Effects of spatial variability on Bayesian model updating using measured excavation responses

1. Home
2. Publications
3. Online Library
4. Effects of spatial variability on Bayesian model updating using measured excavation responses

In view of the challenges in the characterization of soil properties, particularly the spatial variability, through conventional site investigation activities, the measured responses of geosystems have been increasingly used to back-calculate soil properties. However, the presence of spatial variability, which is commonly considered as one of the most critical uncertainties, may add to the challenges in the back-analysis of geosystems using field-response measurements. In this paper, we used a synthetic excavation example to examine the effects of spatial variability on the Bayesian model updating of soil properties using measured excavation responses. Specifically, we adopted the random-field finite-element model to generate synthetic wall deflection measurements. In this way, the spatial variability is manifest in the generated measurements. Bayesian model updating was then carried out in conjunction with the deterministic analysis following the conventional back-analysis procedures adopted by engineering practitioners. We showed that the omission of spatial variability leads to (i) a bias in the posterior distributions of soil properties, and (ii) underestimations in the variations of the predicted wall deflections. The results highlight the critical need of an advanced model updating framework that considers spatially variable soil properties.

10th European Conference on Numerical Methods in Geotechnical Engineering (NUMGE2023)

5. Probabilistic and inverse analysis