Puketoka formation soil deposits predominantly consist of a heterogeneous sequence of interbedded fine-grained sands, silts, and clays derived from volcanic ash and tuff, estuarine peat, and coarse-grained soils. These deposits are present in the southern and western parts of the Auckland Region, New Zealand. The available geological maps and reports indicate that Puketoka Formation contains pumiceous grains, though these are often not picked up or described by engineering geologists. Geology-based liquefaction criteria show that these deposits are unlikely to liquefy. However, the Puketoka formation soils have high pumice contents that may skew the results of traditional liquefaction analyses. Moreover, recent detailed desk studies for liquefaction assessment of this formation showed that the Puketoka Formation loose sands and sandy silts are potentially liquefiable due to ground shaking from earthquakes with a return period as short as 100 years; consequential effects would increase (greater settlements and greater risk of lateral spreading) for earthquakes with a return period greater than 250 years. Thus, more investigation is required to understand the seismic response of the Puketoka Formation and site-specific detailed assessments can help understand the liquefaction vulnerability of these deposits. Toward this end, numerical simulation of a site in South Auckland is performed using the PM4Sand model to investigate the behavior of these soil deposits in terms of stress-strain relation and excess pore water pressure generation. First, the model and input parameters are calibrated using available laboratory undrained cyclic direct simple shear tests on undisturbed Puketoka Formation soils. Next, site-specific 1-D numerical simulations are performed to investigate the liquefaction performance of the soil deposits. The results obtained provide better insights into the liquefaction potential of these deposits.

Tailings dams impound large amounts of saturated soil which can be highly susceptible to liquefaction. Liquefaction results in a severe loss of strength in the retained soil and potentially failure of the dam. If the dam is breached, a massive debris flow of liquefied soil is then released with potentially disastrous consequences downstream. Numerical models are frequently utilized to predict the liquefaction response of tailings dams and the potential runout, and these analyses inform engineering decisions regarding hazard avoidance and mitigation. The Finite Element Method (FEM) is a widespread tool which excels at modeling liquefaction triggering and initial movements, but it quickly loses accuracy when modeling large deformations due to mesh distortion. Conversely, the Material Point Method (MPM), a hybrid Eulerian-Lagrangian method, employs particles that move freely across a background grid and can account for large deformations without losing accuracy. However, issues with the accuracy of MPM’s stress distributions and limits associated with the available boundary conditions impair its ability to predict liquefaction initiation. In this paper, we utilize a sequential hybridization of the FEM and MPM methods as a superior alternative to either individually. To demonstrate the efficacy of this hybrid method to simulate the entire process of tailings dam failures from initiation to runout, we model the 1978 Mochikoshi Tailings Dam failure. In this case, the dam collapsed during the main shaking of an earthquake due to the combined effects of the inertial seismic loading and liquefaction. We initiate this model in FEM to capture the immediate effects of the earthquake: the seismic response and liquefaction triggering. Then, we transfer the model into MPM, inheriting the FEM failure mechanism and capturing the runout behavior without mesh issues. The analysis successfully captures the liquefaction triggering and movements, but underestimates the final runout. Further refinements to the MPM phase of the analysis are required to better capture the large runout response.

The large strain liquefaction behavior was analyzed by focusing on the relationship between the normalized cumulative dissipated energy and the internal friction angle. As a result, the internal friction angle defined by the peak stress ratio showed a decreasing trend , and the decreasing trend became more significant with decreasing D_r. A single-element liquefaction simulation was performed using an effective stress analysis method based on finite deformation theory that incorporates the decreasing trend of f . As a result, the employed effective stress analysis method was able to simulate the liquefaction behavior up to the large strain level which is difficult to perform with conventional methods.

