Identifying Soil Parameters by Inverse Analysis of Field Testing

The identification of constitutive parameters by inverse analysis of limited experimental information is investigated. In order to ensure the quality of measurements, all the data come from a thick wall cylinder test performed on a saturated sand sample. A numerical FEM-simulation considering a linear elastic perfectly plastic model (Mohr-Coulomb) is carried out in order to determine the pressure — cavity deformation relationship. Two kinds of optimization algorithms, a deterministic and a stochastic one, are presented and implemented in the inverse procedure. The performance of theses algorithms taking for identifying three parameters (Young's Modulus, friction angle and dilatancy angle) is compared. The two kinds of algorithms provide a quite similar response for the mathematical optimum, but some parameters sets did not have an acceptable physical meaning. Contrary to the deterministic algorithm, the genetic method provides an assembly of "satisfactory" individuals. Some of the parameters sets have acceptable physical meaning and the expert has to choose the best one considering his scientific and technical background.