A novel multisource-intelligent calibration method for discrete element model parameters and application in macro-mesoscopic strength analysis of slope soil

The mesoscopic parameters of the discrete element method (DEM) are crucial for accurately simulating the strength characteristics of slope soil. Existing DEM parameter calibration methods suffer from insufficient multisource data mining and low levels of intelligence, severely limiting the accuracy and reliability of calibration results. Hence, this study proposes a multisource-intelligent calibration method for DEM parameters. First, the key DEM parameters are determined using multidimensional correlation analysis methods, and a multisource data set is constructed based on these key DEM parameters, test parameters, and inherent soil properties. Then, the Grid Search Cross Validation (GridSearchCV) and eXtreme Gradient Boosting (XGBoost) are used to mine information from multisource data, combined with multiple interpretability analysis algorithms to enhance the model's transparency, developing an interpretable soil strength prediction model. Finally, coupling the genetic algorithm (GA) with the prediction model enables dynamic optimization and intelligent calibration of key DEM parameters. The calibration results are then comprehensively validated and applied to analyze the macro-mesoscopic strength characteristics of slope soil. The results indicate that the GSCV-XGBoost model can effectively integrate multisource data. Its predictions not only meet physical consistency principles but also show excellent prediction accuracy. The proposed intelligent calibration method enhances the accuracy of DEM parameter calibration and has been validated as reliable, which also shows favorable application effects in a specific slope engineering. • Proposing a novel multisource-intelligent calibration method for DEM parameters. • The GSCV-XGBoost model can fuse multisource data and achieve high prediction accuracy. • Interpretability analysis algorithms follow consistent feature importance rankings. • The calibration method provides optimal accuracy and enables effective application.