Probabilistic characterisation of critical state model parameters for saprolitic soil in Hong Kong

Realistic numerical representation of soil behaviour requires both constitutive models that are theoretically sound and a good understanding of the associated parameters for the soil to be modelled. This paper explores the multivariate probability distributions of critical state model parameters of a decomposed granite soil type in Hong Kong, including the correlations among these parameters and the engineering implications of such material variability. The parameters were determined through calibrating the NorSand model against a database of stress path measurements from 97 multi-staged consolidated–undrained triaxial tests of specimens retrieved from a region in Hong Kong. Based on the test data, the vine copula method is utilised to unveil the non-linear dependence structure of various constitutive model parameters. Apart from the probability distributions of model parameters, the approach also reveals the model bias associated with the NorSand model in representing the behaviour of decomposed granite soils, as an indicator of its accuracy based on this large dataset. The value of probabilistic parameter characterisation is demonstrated through settlement analyses of a hypothetical case of a shallow footing founded on this soil type, whereby the probability of failure can be rationally determined with consideration of material variability in dilatancy features and hence the stress–strain response.