To study the mapping relationship between the eccentric compression bearing capacity of recycled aggregate concrete filled steel tubular slender columns and each design variable,and to establish a high-precision prediction model,and to analyze the interpretability of the machine learning model,a database containing 155 samples was established,and the VIF test was performed on the selected input variables of the machine learning model.The RBFNN,RF,LightGBM and XGBoost algorithms were used to establish the eccentric compression bearing capacity prediction model,and the most suitable model was selected based on the coefficient of determination(R2),the root-mean-square error(RMSE),and the mean absolute error(MAE).Using the SHAP method to explain the contribution of input variables to the output bearing capacity and the mechanisms of each input variable,the VIF coefficient between input variables was found to be less than 5,indicating no obvious multicollinearity between variables.Through comparative analysis,the XGBoost algorithm demonstrated superior performance with an R2 of 0.998,RMSE of 16.397,and MAE of 7.76,enabling high-precision prediction of eccentric compression bearing capacity.The eccentricity and diameter of the specimen were identified as key variables affecting capacity.With the increase in recycled aggregate substitution rate,the eccentric compression bearing capacity of circular steel tubular recycled aggregate concrete slender columns decrease.
维普
万方数据