基于数据驱动的岩石微型桩水平极限承载力预测OA

To improve the prediction accuracy of the horizontal ultimate bearing capacity of rock mi-cro-piles in mountainous areas,a comprehensive research method combining numerical simulation and multi-ple data-driven prediction models is proposed.Based on the orthogonal experimental design,the three-dimen-sional finite element model with variables such as pile length,pile radius,and overburden soil thickness are es-tablished,and a dataset of bearing capacity of micro-piles under different service conditions is generated.Us-ing the BP neural network,support vector machine(SVM),extreme gradient boosting(XGBoost),and random forest(RF),the prediction models for bearing capacity were constructed,and the model performance is com-pared and analyzed by the evaluation metrics including the coefficient of determination(R2),mean absolute er-ror(MAE),and root mean square error(RMSE).The results show that the random forest model achieves the best prediction accuracy and generalization ability,with R2 reaching 0.999 5,MAE of 1.2 kN,and RMSE of ap-proximately 0.002 6 kN.Although the BP neural network,SVM,and XGBoost models have slightly lower overall accuracy,they can effectively capture the overall trend of bearing capacity variation and can serve as auxiliary tools for preliminary prediction of the horizontal ultimate bearing capacity of rock micro-piles.