Geo-Congress 2024: Geotechnical Data Analysis and Computation

Improving the prediction of landslides and debris flow is essential to understand and mitigate the risks they pose. The material point method (MPM) is a hybrid Eulerian-Lagrangian approach that uses an Eulerian grid to compute the momentum and collisions while Lagrangian particles are used to track the material state. Thus, it is a methodology well suited to model large deformations without issues such as mesh distortion. However, while existing literature has demonstrated the potential of this method with comparisons to general observations from field data, there is a lack of validation against experimental data of run-out in the field of landslides. In this study, the results from recent centrifuge results idealizing a dam breach are used to illustrate the vastly different run-out behaviors fine-grained materials can show, ranging from static liquefaction and rapid run-outs to seepage and slope instability driven run-outs occurring at a much slower rate. Ultimately, the experiments showed that the state parameter, measured or inferred from CPTs, can be used to distinguish between the dominant deformation and failure mechanisms. This paper presents preliminary results from MPM analyses of element tests and small column collapses with a focus on material properties reflecting the state parameter. This motivates the use of MPM with constitutive models utilizing an effective stress-based interpretation of the material response to be applied to landslides, debris flows, and other run-out problems of civil engineering interest.

The discrete element method (DEM) is widely used in the fields of structural geology and tectonophysics to simulate the formation of fault-induced geologic features. With few exceptions, the vast majority of these simulations have focused on dip-slip fault rupture, with few exceptions of strike-slip fault rupture. Strike-slip fault rupture is particularly difficult to model with DEM, because 3D models often require millions of individual grains to capture the finest details of developing shear zones. Fortunately, high-performance computing makes this entirely possible. This study utilizes high-performance computing with DEM to simulate strike-slip fault rupture in dense and loose analogue soils. The results of the shear zones that develop during rupture are analyzed in plain view from above and from the side view to develop a comprehensive comparison with features observed in nature through post-earthquake reconnaissance activities.

In soil shear strength characterization, particle size is usually much smaller than the testing device size. However, when particle size is similar to the size of the apparatus (e.g., for coarse granular materials), the mechanical response becomes less reliable. To address this, international standards prescribe minimum sample scales based on maximal particle and device sizes. Nevertheless, the influence of the sample scale on the mechanical response is still not well understood. This topic is studied through simple shear simulations in the frame of the discrete-element method, covering a wide range of sample scales and particle size distributions. Micromechanical analyses of force and contact configurations reveal that the stability of parameters is linked to the formation of local rigid structures that carry forces significantly higher than the average. When the height of these structures becomes comparable to that of the sample, macroscopic and microscopic parameters deviate from those found under larger sample scales. Although these findings require further validation, this work suggests that the standards may need to be re-evaluated for an effective material characterization.

Research indicates that heating of soil influences its hydraulic and mechanical response in a coupled manner. While several thermomechanical soil constitutive models are present in the literature, their use in numerical analyses of geotechnical boundary value problems involving thermomechanical loading is rather limited. This paper presents numerical implementation of a thermomechanical soil constitutive model in a finite element analysis (FEA) framework. Building on the basics of the modified cam clay (MCC) model, the discussed constitutive model accounts for thermal influence on soil mechanical behavior through an isotropic hardening mechanism. Upon successful numerical implementation, the constitutive model was calibrated using data from temperature-controlled triaxial compression tests reported in the literature. FE simulations of multiple triaxial tests on different soils and comparison of FEA results with those recorded during laboratory triaxial tests demonstrate successful numerical implementation of the thermomechanical soil constitutive model.

Amending the composition of pore fluids by adding biopolymers can increase soil cohesion and improve the properties of soils. These advancements, however, mostly rely on trial and error in identifying the most suitable biopolymer. In this study, we show that we can model the properties of biopolymer pore fluids using coarse-grained molecular dynamics simulations. We model the effect of adding charged biopolymers, in the presence of nano-clay fillers, on the properties of pore fluids and their effect on improving soil cohesion. Our simulations show that these biopolymers form biogels that can absorb large amounts of water and potentially serve as a sustainable alternative for soil-strengthening purposes. We investigate the effect of charge distribution on gelation and structure formation in our system. Our results show that the physical properties of the gel can be controlled by the types of interactions between chains and nanofillers.

Owing to recent advancements in artificial intelligence, the technology has gained popularity across all possible domains in geotechnical engineering over the past few decades. The present study demonstrates the application of machine learning techniques to predict the undrained shear strength of soft, sensitive clay. Field and laboratory measurements from 24 test locations in Finland are used to create a multivariate database consisting of 384 data points. The main objective of the study is to predict the undrained shear strength using ensemble-based machine learning models, viz., random forest, gradient boosting, extreme gradient boosting, AdaBoost, and light gradient boosting. The undrained shear strength of the clay soil sample is considered as an output parameter for given input features that include pre-consolidation stress, vertical effective stress, liquid limit, plastic limit, and water content. The results from the ensemble-based machine learning models are compared with those from the base models like decision trees. Results show that the ensemble-based extreme gradient boost and light gradient boost models outperform the base models in terms of accuracy and error with the extreme gradient boost model showing the highest coefficient of determination and root mean squared error (indicators of a good prediction model). Furthermore, sensitivity analysis is carried out for the best-performing extreme gradient boosting model, and results show that undrained shear strength is highly influenced by pre-consolidation stress, the existing vertical effective stress, and the plasticity index of the soil.

Geotextile has numerous benefits in improving pavement performance such as drainage, barrier, filtration, and reinforcement. Wicking geotextile is a relatively new geotextile product with the ability to autogenously drain water from soils. This paper introduces the development and implementation of a multiphysics model to simulate the performance of wicking geotextile in pavement system subjected to freezing climate. A variety of environmental effects are taken into account through specific boundary condition settings, from which various climatic factors, ground water level, ground heat, and drainage influence on the system can be captured and simulated by the model. The simulation results are first validated with field data from long-term pavement performance (LTPP) road section in cold region. The validated model is further utilized to analyze the impact of geotextile on the same LTPP section. The simulated results shows that wicking geotextile can generate suction concentration around its installation location, reduces the overall unfrozen water content in pavement, decreases the magnitude of frost heave, postpones the time of frost heave initiation, and has insignificant effect on frost depth.

