Stability analysis and prediction of Bimslope failures using numerical modelling and hybrid meta-heuristic models

Abstract Bimsoil slope (Bimslope) or Soil-Rock-Mixture slopes are complex geological formations made up of geotechnically important ‘blocks’ enclosed in a finer-textured ‘matrix.’ These geomaterials possess heterogeneous properties owing to changes in the block and matrix properties that can arise from natural weathering processes or various anthropogenic activities. The present study focuses on the stability analysis of a Bimslope under varying block properties (orientation and volumetric block proportion), matrix type (Loamy sand, Sandy loam, and Silt loam) and matrix water content. The study indicates that the failure of a Bimslope is not governed by only one physical parameter but rather by a critical combination of block properties, matrix type and matrix water content. When the water content (< 25% saturation) and block content (< 25%) are less, the failure is governed by matrix only. With the increase in block content, the failure is governed by block however under fully saturated (100% saturation) conditions, the failure is solely governed by matrix type irrespective of block properties. In addition to that, the data generated using numerical analysis was used to predict the factor of safety using hybridized ensembled methods considering the XGB model and three other novel meta-heuristic algorithms (MHA): political optimization (POA), the Leopard Seal algorithm (LSA), and the Giant Armadillo Algorithm (GAA). It has been found that among all the models, the GAA-XGB algorithm exhibits better results with R 2 and RMSE values as 0.99628, 0.055436 in training and 0.985664, 0.063092 in the testing stage. Furthermore, SHAP-based analysis has been performed to identify the most influential parameters.