A data-driven approach for predicting the time-dependent settlement of embankments on soft soils

Abstract

Models of soil behaviour, in situ and laboratory testing techniques and system models for describing the geometry and time-dependent processes involved in the settlement of embankments on soft soils have reached a level of maturity where reliable predictions for this problem should be readily achievable. Despite this, there is clear evidence to demonstrate engineering predictions of embankment behaviour can be inaccurate, resulting in construction cost and schedule overruns. One key reason for this is the significant influence that personal judgement plays in processing soil test data to derive engineering parameters. In this paper, a method that makes progress toward overcoming this is developed using oedometer test results directly in a data-driven approach, thus avoiding the need to evaluate profiles of compression indices and yield pressures. The method has been deployed as an easy-to-access cloud-based application and its use demonstrated with a comprehensive case history. Data used in this paper are freely available in electronic format.