Geomaterials such as clays and sands are highly uncertain in both their mechanical properties and spatial variability within a dike or embankment (see Fig. 1). Dike failure poses a great risk to areas around rivers and coasts. Therefore, it is essential to identify the soil strength parameters by assimilating available data from laboratory and field conditions into mathematical or numerical models. Extensive laboratory tests have been conducted to constrain the uncertainty of material (mechanical) parameters. Nevertheless, these laboratory conditions (e.g. triaxial) only represent a handful of the realistic in-situ conditions and identifying soil strength from field data is challenging because the data are only available at discrete locations, often within the shallow subsurface.
Fig. 1 Spatial heterogeneity of the soil within a dike cross section
Assignment
The impact of material and geospatial uncertainty on dike safety is significant not only because it is extremely difficult to fully characterize the subsurface but also due to the inherent heterogeneity within the dike. In this project, you will use advanced Bayesian inference and machine learning techniques available in the software package “GrainLearning” to identify critical parameters for the soil/dike models at the levels of a single material point and/or a geo-structure (dike). The results obtained from this integrated approach will provide indication on the optimal strategies to place and install sensors (e.g., displacement) such that the identification of soil strength parameters is achieved with a reduced uncertainty.
Learning goals:
- Inference of critical state soil parameters from an extensive laboratory database
- Investigating how various sampling/optimization methods affect the inference efficiency
- Assessing the identifiability of soil strength parameters and their spatial variability using in-situ data
- Investigating how the inference quality depends on the locations and number of sensor points
Supervision
Are you interested in this assignment? Contact the Master thesis coordinator:
