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.

Integral abutment bridges are jointless bridges in which the deck is continuous and connected monolithically with the abutment walls supported typically by a single row of piles. This thesis focuses on the effects of two major parameters on the seismic behavior of an integral abutment bridge in Tennessee by considering soil-structure interaction around the piles and in back of the abutments: (1) clay stiffness (medium vs. hard) around the piles, and (2) level of sand compaction (loose vs. dense) of the abutment wall backfilling. Modal and nonlinear time history analyses are performed on a three dimensional detailed bridge model using the commercial software SAP2000, which clearly show that (1) compacting the backfilling of the abutment wall will increase the bridge dominant longitudinal natural frequency considerably more than increasing the clay stiffness around the piles; (2) the maximum deflection and bending moment in the piles under seismic loading will happen at the pile-abutment interface; (3) under seismic loading, densely-compacted backfilling of the abutment wall is generally recommended since it will reduce the pile deflection, the abutment displacement, the moments in the steel girder, and particularly the pile moments; (4) under seismic loading, when the piles are located in firmer clay, although the pile deflection, the abutment displacement, and the maximum girder moment at the pier and the mid-span will decrease, the maximum pile moment and the maximum girder moment at the abutment will increase.

Information presented in site investigation reports has a strong influence in design, project costs and safety. For this reason, both the quality and the reliability of site investigation reports are important. However in our country, geotechnical engineering is relegated to second place and site investigation studies, especially parcel-basis ground investigation works / do not receive the attention they deserve. In this study, site investigation reports, that are required for the license of design projects, are examined and the missing/incorrect site investigations, laboratory tests, geotechnical evaluations and geotechnical suggestions that occur in the reports are catalogued. Also, frequency of each mistake is statistically examined / for geotechnical engineers, recommendations and solutions are presented to help them avoid frequent problems.

This thesis presents the details of designing and commissioning the multi-use static/dynamic large-scale soil testing table. The table was developed with the intention of creating a large scale testing apparatus versatile enough to carry out several different types of testing on a large scale. This report describes the background research done to develop the testing table concept and the thinking that went into each component. The apparatus itself consists of a shake table with a laminar soil container (inside dimensions L:100cm W:50cm H:65cm) and a top which can be lowered to apply overburden pressures on specimens. It is set up to run both static and cyclic tests on large soil samples. The final design allows for performing shaking tests with a non-fixed top, static and dynamic simple shear tests, and direst shear tests with minimal changes to the table configuration. The table has separate control and data acquisition systems which are necessary to run and record tests. All components of the table will be explained thoroughly within the thesis. Preliminary testing was done with the table to determine how well it is functioning and what needs to be done to further improve it. Static simple shear and cyclic simple shear tests were both run, and while the table showed some flaws, the results seem promising. It is determined that with proper instrumentation and after addressing some small issues, the testing table can be a useful and versatile tool in the future. This thesis will outline the strengths and flaws of the table as currently constructed and determine what the future applications for this testing apparatus will be. / text

Modern day bridge foundations are often designed for the maximum loading condition on a single member and each following foundation member often follows the same design. A 4-pile and a 2-pile foundation system were analyzed with an upperbound plasticity model in an attempt to approximately represent the system capacity. A sensitivity analysis was performed to measure the change in the ultimate system capacity by altering the parameters used to define individual pile capacities. The load and the resistance are considered to be normally distributed and are each characterized by a mean and a standard deviation. A first order reliability method was used to measure and account for any system redundancy in the foundation system by predicting a probability of failure based on the load and resistance of the system. Based on the most heavily loaded pile and the analysis performed on the 2-pile and 4-pile foundation systems, the probabilities of failure are lower for the system than for the single pile. This would bring reason to state there is some measurable redundancy in the basic systems analyzed where these systems would once be considered to be non-redundant systems. / text

