Abstract:

 To address the challenge of determining parameters for the hardening soil (HS) model in engineering practice, a database named HS-CLAY/9/196, which includes HS parameters, is established. A multivariate probability distribution for HS parameters is constructed based on the database. The probability distributions of HS parameters are updated using the available measured data of common soil parameters. The effects of the types and quantities of measured soil parameters on the estimation of HS parameters are studied. In addition, the probabilistic transformation models of HS model parameters (i.e., oedometric tangent stiffness  E^ref_oed, triaxial secant stiffness E^ref₅₀, and unloading/reloading stiffness E^ref_ur) are proposed based on the given measured data. The results show that the established multivariate probability distribution model effectively characterizes the statistical characteristics and cross- correlations of HS parameters. Based on the constructed multivariate probability distribution model, various measured data can be integrated to enhance the accuracy of parameter predictions through Bayesian updating. The prediction uncertainty can be reduced as the variety of measured data increases. To achieve accurate predictions forE^ref_oed,E^ref₅₀ , and E^ref_ur with low uncertainty, priority should be given to collecting the measured data of soil parameters that are strongly cross-correlated with the target HS parameters, such as the compressibility modulus E_s1-2, water content w, and void ratio e.

Key words: hardening soil model, soil parameters, database, multivariate probability distribution, Bayesian updating