Gas hydrates are ice-like substances found in deep sediments under high pressures and low temperatures, holding substantial natural gas reserves. Extracting gas from these reservoirs faces challenges due to complex interactions between hydrate-bearing sediments, fluid flow, and thermal processes. To address these challenges, a thermo-hydro-mechanical-chemical (THMC) simulator has been developed in this study. This simulator incorporates the change in mechanical (using advanced elastic-plastic models), thermal, and fluid flow characteristics in response to pressure, temperature, and production changes. A numerical solver for the coupled THMC processes in gas hydrate reservoirs has been created using finite volume method (FVM) and finite element method (FEM) approach within the PETSc framework. Using this simulator, the numerical study explores gas extraction using a single horizontal well, considering different reservoir porosity and hydrate saturation scenarios. The analysis reveals that the reservoirs characterized by higher porosity and hydrate saturation exhibit greater cumulative gas production till 30 days. However, the case with higher porosity and lower hydrate saturation experiences the maximum differential settlement after 30 days.

This study presents the findings of a 3D explicit finite element analysis exploring the effect of nose shape on the depth of burial (DoB) of projectiles in cohesive soil. Large deformation finite element (LDFE) analyses were performed using the commercial finite element code ABAQUS. Euler and Lagrange element interactions were captured using the coupled Eulerian-Lagrangian (CEL) algorithm. Ogive projectiles with various caliber radius heads (CRH), ranging from 9 to 11, dynamically impacted a clay target at an impact velocity of 200 m/s. A constitutive model comprising a Tresca soil model that accounted for strain softening and strain-rate dependency of shear strength through the Herschel-Bulkley model was employed. The results of the simulations were used to calibrate a phenomenological penetration model for predicting the depth of burial of projectiles in soils. The two fitting parameters of the phenomenological model were the drag coefficient C, and the bearing coefficient R, both of which were assumed to be constants. It was found that varying the ogive CRH within the ranges studied resulted in minimal change in the drag coefficient. Furthermore, the DoB was not significantly affected by these variations in the CRH. The bearing coefficient R was found to be constant for all ogives. It was concluded that the DoB was not sensitive to minor changes in the nose shape of ogive ordnances in clayey soils.

In situ soils generally exhibit a structure due to bonding that exists between the soil constituents. The behaviour of the structured clay is different from the reconstituted clay which needs to be modelled accurately. This study aims to understand the accuracy and reliability of two different phenomenological models based on the critical state framework formulated to predict the behaviour of structured clay. These models (Lee and CASM-n) are calibrated with drained conventional triaxial experiments for two different percentages of cement content (high and low cement content) after obtaining the material parameters. A non-linear optimization technique is used to obtain other model-fitting parameters. After obtaining the parameters, a single-element explicit integration scheme is used to simulate the drained triaxial test for different confining pressures. The Lee model predictions deviate from the experimental results for the materials that exhibit a dominant frictional response, such as that exhibited by low-cemented clays, but predicted the behaviour of cohesion dominant response as observed for high-cemented clays reasonably well. However, CASM-n model was able to predict the behaviour of low-cemented clays accurately.

The particle-based discrete element method is used to simulate a block’s failure on an inclined surface. The particles were generated in polygons and bonded using a parallel bond model to simulate blocks with different width-to-length ratios on an inclined surface. The size of the particles can influence block movement (e.g., toppling or sliding) on the inclined surface. By decreasing the particle size, the interlocking forces between the particles are reduced, and slip occurs between the block and inclined surface. Reducing the particle size, however, increases the analysis time, especially when the number of particles is too high. Introducing smooth joints eliminates the effect of overriding particles (dilation) caused by local particle orientations. As a result, the effect of the two surfaces’ asperities will be removed, especially when larger particles are utilized to reduce the analysis time.

The discrete element method (DEM) is a numerical modeling approach used to simulate granular materials, such as soil. Early DEM software used circular or spherical particle shapes for computational efficiency. However, soil particles are not spherical, and particle shape has a significant impact on soil strength and stiffness. More modern DEM software uses polyhedral particle shapes to better represent soil behavior; however, contact detection of polyhedral particles is computationally expensive. This computational cost has limited the size and type of geotechnical engineering problems that can be solved with DEM. Many researchers have developed simulation acceleration algorithms to expand the size of problems DEM can solve; however, there have been no systematic comparisons of the computational speed of available polyhedral-shaped DEM platforms. This study compares the computational speed and memory usage of three DEM software platforms: BLOKS3D, LAMMPS, and YADE-DEM, using simple particle gravity deposition simulations. The simulations involve deposition of particles in a container under gravity and reporting the computational times and memory usage for each simulation. Where possible, the same material properties and simulation settings are used for each software. BLOKS3D performed simulations approximately two orders of magnitude faster than YADE-DEM and LAMMPS. The memory usage was on the same order of magnitude for all platforms; however, LAMMPS and YADE-DEM had more efficient memory scaling than BLOKS3D.

The K_α parameter is a widely used factor that simplifies the interpretation of the influence of static shear stress on cyclic resistance of granular materials. However, it is still unclear whether the inherent fabric resulting from sample preparation protocols affects the K_α parameter. In this study, we use 3D-DEM to construct samples with polydisperse spherical particles with the same density and initial stress conditions, but different inherent fabrics characterized by coordination number and contact-normal fabric anisotropy. We conduct constant-volume cyclic simple shear tests on these samples and determine their cyclic resistance. Our simulation results indicate that inherent fabric does, quantitatively but not qualitatively, alter the K_α effect. Our results also show that inherent fabric influences the evolution of macro- and micro-parameters, such as mean stress drop, coordination number, fabric anisotropy, and coaxiality of principal directions of fabric and stress for different samples.

