基于SMOTE-XGBoost算法的混凝土强度预测  

Prediction of strength of concrete based on SMOTE-XGBoost algorithm

作　　者：薛飞 XUE Fei(No.1 Engineering Co.,Ltd.,Cccc First Highway Engineering,Beijing 102205,China)

机构地区：[1]中交一公局第一工程有限公司,北京102205

出　　处：《混凝土与水泥制品》2024年第8期32-36,共5页China Concrete and Cement Products

摘　　要：基于244组混凝土配合比构建了数据库,采用SMOTE-XGBoost算法对混凝土28 d抗压强度进行了预测。首先通过SMOTE算法对划分的训练集进行平衡处理;然后对比了SMOTE算法平衡前后XGBoost与常用混凝土强度预测模型的评估结果;最后进行了SMOTE-XGBoost算法的实际工程验证。结果表明:SMOTE-XGBoost算法有效解决了数据不平衡问题,提高了预测模型的精度;相较于其他机器学习模型,SMOTE-XGBoost算法的预测结果较好;应用SMOTE-XGBoost算法对无岳高速WYTJ-07标段工程自制花岗岩混凝土的28 d抗压强度进行了预测,预测结果误差较小,该算法在工程混凝土强度预测方面具有广泛的应用前景。A database was constructed based on 244 sets of concrete mix proportions,and the SMOTE-XGBoost algorithm was used to predict the compressive strength of concrete at 28 d.Firstly,the divided training set was balanced using the SMOTE algorithm.Then,the evaluation results of XGBoost and commonly used concrete strength prediction models before and after balancing using the SMOTE algorithm were compared.Finally,practical engineering validation of the SMOTE-XGBoost algorithm was conducted.The results show that the SMOTE-XGBoost algorithm effectively solves the problem of data imbalance and improves the accuracy of the prediction model.Compared to other machine learning models,the SMOTE-XGBoost algorithm has better prediction results.The SMOTE-XGBoost algorithm is applied to predict the 28 d compressive strength of homemade granite concrete in the WYTJ-07 section of the Wuyue Expressway project,and the prediction results show a small error,and this algorithm has broad application prospects in predicting the strength of engineering concrete.

关 键 词：混凝土28 d抗压强度 机器学习 SMOTE-XGBoost算法 预测 

分 类 号：TU528.53[建筑科学—建筑技术科学]