Reliability sensitivity analysis of geotechnical monitoring variables using Bayesian updating

Determining the sensitivity of monitoring variables is essential to field monitoring design for effectively monitoring the safety and reliability levels of geotechnical structures in uncertain environment. Reliability sensitivity analysis of monitoring variables provides a rational approach for identifying sensitive monitoring variables and is capable of accounting for geotechnical uncertainties. It, however, can be computationally expensive, especially when sophisticated numerical models (e.g., finite difference model, FDM) are involved and repeated simulation runs are required. This paper proposes a reliability sensitivity analysis method that leverages on the robustness of direct Monte Carlo simulation (MCS) and the Bayesian Updating with Structural Reliability Methods. The proposed approach allows performing the reliability sensitivity analysis of a monitoring variable by a single run of direct MCS, avoiding repeated simulation runs for different possible observational values of a given monitoring variable. Illustrative examples demonstrate the capability of the proposed approach in identifying the most sensitive monitoring variables among candidates. It is possible to achieve a significant reduction in the number of evaluations of numerical models for reliability sensitivity analysis of monitoring variables using the proposed approach.