This dissertation is concerned with a study of soil nailing, in particular the interaction mechanism between the soil and a nail and the failure mechanism and suitable design procedure for nailed slopes in sand. The interaction mechanism of a nail was studied by carrying out a number of pull-out tests, direct shear tests of nailed sand and interface tests using two uniform sands. Major parameters of the tests were flexibility, surface roughness and diameter of a nail. From the tests, it was found that: (1) flexibility of a nail significantly influences the interaction mechanism. Both the interaction parameter and apparent friction coefficient differ between a flexible and a stiff nail. Theoretical consideration indicates that the mobilization of nail forces is dominated by the relative stiffness between soil and nail. (2) a smooth-surface nail produces smaller bond friction than the critical state friction angle and mineral-to-mineral angle of the soil. This is due to the very thin rupture surface developed around the nail. On the other hand, a rough-surface nail was observed to produce two to four times larger bond friction than the direct shear friction angle of sand, due to the thick rupture surface developed and the dilatancy of the soil. (3) increasing the diameter of a nail produces a smaller apparent friction coefficient. Restrained dilatancy was found to play an important role. (4) the pull-out test, direct shear test of nailed sand and interface test produce different values of apparent friction coefficient , due to the different amount of restrained dilatancy effect around the nail (or reinforcement). The overall behaviour of nailed slopes was studied by carrying out a comprehensive series of centrifuge tests. Excavation of soil was simulated by draining water from two rubber bags in front of the facing wall. The centrifuge tests have provided much useful information on the mechanics of soil nailing. From the tests, it was found that: (1) draining of the water significantly influences both the earth pressure on the facing wall and the displacements of the nailed slope. Horizontal displacements of the facing wall were decreased by increasing the length and/or friction (bond) of the nail. (2) earth pressures on the facing wall do not exhibit a simple hydrostatic distribution. The deviations of the earth pressure are not negligible especially near the top and bottom of the facing wall. (3) roughness and bending stiffness of the facing wall considerably influence the stability and displacement of the nailed slope, respectively. (4) the observed failure surfaces were well described by a logarithmic spiral passing through the toe of the facing wall. (5) fairly good predictions for the failure acceleration were made using stability analysis of the nailed slopes based on the limit equilibrium method, provided an accurate friction angle for the sand and pull-out resistance of each nail could be determined. The factor of safety F5 of the nailed slopes was estimated by comparing the total available force and the total required force based on the observed failure surfaces.

Deep foundations, such as piles and suction caissons, are used throughout an offshore oil and gas production facility in deepwater. Ideally, the values of geotechnical properties for foundation design are determined by results from geotechnical investigation programs performed at the site of the foundation. However, the locations for facilities are not known exactly when soil borings are drilled and the footprint of a facility in deepwater can be very large with numerous foundation elements spread out over miles. Therefore, it is not generally feasible to perform a site-specific investigation for every foundation element. The objective of this research is to assess, analyze and model spatial variability in geotechnical properties for offshore foundations. A total of 97 geotechnical investigations from 14 offshore project sites covering the past twenty years of deepwater development in the Gulf of Mexico are compiled into a database. The geologic setting is primarily a normally to slightly overconsolidated marine clay, and the property of interest for the design of deep foundations is the undrained shear strength. The magnitude and characteristics of variability in design undrained shear strengths are analyzed quantitatively and graphically. Geostatistical models that describe spatial variability in the design shear strength properties to the distance away from the available information are developed and calibrated with available information from the database. Finally, a methodology is presented for incorporating the models into a reliability-based design framework to account for spatial variability in foundation capacity. Design examples are presented to demonstrate the use of the reliability methodology. Based on the design undrained shear strength profiles for the past 20 years in this Gulf of Mexico deepwater area, the design undrained shear strength varies spatially but does not depend on the time or method for site investigations. There are nonlinear spatial relationships in the point shear strength laterally and vertically due to stratigraphy such that depth-averaged shear strengths are correlated over further distances than point shear strengths. The depositional forces are an important factor causing spatial variations in the undrained shear strength, with greater variation and less spatial correlation in the more recent hemipelagic deposits (about upper 60 feet) than the deeper turbidite deposits and along the shelf versus off the shelf. The increased conservatism required in deep foundation design due to spatial variability when site specific strength data are not available is generally small with less than a five percent increase required in design capacity in this geologic setting. / text