The stability evaluation of the Low Lake Level Pumping Station (L3PS) at Lake Mead is presented in this study. The subsurface structure consists of a riser shaft, an access shaft, 34 well shafts, a 107-m-long forebay with a height of 11 m and width of 10 m, a stub connector tunnel, and a future forebay extension. The site geology contains several joint sets, faults, and shear zones intersecting the excavation. The behavior of the entire subsurface structure and the interaction of various components in the vicinity of fault/shear zones were evaluated using a coupled hydromechanical finite element method (FEM) approach. The response of the rock mass was also evaluated using a hybrid finite difference-discrete element method (FDM-DEM) by accounting for major and minor discontinuities within the rock mass. The areas exhibiting significant plastic yielding were identified for a prompt installation of rock support to stabilize the heading.

Pre-support systems are required in New Austrian Tunneling Method (NATM) construction in weak soils, shallow overburden conditions, in areas of settlement-sensitive structures. Umbrella arch method (UAM) is one of these pre-support systems that generate an artificial improved roof, before the excavation. In this study, a section of the Çekmeköy-Sancaktepe-Sultanbeyli Metro Line (with twin tunnels) constructed with NATM and UAM support system is modeled and analyzed using finite element method. The tunnel section was modeled using three approaches (i.e., plate, dummy plate, and individual pipe), and the results were compared with the field measurements. The results of the numerical analysis agreed with the field measurements. The dummy plate approach resulted in more accurate settlement predictions compared to other approaches. The results also indicate that the tunnels were far enough, such that the surface settlement profile did not indicate a clear impact due to the sequence of construction of the twin tunnels.

The study examines the seismic response of tunnels constructed in a layer of cohesionless soil of depths more than 20 m. The soil assembly and tunnel lining were modeled using three-dimensional discrete-element method (DEM) simulations. Soil particles were modeled as rigid spherical particles that may overlap at contact locations. The tunnel lining was modeled as a flexible structure composed of particles bonded together with cementitious bonds to mimic the physical features of a real tunnel. Free-field boundaries were utilized at the model’s lateral sides to prevent the propagating wave’s reflections back to the domain and employ the free-field motion. The DEM computational framework was validated using published centrifuge test results, and the numerical simulation results reveal good agreement with the experimental test results when subjected to the same input motion.

Nano-level particle scale and stoichiometry of clay minerals result in the dominance of interfacial phenomena, including the adsorption of ions, ultimately yielding macroscopic swelling behavior. In this study, the surface interaction properties were compared between two clay minerals classically considered for their diametric swelling behaviors, kaolinite, and montmorillonite, using molecular dynamics (MD) simulation. Minerals were solvated with sodium chloride water solution to compare surface interactions with ions. Density was measured as a function of distance from respective mineral interfaces to compare with similar analyses of adsorbed water density in the literature. Ion radial distribution functions are prepared for each mineral. The results show a local density of water up to 1,900 kg/m³, and bulk density was achieved at approximately 1 nm normal to mineral surfaces; however, surface-ion structure is distinct between mineral surfaces due to crystallographic defects in montmorillonite.

Numerical simulations of cone penetration test (CPT) can provide valuable insights into the mechanical behavior and in situ state of geomaterials. However, adequate simulation of CPT is challenging due to the large deformations occurring during penetration and the need for representative soil constitutive models. This study investigates the use of the material point method (MPM) and a version of the SANISAND family of models as a representative constitutive model to simulate CPTs in dry sand. The numerical model is validated against experimental data on CPT in a calibration chamber. The simulations are done on a range of soil overburden pressures and void ratios. The results illustrate the sufficient adequacy of the numerical configuration in capturing the cone tip resistance for the soil states considered. Additionally, the variation of void ratio versus mean effective stress in selected material points is presented to explore details of material state evolution during the cone penetration.

Wind, wave, 1P, and 3P loading subject the offshore wind turbine (OWT) structure to multidirectional long-term cyclic loads that have varying amplitudes, frequencies, and patterns. Therefore, studying the soil-foundation-structure interaction (SFSI) under realistic loading is required to understand the response of the entire OWT during the service life of the structure. This paper presents the establishment and capabilities of a new large-scale multidirectional offshore wind SFSI testing facility at the Advanced Technology for Large Structural Systems (ATLSS) research center at Lehigh University. The new testing facility has unique multidirectional loading capabilities that allow for simultaneous application of realistic wind, wave, gravity loads, and their induced moments at the top of the OWT foundations (i.e., mudline). The capabilities of the new testing facility also include large-scale real-time hybrid simulation (RTHS) testing. These capabilities provide the ability to evaluate the response of the whole OWT structure under loading. In addition to presenting the design, concepts, and framework of the facility, this paper presents validation results for the RTHS framework of OWTs using small-scale tests. These small-scale RTHS tests are used as a first step to prepare for conducting large-scale RTHS tests.

Traditional methods for slope stability analysis usually do not consider the influence of the intermediate principal stress, though it has certain favorable impacts on the strength of geomaterials, such as soil and rock. Hence, this study presents the unified slip line solution for classical slope stability problems under seismic conditions. The analysis is conducted based on an integrated framework of the unified strength theory and the slip line approach. The unified strength theory incorporates the effect of the intermediate principal stress on slope stability by introducing a specified parameter, b. In contrast, the slip line theory precludes the need for any predefined slope geometry and slip surface in the analysis. Instead, the slope geometry required to sustain the stability with a prescribed factor of safety (FS) is derived for slopes in cohesive-frictional soils. In addition, the earthquake effect is captured using the pseudo-static approach. The slope profiles thus derived from the current formulation are found to have a reasonable morphological similitude with natural slopes. Additionally, the results indicate that the unified strength theory parameter b substantially affects the seismic slope stability. With an increase in b, the slope profiles are generally seen to form a steeper layout, revealing the fact that the potential strength of the slope medium is effectively utilized. Further, a rational comparison of the present results with the solutions available in the literature affirms the legitimacy of the proposed approach in providing practical design guidelines for optimal slope profiles and intimating the need for slope-strengthening measures to prevent any imminent slope failures in advance.

