Advancing Data – Driven Site Characterization Techniques in the Age of Artificial Intelligence

Abstract

Data-driven site characterization is one of the cornerstones of geotechnical engineering design and construction. Current practice of interpreting site investigation data to establish stratification models is inadequate in handling the complexities and uncertainties. As a result, the subsequent geotechnical reliability analysis cannot explicitly consider the impact of these uncertainties. This presentation introduces a novel approach for generating a subsurface stratification model with the ability to quantify model uncertainty using structured site data, stochastic modeling, and Bayesian machine learning. Examples of applications are presented.