Bayesian Updating of Bearing Capacity Models for Individual Stone Columns

Abstract

Stone columns are often employed to increase the bearing capacity of soft clay. Many models—for instance based on plasticity theory or on statistical analysis of empirical data—have been proposed to estimate their bearing capacity. For a single stone column, the empirical expression qult=NpSu is commonly used due to its simplicity and predictive performance. However, a wide range of Np values have been reported in the literature, resulting in widely variable predictions; it is therefore difficult to select good estimates of Np in practice. In this study, such empirical models are comprehensively analyzed in the framework of Bayesian model assessment, which in addition to model parameter estimates, can provide uncertainty estimates of parameters and predictions. In agreement with previous research, a general model with Np=20, which can be employed for a typical soil and construction method, is obtained for a single stone column. Then, the authors illustrate that such general model can be updated to incorporate new project-specific information as it becomes available, reducing the model’s uncertainty and improving its predictive capability. Finally, the authors show how the Bayesian analysis can be incorporated into decision making under uncertainty through its application in risk analyses.