Contaminated sites are found all around the world. In order to contain these contaminants, engineers propose capping the contaminated sediments with a sand cap. When capping these contaminants, the sand causes consolidation to occur and could cause a slope failure if the contaminants were on a slope. Investigating the properties of these contaminated sediments allows for proper analysis of a slope failure. The primary objective of this research was to determine the shear strength of contaminated sediments. Since soil samples from actual contaminated sites are highly variable and difficult to explain, the soil used in this research project was mixed and controlled in the lab. A mixture of Kaolinite, water and mineral oil (NAPL, non-aqueous phase liquid) was used for the specimens. Different oil amounts were placed into the specimens to create different scenarios. The different oil combinations included: 100% water, 100% oil, 90% oil, 70% oil, and 50% oil. All of the specimens were fully saturated, and the specimens that had less than 100% oil contained water in the remaining percentage. Consolidated Undrained and Consolidated Drained triaxial tests were performed on the specimens. The constructed specimens were subjected to consolidation stages ranging from 0.6psi to 29psi in confining pressure. The main focus of the study was on low confining pressures. After consolidation the specimens were sheared either undrained or drained. Both tests were utilized in order to see the difference in the pore pressures generated. Failure envelopes were developed for the different oil contents that contained three dimensions included the shear strength, the effective stress, and the pore pressure difference between the pore oil pressures and the pore water pressures. Also, the behavior of oil-dominated versus water-dominated was determined. Results from the 100% water specimens were comparable to previous data. The shear strength for the 100% oil specimens was higher than the 100% water specimens, but lower than the 90% oil and 70% oil specimens. The 50% oil specimens resulted in a great deal of variability on whether the specimen was water-dominated or oil-dominated. The main conclusion was that the Kaolinite had an increase in strength with the introduction of mineral oil. / text

The compression behaviour, shear strength characteristics, and material properties of dense slurries and soft settled beds of natural fine-grained sediments were studied experimentally. Slurries of varying initial density, initial height, and pore fluid salinity were settled one-dimensionally, by self-weight, in the laboratory. Settling behaviour was studied in terms of slurry appearance, particle segregation, height of surface versus time, sediment surface and element settlement rates, and the redistribution of sediment with respect to height and time. Consolidation behaviour was studied in terms of sediment compressibility and pore fluid flow. Shear strength was examined "in situ" and elated to the parameters effective stress and specific volume. Instruments and techniques were developed to facilitate the measurement of low effective stresses, low strengths, and high specific volumes. A small scale sediment sampler was developed and used in an attempt to study the arrangement of particles within soft sensitive beds. The experimental results revealed numerous fundamental reasons why theoretical models of settling and consolidation are unsatisfactory when applied to dense flocculated slurries and beds of high specific volume, respectively. For the sediment studied, well-defined compressibility and shear strength relationships were observed. Pore fluid flow relationships were non-unique at high specific volumes. The compression behaviour of slurries was found to have a rational basis in terms of electrochemical forces and degrees of particle association. The experimental results are relevant to engineering practice. Recommendations are made regarding future research.

The lack of knowledge about the behaviour of footings for jack-up rigs under storm loads poses a design problem which can be tackled by model testing. The areas of prime concern are the ultimate loads on footings under combined loading, which affects the safety of the rig, and the rotational stiffness, which affects the interaction between the foundation and the structure. A programme of loading tests was performed on model footings on clay, and was divided into two stages: monotonic loading and cyclic loading. The clay samples were obtained by consolidating Speswhite kaolin slurry in cylindrical tanks 450mm in diameter. The strength and compressibility characteristics of the samples were verified by means of standard laboratory tests. The model footings were 50mm and 100mm in diameter and several shapes were tested: circular flat plate, cones of various angles and model spud-cans. Loads and displacements were monitored using appropriate instrumentation and a data logger. A series of central vertical loading tests provided data for comparison with existing bearing capacity theories. Combined loading tests were performed applying a displacement controlled horizontal load at a fixed height above the footing which was also subjected to a fixed vertical load. The main series of tests involved a parametric study of the relevant variables. Special tests allowed the assessment of the effect of embedment of the footing and the interaction of a flexible leg with the foundation. Cyclic loading tests were carried out using a load controlled system which applied a sinusoidal load simulating wave action. Effects of currents were investigated by introducing an offset to the loading cycle. The influence of amplitude and period of loading as well as the influence of vertical load were also investigated. Special tests were carried out to cover some peculiarities of real loading conditions. Fitting of a three-parameter hyperbola to the test results provided a systematic and accurate method of analysis of monotonic loading tests, leading to valuable information involving stiffness and ultimate loads. Analysis of cyclic loading tests yielded useful qualitative information regarding the progress of settlement and the variation of rotational stiffness and damping ratio with the number of cycles.