Updating Assessed Damage Probabilities for Deep Excavations in Soft Soils

This paper presents the results of Bayesian updating of assessed damage probabilities for buildings adjacent to a deep excavation in soft soils. The prediction of building behavior respect to a serviceability limit state regards a damage potential index DPI, which depends on lateral movement and ground settlement induced in buildings as the excavation proceeds. Numerical analyses were performed on synthetic tridimensional deep excavation finite element models, considering constitutive model Hardening Soil Small Strain, and simulating two constitutive parameters as random fields. Models response regarding DPI was used to have an initial assessment of building damage probabilities for each neighboring building and construction stage. Bayesian updating is used to enhance initial predictions integrating DPI-based damage probabilities from numerical models and DPI values from ground movements obtained using semiempirical models KJHH and KSJH. Comparison of prior and updated results show the improvement in performance predictions and a reduction in parameter uncertainty.