Year: 2024
Type: article
Abstract: The Multichannel Analysis of Surface Waves (MASW) method has recently garnered significant attention due to its capability to non-destructively evaluate subsurface ground characteristics.However, the conventional inversion methods for estimating shear wave velocity at different depths have issues due to the non-uniqueness, non-linearity, and ill-posed nature of dispersion curves, leading to reliance on the engineer's subjective judgment and experience.To avoid such issues of inversion methods, this study proposes a deep learning-based approach to directly predict shear wave velocity from MASW data.A Temporal Convolutional Network (TCN), suitable for processing spatiotemporal data, was applied with MASW measurements as an input and shear wave velocity as an output.The experiment results, obtained with 2,778 samples augmented from 926 measurements, show that the proposed method achieved the prediction accuracy of 90.20% and 88.02% for the training and testing datasets, respectively.It can thus be expected that the accuracy of shear wave velocity predictions can be further improved through the advancement of deep learning models and the training of data under various ground conditions.
DOI: 10.55785/jcar.3.3.1
Authors: Aram Seo, SangUk Han
Institution: Hanyang University
Language: English
Topic: Seismic Waves and Analysis
Subfield: Geophysics
Field: Earth and Planetary Sciences
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4402894072