Uncertainty in Liquefaction-Induced Settlement in Numerical Simulations due to Model Calibration

Liquefaction-induced reconsolidation settlements occur as earthquake-generated excess pore pressures dissipate and can lead to significant damage to overlying infrastructure as well as buried structures. Design of earthquake resilient infrastructure in areas with high liquefaction susceptibility requires a methodology to predict these settlements for a variety of soil types and boundary conditions. The state-of-practice empirical models that are commonly used to compute settlements exhibit several limitations that might affect the accuracy of predicted settlement, including an inability to include the effects of partial drainage, non-liquefiable crust, thin layers, and soil fabric. Numerical models can overcome some of these limitations but require proper calibration and validation. Centrifuge and shaking table experiments can be used for this validation, but multiple challenges arise when comparing numerical simulations and experimental results for liquefaction-induced settlements. This paper describes the need for soil-specific calibration of reconsolidation behavior which is one of the key challenges in accurate prediction of settlements. Multiple experiments considering free-field conditions are simulated using the numerical platform FLAC and the constitutive model PM4Sand. The bias in predicted settlement using the default PM4Sand reconsolidation formulation is assessed. A new, tentative relationship for the PM4Sand reconsolidation parameter, f(sed,min), is proposed by incorporating the mean grain diameter, D-50. Settlement results from simulations using the default and modified reconsolidation schemes are compared to the experiments to assess the level of agreement between them.