Assessment of Slope Stability under Bounded Uncertainties Using Differential Evolution and Particle Swarm Optimization

In the assessment of slope stability, the uncertainty of soil properties affects directly the factor of safety, which is the objective function to be minimized. Therefore, we propose an approach that treats the uncertainties as additional variables besides three common variables, i.e. the radius of the critical slip surface R and its centre coordinates X, Y. This approach is also compared with two other approaches: i) minimizing the objective function (with variables R, X and Y) at all vertices of the feasible region defined by lower and upper bounds of uncertainties, and then choosing the best solution from them; ii) using the two-level optimization in which the optimal solution (R, X and Y) satisfies the least favourable condition of uncertainties at these vertices. The proposed approach with extra variables shows the efficiency in finding good solutions with a significant decrease of computation time. The study also shows that both Differential Evolution and Particle Swarm Optimization can find good solutions, although the former is marginally better.