Reliability Analysis of Settlement Using an Updated Probabilistic Unified Soil Compression Model

Settlement of a structure is a matter of great concern. Both excessive and differential settlement can cause expensive damage to buildings and must be avoided. Most methods used to estimate settlement are both deterministic in nature and are based on elastic analysis of soils. To better estimate settlement, a probabilistic estimate that uses a more in depth analysis of the behavior of soil is required. This thesis develops a new probabilistic model for estimating settlement based on a probabilistic unified soil compression model. The model is then used to estimate the settlement of an embankment. Lastly, a reliability analysis of settlement is carried out on the settlement estimate of the embankment.

The new probabilistic unified soil compression model used in this thesis was developed based on a previously developed probabilistic unified soil compression model, accounting for further uncertainties into the model and correcting for errors in the model calibration. This model was calibrated using data from a site on the Venice Lagoon using a Bayesian approach. The model to estimate settlement was developed based on this probabilistic soil compression model and is unbiased in nature. Using this model, unbiased settlement estimates were obtained for an embankment also located in the Venice Lagoon.

Using the developed probabilistic model for settlement, reliability analysis was carried out. This reliability analysis involved assessing the conditional probability that, for a specific load and given soil properties, a specified settlement threshold would be reached or passed. Sensitivity and importance analysis were carried out, determining which parameters and random variables have the largest impact on the fragility estimates. Lastly, a closed-formed approximation based on the Central Limit Theorem was developed to allow for easier fragility estimation.