The material point method (MPM) is an advanced particle-based numerical tool capable of modeling large deformations. However, its abilities to predict seismic site response, ground failure initiation, and post-failure consequences, including undrained conditions, have not been fully realized. This research aims to use recent MPM developments to simulate earthquake-triggered slope failures. As a novelty, an advanced constitutive model such as the intergranular strain anisotropy hypoplastic model is implemented in an effective-stress undrained MPM framework to simulate the hydromechanical response of dense and loose sand. First, an MPM simulation of a level-ground sand column is used to understand the system-level undrained response under simple sinusoidal excitation. Loose sand tends to liquefy under an increase in pore pressure, while dense sand exhibits a decrease in pore pressure. Second, a small-scale slope geometry is adopted whereby the runout investigation is conducted using void ratios ranging from 0.45 to 0.95. The framework shows promising system-level results, needing further validation using centrifuge experiments.

Increment of wildfire causes numerous disturbances and irregularities that jeopardize a sustainable society and reliable infrastructure. Due to the rigorous burns, the soil particles may breakdown due to thermal effects leading to a significant loss of the physical and mechanical properties of the soil mass, including reduced modules and strength and subsequent major problems such as soil erosion and near-surface slope slides. Many ambiguities are tied up with the multi-physics process of wildfire-burnt soil, the vegetation anchoring effect on soil strength, and the alteration of micro-scale soil properties. This study presents an innovative thermal-mechanical coupled model to simulate rock damage and breakage during heating-cooling processes. A series of simulations are carried out to capture the behaviors of rock samples under heating-cooling and subsequent compressive loadings using a two-dimensional discrete element method (DEM) model. The results suggest that the effect of mild wildfire on the strength and modulus reduction of rock is negligible. But the reduction of strength and modulus could be as high as 53% and 12%, respectively, under severe wildfire conditions.

Understanding the deformation mechanisms and the impact forces generated by landslides is important for risk assessment and for implementing adequate mitigation measures. This work studies the effects of the basal surface characteristics on the impact forces exerted by a compound landslide mass on a rigid vertical wall. A generic model of a biplanar compound landslide is numerically analyzed using the material point method (MPM). MPM allows the study of the entire instability process, that is, failure initiation, propagation, and impact on the wall. In the present study, different curvatures of the slope transition along the basal surface are considered, including a sharp transition (i.e., kink). Different geometries develop different internal shearing mechanisms. Impact forces on a vertical rigid wall are compared for the different geometries, and that impact forces decrease as the curvature of the transition increases. Ultimately, the biplanar geometry with a sharp transition presents the lowest impact forces on the wall. This is consistent with the fact that a higher amount of energy is internally dissipated during the sliding process in a sharp biplanar transition compared to a curved one. The presence of friction on the sliding surface is also evaluated. The results show that the higher the basal friction resistance, the lower the maximum impact forces for all the slope geometries, but the differences between rounded vs. sharp transitions are more prominent.

This paper presents a method for three-dimensional slope stability analysis based on finite element technologies. In this method, a potential slip surface is implemented as interface elements coupled with the soil slope as solid elements. Stress and strain mobilization analysis to bring the potential slip surface and the sliding soil mass above the slip surface from an initial static equilibrium state to a limit equilibrium state is carried out by a coupled procedure between the interface elements and the solid elements. The coupled procedure is used to study the stress redistribution in whole soil slope caused by material yielding along a potential slip surface. The interface elements can be created by intersection calculation between the solid finite element mesh and 3D slip surface grid shape models. This paper shows two typical 3D slip surface grid shape models that can be used to simulate most slope sliding patterns under three-dimensional conditions. A three-dimensional slope stability analysis case is also presented in this paper.

The consistent particle method (CPM) is a Lagrangian particle method initially proposed for solving fluid dynamics problems and has recently been expanded to the area of solid mechanics. The governing equations for mass conservation and dynamics of particles are solved with a predictor-corrector approach. The spatial derivatives of variables (e.g., velocities and stresses) are computed by Taylor series expansion. The physical material density is directly calculated without using the number density of particle. In addition, no extra particles are required to model boundary conditions. For partially loaded boundary surface, the stress singularity problem because of sudden stress change is overcome by the inverse distance weighting method. Numerical examples of bearing capacity of strip footing as well as large deformation penetration of strip footing are presented to show CPM’s abilities to handle geomechanics problems.

A numerical reactive transport model was developed to simulate microbially induced desaturation (MID), an emerging bio-mediated ground improvement technique. MID employs nitrate-reducing microorganisms to produce gas bubbles that desaturate granular soils, thereby enhancing the cyclic shear strength of the soil. MID is of particular interest due to its potential to mitigate earthquake-induced soil liquefaction non-disruptively, for example, underneath existing structures. The model couples fluid flow with reactive transport and predicts the distribution of substrates and products and changes in porosity, permeability, and degree of saturation from MID. The numerical model was implemented in COMSOL Multiphysics finite element software and validated using bench-scale tank experiment results. Comparison of changes in the degree of saturation and electrical conductivity, time-lapse images, and the breakthrough curves observed during the experiment with the results of the numerical simulations provides insight into the distribution of biomass and consequent varying reaction rates throughout the tank.

It’s a critical task to ensure the safety of highway slopes constructed on highly plastic clay, given their significant economic, social, and safety impacts and the unique geotechnical challenges that they pose. The factor of safety (FS) for a slope is not constant, and it continues to evolve as a function of the weather condition, soil moisture content, and matric suction. Increasing heavy rainfall driven by climate change puts existing slopes at greater risk of failure. In this research, our goal is to create a particle swarm optimization (PSO)-based neural network (NN) (hybrid) model for predicting the long-term FS of highway slopes constructed on highly plastic clay considering the aforementioned factors. In order to obtain the necessary data for training the model, a series of 2D and 3D finite element analyses were performed, which were validated using a slope monitoring case study. In addition, sensitivity analyses were carried out to examine and assess the key factors that influence the FS of highway slopes. The developed PSO-based NN model can provide accurate FS predictions with a very low value of root mean square error.

This paper summarizes 3D geo-dynamic modeling work performed by Jacobs as part of assessing track embankment stability and ground response under dynamic loading from the high-speed train proposed for use on the California High-Speed Rail (CHSR) Project. The advanced numerical modeling technique was benchmarked with the well-known critical velocity case study site at Ledsgård, Sweden, using linear, equivalent-linear, and full non-linear soil models. The advantages and limitations of the modeling techniques are highlighted, with other design approaches and methodologies also briefly discussed based upon the experience of the Jacobs design team on other high-speed rail projects. This work shows that understanding the practical geotechnical management of Rayleigh waves is still under development. It requires a change away from normal geotechnical methods of static loading toward the use of methods appropriate for repeated loading conditions and the management of dynamic response.

