Método de retroanálise de parâmetros de resistência ao cisalhamento em taludes

Abstract

The back analysis of slopes that have failed is of paramount importance to identify the geotechnical parameters at the moment of sliding, implement corrective actions, and comprehend the conditions during failure to prevent similar accidents. However, the inherent difficulty in utilizing existing methods, many of which are based on trial and error, coupled with scattered and incongruent results, fosters research in this field. Physical models require extensive time and financial resources, while field and laboratory tests present significant uncertainties related to factors such as sample disturbance and mass heterogeneity. Consequently, the core of this study aimed to develop and validate a method for back analyzing shear strength parameters for slopes, homogeneous and heterogeneous, saturated and unsaturated, with circular and non-circular sliding surfaces, two-dimensional and threedimensional. In this method, an objective function named H was defined, composed of the normalized distance between the actual failure surface and a critical sliding surface (dn), as well as the variation of the safety factor (G). Optimization techniques, including Cuckoo Search and Entry and Exit critical surface methods, along with global and local optimization methods (Nelder-Mead), were employed. The effectiveness of the back analysis was demonstrated in efficiently obtaining optimal values; however, caution in the approach and validation with laboratory and field data are necessary. It was noted that the choice of optimization method along with the critical surface location method is crucial for obtaining the desired optimal parameters. For circular surfaces of saturated homogeneous slopes, global optimization methods are dispensable when using the Entry and Exit method, as methods like Nelder-Mead require fewer iterations and efficiently achieve optimal parameters. The optimization of the objective function H for saturated homogeneous slopes with non-circular surfaces required multiple iterations due to the presence of multiple local minima, accentuating the difficulty and complexity of non-circular surfaces. For unsaturated homogeneous slopes, it is recommended to employ global optimization methods, and the Entry and Exit critical surface search method yielded more satisfactory results compared to Cuckoo Search. In two-layer saturated slopes, successful minimization occurs as the most influential parameters approach optimal values. In three-dimensional slopes, optimization results showed a reduction in cohesion values and an increase in friction angles compared to the 2D benchmark.in cohesion values and an increase in friction angles compared to the 2D benchmark.