Metodologia de avaliação de liquefação em barragens de rejeito: uma abordagem probabilística

Abstract

Accidents involving tailings dams failures may have catastrophic consequences. The understanding of the behaviour of large infrastructure projects, which expose society and the environment to risk, must be the duty of all technical and political agents, and, most importantly, of all specialists and economic agents. One of the possible causes of those failures is liquefaction. This failure mode occurs in undrained conditions, in which the soil loses strength and stiffness due to sudden increase in pore pressures. Although several studies regarding dam's liquefaction are being developed, it can be seen a need to develop field studies of reliability. Engineering's good practice and qualitative approaches are common in engineering. However, these methods do not consider the whole failure process and its uncertainties. Considering the importance of the theme for engineering, society and the environment, this work aims to develop an integrated methodology to determine the probability of failure by static liquefaction of tailings dams built using the hydraulic earthfill method. Particularly the ones heightened by the upstream method, which are more susceptible to develop this failure mode. The methodology is based on an event tree analysis (ETA), being the estimation of failure probability of each node of the ETA obtained by failure tree and Monte Carlo methods. Additionally, it was applied the uncertainty estimation for each node and calculated the final probability of the ETA by Monte Carlo method and Integrated Boolean Operators. The choice of the event tree method is justified considering the representativeness of the events chain corresponding to the liquefaction phenomenon. The proposed ETA contains 6 main nodes: Susceptibility; Water Level Management; Trigger; Safety Management System; Beginning and Progress of Liquefaction. The developed methodology presented to be appropriate, replicable and feasible regarding its approach, despite of the limited scenarios studied in this work. The procedure of estimating probable values for an event chain as well as the imprecision of these values is justified by the complexity of the failure mode and the many uncertainties involved. This work aimed to bring greater clarity on probability and liquefaction issues, proposing a reasonably practical and efficient procedure.