Sensitive clays experience significant strain-softening behavior, that is, when subjected to large strains. They disintegrate into a remolded liquid with diminutive shear strength. When a slope begins to fail, the remolded clay keeps moving away from its original position causing subsequent failures, resulting in catastrophic aftermath. The capability to reproduce realistic strain-softening characteristics in the constitutive soil model is necessary for more accurate numerical slope analyses in sensitive clays. This paper illustrates a simple yet practical constitutive model specially developed for simulating the strain-softening behavior of sensitive clays. The model is then implemented in Anura3D, an open-source software that uses the material point method to simulate large deformations. The model is tested using the failure of a previously occurred retrogressive failure in sensitive clay. Finally, the model’s predictions have been compared with the actual post-failure run out of the landslide, and the results show that the model is reasonable and practical.

Constitutive modeling of cyclic liquefaction is challenging due to the complex soil behavior under cyclic loading. Aside from the predictive capability of the constitutive model, precise calibration of the model parameters is critical to accurately simulate the soil behavior in engineering applications. However, the calibration process for advanced constitutive soil models can be time-consuming and requires significant expertise. To address these challenges, an automatic parameter calibration tool (ACT) was developed to simplify the calibration process and reduce application hurdles. In this study, the performance of the ACT was evaluated on two versions of the SANISAND class of models with a focus on cyclic simple shear tests. Experimental data on Ottawa-F65 sand were used to establish calibration and prediction sets for the ACT. The results demonstrate the high efficiency of the ACT in calibrating the parameters within the limits of their simulative capabilities. The ACT significantly reduces the time and expertise required for calibration, making advanced soil models more accessible to engineers and researchers in the field of geotechnical engineering. The study provides a valuable contribution toward improving the accuracy and efficiency of cyclic liquefaction modeling.

Due to the material point method’s hybrid Lagrangian-Eulerian formulation, both conforming and nonconforming boundaries often exist throughout a simulation. Imposing nonconforming boundary conditions in the MPM remains a challenge due to the material boundary continually updating throughout a simulation. Regardless of the computational challenges, imposing Neumann boundary conditions is vital for modeling a wide variety of geotechnical problems. Recently, two variations of the MPM, the CutMesh MPM and the virtual stress boundary (VSB) method, have been proposed for imposing nonconforming Neumann boundary conditions. The current study uses two benchmark problems to directly compare the numerical performance of these newly developed variations of the MPM. For very simple problems, the CutMesh MPM has lower error than the VSB method, but as problem geometry becomes more complex, both methods produce comparable results.

This paper presents a method for slope stability analysis by finite elements. In this method, a potential slip surface is implemented as interface elements coupled with the soil slope as solid elements. The fundamental theory for this method and the general slice method is discussed. Stress mobilization analysis to bring the potential slip surface to a limit state is carried out by a coupled procedure between the interface elements and the solid elements. In order to couple the slip surface as interface elements and the soil slope as solid elements, the principle of energy stationarity must be satisfied in whole procedure to bring the slip surface from initial static equilibrium state to the limit equilibrium state. The principle of virtual displacement was used for both the slip surface as interface elements and the soil slope as solid elements in this method. The factor of safety for a potential slip surface is defined and discussed when applying the principle of virtual displacement on the slip surface and the soil sliding mass above the slip surface. This paper also presents an analysis case and comparison with the general slice methods.

A submerged floating tunnel (SFT) is a newly proposed design for a tunnel that operates while floating in water. The submerged floating tunnel needs to be connected into the shore ground to provide an underwater route between continents. In the shore connection, there is a risk of stress concentration due to the displacement imbalance between the underground tunnel part under high confining stress and the underwater tunnel part under low confining stress. Therefore, it is necessary to analyze the behavior of the shore connection. The shore ground where the tunnel is connected exhibits nonlinear behavior according to the interaction between the ground and tunnel. Therefore, in this study, the nonlinear deformation of the ground and the resulting change in the dynamic behavior of the submerged floating tunnel were analyzed using a numerical method. The nonlinear response of the ground was observed considering various ground types and the design using elastic joints that can improve the dynamic stability of the shore connection. As a result, the ground, showing nonlinearity, affected the dynamic behavior of the underwater tunnel part, acting as a boundary. In the case of shore ground with high stiffness, the ground exhibited a low deformation, which causes a slight nonlinearity. However, the higher the stiffness of the ground, the higher the maximum stress generated in the shore connection. Accordingly, it is considered that using an elastic joint to allow high displacement under the elastic deformation range and to prevent nonlinear behavior of the ground is an appropriate design to ensure dynamic stability.

Reliable evaluations of geotechnical hazards like landslides and debris flow require accurate simulation of granular flow dynamics. Traditional numerical methods can simulate the complex behaviors of such flows that involve solid-like to fluid-like transitions, but they are computationally intractable when simulating large-scale systems. Surrogate models based on statistical or machine learning methods are a viable alternative, but they are typically empirical and rely on a confined set of parameters in evaluating associated risks. Conventional machine learning models require an unreasonably large amount of training data for building generalizable surrogate models due to their permutation-dependent learning. To address these issues, we employ a graph neural network (GNN), a novel deep learning technique to develop a GNN-based simulator (GNS) for granular flows. Graphs represent the state of granular flows and interactions, like the exchange of energy and momentum between grains, and GNN learns the local interaction law. GNS takes the current state of the granular flow and estimates the next state using Euler explicit integration. We train GNS on a limited set of granular flow trajectories and evaluate its performance in a three-dimensional granular column collapse domain. GNS successfully reproduces the overall behaviors of column collapses with various aspect ratios that were not encountered during training. The computation speed of GNS outperforms high-fidelity numerical simulators by 300 times.

