刘聪林,李盛,崔晓宁,蔡磊,张建功

• 1:兰州交通大学土木工程学院
• 2:兰州交通大学道桥工程灾害防治技术国家地方联合工程实验室
• 3:中国铁路兰州局集团有限公司工务部
• 4:中国十七冶集团有限公司

摘要(Abstract)：

【目的】为精准预测混凝土抗压强度、突出XGBoost模型的预测优势和实现XGBoost模型的可解释功能,【方法】构建以水泥、龄期、水等8种影响因素为输入特征和以抗压强度为目标特征的1 030个样本数据集,建立支持向量回归(SVR)、随机森林(RF)和极端梯度提升树(XGBoost)机器学习算法模型,开展混凝土抗压强度预测研究,将XGBoost模型和ACI209公式的预测结果对比,同时,引入SHAP模型对XGBoost模型进行解释和分析。【结果】结果表明:XGBoost模型的预测精度最高,R~2为0.952, MAE为2.48, MAPE为9.16, RMSE为3.58,但XGBoost模型对小于30 MPa的低抗压强度样本的预测误差较大,随着抗压强度的增大,XGBoost模型的预测精度提高,超限样本比例从25%下降到2.7%;与ACI209公式的预测结果相比,XGBoost模型在龄期56 d和100 d样本的预测值绝对误差率均值为4.10%,3.64%,而ACI209公式则为11.27%,17.96%。【结论】XGBoost模型适用于混凝土强度大于30 MPa的样本的预测;SHAP模型不仅可以定量地给出特征重要性排序,还能定性地给出每个特征参数是对抗压强度的影响规律,能为混凝土相关研究及其他需要对机器学习模型进行解释的研究提供参考。

关键词(KeyWords)： 机器学习;XGBoost;SHAP;抗压强度预测;混凝土;力学性能

Abstract:

Keywords:

基金项目(Foundation): 宁夏回族自治区重点研发计划项目(2022BEG02056)

