as a function of the rate and strain-softening parameters. By considering average strength conditions during penetration and extraction of these full-flow penetrometers, an approximate expression was derived that allowed estimation of the hypothetical resistance factor with no strain-softening, and hence an initial estimate of the stain-rate dependency of the soil. Further simulations of cyclic penetration tests showed that a cyclic range of three diameters of the penetrometers was sufficient to avoid overlap of the failure mechanism at the extremes and mid-point of the cyclic range. The ball had higher resistance factors compared with the T-bar, but with similar cyclic resistance degradation curves, which could be fitted accurately by simple expressions consistent with the strain-softening soil model adopted. Based on the curve fitting, more accurate equations were proposed to deduce the resistance factor with no strain-softening, compared with that suggested previously based on the resistances measured in the first cycle of penetration and extraction. The strain-rate dependency was similar in intact or post-cyclic soil for a given rate parameter. The resistance factor for the post-cyclic condition was higher than that for the initial conditions, to some degree depending upon soil sensitivity and brittleness parameter. For the shallow penetration of a cylindrical object, the penetration resistance profile observed from centrifuge model tests was very well captured by the numerical simulation. The mechanism of shear band shedding was reproduced by the numerical technique, although the frequency of the shear band generation and the exact shape of the heave profile were not correctly captured, which were limited by the simple strainsoftening soil model adopted.

Bond degradation is an irreversible phenomenon that, experimentally, appears to be controlled by plastic strain accumulation. Conventional constitutive soil models do not capture the effects of small strain non‐linearity, recent stress history as well as material structure and its consequent reduction due to bond degradation. The aim of this thesis is to investigate the behaviour of a constitutive model that describes the initial structure, on various geotechnical problems. The kinematic hardening structured constitutive model (Rouainia and Muir Wood, 2000) ,formulated within a framework of kinematic hardening and bounding surface plasticity, was implemented into the PLAXIS Finite Element Analysis software package and was used to simulate a variety of boundary value problems. The implementation of the model was validated through a number of single finite element analyses of laboratory tests on natural clay from the Vallericca valley in Italy. The model was further adopted in the finite element analyses of geotechnical problems. The first of these simulated the Self Boring Pressuremeter test in London Clay, with the main focus being the characterisation of the degree of initial structure of London clay, as well as identifying the effect of structure related parameters. The premise that the SBPM is installed without damage was also investigated. The second boundary value problem involved the 2D and 3Danalysis of an embankment situated on soft structured clay in Saint Alban, Canada. The numerical predictions of pore‐water pressures and settlements are also compared with field measurements. The model developed in this work was then adopted in the study of the behaviour of a deep excavation located in Boston, Massachusetts, USA. The numerical simulations were aimed to demonstrate that the added features of the model implemented in this work such as small strain stiffness, structure and anisotropy are vital components to give a good prediction. Comparison of the predicted wall profiles, time dependent dissipation of excess pore water pressures and associated ground heave with field data are provided.

Nowadays, thanks to the increase of computers capability to solve huge and complex problems, and also thanks to the endless effort of the geotechnical community to define better and more sophisticated constitutive models, the challenge to predict and simulate soil behavior has been eased. However, due to the increase in that sophistication, the number of parameters that define the problem has also increased. Moreover, frequently, some of those parameters do not have a real geotechnical meaning as they just come from mathematical expressions, which makes them difficult to identify. As a consequence, more effort has to be placed on parameters identification in order to fully define the problem. This thesis aims to provide a methodology to facilitate the identification of parameters of soil constitutive models by backanalysis. The best parameters are defines as those that minimize an objective function based on the differences between measurements and computed values. Different optimization techniques have been used in this study, from the most traditional ones, such as the gradient based methods, to the newest ones, such as adaptive genetic algorithms and hybrid methods. From this study, several recommendations have been put forward in order to take the most advantage of each type of optimization technique. Along with that, an extensive analysis has been carried out to determine the influence on soil parameters identification of what to measure, where to measure and when to measure in the context of tunneling. The Finite Element code Plaxis has been used as a tool for the direct analysis. A FORTRAN code has been developed to automate the entire backanalysis procedure. The Hardening Soil Model (HSM) has been adopted to simulate the soil behavior. Several soil parameters of the HSM implemented in Plaxis, such as E_50^ref, E_ur^ref, c and f, have been identified for different geotechnical scenarios. First, a synthetic tunnel case study has been used to analyze all the different approaches that have been proposed in this thesis. Then, two complex real cases of a tunnel construction (Barcelona Metro Line 9) and a large excavation (Girona High-Speed Railway Station) have been presented to illustrate the potential of the methodology. Special focus on the influence of construction procedures and instruments error structure has been placed for the tunnel backanalysis, whereas in the station backanalysis, more effort has been devoted to the potential of the concept of adaptive design by backanalysis. Moreover, another real case, involving a less conventional geotechnical problem, such as Mars surface exploratory rovers, has been also presented to test the backanalysis methodology and the reliability of the Wong & Reece wheel-terrain model; widely adopted by the terramechanics community, but nonetheless, still not fully accepted when analyzing lightweight rovers as the ones that have been used in recent Mars exploratory missions.<br>Actualmente, gracias al aumento de la capacidad de los ordenadores para resolver problemas grandes y complejos, y gracias también al gran esfuerzo de la comunidad geotécnica de definir mejores y más sofisticados modelos constitutivos, se ha abordado el reto de predecir y simular el comportamiento del terreno. Sin embargo, debido al aumento de esa sofisticación, también ha aumentado el número de parámetros que definen el problema. Además, frecuentemente, muchos de esos parámetros no tienen un sentido geotécnico real dado que vienen directamente de expresiones puramente matemáticas, lo cual dificulta su identificación. Como consecuencia, es necesario un mayor esfuerzo en la identificación de los parámetros para poder definir apropiadamente el problema. Esta tesis pretende proporcionar una metodología que facilite la identificación mediante el análisis inverso de los parámetros de modelos constitutivos del terreno. Los mejores parámetros se definen como aquellos que minimizan una función objetivo basada en la diferencia entre medidas y valores calculados. Diferentes técnicas de optimización han sido utilizadas en este estudio, desde las más tradicionales, como los métodos basados en el gradiente, hasta las más modernas, como los algoritmos genéticos adaptativos y los métodos híbridos. De este estudio, se han extraído varias recomendaciones para sacar el mayor provecho de cada una de las técnicas de optimización. Además, se ha llevado a cabo un análisis extensivo para determinar la influencia sobre qué medir, dónde medir y cuándo medir en el contexto de la excavación de un túnel. El código de Elementos Finitos Plaxis ha sido utilizado como herramienta de cálculo del problema directo. El desarrollo de un código FORTRAN ha sido necesario para automatizar todo el procedimiento de Análisis Inverso. El modelo constitutivo de Hardening Soil ha sido adoptado para simular el comportamiento del terreno. Varios parámetros del modelo constitutivo de Hardening implementado en Plaxis, como E_50^ref, E_ur^ref, c y f, han sido identificados para diferentes escenarios geotécnicos. Primero, se ha utilizado un caso sintético de un túnel donde se han analizado todas las distintas técnicas que han sido propuestas en esta tesis. Después, dos casos reales complejos de una construcción de un túnel (Línea 9 del Metro de Barcelona) y una gran excavación (Estación de Girona del Tren de Alta Velocidad) se han presentado para ilustrar el potencial de la metodología. Un enfoque especial en la influencia del procedimiento constructivo y la estructura del error de las medidas se le ha dado al análisis inverso del túnel, mientras que en el análisis inverso de la estación el esfuerzo se ha centrado más en el concepto del diseño adaptativo mediante el análisis inverso. Además, otro caso real, algo menos convencional en términos geotécnicos, como es la exploración de la superficie de Marte mediante robots, ha sido presentado para examinar la metodología y la fiabilidad del modelo de interacción suelo-rueda de Wong y Reece; extensamente adoptado por la comunidad que trabajo en Terramecánica, pero aún no totalmente aceptada para robots ligeros como los que se han utilizado recientemente en las misiones de exploración de Marte.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Civil and Environmental Engineering, 2017.<br>Cataloged from PDF version of thesis.<br>Includes bibliographical references (pages 269-281).<br>Although finite element (FE) methods are well established for modeling geotechnical problems in soil masses and soil-structure interaction, most prior research on large deformation problems has been limited to simplified assumptions on drainage conditions and constitutive behavior. This thesis investigates two large deformation problems in soft clay and proposes a methodology for performing coupled flow and deformation analyses with advanced effective stress models. The first part of the research focuses on realistic 3-D finite element analyses (using AbaqusTM Standard) of a conductor (steel pipe pile) embedded within soft marine clay subjected to large lateral deformations caused by drift/drive-off of a drilling vessel. The proposed analyses use coupled pore pressure-displacement procedures together with the MIT-E3 soil model to represent the anisotropic, non-linear and inelastic effective stress-strain-strength properties of deepwater marine sediments with input parameters derived from a series of laboratory element tests performed on reconstituted Gulf of Mexico (GoM) clay. The numerical predictions are evaluated through comparison with experimental results from centrifuge tests with a well-instrumented model conductor. The FE results accurately predict the measured bending moment distribution along the length of the conductor and the spread of plastic strains within the conductor itself. The study has also shown the effects of soil behavior on local pile-soil interactions, enabling simplified analyses using macro-elements. The FE results have been used to calibrate input parameters for BWGG framework (Gerolymos & Gazetas, 2005), the Bouc-Wen (BW) model extended by Gerolymos and Gazetas (GG), that simulates generalized hysteretic pile-soil interactions and allows for degradation in soil resistance associated with geometric non-linearities. The second application considers the effects of partial drainage for large deformation, quasi-static piezocone penetration in clay. The proposed axisymmetric FE analysis procedure introduces automated remeshing and solution mapping technique (similar to RITSS; Hu & Randolph, 1998) within a commercial FE solver. We have analyzed the penetration resistance for a piezocone device using two elasto-plastic soil models (MCC, MIT-E3) and the recent elasto-viscoplastic MIT-SR soil model (Yuan, 2016) over a range of steady penetration velocities. The MCC predictions are in very good agreement with laboratory measurements of tip resistance and penetration pore pressures measured in centrifuge model tests in reconstituted kaolin. The results from more advanced soil models illustrate the impacts of anisotropic, rate dependent soil behavior on penetration tests in natural clays and are within the range of empirical measurements. The proposed analyses provide a complete framework that can now be used to investigate effects of partial drainage that occurs in piezocone tests for soils (such as silts) of intermediate permeability.<br>by Zhandos Y. Orazalin.<br>Ph. D.

