Integrated approach for the MASW dispersion analysis using the spectral element technique and trust region reflective method

Abstract

Abstract In this paper, a semi-analytical elastodynamic forward solver was proposed for the Multichannel Analysis of Surface Waves (MASW) using the spectral element technique for the first time. A root finding method based on Brent's algorithm was proposed to accurately extract the dispersion curves. The trust region reflective method, an effective bound-supported least square algorithm, was applied, for the first time, for the inversion analysis of MASW data. In comparison to the commonly used neighborhood algorithm, the proposed solver converges rapidly; the Euclidean distance between the measured and numerically calculated dispersion curves are significantly reduced within only 300 runs of the forward solver in comparison to over 10,000 runs using the neighborhood algorithm. Several numerical case studies were performed to demonstrate the selection of initial guesses as well as the bounds of each optimization parameter. Also, a parametric study was performed to determine the effect of discontinuity in the soil stratigraphy on the dispersion curves. Finally, the integrated approach developed in this study was used to analyze the MASW data collected in a site in south Iceland. It was concluded that the proposed approach determines effectively and efficiently the soil stratigraphy as well as soil properties.