Efficient and flexible Bayesian updating of embankment settlement on soft soils based on different monitoring datasets

In situ monitoring provides valuable information to update the predictions of the embankment settlement on soft soils. Observational data obtained at different monitoring moments can be incorporated into the updating under a Bayesian framework for rational treatment of various geotechnical-related uncertainties. Nevertheless, as complicated models (e.g., finite element model) are involved to link monitoring data with uncertain parameters, Bayesian updating can be computationally demanding. This becomes more challenging if multiple different datasets (e.g., datasets obtained at different monitoring moments) are concerned, for each of which a Bayesian updating run shall be performed. This paper develops a novel Bayesian framework that allows efficient and flexible updating of the embankment settlements based on different datasets. It contains two major components: (1) driving Bayesian analysis to generate problem-specific information on the embankment settlement based on the observational data obtained at early monitoring moments, and (2) target Bayesian analysis to update the embankment settlement given new datasets obtained at latter monitoring moments by making use of the information generated in the first step, which requires negligible computational efforts. The proposed approach not only improves significantly the computational efficiency for updating the embankment settlements given different monitoring datasets, but also provides the flexibility to shed lights on effects of different updating strategies (or datasets) on the updated embankment settlements. An embankment settlement monitoring example is adopted to demonstrate the efficiency and flexibility of the proposed approach.