Numerical modeling using finite element method (FEM) is well-recognized as a powerful method for both engineers and researchers to solve boundary value problems. In the modeling of geotechnical problems, the analyses are often limited to simple static problems with either steady-state effective or total stress approach while the transient response (development and dissipation of excess pore water pressure, uex) is seldom considered. Besides, infinitesimal small soil deformation is usually assumed. The simulation is further complicated when the soil-structure interaction problems involve significant soil displacements; like a pile subject to negative skin friction (NSF) and a cone/pile penetration. However, conventional FEM analysis prematurely terminates due primarily to excessive mesh distortion. One could see that simulating a transient problem with large deformation and distortion remains a great challenge. In this study, advanced FE simulations are performed to give new insights into the problems of (1) a pile subject to NSF; and (2) a cone penetration. The transient response of the NSF problem is modeled with the fluid-coupled consolidation technique and geometric nonlinearity. The fluid-coupled cone penetration problem is modeled with a newly developed adaptive approach. The NSF and cone penetration simulations involve complex soil-structure interface modeling. Two types of modified interface responses are developed and verified which consider fluid coupling. The developed algorithm is applied to back analyze a case history of a pile subject to NSF induced by surcharge loading. Promising results were shown. Development of dragload and neutral plane (NP) with time is studied. NP locates at 75% of the pile embedded length (D) in long-term. Next, a parametric study is performed to investigate the influences of pile geometries, ground compressibility and loading conditions towards the pile responses. The long-term NP locates at around 0.55D to 0.65D in the studied engineering scenarios. The maximum downdrag can be up to 10% of the pile diameter. NP shifts upward when the head load increases. A simple design chart is proposed which helps engineers to estimate the long-term axial load distribution. An illustrative example is given to demonstrate the application and performance of the chart. The study is extended to investigate the cone penetration problem. An advanced adaptive method is developed and implemented into the FE package ABAQUS to resolve the problems of numerical instability, excessive mesh distortion and premature termination. The proposed method is verified by modeling a ground consolidation problem. Next, total stress back analysis of cone penetration is conducted with the proposed method. The development of cone factor predicted by the proposed method gives a better match with the laboratory result when comparing with the built-in ALE method. Next, the development and dissipation of uex during cone advancing with the proposed method and fluid-coupled technique is investigated. uex develops dramatically around the cone tip. The soil permeability is back calculated from the dissipation test and agrees well with the input value. It is believed that the construction effects of a press-in pile and the subsequence NSF on that pile can be modeled by utilizing the finding of this study.<br>published_or_final_version<br>Civil Engineering<br>Doctoral<br>Doctor of Philosophy

Soil behaviors are quite complex under dynamic loadings, such as wave loading, earthquake loading, etc, but they share common characteristics that the soil is subjected to considerable principal stress rotations (PSR). PSR can generate plastic deformation even without a change of principal stress magnitudes. Continuous PSR can also generate excess pore water pressures and cumulative shear strains in undrained condition. Therefore, the PSR from the dynamic loadings can accelerate undrained soil liquefaction because it can cause cumulative plastic volumetric deformations. Ignoring PSR induced deformation may lead to unsafe design. It is therefore important to understand the soil behaviors under cyclic loadings with the PSR and take account of this PSR impact in the numerical simulations of corresponding geotechnical problems. Although researchers have recognized the importance of the PSR in real geotechnical problems under diverse loading conditions and conducted extensive experimental studies, there are limited considerations of the PSR impact on numerical simulations of boundary value problems. Moreover, most of the constitutive models widely-used in the numerical investigations at present cannot simulate this PSR behavior properly. Therefore, a new kinematic hardening soil model (PSR model) developed on the basis of a well-established model with bounding surface concept is used to simulate the PSR behavior in this research. It can take account of the PSR impact by treating the stress rate generating the PSR independently. To investigate the impacts of PSR in numerical simulations of geotechnical problems, the PSR model is implemented into both the single element and finite element analysis of a series of geotechnical problems by a constitutive model subroutine written in Fortran. In this subroutine, an explicit substepping integration algorithm with automatic error controls is used to perform the constitutive formulations. The imposed strain increment can be automatically divided and the sizes of the sub-increments are also automatically determined based on the prescribed error tolerance in this numerical integration scheme. The single element analyses include the simulations of the triaxial and hollow cylinder tests with monotonic, rotational and torsional loading paths, while the finite element analyses consist of the simulations of the centrifuge experimental tests under wave loadings and earthquake loadings. The predicted results by using the soil model with and without considering the PSR impact, as well as the experimental results will be compared. From these single element and finite element analyses, it is evident that the rotational, torsional and dynamic loadings such as wave and earthquake loadings can produce the PSR and non-coaxiality in the soil. The comparisons between the predicted results from the modified PSR model, the original model, and the laboratory results from these experimental tests all show that although the original model can reflect some non-coaxiality, it can produce very limited build-up of pore water pressure and cumulative shear strain under cyclic loading path. However, due to consideration of the PSR impact, the modified PSR model can generate higher pore water pressure and shear strain than the original model, thus bringing the soil to the liquefaction and agrees better with the experimental results. Therefore, it is important to consider the PSR effect in the simulation of geotechnical problems such as wave-seabed interactions and the earthquake induced liquefactions, and the PSR model presented in this research has a great ability and plays an important role in these numerical simulations.

