A site-retrieval approach for the prediction of jet-grouted parameters

Abstract

Jet grouting is a commonly employed method for improving soil conditions, yet its design frequently involves uncertainties that might increase cost during construction projects. At preliminary phase design, it is common practice to infer local mechanical and geometrical prediction of jet-grouted columns from theoretical or empirical-based correlations. However, theoretical approaches often require parameters that are not typically available at this stage. Consequently, there is a growing trend towards adopting more advanced data-driven or probabilistic approaches to achieve more accurate and precise predictions. In this context, this work presents a novel preliminary data set from previous regional projects. The data set covers a wide range of soil types, and it is compiled by ten different parameters in terms of soil properties, jet-grouting system’s parameters as well as unconfined compressive strength (UCS) and column’s diameter ([Formula: see text]). Based on such information, a probabilistic approach for joint predictions of local UCS and [Formula: see text] is proposed. The method is based on an already consolidated Bayesian formulation, which combines local and most-similar site information to derive local prediction of UCS and column’s diameter and assess local data set incompleteness. Results for two different sites demonstrate that the proposed method can well predict local UCS and [Formula: see text].