Abstract: The assessment of model parameters is crucial in developing constitutive models. However, the results of parameter assessment for these models are inevitably subject to errors. Hence, a Bayesian framework utilizing structural reliability, subset simulation, and adaptive conditional sampling methods is employed to assess the uncertainty of constitutive model parameters through test data inversion. Using the joint shear constitutive model for shear stress-shear strain characterization of rock-soil mass and the hypoplastic clay constitutive model for nonlinear soil stress-strain increment description as case studies, this study investigates the uncertainties in model parameters, shear stress-strain curves, and pore ratio-pressure curves. Furthermore, it analyzes the sensitivities of model parameters to test outcomes. The research demonstrates that this approach evaluates parameter uncertainty driven by test data. The joint shear constitutive model parameters are primarily influenced by  in stress-strain curve representation, while most parameters of the hypoplastic model have a notable impact. These results enhance comprehension and enhance the predictive reliability of these two constitutive models.

Key words: shear stress constitutive model, hypoplastic model, Bayesian framework, uncertainty quantification