The geotechnical characterization of tailings dams presents significant challenges due to the spatial heterogeneity of the materials and the limited amount of available test data. Bayesian statistical methods provide a robust framework to represent and update uncertainty in the definition of geotechnical parameters and hydrogeological conditions by integrating prior information with new experimental data.
This work presents a practical application of Bayesian inference for estimating the undrained shear strength of tailings in an operating dam. Results from cone penetration tests and laboratory tests were combined, incorporating empirical correlations and expert judgment within a unified probabilistic model.
Through the Bayesian updating process, predictive posterior distributions were obtained that captured both material variability and measurement uncertainty, allowing strength estimates to be refined based on the available evidence while reducing reliance on subjective criteria. As a result, a characteristic strength was derived in accordance with the definition proposed in the second generation of Eurocode 7.
This procedure proved to be an effective tool to improve the design and stability assessment process for tailings dams within a reliability-based framework.
