Behaviour of frozen soils from micro to macro: new insights, database and numerical techniques – REFROZEN

In this project we will investigate the behaviour of frozen soils, both natural and artificially frozen, from the micro-scale at particle level to macro-scale in case studies.
The project will use the latest MRI and µCT imaging technology to provide a comprehensive dataset of frozen soils under different temperature and loading regimes at particle level. The results will be compared to element tests carried out in a freezing chamber.
The new database of observed behaviour will be used to develop and validate novel constitutive relationships including Fast-Fourier-Transform homogenisation and Machine Learning. These will be used in boundary value problems which will be solved in finite element models covering multi- scale and poromechanical approaches, as well as in physics-informed neural networks (PINNs). The final step towards macro-scale will be taken using these newly developed techniques to calibrate them against real case studies at two levels, one in a controlled physical model in the laboratory and two real case studies: a slope failure in Mount Fuji, Japan, and the artificially frozen ground campaign carried out in Berlin during the construction of the underground works at the famous Unter den Linden street.
The project will provide a unique and novel database and insights from the observed behaviour with an unprecedented level of detail as well as novel numerical approaches to frozen soils, never attempted to date.