Search for: All records

Non-linear dynamic analyses (NDAs) can capture the complex dynamic behavior of the soil using properly calibrated constitutive models. However, the quality of results from an NDA tudy hinges on several factors. Validation, which involves comparing numerical results to physical measurements, can assess the ability of an NDA to capture key responses through selected metrics. This study presents the application of a time history-based validation metric for evaluating the performance of numerical simulations. The centrifuge experiment conducted at UC Davis under the LEAP-2017 project, along with simulations performed using the PM4Sand constitutive model, provides the experimental and numerical data, respectively. The validation of the simulations against experimental measurements using the proposed metric is followed by a discussion on the potential experimental and numerical sources causing the quantified discrepancies. Conclusions are drawn on the effectiveness of the investigated metrics in facilitating the performance evaluation of numerical simulations and enhancing their reliability. 

Non-linear dynamic analyses (NDAs) are widely used in engineering practice to evaluate the seismic performance of geotechnical structures affected by liquefaction or cyclic softening. The quality of results from an NDA study depends on several technical and nontechnical factors. Validation, wherein a numerical prediction is compared to its physical counterpart, can assess the ability of an NDA to capture the various metrics of the response and potentially provide guidance toward improving the prediction. This study investigates select methodologies and validation metrics commonly used in signal processing problems to assess their effectiveness in capturing discrepancies between experimental and simulation results for a specific response of interest. Three simple problems are initially evaluated to analyze the metrics’ capabilities and identify necessary modifications. Then, the metrics are applied to nine sets of experimental and simulation time series, focusing on one response of interest (pore water pressure). It is found that cross-correlation successfully captures the lag in the initiation of liquefaction triggering, while Russell’s error metric captures magnitude and phase discrepancies. 

The abundance of post-earthquake data from the Canterbury, New Zealand (NZ), area can be leveraged for exploring machine learning (ML) opportunities for geotechnical earthquake engineering. Herein, random forest (RF) is chosen as the ML model to be utilized as it is a powerful non-parametric classification model that can also calculate global feature importance post-model building. The results and procedure are presented of building a multiclass liquefaction manifestation classification RF model with features engineered to preserve special relationships. The RF model hyperparameters are optimized with a two-step fivefold crossvalidation grid search to avoid overfitting. The overall model accuracy is 96% over six ordinal categories predicting over the Canterbury earthquake sequence measurements from 2010, 2011, and 2016. The resultant RF model can serve as a blueprint for incorporation of other sources of physical data such as geological maps to widen the bounds of model usability. 

Buried water reservoirs are increasingly being built to replace open aboveground municipal water supply reservoirs in urban areas to enhance water quality and utilize their surface footprint for other purposes such as public parks or placement of solar arrays. Many of these lifeline structures are in seismically active regions and, as such, need to be designed to remain operational after severe earthquake shaking. However, evaluating their seismic response is challenging and involves accounting for the interaction of the structure with the stored fluid and the retained soil; in other words, accounting for fluid–structure–soil interaction (FSSI). This paper presents a combined experimental–numerical study on the seismic behavior of buried water reservoirs while considering FSSI. Two series of centrifuge model tests were performed at different reservoir orientations to investigate one-dimensional (1D) and two-dimensional (2D) motion effects under full, half-full, and empty reservoir conditions. Corresponding numerical models were developed whereby the structure and the soil were represented by continuum Lagrangian finite elements, while the fluid was modeled via Arbitrary Lagrangian Eulerian formulation. Soil–structure and fluid–structure interface parameters were calibrated using the experimental measurements. The simulations successfully captured the measured reservoir responses in terms of accelerations, bending moment increments, and water pressures. The study found that the common assumption of plane strain is not applicable for reservoirs because their behavior was found to be truly three-dimensional (3D) whereby stresses accumulated at the corners. Furthermore, the full reservoir resulted in the highest seismic demands in the reservoir walls and roof while the empty reservoir yielded the highest base slippage. The study demonstrates that the complex reservoir seismic response is best captured by carrying out a 3D FSSI numerical simulation. 

The abundant post-earthquake data from the Canterbury, New Zealand (NZ) area is poised for use with machine learning (ML) to further advance our ability to better predict and understand the effects of liquefaction. Liquefaction manifestation is one of the identifiable effects of liquefaction, a nonlinear phenomenon that is still not well understood. ML algorithms are often termed as “black-box” models that have little to no explainability for the resultant predictions, making them difficult for use in practice. With the SHapley Additive exPlanations (SHAP) algorithm wrapper, mathematically backed explanations can be fit to the model to track input feature influences on the final prediction. In this paper, Random Forest (RF) is chosen as the ML model to be utilized as it is a powerful non-parametric classification model, then SHAP is applied to calculate explanations for the predictions at a global and local feature scale. The RF model hyperparameters are optimized with a two-step grid search and a five-fold cross-validation to avoid overfitting. The overall model accuracy is 71% over six ordinal categories predicting the Canterbury Earthquake Sequence measurements from 2010, 2011, and 2016. Insights from the SHAP application onto the RF model include the influences of PGA, GWT depths, and SBTs for each ordinal class prediction. This preliminary exploration using SHAP can pave the way for both reinforcing the performance of current ML models by comparing to previous knowledge and using it as a discovery tool for identifying which research areas are pertinent to unlocking more understanding of liquefaction mechanics. 

The rise of data capture and storage capabilities have led to greater data granularity and sharing of data sets in geotechnical earthquake engineering. This broader shift to big data requires ways to process and extract value from it and is aided by the progress in methodologies from the computer science domain and advancements in computer hardware capabilities. General machine learning (ML) models typically receive a set of input parameters and run them through an algorithm to gain outputs with no constraints on the parameters or algorithm process. Three topic areas of ML applications in geotechnical earthquake engineering are reviewed and summarized in this paper: seismic response, liquefaction triggering analysis, and performance-based assessments (lateral displacements and settlement analysis). The current progress of ML is summarized, while the challenges and potential in adopting such approaches are addressed. 

This paper investigates and presents the numerical modeling and validation of the response of a uniform clean sand using monotonic and cyclic laboratory tests as well as a centrifuge model test comprised of a submerged slope. The dynamic response of the sand is modeled using a critical state compatible, stress ratio-based, bounding surface plasticity constitutive model (PM4Sand), implemented in the commercial finite-difference platform FLAC, and PM4Sand’s performance is evaluated against a comprehensive testing program comprised of laboratory data and a well-instrumented centrifuge model test. Three different calibrations informed by the lab and centrifuge data are performed and the goodness of the predictions is discussed. Conclusions are drawn with regards to the performance of the simulations against the laboratory and centrifuge data, and recommendations about the calibration of the model are provided. 

A broad spectrum of well-graded, coarse-grained soils are commonly present in natural deposits, though characterization of these materials has been approximated using sand-based engineering methods in liquefaction evaluations. Through combined results of 31 constant stress direct simple shear and drained triaxial compression tests, this study experimentally investigates the effect of mean grain size (D50) and gradation (Cu) on the drained monotonic strength and stress-dilatancy of poorly- to well-graded, coarse-grained soils. Coarse-grained mixtures of varying D50 and gradations were prepared to relative densities of 20%–75% and tested under a range of overburden stresses. Results are analyzed in terms of the frictional resistance and dilative contributions to the shear strength of soils with varying gradations, as compared to clean sands, using different shearing modes. It is shown that (1) increased gradation of soils increases the peak shear strength and frictional resistance due to a greater initial rate of dilation exhibited in well-graded, coarse-grained soils; and (2) current stress-dilatancy relationships underestimate the dilative behavior of well-graded test materials.