Calibration of Soil Constitutive Models with Spatially Varying Parameters

Abstract

Soil constitutive models are frequently calibrated from laboratory tests that utilize global boundary measurements, which necessarily relegate soil response to that of a homogenized equivalent medium. This paper demonstrates the applicability of advanced experimental technologies to enhance the state of model-based predictions in soil mechanics by taking into account the possibility of material heterogeneity during model calibration. By utilizing the full-field displacement measurement technique of three-dimensional digital image correlation, displacements of the surfaces of deforming triaxial sand specimens are measured throughout deformation. These displacements are assimilated into finite-element (FE) models of the test specimen through solution of an inverse problem. During optimization, in which the difference between measured and predicted displacements across the specimen surface form the basis for the objective function, model parameters are allowed to vary spatially throughout the specimen volume. FE models allowing three different levels of spatial variability are tested. Results indicate that accommodating consideration of material heterogeneity during calibration leads to more accurate predictions of global stress-strain behavior that are more faithful to observed full-field response.