Automated reconstruction of soil stress-strain response from full field displacement measurements and loading data using optimisation

The significant amounts of image data that can be generated during physical model tests can provide a useful alternative and direct route to determining the stress-strain response characteristics of the soil used in the model without recourse to e.g. sampling and triaxial testing. Using optimisation, it is possible to find the stress-strain curve such that internal work calculated using the Particle Image Velocimetry (PIV) derived full-field displacement data is equal to external work. Whereas previously published work by the authors optimized for the unknown stress-strain curve by splitting it piecewise into many hundreds of segments, an alternate formulation presented in this work instead allows the description of the curve based on constitutive models, with only a handful of optimization variables necessary. FEA derived “perfect” data sets are reused to validate the new formulation. The eventual goal of the methodology is the robust application of the method to physical modelling test data, this paper presents a milestone towards this goal.