Probabilistic Analysis of Slope Using Bishop Method of Slices with the Help of Subset Simulation Subsequently Aided with Hybrid Machine Learning Paradigm

The use of probabilistic analysis of slopes as a useful technique for determining the level of uncertainty in various variables has grown. In this study, a probabilistic analysis of a representative embankment with a height of 12 m under seismic conditions was carried out utilizing the UPSS ADD-INs 3.0 and Subset Simulation methodologies. The seismic coefficients kh of 0.12, 0.14, and 0.18 were all taken into consideration. In order to account for uncertainty in soil properties including cohesiveness, angle of internal friction, and unit weight of soil, the study used lognormal random fields and Cholesky matrices. The use of Subset Simulation enabled the fast calculation of the reliability index and the probability of failure, providing significant insights into the embankment's failure risk. In addition, a hybrid computational technique was used to optimize the worst-case scenario for failure probability. To address a gap in the literature, this work focused on developing a probabilistic analysis using subset simulation and a hybrid Artificial Neural Network-Teaching Learning-Based Optimization model. The performance of this model was evaluated and compared to existing hybrid models built using seven different swarm intelligence methods. During the validation phase, it was observed that the proposed Artificial Neural Network-Teaching Learning-Based Optimization model outperformed other hybrid models, exhibiting a high determination coefficient value of 0.9974 and a low root mean square error value of 0.0226. This superiority can be attributed to the Teaching Learning-Based Optimization component, which emphasizes global search by incorporating a teaching phase to improve less optimal solutions.