Revisiting the calibration philosophy of constitutive models in geomechanics

Conventional laboratory tests in geomechanics are inherently designed to create a uniform stress-strain field whereby relative homogeneity in data prevails at the expense of information diversity. The one-element convenience using stress homogeneity lends itself well to constitutive model calibration. From a mathematical standpoint, this approach potentially promotes uncertainty and error propagation. More recently, there has been a growing effort to utilize all data of conventional tests, including the effect of end restraints, to improve model calibration. This conceptual study revisits the notion of constitutive model calibration holding the perspective of inverse problems. For calibration purposes, nonconventional loading and complex boundary conditions were deemed to create a rich (nonuniform) strain-stress field. In this study, flexure excitation was investigated as a plausible test configuration in experimental geomechanics. An attempt was made to gauge the merits of flexural loading of a cylindrical soil specimen in providing independent data (information-rich test). The proposed test configuration was experimentally conducted and numerically optimized for assessment of model calibration.