Most downhole tools in use today measure properties immediately adjacent to the borehole, and as such, only a small portion of the subsurface volume is known with any degree of certainty. When dealing with geologic situations which are characteristically heterogeneous, the engineer often requires more information than what present tests can provide. Tomography is an in-situ testing method that allows the generation of a two dimensional subsurface image by reconstructing material property variations between boreholes. It is essentially a solution to the inverse problem where signals are measured and, through computer manipulation, are used to infer material contrasts in the subsurface. For the purposes of this thesis, a two dimensional configuration is used to demonstrate and evaluate the tomographic technique with source and receiver locations positioned at intervals down adjacent and nearly vertical boreholes. Both iterative and direct matrix solution methods are used to evaluate the use of seismic and groundwater flow data for subsurface tomography. The iterative methods include a variation of the classical algebraic reconstruction technique (CART), a modified version of the ART algorithm (MART), and a modified version of the ART algorithm using the Chebyshev norm criterion (LART). The purpose of the iterative tests is to determine the best algorithm for signal reconstruction when data noise and different damping parameters are applied. The matrix methodologies include a constrained L¹ linear approximation algorithm and singular value decomposition routines (SVD). These methods solve the set of linear equations (Ax = b) which the tomographic techniques produce. The purpose of this stage of testing is to optimize a direct method of solution to the sets of linear equations such that different forms of anomaly can be discerned. Numerous synthetic seismic and groundwater data sets are used by both iterative and matrix algorithms. Seismic test data sets are generated by calculation of transit times through materials of known seismic velocity. Groundwater test data sets are generated by drawdown analyses and finite element procedures. All algorithms demonstrate a reasonable ability at reconstructing sections which closely re-sembled the known profiles. Vertical anomalies, however, are not as well defined as horizontal anomalies. This is primarily a result of incomplete cross-hole scanning geometry which also affects the rank and condition of the matrices used by the direct forms of solution. The addition of Gaussian noise to the data produces poor reconstructions regardless of the type of algorithm used. This emphasizes the fact that tomographic techniques require clear and relatively error-free signals.<br>Science, Faculty of<br>Earth, Ocean and Atmospheric Sciences, Department of<br>Graduate

Tese (doutorado)—Universidade de Brasília, Faculdade de Tecnologia, Programa de Pós-Graduação em Geotecnia, 2016.<br>Submitted by Fernanda Percia França (fernandafranca@bce.unb.br) on 2016-11-24T15:02:37Z No. of bitstreams: 1 2016_MarceloAlejandroLlanoSerna.pdf: 4597920 bytes, checksum: d1815fd743fd8c6b5479aca9482e7686 (MD5)<br>Approved for entry into archive by Raquel Viana(raquelviana@bce.unb.br) on 2017-01-28T19:32:06Z (GMT) No. of bitstreams: 1 2016_MarceloAlejandroLlanoSerna.pdf: 4597920 bytes, checksum: d1815fd743fd8c6b5479aca9482e7686 (MD5)<br>Made available in DSpace on 2017-01-28T19:32:06Z (GMT). No. of bitstreams: 1 2016_MarceloAlejandroLlanoSerna.pdf: 4597920 bytes, checksum: d1815fd743fd8c6b5479aca9482e7686 (MD5)<br>O objetivo deste trabalho é investigar os mecanismos de alguns problemas geotécnicos submetidos a grandes deformações, e mais especificamente o cone de penetração e escorregamentos na área de estabilidade de taludes. O fenômeno de grandes deformações em Geotecnia pode ser observado em problemas de ensaios de campo como SPT, CPT, DMT; ensaios de laboratório como o ensaio de cone e de palheta; em aplicações práticas como a cravação de estacas e em encostas após a ruptura de um talude. Uma das principais limitações na prática da engenharia geotécnica é que as formulações tradicionais para o cálculo de estruturas dependem da hipótese de pequenas deformações. Na última década, com o aumento da capacidade computacional e surgimento de novos métodos numéricos, tornou-se factível a modelagem numérica de problemas de grandes deformações, gerando a possibilidade de estudá-los em maior detalhe. Este trabalho centra-se na aplicação do Método do Ponto Material (MPM). O MPM é uma ferramenta numérica que adota um esquema de discretização Euleriano-Lagrangiano, o que fornece um esquema sofisticado para resolver o balanço de momento linear quando se observam grandes deformações. O método foi aplicado à análise de ensaios de penetração de cone em laboratório e a problemas reais de escorregamentos de taludes com grandes movimentos de massa. Inicialmente, foram feitos ensaios diretos e indiretos de resistência ao cisalhamento em amostras de caulim. O programa de ensaios de laboratório inclui o ensaio de palheta, ensaio de cone, ensaio de compressão oedométrica e ensaio de compressão triaxial convencional. Como produto dos ensaios de laboratório, foram propostas algumas relações entre parâmetros de estados críticos e o ensaio de queda de cone. Também baseado nos ensaios de laboratório, o programa NairnMPM foi testado e calibrado para resolver problemas geotécnicos simples como o ensaio de cone e o colapso de uma coluna de solo. Depois disso e com o intuito de verificar a capacidade do MPM para resolver problemas de grande escala, foram simulados os escorregamentos de taludes na barragem de Vajont, na Itália, e na rodovia Tokai-Hokuriku, no Japão. Finalmente, foi testado o processo de modelagem do escorregamento de Alto Verde, na Colômbia, e as variáveis dinâmicas previstas no modelo foram usadas no cálculo de risco. Os resultados se ajustaram muito bem às observações de campo, destacando a potencialidade do MPM como ferramenta prática na modelagem de vários problemas de grandes deformações na engenharia geotécnica.<br>The goal of this work is to investigate the mechanisms of various geotechnical problems subjected to finite strains, more specifically the fall cone test and run-out process during landslides. Large deformation phenomena may be observed in field testing such as SPT, CPT, DMT; laboratory testing such as fall cone test, mini-vane test, and practical problems such as pile driving and run-out process during landslides. The main limitations in the practice of geotechnical engineering are due to the fact that a wide number of design frameworks are based on the small strain hypothesis. In the last decade, with the increasing computational capacity and the development of novel numerical methods; solving large deformation models have become feasible. This fact allows studying in detail a wide number of phenomena in geotechnics. This work focuses on the application of the Material Point Method (MPM). The MPM is a numerical tool that adopts a Eulerian-Lagrangian scheme. Moreover, it allows a solid framework to solve the linear momentum balance when finite strains are observed. The method was used in the simulation of the fall cone test and real scale mass movements in landslides. Initially, direct and indirect shear strength measurements on kaolin clay were performed. The laboratory testing program included mini-vane shear test, fall cone test, oedometric compression, and conventional triaxial compression test. As a result of the laboratory testing, interesting relationships between the critical state parameters and the fall cone were established. Furthermore, NairnMPM open source code was tested and calibrated using the laboratory results to later solve simple geotechnical problems such as fall cone test and the collapse of a soil column. Afterwards, the possibility of simulating real-scale problems in landslides was addressed. The slope failure in Vajont, Italy, and Tokai-Hokuriku Expressway, Japan, were considered. Finally, the framework was tested in a landslide in Alto Verde, Colombia. The computed dynamic quantities were used in risk assessment of landslides. The results matched very well with field observations highlighting the potential of using MPM as a practical tool for modelling various problems involving large strains in geotechnical engineering.