Earthen flood control structures such as dams and levees are vital for managing flood risk. In the United States alone, the United States Army Corps of Engineers manages approximately 14,000 mi of levees, with an average age of 59 years. Current inspection standards consist of walking or driving along the levee looking for important performance indicator signs. Manually inspecting levees for such conditions is time consuming, in certain instances possibly dangerous to inspectors, and is susceptible to human subjectivity in some instances with the sheer quantity of levees to inspect. In recent years across the broader civil infrastructure industry, remote inspection technologies, such as unmanned aerial vehicles, have begun seeing use and allowing for systems to be inspected faster and safer than traditional methods. The data obtained from these remote inspection tools can then be utilized with machine learning algorithms to classify and identify both failures, and potential future failure locations. This combination of rapid remote inspection for data collection, and machine learning for data processing, provides a powerful tool to enhance the condition monitoring of levees rapidly, efficiently, and remotely. This paper outlines some current technologies and use cases of these tools for levee inspections and discusses ongoing work related to machine vision and machine learning automation of visual imagery applied to crack detection using convolutional neural networks, where an accuracy of up to 99.76% is achieved for crack detection using a database of 20,800 images of concrete.

Characterizing seabed sediment is vital for offshore and nearshore engineering projects, often relying on in situ tests due to challenges in obtaining seabed samples. While cone penetration testing (CPT) is the conventional choice for seabed characterization, portable free fall penetrometers (PFFP) offer a cost-effective alternative in challenging environments, focusing on surficial seabed data. However, streamlined soil classification methods for PFFP data are lacking, as CPT-based approaches may not directly apply due to different penetration rates. This study introduces a 1D convolutional neural network (CNN) to predict sediment classes using the deceleration curve from a PFFP. Soil samples from deployment sites (Sequim Bay, WA; York and Potomac Rivers, VA) underwent laboratory testing for true classification labels. Initial results exhibit an 80% accuracy, suggesting potential practical applicability with further refinement and additional training data. This approach holds promise for reliable and practical seabed sediment classification in geotechnical applications.

California Bearing Ratio (CBR) test provides a measure of soil subgrade strength, and thus, it is widely used for pavement design. However, in practice, the CBR value is often estimated through dynamic cone penetration (DCP) testing as its equipment is portable and easier to set up compared to the CBR testing equipment. Over years, a lot of effort has been put into developing CBR-DCP correlation models, most of which are based on regression analysis. Moreover, most of the models developed are based on a wide dataset compiled from different regions, which makes them incapable of accounting for the spatial variability of soil. In this paper, hierarchical Bayesian modeling (HBM) approach is adopted to develop region-specific CBR-DCP models considering spatial variability. For this purpose, a dataset of 783 CBR-DCP data pairs from 32 locations in Ohio is compiled from the LTPP InfoPave database. The developed models using HBM approach are then compared to a conventional regression model which is developed based on the entire dataset from all regions. The analysis showed that the region-specific HBM models outperform the conventional regression model. Additionally, because data sparsity is a major concern in geotechnical engineering, the performance of the developed HBM models is investigated in a new region where only a few data points are available. The results indicated that the HBM models outperform conventional regression models in new regions with very limited data.

Underground duct banks carrying power cables dissipate heat to the surrounding soil. The amount of heat dissipated determines the current rating of cables which in turn affects the sizing of the cables. The dissipation of heat through the surrounding soils happens through conduction and convection. The mode of heat transfer depends on the soil’s thermal and hydraulic properties like diffusivity and permeability. The soil surrounding the cables could be designed to have maximum heat dissipation so as to have an improved current rating of cables. Differentiable programming is a novel technique that combines automatic differentiation with gradient-based optimization to minimize a loss function. Hence, differentiable programming can be used to evaluate input parameters based on output results. Given a desired heat distribution in the soil and a temperature source, we use differentiable programming to solve the inverse problem of estimating the soil permeability. In the present study, we employ differentiable programming to optimize the design of buried duck bank and the backfill soil to improve the heat dissipation. The design involves optimizing the permeability and size of the fill material compared to the surrounding natural soil. We develop an inverse finite difference code in the Julia programming language and ForwardDiff package to implement automatic differentiation. We demonstrate the design capabilities of the differentiable programming technique to obtain the optimum permeability of the back fill material from the norm of the temperature distribution in the surrounding soil.

The resilient modulus (M_R) of subgrade soil is an important factor in pavement design, which can be determined through repeated load laboratory testing. However, due to its high cost, time-intensive nature, and intricate process, geotechnical engineers and researchers have consistently expressed interest in creating an M_R prediction model using alternative soil test data. With advancements in prediction tools using machine learning (ML) algorithms, during the past decade, much research has been focused on implementing ML algorithms for the development of M_R prediction models. Many recent studies have been focused on using what is referred to as “black box” models, such as artificial neural networks (ANN). While these black box models provide high prediction accuracy, they are not explainable. In other words, there is no closed-form solution in which the relationship between the predicted label and the input features is explicitly stated. Moreover, these black box models are computationally complex and require advanced coding knowledge, making them inaccessible to many design professionals. In this paper, for illustration purposes, a dataset of 279 M_R test results that have been extensively used for ML-based model development of M_R is adopted from the literature. Different regression approaches, from simple regression models to advanced ML algorithms, are applied to this dataset, and the developed models are compared in terms of accuracy, simplicity, and explainability. Additionally, the bias and variance in error are investigated for different algorithms. It is concluded that the improved accuracy of the developed models is not always adequate to justify the effort, time, and complexity of adopting complex advanced algorithms instead of simple regression models. Moreover, it is shown that overly complex algorithms could lead to high variance, causing the algorithm to model random noise in the training data.

Many problems in geotechnical engineering are prime candidates for algorithmic learning, or “artificial intelligence” (AI), given the complex, interrelated, multivariate relationships needed to define the behaviors and responses of geo-materials. There has recently been a rapid increase in the use of AI in developing geotechnical prediction models. This is not surprising, considering the ease of implementing AI, its potential advantages over traditional statistical methods, and the immense attention it receives in the media and popular culture. However, despite a growing presence in the literature, AI models are still widely ignored by geotechnical practitioners. This paper aims to investigate the reasons behind this underutilization through the representative lens of one important topic in the geotechnical profession: soil liquefaction. The authors conducted a literature review of 75 AI liquefaction prediction models to identify why these and other AI models are not being adopted. This review reveals several reasons, including the lack of comparison to existing, familiar models, departure from established best practices, questionable uses of AI, overly complex presentation styles, and failure to provide models that can actually be used. These recurrent shortcomings must be addressed to enable the effective use, and realize the full potential, of AI in the geotechnical engineering profession. It is important to note that these shortcomings do not indicate that AI itself is inherently problematic or that all previous efforts have been in vain. The lessons learned from this review are relevant to all areas of geotechnical engineering and could help to improve the future direction and perceptions of AI in the profession.

