Comparison of simple stress-strain models in the moderate strain range for fine-grained soils: A review

Abstract

The prediction of stress-strain behaviour in soils is a problem that can be approached in different ways depending on the design scenario. In some cases, a multi-parameter constitutive model calibrated with non-routine soil tests may be appropriate, for example, where a model has been developed for the relevant soil at a building site subject to complex loading. However, simple characteristic parameters are desirable for examining the variability of soil behaviour especially at regional scales. This paper describes a method of assessing the suitability of simple models for simulating non-linear undrained soil stress-strain behaviour in the moderate strain range. The moderate strain range is defined by a soil strength mobilisation of 20% to 80%. Three simple stress-strain models are compared. A published database of reconstituted triaxial tests is used to evaluate the three models with selected statistical tools that quantify errors associated with the simple model approximation of the relationship between stress and strain. The paper discusses the value of computing the model error and the trade off to make between introducing a greater number of parameters (and tests) for model precision and limiting the complexity of the variability characterisation.

Keywords

• mobilisation strain framework
• variability characterisation
• model error
• design