Seismic events around the globe directly affect all ranges of structures, from complex and expensive ‘skyscrapers’ to simple frame structures, the latter making up a higher proportion of the number of structures affected as they are a much more common type of structure. The impact of a seismic event can be devastating, especially if adequate predictions of their impact and imposed structural response are not made during the design stage of the structure. Knowing what response to expect allows the engineer to design the structure to survive an event and protect the occupants. The structural response to a seismic event is very complex and can be affected by a wide range of structural, geotechnical and environ parameters. While larger, expensive structures make use of expensive, time consuming, finite element analytical procedures to determine their response the cheaper, simpler, frame structures have to make do with existing, simplified, spectral method predictions. This research firstly involves finite element analysis of simple frame structures, considering different structural and geotechnical parameters which may influence the seismic response, namely the stiffness of the structural joints, the geometry of the structure (influencing the individual structural element flexibility) and the foundation conditions (fixed base or shallow foundations with soil structure interaction). A range of frames, of varying geometry, are considered which mobilise different amounts of inter-storey drift, local rotation and global rotation response. The influence of soil structure interaction (SSI) and frame rigidity (i.e. the properties of the joints) on the response behaviour is investigated. The finite element database is then used to validate improved methods for predicting the spectral response parameters, specifically the natural period and damping of equivalent single degree of freedom (SDOF) systems, which include the effects of frame rigidity, geometry and SSI. Dynamic centrifuge testing is also carried out in order to further validate the improved spectral model for the case of real soil with shear dependant stiffness. The physical model testing is also extended to consider how environs, such as other structures in close proximity, influence the response of a structure.

Karst bedrock conditions and deterioration of the lock and dam structures have resulted in significant leakage through, underneath, and around Lock and Dam No. 8 on the Kentucky River. During severe droughts, the water surface in Pool No. 8 has been observed to drop below the crest of the dam, resulting in water supply shortages and water quality issues for surrounding communities reliant on the pool. Presently, the primary purpose of Lock and Dam No. 8 is water supply. Pool No. 8 is currently where the cities of Nicholasville (Jessamine County, KY) and Lancaster (Garrard County, KY) draw their water. Due to the age and condition of the structures, and the criticality of the retained water supply, the project Owner commissioned a replacement dam to be built. One major component of the replacement dam was a foundation improvement program. The foundation improvement program was designed to address the karst bedrock conditions at the site. The foundation improvements included a secant pile cutoff wall and a double-row grout curtain. The grout curtain at Lock and Dam No. 8 was evaluated based on the metrics presently available.

Tieback walls are typically design based on predetermined pressure distribution; however, these pressures were proposed based on performance of excavations. For retaining walls used in slope remediation, the application of these pressures might not be adequate; the construction procedure; therefore, a different response of the wall is expected. This document, presents the performance of two tieback walls installed in a shale stratum. Monitored responses is correlated with construction activities; these activates implied excavation and backfilling in both of the tieback walls. In addition, this research shows a numerical procedure to evaluate the anchor capacity based on the t- z approach. Finally, this study introduces an empirical method to estimate lateral wall deformation profiles and internal bending moments along a retaining wall installed in a clay stratum.

Large diameter piles are widely used as foundations to support buildings, bridges, and other structures. As a result, it is critical for the field to have an optimized approach for quality control and efficiency purposes to measure the suggested number of load tests and the required measured capacities driven piles. In this thesis, an analysis of a load test program designed for proposed bridge replacements at Kentucky Lake is performed. It includes a detailed site exploration study with in-situ and laboratory testing. The pile load test program included monitoring of a steel H-pile and steel open ended pipe pile during driving and static loading. The pile load test program included static and dynamic testing at both pile testing locations. Predictions of both pile capacities were estimated using commonly applied failure criterion, and a load transfer analysis was carried out on the dynamic and static test data for both piles. The dynamic tests were then compared to the measured data from the static test to examine the accuracy. This thesis concludes by constructing t-z and q-z curves and comparing the load transfer analyses of the static and dynamic tests.

Land loss in South Louisiana is increasing at a fairly rapid rate. In an effort to reduce land loss and save the marshes of Louisiana, marsh creation projects have been proposed in carefully selected regions around the coast as part of the CPRA Coastal Master Plan 2017. Properties and characteristics of the soil obtained from soil borings were analyzed and used to determine the various design parameters that allow the marsh creation process to occur. Other properties that were taken into consideration for Louisiana coastal sediment are the geothermal properties. This research analyses those different properties obtained from geotechnical reports from CPRA and other data bases, in order to find correlations between the different soil characteristics, specifically between the soil’s compressive strength, consolidation properties, Atterberg Limits and moisture content. Furthermore, this research also studies the geothermal properties of selected Louisiana soils and the correlation between moisture content and thermal conductivity.