CONSELHO NACIONAL DE DESENVOLVIMENTO CIENTÍFICO E TECNOLÓGICO<br>O presente trabalho trata da aplicação da análise limite numérica (ALN) a problemas geotécnicos. Os meios (solo ou rocha) são considerados como contínuos convencionais e como contínuos de Cosserat. Da aplicação da formulação mista da análise limite e da discretização do meio por uma malha de elementos finitos é obtido um problema de programação matemática (PM). A aplicação desta metodologia nos contínuos de Cosserat (2D) fornece problemas de programação linear (PL) e nos contínuos convencionais (2D e 3D), problemas de programação não-linear (PNL). A solução do problema de PM foi através dos programas de otimização: LINDO (PL), LINGO (PNL), MINOS (PNL) e LANCELOT (PNL). Também foram implementados os algoritmos não lineares -Quase Newton com deflexão- e -Han-Powell-. A formulação é validada em problemas cuja solução analítica é conhecida ou em dados experimentais. Estes exemplos mostram a rapidez e a eficácia da ALN para a determinação da carga de colapso e do mecanismo de ruptura do problema.<br>The present work treats of the application of the numerical limit analysis (NLA)to geomechanics problems. The soil or rock mass is considered as conventional continuous and Cosserat continuous. A mathematical programming (MP) problem is obtained through the application of the mixed formulation of limit analysis and the finite elements mesh. The application of this methodology in the Cosserat continuous (2D) supplies linear programming (LP) problems and in the conventional continuous (2D and 3D) nonlinear programming (NLP) problems. The solution of the problem of MP was through the LINDO (LP), LINGO (NLP), MINOS (NLP) and LANCELOT (NLP) programs. It was also implemented nonlinear algorithms -Quasi-Newton feasible point method- and -Han-Powell-.The formulation is validated in problems whose analytic solution is known or in experimental data. These examples show the speed and the effectiveness of NLA for the determination of the collapse load and of the mechanism of rupture of the problem.<br>EL presente trabajo trata de la aplicación del análisis límite numérica (ALN) a problemas geotécnicos. Los medios (suelo o roca) son considerados como contínuos convencionales y como contínuos de Coserat. De la aplicación de la formulación mixta del análisis límite y de la discretización del medio por una malla de elementos finitos se obtiene un problema de programación matemática (PM). La aplicación de esta metodología en los contínuos de Coserat (2D) nos lleva a problemas de programación lineal (PL) y en los contínuos convencionales (2D y 3D), problemas de programación no lineal (PNL). La solución del problema de PM fue a través de los programas de optimización: LINDO (PL), LINGO (PNL), MINOS (PNL) y LANCELOT (PNL). También fueron implementados los algoritmos no lineares quase- Newton con deflexión y Han Powell . Se evalúa la formulación propuesta en problemas donde se conoce la solución analítica o en datos experimentales. Estos ejemplos muestran la rapidez y la eficacia de la ALN para la determinación de la carga de colapso y del mecanismo de ruptura del problema.

PONTIFÍCIA UNIVERSIDADE CATÓLICA DO RIO DE JANEIRO<br>COORDENAÇÃO DE APERFEIÇOAMENTO DO PESSOAL DE ENSINO SUPERIOR<br>PROGRAMA DE EXCELENCIA ACADEMICA<br>Em problemas geotécnicos podem ocorrer grandes deformações devido a chuvas prolongadas, sismos, deslizamentos de encostas, etc. Material point method (MPM) é um método de solução baseado no Método dos Elementos Finitos (MEF) que oferece vantagens para o cálculo estático e dinâmico que envolve deformações desse tipo. O objetivo desta dissertação é utilizar o MPM em problemas geotécnicos em condições estáticas e dinâmicas. Esta pesquisa mostra o procedimento de analises do MPM para a condição não acoplada (só solido sem presença de água) e depois acoplada. Para a revisão matemática de MPM se faz antes um resumo da teoria do MEF na metodologia de conservação de quantidade de movimento. Nestas duas formas de resolver os problemas geotécnicos foram expostos três exemplos simples. O primeiro é uma coluna de solo simulado sob os fundamentos da elasticidade, com o objetivo de verificar o deslocamento vertical pelo peso próprio. Isto foi resolvido mediante quatro diferentes métodos: analítica, MEF por resíduos ponderados, MEF por conservação de quantidade de movimento, e MPM. Todos eles consideram somente a fase solida. No segundo exemplo, tem-se solo na geometria de quadrado de lado 1 metro, onde busca-se obter as poropressões quando atingir a condição permanente enquanto os deslocamentos ocorrem ao longo do tempo; ou seja, a análise é acoplada e é resolvida pelo método MPM. Para uma aplicação mais realista, foi feita a análise (não acoplada) da barragem Palo Redondo, pertencente ao projeto Chavimochic, localizada na região de La Libertad, Perú. Nesta análise dinâmica considerou-se dois modelos constitutivos: Elástico e Mohr Coulomb, além de um sismo.<br>In geotechnical problems can happen large strains because of prolonged rains, earthquake, slide slope, etc. Material point method is a solution method based on the finite element method (FEM) which offers advantages for static and dynamic calculation that involve that kind of strains. The objective in this dissertation is to use the MPM in geotechnical problems in statics and dynamics conditions. This research shows the analysis procedure of MPM for uncoupled condition (only solids, without water) and then coupled. Before the mathematical theory of MPM, a review of the theory of FEM in the conservation of quantidade de movimento is done. In this two methodology were raised three examples. The first one is a soil column that was modeled elastically, in which the main objective in to analyze the vertical displacement because of own weight. This was solved by four different methods: analytically, FEM weighted residual, FEM conservation of momentum, and MPM. All of them consider only the solid phase. The second example is a square of soil with side 1 meter, where the objective is to know the pore-pressure in the permanent condition and at the same time the vertical displacement were generated, it means that the analysis is coupled and were solved by MPM. In order to make a more realistic application, Palo Redondo dam is analyzed (uncoupled condition), which belongs to the Chavimochic project located in La Libertad, region of Perú. This dynamic analysis was done considering two constitutive models: Elastic and Mohr Coulomb, additionally seismic forces.

