Loess soils cover approximately 10% of the world's landmass. This quaternary deposit has been studied for many decades, but predicting the collapse behaviour still remains an active subject for discussion and research. For this project, collapse mechanisms have been studied using two techniques; artificial loess, (the physical model) and a numerical model. The physical model is a continuation of artificial soil techniques developed in the literature. The artificial soil is made from silt and clay particles and has been developed to reproduce the collapse characteristics found in natural loess. This artificial loess has been tested using oedometer and triaxial equipment and will be used as the foundation material in a scale model of a footing. Results show that the physical model can reproduce collapse behaviour very well. Secondly, a numerical model has been developed to simulate the collapse behaviour observed in the laboratory tests. This model has been implemented into the finite element program CRISP. Oedometer tests have been accurately simulated and the extension of both of these models to model the behaviour of a footing in plane strain will be discussed.
The abundant nutrients and low density of loess make it an ideal agricultural soil. These bonding systems include clay bonding, carbonate cementation and suction (Barden et al., 1973). These components interact in a complex way and therefore modelling the behaviour of these soils is not easy. The relative roles of the bonding components are not clearly understood and models that are based solely on suctions do not account for the various cementing components in natural collapse systems such as loess. Neglect of these various elements has in the past caused considerable problems to the built environment and other associated works, such as the Teton dam collapse in 1976, in Idaho, USA (Smalley, 1992).