<p>There is consistent research evidence that shows improvement of the engineering properties of subgrade soils treated with lime or cement. However, limited information is available on the effect of climatic changes on the subgrade stiffness. The thesis studies the effects of changes in soil moisture content and temperature on the resilient modulus of treated and untreated subgrades in Indiana. Two types of soils were tested: A-6 and A-7-6, from two locations in Indiana: Hartford City and Bloomington, respectively. When existing standards ASTM D559/559-15 and ASTM D560/560-16 for wetting/drying (WD) and freezing/thawing (FT) processes, respectively, were followed, the treated and untreated samples failed through the process of preparation due to the stringent procedures in the standards. Appropriate test conditions were investigated, as part of the research, to develop new protocols more appropriate to the field conditions in Indiana. Two new test protocols were developed and successfully applied to the treated soils. A total of 26 resilient modulus, M_R, tests were conducted following the standard AASHTO T307-99. The M_R results showed that the repeated action of WD and FT cycles reduced the stiffness of the chemically-treated soils down to values similar to or lower than those of the untreated soils. However, when the amount of chemical was doubled, with respect to the optimum, the M_R of the treated soils improved over that of the untreated soils, even after the wetting-drying cycles.</p>

<div>The Indiana Department of Transportation implements maximum soil fill covers to ensure the safe installation and operation of buried pipes. Historically, fill cover tables are provided by INDOT, but the methodology for calculating these covers is not well documented. The finite element method enables a comprehensive analysis of the soil-pipe system taking into account soil conditions, pipe type and geometry, and conditions on the pipe-soil interface. </div><div><br></div><div>This thesis discusses the calculation of maximum fill covers for corrugated and structural plate pipe-arches using the finite element software CANDE and compares the results with previous estimates provided by INDOT. The CANDE software uses the Finite Element Method, and the Load and Resistance Factored design based on a two-dimensional culvert installation in a soil-pipe model. The model is set up under plain strain conditions and is subjected to factored dead and live load, and provides an analysis of the structure based on safety measures against all factored failure modes associated with the structural material.</div><div><br></div><div>Significant issues were encountered when calculating the maximum fill covers for pipe-arches in CANDE, including the inability of standard CANDE (Level 2 mesh) to model pipe-arches, lack of convergence for nonlinear analysis, and fill cover results higher than expected. To solve these issues, the pipe-arches were modeled using Level 3 solution in CANDE. The CANDE analyses were run using small-deformation analysis after buckling was eliminated as a governing failure mode using parallel simulations in Abaqus. Numerical results were compared to analytical solutions following ASTM standards.</div><div><br></div><div>The results showed that CANDE and INDOT calculations differ significantly, with the CANDE results yielding higher fill covers than those provided in INDOT specifications. These differences are attributed to the assumed loading pattern at failure. While the CANDE results assume that the maximum fill cover height is defined by the failure of the pipe considering the radial pressure (Pv), the INDOT results are consistent with results obtained by limiting the bearing capacity of the soil around the corner radius (Pc).</div>

Underground structures perform an important role in transportation systems in many seismically active regions around the world, but empirical data regarding the seismic behavior of these structures is limited. This research works towards filling that empirical gap through the use of scale model shake table testing. Underground seismic soil-structure interaction (USSSI) effects were investigated for a stiff rectangular tunnel cross-section embedded within soft clay. San Francisco Young Bay Mud was used as a prototype soil for developing a scale model soil mixture consisting of kaolinite, bentonite, class C fly ash, and water. A single cell Bay Area Rapid Transit (BART) cut-and-cover subway tunnel was used as the prototype for the 10th scale model subway cross-section. A flexible walled test container originally developed for a pile study at UC Berkeley was modified for use at Cal Poly, San Luis Obispo. The flexible container allows for close approximation of one-dimensional (1D) free-field site response by significantly limiting the rigidity of the boundary conditions and allowing the soil to deform under simple shear. The study was conducted over two shake table testing phases: Phase I consisted of shaking a model soil column to evaluate the ability of the test container to produce adequate 1D free-field site response, and Phase II tests explored the horizontal racking distortion of a shallow rectangular tunnel cross-section subjected to strong transverse ground shaking. Phase I test results and comparison with SHAKE models indicate that the test container can sufficiently mimic 1D free-field conditions, specifically for the primary shear deformation mode. Similarly, the equivalent linear soil-structure interaction code FLUSH was found to adequately model site response for the Phase II soil-structure system. Comparison of recorded horizontal racking distortions of the model structure with those from numerical modeling suggest that current simplified design methods may overestimate distortions to some degree for cases similar to those examined in this research. Overall, the flexible wall testing container shows promise as a viable means for gaining further insight into USSSI topics, as well as various other geotechnical and soil-structure interaction problems.

The U.S. Congressional Research Service issued a report for Congress in May 2016, entitled” Earthquake Risk and U.S. Highway Infrastructure: Frequently Asked Questions” which highlighted the absence of a national database on the status of seismic vulnerability of bridges or other infrastructure, and thus no estimate of costs to retrofit vulnerable bridges. Low to moderate seismicity regions exist in each of the continental United States, with over 30 states having mostly or entirely low-to-moderate seismicity. Resources at state transportation agencies and municipalities are focused on higher seismicity regions, creating a gap in quantifying the system-wide seismic vulnerability despite an overall aging bridge inventory, much of which was built before current seismic design standards. This research addressed this data gap and reduces barriers to quantifying seismic vulnerability of existing bridges in low-to-moderate seismicity regions. The work included nonlinear dynamic numerical modeling of typical multiple span bridge configurations in both pristine and deteriorated conditions, by subjecting them to seventy ground motions across four low-to-moderate seismic hazard levels, to evaluate their seismic performance. These typical bridge configurations represent over 160,000 bridges, which comprise 55 % of the multiple span bridges nationwide. The research results indicate that there is an overall low probability of significant seismic damage to these typical bridges in such regions. The results also show that current seismic hazard thresholds used for the design of new bridges, and for retrofit of existing bridges, which provide the basis for exempting some bridges from specific seismic analysis and design, can underestimate the expected seismic forces. Those results can be used to refine those exemption thresholds to provide appropriate protection against potential seismic damage in those cases. The study results also formed the basis for a system-wide rapid seismic vulnerability screening algorithm developed for the Vermont bridge inventory, which is applicable to other states with low to moderate seismicity regions.

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.