A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy in Science to the Faculty of Engineering and the Built Environment, University of the Witwatersrand, Johannesburg, 1985<br>This volume contains a record of research carried out over the past two decades into problems associated with civil engineering and mining construction materials. Although the various parts of the work were initiated as a result of problems that arose in the Witwatersrand region, the results of the research have in many cases evoked intense interest from other parts of the world. For example, the work on soluble salts in road bases has been taken up in Saudi Arabia and other desert regions, while that on the stability of waste rock dumps has been adopted as a basis for rock dump design in the United States of America. The research revolves entirely about materials, usually, waste, either produced by the mines and reused or disposed of by civil engineers; or reused to provide support by the mines · themselves. The main aim and end result of the research has been a more effective and efficient use of materials and better protection and control of the local environment.<br>Andrew Chakane 2021

博士<br>國立交通大學<br>土木工程系所<br>101<br>Direct current (DC) electrical resistivity method has been developed for almost a century. It has evolved from the 1D, 2D, to more recently the true 3D method. However, field conditions often obstruct 3D surveys and 2D electrical resistivity tomography (ERT) remains to be the state of the practice in geotechnical investigation due to its simplicity in field works and less space requirement. The 2D ERT has been widely applied, but the uncertainty behind the obtained vivid resistivity image is not clear. Its accuracy, spatial resolution, and possible pitfalls should be understood to avoid misinterpretation. The objectives of this study are to investigate the spatial resolution of 2D ERT measurements and its potential limitations. The challenges for engineers to interpret the 2D ERT results are manifested via two case studies including the Hsinchu fault and Hsinsan reservoir leakage investigations. The spatial resolution, reliability of inverted resistivity section, boundary effect, and 3D effect are identified as unclear problems which may significantly affect the interpretation of 2D ERT results. The resolving ability of 2D ERT were qualitatively studied by numerical simulations of 2D ERT surveys in various geological conditions including single horizontal layer, horizontal sandwiched layer, single vertical layer, vertical sandwiched layer, inclined layer, and block-in-matrix structure. The results of 2D survey may be distorted in conditions which violate the assumption of 2D structure and infinitive boundary. This study further used 3D numerical modeling to assess how 3D effects might distort the 2D inversion in some typical scenarios. The Hsinchu fault and Hsinsan reservoir field cases were re-visited by 3D modeling to verify the results of 2D ERT under 3D conditions. From the case studies and numerical simulations, it was shown that 3D and boundary effects may significantly influence the results of 2D ERT resulting in false interpretations. The behavior and pattern of 3D and boundary effects are revealed by the 3D modeling. Suggestions are made accordingly to facilitate reasonable interpretations of 2D ERT and avoid false or over interpretations.

The introductions of large digital computers in the field of engineering have rendered possible the solution of a wide variety of problems without the need to violate the equilibrium and compatibility. The major requirement for such analysis is a good constitutive model that represents the stress strain behaviour of the materials in an accurate way. Nowadays for most of the geotechnical engineering applications the elastoplastic models like Mohr Coulomb model are widely used. All the existing constitutive models which represent the plastic behaviour of soil are developed from the fundamentals of classical theory of plasticity. The classical theory of plasticity is always associated with the concept of yield surface and potential surface to represent the plastic behaviour. The definition of yield surface depends on the location of the yield point. But in practical sense it is very difficult to find out the exact yield point for a material. The expression for yield and potential surfaces are simply mathematical expressions formulated for computational efficiency. Experimentally it is very difficult to find out the yield surface in the case of three dimensional stress spaces. The classical theory of plasticity is developed based on the mechanical process. It is believed that a theory which violates the thermodynamic principle is not able to represent the material behaviour accurately. the initial stage and combined to give the final state of stress. It was proved that the equation proposed by Wu and Wang (1983) can be used to represent the triaxial behaviour of sand very well. The dilation and densification behaviour can be predicted very well with the endochronic constitutive equations. The principal aim of this work is to implement the endochronic constitutive equation in the FLAC3D model library like any other constitutive model and validate it with the triaxial test data. After implementation and validation, the application of the particular constitutive model is extended to some practical geotechnical engineering problems like the stresses and displacements around an underground opening such as tunnels, surface settlement due to shallow tunneling, stress distribution below the footing, settlement analysis of footing on various foundation beds such as sand, clay and sand overlying clay bed, lateral displacement of the secant pile wall due to excavation and the force developed in the horizontal support etc. All the three problems validate the model with the analytical, experimental and field data respectively. The equation proposed by Wu and Wang (1983) is used for the present study. In order to validate the equation proposed by Wu and Wang (1983), MatLab programming is used. The hydrostatic, deviatoric and volumetric behaviour is obtained separately using the concerned equations. The equation is coded in the MatLab and analysis is done for a triaxial element test. Both drained and undrained analyses were done in order to study the behaviour. The pore pressure developed is captured separately with the equation proposed by Geoffrey et al (1975). The results obtained from the analysis of the MatLab were compared with that of the experimental results. The analysis shows that the equation captures the least plastic behaviour well for the triaxial element test. The dilation and densification behaviour obtained using the respective equation shows that it matches well with the experimental results. A parametric study is also conducted in MatLab to see how the parameters affect the stress strain and volumetric behaviour of the sand. The parametric study conducted with the MatLab shows that most of the parameters involved in the equation affect the plastic part of the stress strain curve rather than the initial elastic part. User defined constitutive model was written in visual C++ and compiled as DLL (Dynamic Link Library) files that will be loaded whenever it is needed in FLAC3D. In visual C++, header and source files were written by incorporating the constitutive equation proposed by Wu and Wang (1983), defining the variables and other functions etc, and a dynamic link library is created, which is then integrated into the 3D finite difference code FLAC3D using the CPPUDM module to simulate the stress strain behaviour of the materials. CPPUDM module is an additional option in FLAC3D to implement the user defined constitutive models. The visual C++ code was written in the form of incremental stress strain relationship. The model acts like any other constitutive model in the FLAC3D model library and can be loaded whenever it is required. For the validation of the model in FLAC3D, the data for the MatLab simulation were used. Both drained and undrained tests were simulated with the model. The results obtained from the analysis shows that by suitably selecting the parameters the model can simulate the stress strain behaviour of sand very well. The volumetric and deviatoric behaviour were observed and is matching well with the experimental data. In the case of the undrained test the pore pressure generation is well captured by the equation proposed by Geoffrey et al (1975). In urban areas the construction of shallow tunnels results in excessive settlements of the ground surface and thereby causes damage to the existing above ground structures. In order to minimize the settlements and to reduce the impact due to that, a prior analysis of the displacements and stresses around the opening is very important. Nowadays numerical analysis is widely used for the analysis of such structures. The most important requirement of such analysis is a constitutive model that can represent the unloading behaviour around the tunnel opening of sand very well. Here the endochronic constitutive model implemented in the FLAC3D model library is used to evaluate the stresses and displacements around the tunnel. In the analysis the tunnel is simulated as a cylindrical hole in an infinite medium with the in situ stress. The stresses at the springing line was observed and compared with the analytical solution. The results show that the results are matching well with the analytical results. The comparison of the results with that obtained using the Mohr Coulomb model is also done to see how the model differs from a widely used plastic model. By slightly adjusting the parameters the results obtained from both the models are in well agreement. The strain softening effect which is predominant around an underground opening due to the loosening of soil mass is well captured by the endochronic model compared to the Mohr Coulomb model. The settlement analysis shows that the model is almost in close agreement with the closed form solution proposed by Oteo & Sagaseta (1982) and the results obtained with the Mohr Coulomb model. The settlement trough formed for various shapes is wider and deeper than the Mohr Coulomb model. The vertical stress distribution around the opening of the tunnel is studied with varying the shape of the openings using the proposed constitutive model. The results obtained were compared with that of the Mohr Coulomb model. The slightly higher values in the case of endochronic model are basically due to its plastic nature. The displacement and stresses in the axial direction (along the excavation) is observed with the model. In the case of shallow tunnel the surface get influenced by the loosening of the soil mass which necessitates the use of the support system. The study shows that the model can be used for the simulation of underground opening like tunnel and will capture the behaviour well. Footings are structures used to support the buildings constructed above the ground. The settlement analysis of footings is very important when we consider the stability of the structures supported by it. The vertical stress distribution below the footing is studied using the endochronic constitutive model and compared with the analytical solution proposed by Boussinesq (1885). In the elastic range the model shows matching results with the Boussinesq’s solution. The settlement analysis of footing on various foundation beds such as sand, clay and sand overlying the clay bed were studied using the endochronic constitutive model implemented in the FLAC3D model library. The experimental data conducted in our lab (Sireesh (2006)) was used for the study. The results show that with the chosen parameters the results obtained with the endochronic model are in good agreement with the experimental data. The Mohr Coulomb model over predicts the results. This shows higher modulus value for the Mohr Coulomb model. By conducting the parametric study it was seen that by reducing the value of modulus for the Mohr Coulomb model, the results are in good agreement with the experimental value. The displacement and stress contours obtained for the two models were compared. By analyzing the displacement contours it is seen that the Mohr Coulomb model shows uniform settlement. In the case of endochronic model uniform settlement is observed for about 5% settlement that is in the elastic range. After a certain strain level the displacement contours are tilted more towards one side showing the rotational failure. Here the endochronic model captures the anisotropic behaviour associated with the materials like sand at higher strain level. This result is a concrete evidence that the model can capture the realistic behaviour very well compared to any other model. Even though the model developed is for sand its application can be extended to clay also. The size and shape of the footing is varied to study its effect on the pressure settlement curve. The analysis is done with square shape of 150mm side and circular shape of 150mm diameter. As there is not much variation in the area of influence, the shape has little influence on the pressure settlement curve. As the size of the footing increases the settlement increases for a given pressure. A parametric study is conducted by varying the modulus value used. The study shows that as the modulus value increases, the settlement reduces for a given bearing pressure. The endochronic model analysed with the lower modulus value shows that the model predicts the perfectly plastic behaviour, here the settlement increases for low value of bearing pressure. The application of endochronic model for the simulation of complex geotechnical engineering problems like footings is highly explored in the study. Nowadays most of the infrastructure facilities are concentrated towards the underground space. The excavation and construction of such structures in the urban areas results in damage to the existing above ground structures if the construction is done in close proximity to the structures. In the present study a staged excavation of an underground construction for the Bangalore metro project is simulated with the endochronic constitutive model. In the Bangalore metro project the excavation for the underground station is done at the cricket stadium site. At the site there are two major buildings such as the six storied Hindustan Aeronautical Limited building and 100 years old BSNL masonry building. To minimize the impact on these structures were a major concern during the construction of the work. A robust support system consists of secant pile walls, soldier piles and horizontal struts are installed at the site. The OSV method known as the Onsite Visualization and monitoring is conducted to minimize the damage to the existing structures and the accidents at the construction site. Sensors are connected to LEDs which show change in color when the displacements and forces cross the triggered value. The field instrumentation is done with inclinometers, tilt meters and strain gauges connected to the sensors to observe the lateral deformation of the secant pile wall, tilt of the HAL building and the forces developed in the horizontal struts. The monitoring of field data is done for a period of five months from July to November. From the analysis of the field observed data it is clear that the support system provided were strong enough to resist the forces due to unloading. The lateral deformation of the secant pile wall and the forces developed in the strut were numerically analysed using the endochronic constitutive model and the results were compared with the field monitored data. The results show that the model captures the behaviour very close to the field data for a slightly higher modulus than that reported in the geotechnical report (BMRC report). This may be due to the fact that the value of modulus calculated experimentally might have some error. The analysis with the Mohr Coulomb model shows that the model over predicts the results very close to the surface of the excavation. This indicates that the influence of load is more on that particular depth for the Mohr Coulomb model. But the stiffness of the lateral support system is uniform throughout the depth; the endochronic model predicts the result more accurately than the Mohr Coulomb model. The strut forces developed in the horizontal support system is observed using the two models. The strut forces in the field is affected by so many factors such as the temperature variation, stages of excavation and other live loads acting on the site, so an exact comparison with the field data is quite difficult. The analysis shows that even though it is difficult to simulate the exact three dimensional nature of the problem in the present study the endochronic constitutive model captures the behaviour very well. The results obtained shows that the endochronic constitutive model implemented in the FLAC3D model library provides a very promising solution like any other constitutive model. As the theory is based on the irreversible law of thermodynamics and the formulation of the constitutive equation are based on the internal energy concept it can represent the material behaviour in accordance with the laws of continuum mechanics. The anisotropic behaviour of soil at higher strain level is well represented through the footing problem. The endochronic constitutive model is a very simple one to simulate the stress strain behaviour of the materials without the concept of yield surface; the parameters used in the equation can be obtained directly from a single triaxial stress strain plot. This study highlights the importance of a model without the concept of yield surface to capture the stress strain behaviour of any materials. Since the model is of completely plastic nature it has its own uniqueness in capturing the material behaviour due to loading and unloading.

