Probabilistic approach in tailings dam break studies aiming the definition of a consistent methodology

Abstract

Tailings dams are structures susceptible to failures and present extraordinary risks to society, infrastructure and environment. Due to such great risks associated to tailings dams and recent disasters (Mount Polley (2014), Mariana-MG (2015) and Brumadinho-MG (2019)), maintenance of these structures need to be inevitably intensified and numerical studies of hypothetical ruptures should be executed to estimate flooded area, damaged area and to determine emergency and mitigation plans. Moreover, no specific methodologies are normatized to these studies and physical characteristics of tailings introduce significant uncertainties in the study of propagation of such materials in the valley downstream. In the absence of explicit models to determine uncertain parameters, e.g. released volume of tailings, rupture hydrograph and rheology of tailings, the probabilistic approach is a viable option for dam break studies. Following this motivation, the main objective of this project is to propose a more assertive numerical methodology to assess flows produced by tailings dam break based on pre-rupture data, emphasizing uncertainty quantification and global sensitivity analysis of input parameters of the simulation through a probabilistic model. Considering the cases of Mariana-MG (2015) and Brumadinho (2019), HEC-RAS 6.2 code will be used as a deterministic model to simulate the samples (around 700 combination of input parameters per scenario) and produce output data, e.g. flooded area, flow depths, mean velocity and arrival time. From output data from HEC-RAS 6.2, a surrogate model will be constructed and validated through the data-driven Polynomial Chaos Expansion (PCE) using the UQLab package (MATLAB), enabling both global sensitivity analysis (Sobol's Indices method, which will ultimately quantify the contribution of each input parameter over final results) and drastic reduction of computational costs to obtain results. Finally, the results of this methodological path should provide subsidies with statistical basis to elaborate hypothetical dam break studies and the future thesis should contribute to stablish an assertive methodology to study tailing dam failures.