Liquefaction potential is one of the most important measures when assessing soil liquefaction hazard. With advancements in prediction tools, over the past decade, many researchers have started applying machine learning algorithms to obtain a more accurate assessment of soil liquefaction potential. The adopted algorithms often require rigorous computational efforts to select the optimum set of parameters (referred to as hyperparameters) that leads to a reliable and accurate prediction. This paper presents a hyperparameter tuning method based on a probabilistic technique known as Bayesian learning, in which the optimum set of hyperparameters is selected providing a high prediction accuracy with less computational effort. For illustration purposes, a dataset of 170 case histories of CPT-based soil liquefaction from previous earthquakes is adopted. The XGBoost algorithm, which is known for providing high accuracy models for regression and classification purposes, is adopted herein for prediction of liquefaction potential. Bayesian hyperparameter optimization method is adopted for selecting the optimum set of parameters. The liquefaction potential prediction obtained through this approach is compared with other machine learning models developed by other researchers on the same dataset. The Bayesian hyperparameter optimization is shown to improve the prediction accuracy.

Numerous studies are currently underway to investigate the safety management and evaluation of underground structures, as they require continuous management beyond their initial construction. Although such structures are regularly evaluated in accordance with safety regulations, measurements of degradation factors are typically performed manually by workers. To improve this process, many researchers have turned to computer vision, which offers an objective and automated method for evaluation. However, the performance of many deep learning models in this field is heavily dependent on training data, which may hinder their effectiveness in new domains. To address this issue, this study proposes a new method for detecting cracks on the surface of concrete underground structures based on computer vision. The proposed method achieves high performance with small additional labeling processes by utilizing both labeled and unlabeled datasets from the new domain. The results show that it contributes to objective and effective management with small additional processes in new domains.

This paper explores the deployment of deep reinforcement learning (DRL), a subset of machine learning for automated decision-making, in a lab-scale geosystem. The developed lab-scale geosystem is a slope equipped with a pumping system, a rainfall simulator, a microcontroller, and a water level sensor. The main goal of this study is to establish an intelligent control of the groundwater in the slope. The DRL agent, deep Q-network (DQN), was first trained using the virtual environment to reduce the risk of failure and increase safety. Pre-training helps the agent learn the policy and goal of the system before deploying it in the real world. When the pre-training is complete, the agent was fine-tuned in the lab to adjust the DQN hyperparameters. The results showed that the pre-trained and fine-tuned DRL agent can autonomously regulate a pump to control the groundwater in the geosystem.

Machine learning (ML) is widely used in many industries to develop predictive models and gain new insights. This paper presents an application of ML to a geotechnical problem: estimating shear wave velocity (Vs) from cone penetration testing (CPT) data where no direct Vs data was obtained. This application is valuable when using historical CPT data obtained prior to the widespread availability of seismic CPT (SCPT) equipment or in situations where SCPT is difficult, such as below the water surface in impounded tailings and coal combustion residual (CCR) ponds. We developed both material-specific and universally applicable ML models to estimate Vs based on conventional CPT measurements. The models were built in Python using two tree-based algorithms: random forest and gradient-boosted trees. The ML models were compared to both existing widely accepted models and a linear regression technique to assess if the performance gains justify the increased model complexity. The ML models show more accurate estimates of Vs than other available methods.

While one-dimensional consolidation tests remain the benchmark for performing site-specific evaluations of consolidation characteristics, the time required to perform these tests has resulted in significant research interest in correlating consolidation and recompression indices to more readily available properties such as natural moisture content or liquid limit. Recent advances in machine learning provide the ability to analyze and draw inferences from a combined range of soil properties to better estimate the consolidation and recompression indices. This study utilizes a dataset of consolidation tests (n = 1,261) performed primarily on Holocene age clay and organic materials within the New Orleans area to evaluate the applicability and use of various machine learning algorithms, including multivariate ordinary least squares and random forest regressors. The results of these models are compared to traditional univariate linear regressions and provide commentary on the use of machine learning to statistically treat the variability inherent in geotechnical testing.

Current levee inspection practices consist mainly of visual inspections of the levee surface and sparse, instrumentation monitoring of a single data point at the location of installation. With an assigned grade of “D” from the United States Army Corps of Engineers for levees in the United States, many of these critical systems are in poor health and with significant deterioration putting millions of people at flood risk. There is a need for better levee inspection practices to assess the health of these systems and to create more standardized inspection practices among different organizations as these systems continue to age and deteriorate. By using a combination of UAV-enabled sensing techniques and geophysical techniques, both the surface and subsurface of levees can be spatially and temporally mapped in a non-invasive way, significantly improving current inspection practices. UAV LIDAR, in combination with two geophysical techniques, multichannel analysis of surface waves (MASW) for shear wave velocity and electromagnetic induction (EMI) for apparent resistivity, was used along a stretch of levee on Grand Island in the Sacramento-San Joaquin Delta in California, just south of Sacramento, to assess the feasibility of using these methods to map the health of sandy levees. The subsurface data was used as inputs into unsupervised machine learning algorithms to assess the health and structures of the subsurface. Preliminary analyses have shown that unsupervised machine learning can be used to cluster subsurface data by depth below ground surface.

Functionality and serviceability of surface and buried infrastructure play a fundamental role in modern living standards. A substantial portion of urban utilities are buried under the pavements; replacement and maintenance of which require extensive streetworks with both short- and long-term costs. This study proposes an asset management framework to estimate the direct and indirect costs of trenching operations using digital technologies including condition monitoring, image processing, big data collection, and analysis using machine learning techniques. The framework’s cost model element was initially developed using the existing literature, which would then be updated with collected data. This would allow the asset owners to estimate the long-term costs of the trenching method to inform their decision making. Application of the existing technologies was demonstrated through a case study. The framework can ultimately inform data-driven decision-making and supporting sustainable and effective urban infrastructure management practices.

