PARTICLES 2025

Two-Phase Material Point Method for Simulating Liquefaction-Induced Large Deformations in Sandy Soils

  • Kurima, Jun (Institute of Industrial Science, The Universi)
  • Chandra, Bodhinanda (University of California, Berkeley)
  • Soga, Kenichi (University of California, Berkeley)

Please login to view abstract download link

Liquefaction-induced ground deformation often involves large deformations, which is challenging to model with conventional mesh-based methods due to issues like mesh entanglement. This study employs the Material Point Method (MPM), which combines Lagrangian particles with an Eulerian background grid, to analyze seismic liquefaction and subsequent flow deformation. Our methodology considers a two-phase, single-point u-v-p MPM formulation to model coupled fluid-soil behavior. A semi-implicit time integration scheme (Kularathna et al., 2021) is used to separate the pore-pressure field from the kinematics, solving it implicitly through a Poisson-type equation which contributes to numerical stability for liquefaction analysis. Furthermore, the CycLiq bounding surface model (Wang et al., 2014) is implemented to describe sand behavior under cyclic loading. The framework was validated through two centrifuge experiment simulations. In a 25G test on level Toyoura sand, the model successfully reproduced key liquefaction phenomena, including excess pore pressure generation and cyclic mobility. Pore pressure response and ground acceleration showed good agreement with experimental measurements. Moreover, for a 50G test featuring a sloping embankment, the simulation model is able to capture large deformation of flow failure, including embankment settlement (≈0.5m prototype scale) and lateral spreading (≈2.0m prototype scale). While qualitative behavior patterns were well reproduced, some quantitative discrepancies are still observed in the measured deformation magnitudes. This study highlights the capability and potential of the two-phase MPM in simulating both liquefaction triggering and subsequent large-deformation flow. Results confirm the approach's suitability for analyzing these complex phenomena while elaborating on the need for further refinement in quantitative predictions. Future work should address numerical artifacts, such as stress oscillations and constitutive parameter calibration, for large strain conditions.