Ability of three different soil constitutive models to predict a tunnel's response to basement excavation

Many constitutive models are available nowadays to predict soil-structure interaction problems. It is sometimes not very easy for engineers to select a suitable soil model to carry out their design analyses in terms of complexity versus accuracy. This paper describes the application of three constitutive models to back-analyse a well-instrumented centrifuge model test, in which the effect of basement excavation on an existing tunnel was simulated. These three models include a linear elastic perfectly plastic model with the Mohr-Coulomb failure criterion (called MC model), a nonlinear elastic Duncan-Chang model (DC), and a hypoplastic model (HP), the last of which can capture the state-, strain-, and path-dependent soil stiffness even at small strains and path-and state-dependent soil strength. By comparing with measured data from the centrifuge model test, it is found that the HP model yielded the best predictions of tunnel heave among the three models. Not only the gradient, but also the magnitude of tunnel heave is well predicted by this HP model. This can be explained by the fact that the HP model can capture the state-, strain-, and path-dependent soil stiffness even at small strains and path-and state-dependent soil strength, but not the MC and DC models. However, all three models underestimated the change in tunnel diameter and the maximum tensile bending strain in the transverse direction.