A physics-informed semi-empirical probabilistic model for the settlement of shallow-founded structures on liquefiable ground

Abstract

This paper presents a predictive model for the settlement of shallow-founded structures on liquefiable ground during earthquakes. The model is based on the results of an extensive fully coupled, three-dimensional numerical parametric study of soil–structure systems, validated with centrifuge experiments as well as a database of case history observations. The results of the numerical study provided insight into the relative importance and influence of each input parameter and the functional form of the predictive model for a structure's permanent settlement. The case history database helped validate and refine the predictive model, accounting for complexities of the ground motion and site conditions in the field. Non-linear regression and latent variable analysis were used to develop model coefficients. The uncertainty around model estimates was modelled by a lognormal distribution. An additional logistic model was provided to estimate the probability of insignificant settlement (defined as less than 1 cm). The proposed probabilistic procedure considers variations in site conditions as well as the presence and properties of a building in three dimensions. By including the case history database in its validation and adjustment, the model captures all mechanisms of settlement below the foundation, including volumetric and deviatoric strains as well as ejecta. The total uncertainty around its predictions is rigorously characterised, which is a necessary step before the benefits of performance-based seismic design can be realised in the evaluation and mitigation of the liquefaction hazard.