Using Bayesian Updating to Improve Forecasts of Embankment Settlements on Soft Soils

In this paper, Bayesian updating is combined with a data-driven approach known as CRACA to provide settlement forecasts for embankments on soft soils. Oedometer test data are ingested directly in the CRACA method and processed automatically to ensure minimal subjective intervention. The uncertainties involved in the laboratory measurements and the automated interpretation process are represented using scaling factors with an initial value of unity (returning the prior). These scaling factors are treated as random variables in Bayesian updating using early-stage settlement monitoring data to improve future forecasts. This method is applied to an embankment constructed at the Bloemendalerpolder site, Netherlands, and is shown to provide a significant improvement in the accuracy of the four-year surface settlement forecast using settlement monitoring data over the first year, reducing the different between model and measured settlement from 8 % (prior) to 2%. Using 600 days of data, the forecast settlement matches the 4-year measured settlement almost precisely.

Keywords: Data-driven, Bayesian updating, Embankment, Settlement, Consolidation, Creep.