Embankment prediction using testing data and monitored behaviour: A Bayesian updating approach

Settlement prediction is critical for soft ground projects. Traditional predictions using laboratory and field test data, however, can suffer from a lack of accuracy, which results in a lack of confidence by the designer. This paper employs the Bayesian approach with laboratory data, field test data, and monitoring data to yield accurate predictions during the construction and consolidation periods for the test embankment built at Ballina, New South Wales, Australia. We show that surface settlement can be well predicted using 116 days of observed settlements, while the pore pressure can be predicted using 292 days of pore pressure measurements. The predictions are shown to converge to the field measurements, regardless of some assumptions about the measurement errors. Finally, it is demonstrated that incorporating more monitoring data into the Bayesian updating process enables more accurate predictions. (C) 2017 Elsevier Ltd. All rights reserved.