Identification of soil dynamic properties through an optimization analysis

An identification technique is used to estimate the nonlinear dynamic properties of a soil deposit using the acceleration records provided by a vertical (downhole) array along with a nonlinear least squares optimization algorithm. The technique employs non-parametric estimates of the shear stresses derived from the recorded accelerations; therefore, it does not require the forward modeling of the whole soil deposit. Soil properties are described by a hyperbolic shear stress-strain relation. A multi-surface plasticity approach is used to model the stress-strain relation. Convergence and accuracy of the identification technique are assessed using numerical simulations. Two centrifuge experiments are used to validate and demonstrate the capabilities of the technique to estimate the stiffness and damping ratio profiles of a site subjected to base excitation. Performance of the technique is also evaluated using field data corresponding to the 1987 Superstition Hills earthquake recordings from the Wildlife Liquefaction Array at the Imperial Valley in Southern California. The identification results were found to be in good agreement with direct measurements of shear wave velocities.