碩士<br>國立臺灣科技大學<br>營建工程系<br>106<br>Probabilistic Machine Learning Model: Extended Monte Carlo Simulation for Solving Geotechnical Engineering Problems Thesis Advisor: Jui-Sheng Chou Graduate Student: Lucia Dewi Santoso ABSTRACT Machine learning (ML) is a data mining technique that integrates the principles of statistics, pattern recognition in machine learning, and data base systems. However, prediction, a powerful function of ML, mostly uses deterministic inputs to develop a deterministic prediction model; this model definitely cannot deal with uncertainty of the deterministic output result. Input data commonly includes outliers that contaminates the predictors, raising a question: with what do the inputs support the probability or certainty of the predictions? Monte Carlo simulation (MCS) is a probability-informed approach that is used in analyzing system reliability and risk because most of expected assumptions are taken in account in simulating the risk of the system that reasonably to be believed. Owing to the great effectiveness of machine learning in forecasting, this study attempts to construct a method that is based on Monte Carlo simulation to demonstrate a probabilistic machine learning model. This novel approach integrates the pattern recognition aspect of machine learning, empirical statistical rules, curve fitting, and data base systems, which are constructed by random samplings in the MCS to predict the probabilities of the outputs. The results indicate that probabilistic machine learning not only can gives the output variable of interest, along with a probability distribution, but it also performs faultless classification prediction that is based on a resampled dataset. Three case studies in the field of geotechnical engineering, which involve the peak shear strength of reinforced soil, factor of safety of slope, and the stability of a slope are presented to establish the effectiveness of this simulation method. The results reveal that, in addition to linking probabilistic information with outputs, the probabilistic machine learning model can significantly improve all of the prediction models that are used in demonstrated geotechnical cases. Keywords: probabilistic machine learning; Monte Carlo simulation; data mining; engineering prediction; regression; classification.

