In-Situ Viscoelastic Soil Parameter Estimation Using Love Wave Inversion

摘要

Industry best practices for estimating viscoelastic soil properties employ either crosshole seismic surveys, or ex-situ laboratory testing. The former method can be costly, and its area of investigation limited to a few meters. The latter method samples only a tiny volume from the research area, and requires that samples be disturbed from their native condition. We investigated an alternative method that uses Love wave inversion to estimate layer geometry, shear modulus, and soil viscosity. We derived a method for determining the complex velocity of Love wave modes in horizontally layered viscoelastic media, and used the method to investigate the behavior and propagation of the Love wave fundamental mode and first three overtones in one, two, and three layered media. We studied the evolution of Love wave modes with increasing frequency, and found that roots representing the complex velocities of Love wave modes evolve in pairs, with one root originating from along the real axis, and the other root originating from along the imaginary axis. In all cases studied, we observed that only one root from each pair was expressed as a propagating wave.The simultaneous propagation of multiple Love wave modes poses a challenge to their separation and analysis. We developed a technique for separating Love wave modes and used the information thus obtained to produce dispersion and attenuation relationships. We characterized the technique, and demonstrated its viability using synthetic data.Using these dispersion and attenuation relationships, we used the Gauss-Newton inversion method to deduce best-fit layer property models. We investigated the methodu27s sensitivity to constraints on model properties, and to the types of data used in inversion.For a single-layer soil model, we found that the method gives reasonable estimates of layer thickness, shear modulus, and viscosity. For two and three-layer systems, we found it necessary to constrain layer thickness in order to obtain consistent estimates of shear modulus and velocity. We thus conclude that the Love wave method is best used for extending estimates obtained using crosshole or downhole information as a control.In some cases, we found it difficult to ascertain, a-priori, which of the modes from each pair was manifested in the data. When the wrong root is chosen, the model may converge to erroneous soil property estimates. We recommend that future work be directed at developing techniques for ensuring that the correct root is used in the inversion model. So long as the correct root is used, Love wave inversion offers a viable in-situ method for estimating viscoelastic soil properties.