Predicting Settlements Development Under Road Embankment Using Probabilistic Models

The road infrastructure project E6 Kvithammar-Åsen in the north of Trondheim, Norway, contains several large road embankments on sensitive clay. To improve the properties of such clays to sustain such embankments, lime-cement columns could have been installed, however, this would have led to high CO2 emissions. Therefore, preloading and vertical drains have been used as alternative soil improvement measures to reduce the extent of traditional lime-cement columns. The soil improved areas have been monitored by deformation sensors and piezometers during the 1.5 years since construction of the embankments. During this time, continuous assessments of the settlements have been performed showing that physics-based deterministic models underestimated the final total settlements. In this paper, the settlement measurements have been combined with available site data to gain new insights on how vertical drains affect the sensitive clays. Regression and probabilistic models have been applied on the time series data from the settlement sensors to see whether probabilistic methods can give a good forecast on how large the final settlements will be and when an acceptable rate of settlements will be reached.

Keywords: Vertical drains, Embankments, Sensitive clay, Time series forecasting, Settlements, Sensor data.