Landslide mobility is essential for landslide risk assessment because impact scales with the distance traveled by the landslide mass in many cases. This paper aims to quantify the mobility of landslides triggered by the M_w 6.5 Lefkada earthquake on November 17, 2015, and develop regression models for landslide travel distance. The study leverages a high-quality landslide inventory that includes the location, area, and volume of 716 landslides, with the source and entire area for each landslide mapped separately. Multivariate linear and machine learning models are used to predict landslide travel distance. The dependent variable, the 3D landslide travel distance, is calculated from the inventory with the digital elevation model. Independent variables include terrain characteristics, material strength, and permanent seismic displacements estimated from seismic displacement models. The results show that terrain characteristics correlate most strongly with landslide travel distances. Furthermore, the multivariate linear regression, random forest regression, and stochastic gradient descent regression have better prediction capacity than k-nearest neighbors’ and XGBoost regression.

Machine learning methods have been increasingly adopted to predict the diameter of jet grout columns. However, a comparative study on the accuracy and reliability of various machine learning models has not been performed. This paper presents a comparison of the performance of the various machine learning models including K-nearest neighbors (KNN), artificial neural networks (ANN), support vector machine (SVM), and long short-term memory networks (LSTM) for the prediction of jet grout diameter. A database of published case studies consisting of soil type, properties of jet grout, and diameter was used to train the machine learning models for the prediction of the diameter of the jet grout column. The performance of the optimized machine learning algorithms was evaluated and compared. The ANN model was found to provide the best performance. K-fold validation (i.e., leave-one-out cross-validation) was performed for the ANN algorithm, and an ensemble model was developed for the prediction of diameter of the jet grout column.

Ground characterization is instrumental to managing ground risks in the built environment. Several variables affect the appropriate characterization of the ground, including the early desk study, site investigation, and ground data management. During project planning and design stages, the quality, management, and interpretation of collected ground data are imperative to develop reliable ground models that accurately represent the subsurface conditions. Owners and asset managers are increasingly recognizing the benefits of collecting digital data, or digitizing existing data, and storing it into centralized data repositories. Creating an up-to-date and centralized library of project information increases efficiency by reducing the repetition of costly investigations and allows for changes to be tracked over the course of the project, from inception to decommissioning. The adoption of advanced digital technology and state-of-the-art tools such as three-dimensional (3D) ground modelling software in tandem with traditional methods of ground data management is required to streamline the current industry practice of ground modelling at local and national levels. This paper presents a digital workflow for efficient and reliable ground characterization that can optimize the planning, design, construction, and operations of ground-related projects. Best practices associated with ground data collection, management, processing, and interpretation are reviewed and documented. Data visualization tools are also critical to help users develop reliable ground models. Therefore, the benefits of using 3D ground modelling to effectively visualize variability in geological conditions and project locations, along with its challenges and limitations, are discussed. The ground industry is quickly evolving and recognizing the unparalleled benefits of fully integrated data management workflows. It is intended that the recommended workflow provided in this paper can be referred to in the development and design of future projects.

Engineers need to know the strength of soil to support current and future infrastructure, that is, roads and buildings, to determine if the infrastructure will persist or fail. Geophysical tests offer a nondestructive standoff method that can be used to characterize and monitor soil properties. To begin this process, a database of geophysical and geotechnical values for common geotechnical soil types is needed to aid in site characterization. This project’s research goal was to record, collect, and share experimental geophysical and geotechnical data for publication in a reference database with Unified Soil Classification System (USCS) soil types. We performed geophysical field tests, near the Pemigewasset River near Bristol, NH, on unfrozen soil utilizing seismic methods, ground-penetrating radar, electrical resistivity tomography, electromagnetic induction, standard penetration tests (in progress), and boreholes (in progress) to obtain laboratory samples and ground truthing. We are compiling a database, to aid researchers and engineers when characterizing site soils from ground response information and provide a base dataset for modeling soil behavior.

Persistent long-term monitoring of civil infrastructure such as pavements, highway embankments, and earth retaining systems is of significant interest to asset owners and stakeholders. The use of orbital synthetic aperture radar (SAR), a remote sensing technique, to assess the condition and performance of such assets has been attracting major interest from transportation agencies over the last decade. Currently, operational orbital SAR remote sensing systems have a high spatial and temporal resolution and interferometric capabilities that facilitate spatiotemporal monitoring of asset conditions along with centimeter-level deformation monitoring. This paper aims to provide a review of the use of SAR and interferometric SAR (InSAR) data in monitoring transportation/geotechnical assets. The review also sheds light on the recent explosion of the commercial SAR sector and its potential contribution to near real-time monitoring transportation assets in the near future. The basics of SAR and InSAR data and their relevance in assessing the condition of embankment slopes including a case study in Texas are presented here, along with the challenges and limitations of this technology specific to civil infrastructure. This study would benefit transportation/geotechnical academics and practitioners in understanding the advantages of using these novel remote sensing techniques as a reliable data source for asset management.

Ground movements resulting from landslides are a frequent natural phenomenon that often occur in different regions of the state of Alabama. These movements pose threats to human life, causing infrastructure damage, and disrupting the highway network. Traditional approaches to detecting landslides require manual observations of settlement or cracking, which can be effective only after signs of distress can be observed. This research focuses on utilizing interferometric synthetic aperture radar (InSAR) deformation time series analysis along with GIS-derived geospatial information to study landslides and their cause in the southern regions of the US. In this research, deformation time series and mean velocity were estimated from September 2016 to February 2022 using Sentinel-1 InSAR (COMET-LiCSAR) product in a region that has recently experienced significant deformation caused by landslides, in the North Alabama highway located in Morgan County. Results of ground deformations obtained from InSAR were then combined with geotechnical, geospatial, and climate data such as precipitation, topography of the region, and soil moisture data to comprehensively understand the underlying causes of failure.