A Machine Learning Based Quantitative Integration of Geotechnical and Geophysical Data for Deepwater Site Characterization-Paper

Paper

Pdf Summary

The paper presents a framework for estimating offshore soil strength by quantitatively integrating geophysical data with geotechnical measurements. Its focus is high-resolution 3D seismic data, which offer broad spatial coverage and deeper penetration than sub-bottom profiler (SBP) data, but lower vertical resolution.<br /><br />The workflow inverts seismic reflections to obtain acoustic impedance, converts impedance to soil porosity using empirical relationships, and uses porosity-related parameters and effective stress in an artificial neural network (ANN) trained on Gulf of Mexico geotechnical data. The database includes 903 direct simple shear and cone penetration test measurements. In a deepwater example, impedance profiles derived from seismic data generally matched core measurements and captured major sediment boundaries. At a test boring excluded from model training, predicted undrained shear strength agreed satisfactorily with CPT measurements, with a coefficient of variation of 0.19.<br /><br />The approach could support preliminary design for deep foundations, wellbore fracture-pressure assessment, and deep-seated slope-failure analysis, potentially helping optimize site investigations. However, 3D seismic resolution is insufficient for many shallow-foundation applications, while SBP data offer higher resolution but typically require full-waveform data and nearby ground-truth measurements. Predictions may also be unreliable where excess pore pressure or complex geological conditions invalidate the assumed relationships.<br /><br />The authors recommend further development of deepwater SBP processing and inversion, routine delivery of full-waveform SBP data, improved machine-learning methods for uncertainty assessment, and additional research on areas with elevated excess pore pressure.

Keywords

offshore soil strength

3D seismic data

acoustic impedance

soil porosity

artificial neural network

geotechnical measurements

undrained shear strength

cone penetration test

deepwater geotechnical analysis

sub-bottom profiler