This study uses two-dimensional (2D) nonlinear dynamic analyses and one-dimensional (1D) inverse analyses of accelerations from virtual vertical arrays to examine the effect that a flexible shear beam model container has on the dynamic stresses that developed within centrifuge models of level saturated sand. The 2D dynamic analyses are performed using the finite difference program FLAC2D 9.0 with the user-defined constitutive model PM4Sand to obtain cyclic stress ratios on horizontal planes (CSR_xy) in the soil. The 1D inverse analyses, which assume 1D shear beam responses, are performed using virtual vertical array data extracted from the 2D dynamic models (i.e., horizontal nodal accelerations along columns) to also obtain cyclic stress ratios on horizontal planes (CSR_1D,Σma). Temporal and spatial differences between the CSR_1D,Σma and CSR_xy from the same analysis model provide a basis for estimating potential errors in the CSRs from either 1D dynamic or 1D inverse analyses of centrifuge array data. The 2D nonlinear dynamic and 1D inverse analysis methods are described, followed by results for three different centrifuge model scenarios. The container effects on cyclic stress ratios for this class of problem are summarized and discussed.

Saturated soil is generally modeled in terms of solid (soil skeleton) and liquid (pore fluid) phases. A model for the reciprocating phase transition between solid and fluid in the liquefied ground is proposed based on the three-phase model for saturated soil, which considers floating soil particles. The model considers changes in the soil skeleton's porosity, which is a function of the effective confining pressure. We introduced stiffness degradation and recovery mechanisms into the existing elasto-plastic model to account for the phase transitions. Numerical simulations from both cyclic loading tests and centrifuge tests demonstrate that our model closely aligns with experimental results. The results clearly indicate our model's capability to accurately simulate the dynamic behavior of liquefied ground.

The concern regarding the overestimation of liquefaction resistance in granular materials arises from the challenge of achieving a zero effective stress state in laboratory element tests. The influence of particle self-weight due to the gravity effect, is considered a significant contributing factor to the non-zero effective stress issue. To investigate the impact of gravity on liquefaction behavior, this study employed the discrete element method (DEM) to conduct a series of bi-axial loading simulations. A single layer of spherical particles was initially consolidated under normal gravity (1g) condition at a stress level of 100 kPa. Subsequently, constant-volume cyclic loading tests were performed at varying gravity levels ranging from 0g to 1g, with a cyclic deviator stress of 60 kPa. The findings demonstrate the substantial role of gravity in liquefaction behavior. Lower gravity conditions exhibited reduced stiffness, increased contractive behavior, and lower resistance to liquefaction. In the absence of gravity, specimens with initially highly mobile particles experienced a temporary loss of inter-particle contacts, resulting in a complete absence of frictional interactions at the zero effective stress state. This highlights vulnerability to liquefaction onset, significant accumulation of strain during liquefaction, and diminished resistance in the post-liquefaction stage. Conversely, the presence of gravitational forces enabled the development of frictional interactions between particles, even during liquefaction. This gradually evolving deformation process exhibited a lower likelihood of liquefaction and higher resistance in the post-liquefaction stage under higher gravity conditions. These findings underscore the importance of considering the potential influence of gravity in laboratory-based assessments of liquefaction resistance of soil.

The earthquake-induced pore pressure response of a 5-m clean Ottawa F-65 sand layer of 45% relative density, under an effective overburden of 1 atm and subjected to earthquake base excitation, is numerically simulated. The layer sits on an impervious boundary and is free to drain at the top to simulate a possible partial drainage condition in the field. Numerical software FLAC and constitutive model PM4Sand are used for the simulation. The model is calibrated using the results of an available high-quality centrifuge experiment conducted in the Geotechnical Centrifuge Facility at Rensselaer Polytechnic Institute. Extensive instrumentation was utilized in the experiment to measure the dynamic soil response, including time histories of acceleration, pore water pressure and displacement, as well as bender elements measurements of shear wave velocity. Furthermore, System Identification is used to back-figure time histories of shear stress and shear strain. The instrumentation at different elevations in the experiment provided a comprehensive coverage of the soil response, which supported a system-level model calibration. A brief parametric study of the model’s input parameters preceded the model calibration in order to investigate how they affect the system-level behavior and to establish a proper calibration procedure. The predictions of the calibrated model are in excellent agreement with the records from the centrifuge, providing a strong basis for using the calibrated model to extend the results of the centrifuge experiment to other earthquake input motions, sand layer thicknesses and drainage boundary conditions. The calibrated numerical simulation was repeated after changing the layer thickness from 5 m to 10 m. The results of the new simulation are discussed to illustrate the possibilities of the new calibrated numerical tool.