Improving Drained Soil Strength Parameter Prediction in Geotechnical Engineering: Leveraging Triaxial Stress Measurements with Hierarchical Bayesian Models

In offshore geotechnics, empirical correlations from the literature are often used to interpret CPT data supplementing measurements on shear parameters. However, it has been demonstrated that site-specific correlations should be employed whenever possible, as they better account for the unique soil behavior at a given project site. For sandy soils, the friction angle is a parameter with a significant impact on a design and is typically derived from triaxial tests conducted on at least three samples per soil type, followed by linear regression analysis. These friction angles are then used to establish site-specific correlations with CPT parameters. This process inherently introduces uncertainty into the used friction angles. In this study, a new model is proposed that integrates the linear regression of sample data with a regression linking the CPT parameter and the friction angle, utilizing various Bayesian modeling approaches. The findings indicate that the hierarchical Bayesian model applied to the triaxial test samples provides the highest predictive accuracy. However, site-specific correlations derived from this method can still be readily adopted by practitioners, enhancing the reliability of project-specific soil behavior predictions.

Keywords: Hierarchical Bayesian Model (HBM), Site-specific correlation, Transformation model, Triaxial test, Friction angle prediction, Offshore data, Transformation model.