Bayesian constitutive model parameter calibration for strain softening soils

Conventional element tests are limited in their ability to characterise certain soil behaviour such as strain-softening, which is particularly relevant to certain geotechnical engineering sub-disciplines such as offshore geotechnical engineering. In this paper, an alternative approach to parameter calibration, incorporating full-field measurements, was expanded using Bayesian methods. A constitutive model’s strain softening parameters were calibrated using numerical biaxial compression test datasets, augmented with noise representative of real-world measurement noise that is anticipated from image-based measurement methods such as Particle Image Velocimetry (PIV) or Digital Image Correlation (DIC). Requirements regarding the quality and quantity of data are recognised with a view to ultimately performing experimental trials of the proposed calibration approach.