Computer Aided Slope Stability Analysis Using Optimization And Parallel Computing Techniques

Slope stability analysis is commonly performed using limit equilibrium methods (LEM). In LEM, factor of safety (FS) is calculated for different trial slip surfaces and the one with the minimum FS is reported as the critical slip surface. Since locating the critical slip surface is believed to be an NP-hard (non-deterministic polynomialtime) problem, heuristic global optimization techniques are employed. Although these techniques have usually produced good results, “No Free Lunch” (NFL) theorems seem to have made the problem of locating the critical slip surface an endless research. According to the NFL theorems, no heuristic optimization technique can perform well for all problems. On the other hand, there may exist other slip surfaces that are as important as the critical slip surface in practical analyses. A slip surface is important, if it is located far away from the critical slip surface, but gives FS close to the minimum FS or the consequences of failure along the slip surface is serious. Therefore, there is a need to constantly implement and test different optimization techniques. However, implementation of optimization techniques in slope stability analysis is often not straightforward because it requires internal links to LEM. Firstly, the present study resolves this issue by developing a decoupled algorithm that allows for easy implementation of optimization techniques. Then, to demonstrate the simplicity of this algorithm and to promote the latest research on slope stability, three state-of-the-art optimization techniques are implemented, and their effectiveness and efficiency in detecting single/multiple global and local minima is investigated on a series of test problems