In numerical simulations of geomechanics problems, a grand challenge consists of overcoming the difficulties in making accurate and robust predictions by revealing the true mechanisms in particle interactions, fluid flow inside pore spaces, and hydromechanical coupling effect between the solid and fluid constituents, from microscale to mesoscale, and to macroscale. While simulation tools incorporating subscale physics can provide detailed insights and accurate material properties to macroscale simulations via computational homogenizations, these numerical simulations are often too computational demanding to be directly used across multiple scales. Recent breakthroughs of Artificial Intelligence (AI) via machine learning have great potential to overcome these barriers, as evidenced by their great success in many applications such as image recognition, natural language processing, and strategy exploration in games. The AI can achieve super-human performance level in a large number of applications, and accomplish tasks that were thought to be not feasible due to the limitations of human and previous computer algorithms. Yet, machine learning approaches can also suffer from overfitting, lack of interpretability, and lack of reliability. Thus the application of machine learning into generation of accurate and reliable surrogate constitutive models for geomaterials with multiscale and multiphysics is not trivial. For this purpose, we propose to establish an integrated modeling process for automatic designing, training, validating, and falsifying of constitutive models, or "metamodeling". This dissertation focuses on our efforts in laying down step-by-step the necessary theoretical and technical foundations for the multiscale metamodeling framework. The first step is to develop multiscale hydromechanical homogenization frameworks for both bulk granular materials and granular interfaces, with their behaviors homogenized from subscale microstructural simulations. For efficient simulations of field-scale geomechanics problems across more than two scales, we develop a hybrid data-driven method designed to capture the multiscale hydro-mechanical coupling effect of porous media with pores of various different sizes. By using sub-scale simulations to generate database to train material models, an offline homogenization procedure is used to replace the up-scaling procedure to generate path-dependent cohesive laws for localized physical discontinuities at both grain and specimen scales. To enable AI in taking over the trial-and-error tasks in the constitutive modeling process, we introduce a novel “metamodeling” framework that employs both graph theory and deep reinforcement learning (DRL) to generate accurate, physics compatible and interpretable surrogate machine learning models. The process of writing constitutive models is simplified as a sequence of forming graph edges with the goal of maximizing the model score (a function of accuracy, robustness and forward prediction quality). By using neural networks to estimate policies and state values, the computer agent is able to efficiently self-improve the constitutive models generated through self-playing. To overcome the obstacle of limited information in geomechanics, we improve the efficiency in utilization of experimental data by a multi-agent cooperative metamodeling framework to provide guidance on database generation and constitutive modeling at the same time. The modeler agent in the framework focuses on evaluating all modeling options (from domain experts’ knowledge or machine learning) in a directed multigraph of elasto-plasticity theory, and finding the optimal path that links the source of the directed graph (e.g., strain history) to the target (e.g., stress). Meanwhile, the data agent focuses on collecting data from real or virtual experiments, interacts with the modeler agent sequentially and generates the database for model calibration to optimize the prediction accuracy. Finally, we design a non-cooperative meta-modeling framework that focuses on automatically developing strategies that simultaneously generate experimental data to calibrate model parameters and explore weakness of a known constitutive model until the strengths and weaknesses of the constitutive law on the application range can be identified through competition. These tasks are enabled by a zero-sum reward system of the metamodeling game and robust adversarial reinforcement learning techniques.

Limit analysis is a very powerful tool to find accurate solutions of several geotechnical stability problems. This analysis is based on the theory of the plasticity and it provides two limiting solutions within lower and upper bounds. With the advancement of the finite elements and different robust optimization techniques, the numerical limit analysis approach in association with finite elements is becoming quite popular to assess the stability of various complicated structures. The present thesis deals with the formulations and the implementation of the finite element limit analysis to obtain the solutions of different geotechnical axisymmetric stability problems. The objectives of the present thesis are twofold: (a) developing limit analysis formulations in conjunction with linear and nonlinear optimizations for solving axisymmetric stability problems related with soil and rock mechanics, and then (b) implementing these axisymmetric formulations for solving various important axisymmetric stability problems in geomechanics. Three noded linear triangular elements have been used throughout the thesis. In order to solve the different problems, the associated computer programs have been written in MATLAB. With reference to the first objective of the thesis, the existing finite element lower bound axisymmetric formulation with linear programming has been presented. A new technique has also been proposed for solving an axisymmetric geomechanics stability problem by employing an upper bound limit analysis in combination with finite elements and linear programming. The method is based on the application of the von-Karman hypothesis to fix the constraints associated with the magnitude of the circumferential stress (), and finally the method involves only the nodal velocities as the basic unknown variables. The required computational effort becomes only marginally greater than that needed for an equivalent plane strain problem. The proposed methodology has been found to be computationally quite efficient. A new lower bound axisymmetric limit analysis formulation, by using two dimensional finite elements, the three dimensional Mohr-Coulomb (MC) yield criterion, and nonlinear optimization has also been presented for solving different axisymmetric stability problems in geomechanics. The nonlinear optimization was carried out by employing an interior point method based on the logarithmic barrier function. The yield surface was smoothened (i) by removing the tip singularity at the apex of the pyramid in the meridian plane, and (ii) by eliminating the stress discontinuities at the corners of the yield hexagon in the plane. No inherent assumption concerning with the hoop stress needs to be made in this formulation. The Drucker-Prager (DP) yield criterion was also used for computing the lower bound axisymmetric collapse load. The advantage of using the DP yield criterion is that it does not exhibit any singularity in the plane. A new proposal has also been given to simulate the DP yield cone with the MC hexagonal yield pyramid. The generalized Hoek-Brown (HB) yield criterion has also been used. This criterion has been smoothened both in the meridian and  planes and a new formulation is prescribed for obtaining the lower bound axisymmetric problems in rock media in combination with finite elements and nonlinear optimization. With reference to the second objective, a few important axisymmetric stability problems in soil mechanics associated with footings and excavations have been solved in the present thesis. In all these problems, except that of a flat circular footing lying over either homogeneous soil or rock media, it is assumed that the medium is governed by the MC failure criterion and it follows an associated flow rule. For determining the collapse loads for a circular footing over homogenous soil and rock media, the problem has been solved with the usage of Drucker-Prager, Mohr-Coulomb and Hoek-Brown criteria. The bearing capacity of a circular footing lying over fully cohesive strata, with an inclusion of a sand layer is evaluated. The effects of the thickness and internal friction angle of the sand layer () on the bearing capacity have been examined for different combinations of cu/(b) and q; where (i) cu defines the undrained shear strength, (ii)  is the unit weight of sand, (iii) b corresponds to the footing radius, and (iv) q is the surcharge pressure. The results have been presented in the form of a ratio () of the bearing capacities with an insertion of the sand layer to that for a footing lying directly over clayey strata. It is noted that an introduction of a layer of medium dense to dense sand over soft clay improves considerably the bearing capacity of the foundation. The improvement in the bearing capacity increases continuously (i) with decreases in cu/(b), and (ii) increases in  and q/(b). The bearing capacity factors, Nc, Nq and N, for a conical footing are obtained in a bound form for a wide range of the values of cone apex angle () and with  = 0, 0.5 and . The bearing capacity factors for a perfectly rough ( = conical footing generally increase with a decrease in . On contrary for  = 0, the factors Nc and Nq reduce gradually with a decrease in . For  = 0, the factor N for  ≥ 35o becomes minimum for  approximately equal to 90o. For  = 0, the factor N for  ≤ 30o, like in the case of  = , generally reduces with an increase in . It has also been intended to compute the bearing capacity factors Nc, Nq and N, for smooth and rough ring footing for different combinations of ri/ro and ; where ri and ro refer to inner and outer radii of the ring, respectively. It is observed that for a smooth footing, with a given value of ro, the magnitude of the collapse load decreases continuously with an increase in ri. On the other hand, for a rough base, for a given value of ro, hardly any reduction occurs in the magnitude of collapse load up to ri/ro ≈ 0.2, whereas beyond this ri/ro, the magnitude of the collapse load, similar to that of a smooth footing, decreases continuously with an increase in ri/ro. An attempt has also been made to determine the ultimate bearing capacity of a circular footing, placed over a soil mass which is reinforced with horizontal layers of circular reinforcement sheets. For performing the analysis, three different soil media have been separately considered, namely, (i) fully granular, (ii) cohesive frictional, and (iii) fully cohesive with an additional provision to account for an increase of cohesion with depth. The reinforcement sheets are assumed to be structurally strong to resist axial tension but without having any resistance to bending; such an approximation usually holds good for geogrid sheets. The shear failure between the reinforcement sheet and adjoining soil mass has been considered. The increase in the magnitudes of the bearing capacity factors (Nc and N) with an inclusion of the reinforcement has been computed in terms of the efficiency factors c and . The critical positions and corresponding optimum diameter of the reinforcement sheets, for achieving the maximum bearing capacity, have also been established. The increase in the bearing capacity with an employment of the reinforcement increases continuously with an increase in . The improvement in the bearing capacity becomes quite extensive for two layers of the reinforcements as compared to the single layer of the reinforcement. The stability of an unsupported vertical cylindrical excavation has been assessed. For the purpose of design, stability numbers (Sn) have been generated for both (i) cohesive frictional soils, and (ii) pure cohesive soils with an additional provision to account for linearly increasing cohesion with depth by using a non-dimensional factor m. The variation of Sn with H/b has been established for different values of m and ; where H and b refer to height and radius of the cylindrical excavation. A number of useful observations have been drawn about the variation of the stability number and nodal velocity patterns with changes in H/b,  and m. In the last, by using the smoothened generalized HB yield criterion, the ultimate bearing capacity of a circular footing placed over a rock mass is evaluated in a non-dimensional form for different values of GSI, mi, ci/(b) and q/ci. For validating the results, computations were exclusively performed for a strip footing as well. For the various problems selected in the present thesis, the failure and nodal velocity patterns have been examined. The results obtained from the analysis have been thoroughly compared with that reported from literature. It is expected that the various design charts presented here will be useful for the practicing engineers. The formulations given in the thesis can also be further used for solving various axisymmetric stability problems in geomechanics.

