Development of a framework for automatic quantification of uncertainty in seismic cone penetration testing

Abstract

Accurate quantification of the shear wave velocity, Vs, of geo-materials is an important consideration in geotechnical design. Seismic Cone Penetration Testing (SCPT) measures shear wave travel times from a source to in situ receivers along assumed travel paths to calculate Vs. Despite complexities and uncertainties associated with obtaining Vs, results are often reported to designers as a single deterministic profile without an intuitive measure of uncertainty that can be incorporated into the design process. A rigorous workflow to rapidly obtain uncertainty-quantified profiles from SCPT using a Bayesian inversion approach is developed. While similar approaches have been documented, this inversion approach explicitly considers sources of measurement error which are generally neglected (i.e., assumed to be low) in order to deliver more realistic probability distributions of true Vs and improve robustness against imperfect data. Such errors can remain undetected when using traditional approaches, despite potentially leading to inaccuracy. Additionally, an outlier detection framework is incorporated into the workflow to improve accuracy. The workflow is demonstrated by application to a large database of SCPT data. The results show significant improvement over existing methods in terms of robustness and validity, and therefore that the workflow is a valuable tool for practical analyses. Further, they provide crucial insight into the prevalence and magnitude of key errors which are traditionally present but undetected.

Full Paper