Análise de incerteza de parâmetros reológicos aplicada à modelagem hidrodinâmica de inundação em estudos de ruptura hipotética de barragens de rejeitos de lavagem de bauxita

Abstract

The hydrodynamic modeling of the flood wave propagation is an essential component of tailings dam break analysis, usually modeling the tailings flow with rheological models. Rheological models have rheological parameters as their input, and these parameters are specific to the mobilized material studied, heavily depending on its physicochemical properties. Mining tailings, however, may exhibit high heterogeneity, resulting in a significant variability of possible rheological parameters leading to uncertainties that propagate to the hydrodynamic modeling results. This research aims to analyze the influence of parametric uncertainties in the rheological characterization of bauxite washing tailings on the results of the flood wave propagation hydrodynamic modeling in dam break studies. The rheological behavior of bauxite washing tailings was evaluated using hierarchical Bayesian modeling applied to a database of rheometric characterization tests using a vane rotational rheometer. Through hierarchical Bayesian modeling, credibility intervals for the prediction of shear stress in function of the applied shear rate on the material were obtained. The rheological parameters vectors for the material were obtained in function of their non-exceedance probabilities, characterizing the tailings’ probabilistic rheological behavior. A strong dependency between the rheological parameters and the solids content of the material was observed, and it was found that the yield stress has an exponential dependency of the solids content. Heterocedastic behavior was observed for the rheometric data. Using the probabilistic characterization of the rheological parameters, the influence of these parameters on the hydrodynamic modeling results (maximum flow velocities and depths) and the propagation of uncertainty from rheological parameters to the hydrodynamic results were analyzed. The outputs of the hydrodynamic modeling are highly dependent on hydraulic controls imposed on the flow by the downstream valley topography, making it difficult to generalize observed patterns of correlation and uncertainty propagation between model input and output data. Aiming for greater generalization potential for the results, different geometric conditions for the downstream valley were simulated, through the creation of synthetic digital elevation models, using prismatic channels with varied geometry parameters. Using the synthetic digital elevation models, the correlations and propagation of uncertainties between input and output data of the hydrodynamic model were assessed, without conditioning the results to specific hydraulic controls of a given case study valley topography. The hydrodynamic simulations were conducted using two-dimensional modeling using the HEC-RAS software. It was found that the influence of rheological parameters is more pronounced in the variation of maximum depth results than in the variation of maximum velocity results. Furthermore, the variability of hydrodynamic modeling results, as a function of the variation of input rheological parameters, was found to be higher for lower longitudinal valley slopes and higher for more "open" valleys (with milder transversal slopes).