This thesis presents the implementation of the upper bound limit analysis in combination with finite elements and linear optimization for solving different stability problems in geomechanics under plane strain conditions. Although the nonlinear optimization techniques are becoming quite popular, the linear optimization has been adopted due to its simplicity in implementation and ease in attaining the convergence while performing the analysis. The objectives of the present research work are (i) to reduce the computational effort while using an upper bound finite element limit analysis with linear programming in dealing with geotechnical stability problems, and (ii) to obtain solutions for a few important geotechnical stability problems associated with reinforced earth, unsupported tunnels and a group of anchors. It is also intended to examine the developments of the failure patterns in all the cases. For carrying out the analysis for different stability problems, three noded triangular elements have been used throughout the thesis. The nodal velocities are treated as basic unknown variables and the velocity discontinuities are employed along the interfaces of all the elements. The soil mass is assumed to obey the Mohr-Coulomb’s failure criterion and an associated flow rule. The Mohr-Coulomb yield surface is linearized by means of an exterior regular polygon circumscribing the actual yield circle so that the finite element formulation leads to a linear programming problem. A simple technique has been proposed for reducing the computational effort while solving any geotechnical stability problem by using the upper bound finite element limit analysis and linear optimization. In the proposed method, the problem domain has been discretized into a number of different regions in which a particular order (number of sides) of the polygon has been specified to linearize the Mohr-Coulomb yield criterion. A greater order of the polygon needs to be chosen only in that part of the domain wherein the rate of the plastic strains becomes higher. The computational effort required to solve the problem with this implementation reduces considerably. By using the proposed method, the bearing capacity has been computed for smooth as well as rough strip footings and the results obtained are found to be quite satisfactory. The ultimate bearing capacity of a rigid strip footing placed over granular, cohesive-frictional and purely cohesive soils, reinforced with single and a group of two horizontal layers of reinforcements has been determined. The necessary formulation has been introduced to incorporate the inclusion of reinforcement in the analysis. The efficiency factors, and , to be multiplied with Nc and Nγ for finding the bearing capacity of reinforced foundations, have been established. The results have been obtained (i) for different values of soil friction angles in case of granular and cohesive-frictional soils, and (ii) for different rates at which the cohesion increases with depth for purely cohesive soil under undrained condition. The optimum positions of the reinforcements' layers corresponding to which and becomes maximum, have been established. The effect of the length of the reinforcements on the results has also been analyzed. As compared to cohesive soil, the granular soils, especially with greater values of frictional angle, cause much more predominant increase in the bearing capacity. The stability of a long open vertical trench laid in a fully cohesive and cohesive-frictional soil has been determined with an inclusion of single and a group of two layers of horizontal reinforcements. For different positions of the reinforcement layers, the efficiency factor (ηs), has been determined for several combinations of H/B, m and where H and B refer to height and width of the trench, respectively, and m accounts for the rate at which the cohesion increases linearly with depth for a fully cohesive soil with = 0. The effect of height to width of the long vertical trench on the stability number has been examined for both unreinforced and reinforced soils. The optimal positions of the reinforcements layers, corresponding to which becomes maximum, have been established. The required length of reinforcements to achieve maximum efficiency factor corresponding to optimum depth of reinforcement has also been determined. The magnitude of the maximum efficiency factor increases continuously with an increase in both m and . The effect of pseudo-static horizontal earthquake body forces on the stability of a long unsupported circular tunnel (opening) formed in a cohesive frictional soil has been determined. The stability numbers have been obtained for various values of H/D (H = tunnel cover, D = diameter of the tunnel), internal friction angle of soil, and the horizontal earthquake acceleration coefficient The computations revealed that the values of the stability numbers (i) decreases quite significantly with an increase in , and (ii) become continuously higher for greater values of H/D and . The failure patterns have also been drawn for different combinations of H/D, and . The geometry of the failure zone around the periphery of the tunnel becomes always asymmetrical with an inclusion of horizontal seismic body forces. The interference effect on the stability of two closely spaced parallel (twin) long unsupported circular tunnels formed in fully cohesive and cohesive-frictional soils has been evaluated. The variation of the stability number with S/D has been established for different combinations of H/D, m and ; where D refers to the diameter of each tunnel, S is the clear spacing between the tunnels, and is the internal friction angle of soil and m accounts for the rate at which the cohesion increases linearly with depth for a soil with = 0. On account of the interference of two tunnels, the stability number reduces continuously with a decrease in the spacing between the tunnels. The minimum spacing between the two tunnels required to eliminate the interference effect increases with (i) an increase in H/D and (ii) a decrease in the values of both m and . The failure patterns have also been generated for a few cases with different values of S/D. The size of the failure zone is found to become smaller for greater values of m and . The horizontal pullout capacity of a group of two vertical strip anchors embedded, along the same vertical plane in sand, at shallow depths has been determined. At collapse, it is assumed that the anchor plates are subjected to the same uniform horizontal velocity without any bending or tilt. The pullout resistance increases invariably with increases in the values of embedment ratio, friction angle of the sand mass and anchor-soil interface friction angle. The effect of spacing (S) between the anchors on their group collapse load is examined in detail. For a given embedment ratio, the total group failure load becomes maximum corresponding to a certain optimal spacing (Sopt). The values of Sopt increases with an increase in the value of , but the changes in the value of H/B and do not have any significant effect on Sopt. The vertical uplift capacity of a group of two horizontal strip plate anchors with the common vertical axis buried in purely cohesive as well as in cohesive frictional soil has been computed. The variation of the uplift factors Fc, Fq and F , due to the contributions of soil cohesion, surcharge pressure and unit weight, respectively, has been evaluated for different combinations of S/B and H/B. As compared to a single isolated anchor, the group of two anchors generates significantly greater magnitude of Fc. On the other hand, the factors Fq and F , for a group of two anchors are found to become almost equal to that of a single isolated anchor as long as the levels of the lower plate in the group and the single isolated anchor are kept the same. For the group of two horizontal strip plate anchors in purely cohesive soil, an increase of cohesion of soil mass with depth and the effect of self weight of the soil have been incorporated. The uplift factor Fcy both due to cohesion and unit weight of the soil has also been computed for the anchors embedded in clay under undrained condition. For given embedment ratios, the factor Fcy increases linearly with an increase in the normalized unit weight of soil mass upto a certain value before attaining a certain maximum magnitude. The computational results obtained for different research problems would be useful for design.