Geophysical data (e.g., high-resolution 2D/3D seismic data, ultra-high frequency sub-bottom profiler data) play a key role in characterizing shallow subsurface conditions and identifying near-seafloor geohazards in offshore environments. Modern deepwater geotechnical site investigations utilize geophysical data for planning and execution to comprehensively characterize the target site and ensure geotechnical data can be properly extrapolated and used in a cost-efficient way. However, geophysical data usually only provides qualitative information (e.g., soil stratigraphy) and the quantitative evaluation of the engineering properties of soils for design almost exclusively relies on ground truth methods, e.g., samplings, borings, and penetrometer tests. As such, heterogeneity in the soils may not be accurately captured. In this paper, a framework for quantitatively integrating geophysical and geotechnical data is established with the motivation of extracting quantitative soil properties directly from geophysical data for preliminary foundation design. The framework consists of three key components: 1) acoustic impedance inversion using high-resolution 3D seismic data; 2) linkage between the acoustic impedance and the soil porosity; and 3) a machine learning algorithm to predict the soil undrained shear strength based on the existing database. An example application of the proposed framework is presented and satisfactory agreement is achieved between the blindly predicted and measured soil undrained shear strength.
Jinbo Chen, Liqin Sang, Bryce Willems, Jason Newlin, Øyvind Ruden, Ritu Sarker and Shuang Hu
Keywords: Geophysical, Geotechnical, 3D Seismic, Acoustic Impedance, Inversion, Near-Seafloor, Machine Learning
Publisher: DFI