Journal of Korean Society of Agricultural Engineers ISSN:1738-3692(Print) 2093-7709(Online) 한국농공학회논문집

The determination of soil dry unit weight is crucial for assessing the stability of diverse geotechnical structures. While empirical regression methods have been proposed to predict dry unit weight using data from various sounding techniques (e.g., CPT, SPT, and DCPT), these approaches may exhibit uncertainties when applied across different regions or soil types, thus compromising their reliability. This study investigates the efficacy of Bayesian regression analysis in conjunction with outlier processing techniques for the probabilistic prediction of dry unit weight, utilizing Dynamic Cone Penetration Test (DCPT) and in-situ density test data. The findings demonstrate that Bayesian methods, coupled with outlier processing, effectively quantify prediction uncertainty and leverage prior information to yield more robust and accurate predictions. The proposed methodology facilitates reliable and precise probabilistic estimations of dry unit weight using the cost-effective and straightforward DCPT. Moreover, this approach offers the advantage of deriving site-specific prediction models by integrating both the distribution of data from previous studies and a limited amount of field data, as opposed to directly applying generalized formulas. The results underscore the potential of this methodology in enhancing the accuracy and reliability of dry unit weight predictions in geotechnical engineering. Further research exploring the influence of varied soil conditions and the quantity of field data on prediction outcomes may enhance the generality and applicability of these findings.