Stress release coefficient prediction of sandy-gravel soil by extra tree algorithms

One of the most significant geotechnical issues that must be addressed during the dams' first impoundment phase is the collapse settling of embankment dams. Due to the fact that it leads dams to undergo massive deformations and serious damage, it is a significant phenomenon that must to be investigated at the various stages of design. Moreover, the presence of a number of elements that play a role in this situation makes it hard if not impossible to come up with a precise prediction of a settlement's collapse potential. When it comes to sandy-gravel soil, it is very likely that other approaches like machine learning (ML) or laboratory techniques will be considered to determine the Collapse settlement factor (CSF). CSF represents a ratio obtained by the collapse shear stress to the specimen's dry shear stress before to the impounding stage at the conclusion of the impounding stage. In this work, extensive data gathered from studies was applied to predict the CSF with the extra tree regression (ETR). One of the main elements affecting the model's dependability is the ETR hyperparameters that must be chosen with the aid of metaheuristic optimization techniques. The two techniques employed are Sand Cat Swarm Optimization (SCSO) and Horned Lizard Optimization (HLO). The four input factors for the CSF are calculated by a mathematical database consisting of 180 test results—75 percent for training and 25 percent for testing—derived from the literature. This study introduces reliable methods for estimating CSF in sandy-gravel soils. The HLO-enhanced model (HE) being the most effective model, is not only a great asset in assessing the deformation risks of embankment dams during the design and initial filling stages but also contributes to the overall safety and efficiency of dam construction and maintenance. • ETR model predicts Stress Release Coefficient in sandy-gravel soils. • Uses 180-test dataset with 75/25 train–test split for validation. • SCSO and HLO optimize ETR hyperparameters for accuracy. • HLO-enhanced model outperforms in reliability and prediction strength. • Aids dam design by assessing deformation and collapse risk effectively.