Predicting soil settlement with quantified uncertainties by using ensemble Kalman filtering

Abstract

Soil settlement is one of the most common and critical issues in geologic and geotechnical engineering. Due to various sources of uncertainties, it is hard to predict soil settlement accurately. An inverse analysis using the information provided by field measurements is desirable for prediction with higher confidence. In this study, an inverse framework based on ensemble Kalman filtering (EnKF) is proposed to evaluate the soil settlement with quantified uncertainty. The theoretical and practical effectiveness of this scheme is demonstrated through synthetic and realistic tests to predict soil settlement of embankment roads. Inferred results including quantified uncertainties are obtained based on Bayesian theory, which makes a distinction between this method and conventional settlement prediction methods. The results of two synthetic tests show the parameters inferred by the EnKF converge to true values, which verify the satisfactory performance of the proposed scheme. A realistic application of Saga airport road is investigated, and the simulated settlement results are consistent with the field measurements. Moreover, Sobol method is adopted to study the sensitivity of model parameters, and detailed parameter studies are conducted to estimate the influence of ensemble size, the value range of prior distribution and observation error.