作者(Author): 刘聪林,李盛,崔晓宁,蔡磊,张建功

DOI: 10.13928/j.cnki.wrahe.2025.02.020

参考文献(References)：

• [1]罗广彬,洪成雨,程志良,等.基于BP和GA-BP神经网络的混凝土抗压强度预测研究[J].混凝土, 2023(3):37-41.LUO Guangbin, HONG Chengyu, CHEN Zhiliang, et al. Study on prediction of concrete compressive strength based on BP and GA-BP neural network[J]. Concrete, 2023(3):37-41.
• [2] ENGUYEN N H, ABELLáN-GARCíA J, LEE S, et al. Efficient estimating compressive strength of ultra-high performance concrete using XGBoost model[J]. Journal of Building Engineering, 2022,52:104302.
• [3] MAD W, MOHAMMED S, KURDA R, et al. Prediction of concrete materials compressive strength using surrogate models[J].Structures, 2022, 46:1243-1267.
• [4] ZENG Z, ZHU Z, YAO W, et al. Accurate prediction of concrete compressive strength based on explainable features using deep learning[J]. Construction and Building Materials, 2022, 329:127082.
• [5]徐国强,苏幼坡,韩佃利等.基于BP神经网络的绿色混凝土抗压强度预测模型[J].混凝土, 2013(2):33-35, 49.XU Guoqiang, SU Youpo, HAN Dianli, et al. Prediction model of green concrete compressive st rength based on BP neural net work[J]. Concrete, 2013(2):33-35, 49.
• [6]张玉强,蔡自伟,张智,等.再生细砂超高性能混凝土力学性能研究[J].混凝土, 2023(2):92-99.ZHANG Yuqiang, CAI Zhiwei, ZHANG Zhi, et al. Mechanical properties of ultra-high performance concrete incorporating recycled fine aggregate[J]. Concrete, 2023(2):92-99.
• [7] MARTA S, IAKI F O, IRABIEN A. Neural network prediction of unconfined compressive strength of coal fly ash-cement mixtures[J].Cement&Concrete Research, 2003(8):1137-1146.
• [8] MIRZAHOSSEINI M, JIAO P, BARRI K, et al. New machine learning prediction models for compressive strength of concrete modified with glass cullet[J]. Engineering Computations, 2019, 36(3):876-898.
• [9] HUU M N, HI V H M, HOANG S T, et al. A comparative assessment of tree-based predictive models to estimate geopolymer concrete compressive strength[J]. Neural Computing and Applications, 2022, 35(9):6569-6588.
• [10]姚小俊,吴迪.基于贝叶斯优化高斯过程回归法的再生混凝土力学性能预测[J].科学技术与工程, 2023, 23(7):2968-2975.YAO Xiaojun, WU Di. Prediction of mechanical properties of recycled concrete using bayesian optimization-based gaussian process regression method[J]. Science Technology and Engineering, 2023,23(7):2968-2975.
• [11]王振波,周潮,王家赫,等.高延性水泥基材料(ECC)开裂后渗透性研究综述[J].水利发展研究, 2024, 24(6):59-64.WANG Zhenbo, ZHOU Chao, WANG Jiahe, et al. Review of permeability research on high ductility cement-based materials(ECC)after cracking[J]. Water Resources Development Research, 2024,24(6):59-64.
• [12]张紫键,姚占全,马快乐.钢纤维类型对玄武岩纤维混凝土力学性能的影响[J].排灌机械工程学报, 2024, 42(6):598-604.ZHANG Zijian, YAO Zhanquan, MA Kuaile. Influence of steel fiber type on mechanical properties of basalt fiber concrete[J]. Journal of Drainage and Irrigation Machinery Engineering, 2024, 42(6):598-604.
• [13] LUNDBERG S M, LEE S I. A unified approach to interpreting model predictions[C]∥Advances in Neural Information Processing Systems.Long Beach, CA:NIPS, 2017:4765-4774.
• [14] ALABDULLAh A A, IQBAL M, ZAHID M, et al. Prediction of rapid chloride penetration resistance of metakaolin based high strength concrete using light GBM and XGBoost models by incorporating SHAP analysis[J]. Construction and Building Materials, 2022, 345:128296.
• [15] ULLAH I, LIU K, YAMAMOTO T, et al. Modeling of machine learning with SHAP approach for electric vehicle charging station choice behavior prediction[J]. Travel Behaviour and Society, 2023,31:78-92.
• [16]汪声瑞,胡畔,陈思宝,等.基于耦合BAS-MLP的混凝土抗压强度预测[J].建筑材料学报, 2023, 26(7):705-715.WANG Shengrui, HU Pan, CHEN Sibao, et al. Prediction of concrete compressive strength based on coupled beetle search algorithm[J]. Journal of Building Materials, 2023, 26(7):705-715.
• [17]李杨,刘庆华,郭天添.基于CART-SVR模型的混凝土抗压强度预测研究[J].混凝土, 2022(8):40-44.LI Yang, LIU Qinghua, GUO Tiantian. Prediction method of concrete compressive strength based on CART-SVR[J]. Concrete, 2022(8):40-44.
• [18]汪金胜,李永乐,杨剑,等.基于贝叶斯支持向量回归机的自适应可靠度分析方法[J].计算力学学报, 2022, 39(4):488-497.WANG Jinsheng, LI Yongle, YANG Jian, et al. Adaptive algorithm based on Bayesian support vector regression for structural reliability analysis[J]. Chinese Journal of Computational Mechanics, 2022, 39(4):488-497.
• [19]张金喜,郭旺达,宋波,等.基于随机森林的沥青路面性能预测[J].北京工业大学学报, 2021, 47(11):1256-1263.ZHANG Jinxi, GUO Wangda, SONG Bo, et al. Performance prediction of asphalt pavemen based on random forest[J]. Journal of Beijing University of Technology, 2021, 47(11):1256-1263.
• [20]秦元,余宏淦,陶建峰,等.基于PSO-SVR的土压平衡盾构施工进度优化[J].浙江大学学报(工学版), 2022, 56(8):1523-1532.QIN Yuan, YU Hongan, TAO Jianfeng, et al. Advance rate optimization of earth pressure balance shield based on PSO-SVR[J].Journal of Zhe Jiang University(Engineering Science), 2022, 56(8):1523-1532.
• [21]王红爱.基于XGBoost算法的铁路旅客退票率预测研究[J].铁道学报, 2019, 41(12):19-25.WANG Hongai. Research on predicting model of railway passenger ticket refund rate based on XGBoost Algorithm[J]. Journal of the China Railway Society, 2019, 41(12):19-25.
• [22]冯德成,吴刚.混凝土结构基本性能的可解释机器学习建模方法[J].建筑结构学报, 2022, 43(4):228-238.FENG Decheng, WU Gang. Interpretable machine learning-based modeling approach for fundamental properties of concrete structures[J]. Journal of Building Strutures, 2022, 43(4):228-238.
• [23]崔晓宁,王起才,张戎令,等.基于随机森林的高性能混凝土抗压强度预测[J].兰州交通大学学报, 2021, 40(6):1-6, 14.CUI Xiaoning, WANG Qicai, ZHANG Rongling, et al. Prediction of compressive strength of concrete based on random forest[J]. Journal of Lanzhou Jiaotong University, 2021, 40(6):1-6, 14.
• [24] ACI Committee 209. Prediction of creep and shrinkage and temperature effects in concrete structures[S]. American Concrete Institute, Farmington Hills, MI, 1999.
• [25]申嘉荣,徐千军.建筑材料微观结构的人工识别研究[J].水利发展研究, 2023, 23(10):72-79.SHEN Jiarong, XU Qianjun. Research on artificial identification of microstructure of building materials[J].Water Resources Development Research, 2023, 23(10):72-79.
• [26]韩长君,周海龙,陈岩,等.粉煤灰对高强混凝土力学性能及孔隙结构影响[J].排灌机械工程学报, 2024, 42(4):410-417.HAN Changjun, ZHOU Hailong, CHEN Yan, et al. Effect of fly ash on mechanical properties and pore structure of high-strength concrete[J]. Journal of Drainage and Irrigation Machinery Engineering,2024, 42(4):410-417.
• [27]王甲春,阎培渝.基于等效龄期的粉煤灰混凝土抗压强度计算模型[J].中山大学学报(自然科学版), 2014, 53(4):83-87, 93.WANG Jiachun, YAN Peiyu. Computational model of fly ash concrete compressive strength based on the equivalent age[J]. Acta Scientiarum Naturalium Universitatis Sunyatseni, 2014, 53(4):83-87.
• [28] CUI X N, WANG Q C, ZHANG R L, et al. Machine learning prediction of concrete compressive strength with data enhancement[J]. Journal of Intelligent&Fuzzy Systems:Applications in Engineering and Technology, 2021(41):7219-7228.