基于多种机器学习算法的CFRP加固损伤再生骨料混凝土抗压强度预测
Prediction of compressive strength of CFRP reinforced damaged recycled aggregate concrete based on multiple machine learning algorithms
关键词：
碳纤维增强聚合物（CFRP）；损伤再生骨料混凝土；抗压强度；机器学习；特征分析；可解释性
Key words：
Carbon fiber reinforced polymer (CFRP); Damaged recycled aggregate concrete; Compressive strength; Machine learning; Feature analysis; Interpretability
刊期：
2025年第10期
DOI：
10.19761/j.1000-4637.2025.10.130.07
文章编号：
基金项目：
湖北省自然科学基金资助项目（2022CFB662）；武汉市知识创新专项曙光计划项目（2023020201020403）。
作者：
彭 胜，刘 广
单位：
武汉科技大学 城市建设学院，湖北 武汉 430081
摘要
摘 要：对比研究了4种机器学习算法对碳纤维增强聚合物（Carbon fiber reinforced polymer，CFRP）加固损伤再生骨料混凝土抗压强度的预测能力，包括决策树（Decision tree，DT）、随机森林（Random forest，RF）、梯度增强回归树（Gradient boosting regression tree，GBRT）和极限梯度提升（Extreme gradient boosting，XGB）算法，建立了DT、RF、GBRT和XGB模型；采用SHAP方法对GBRT模型进行了解释，分析了输入特征与预测抗压强度之间的关系。结果表明：GBRT模型的性能相对最好，其训练集和测试集的决定系数R2分别为0.996和0.963，能够较为准确地预测CFRP加固损伤再生骨料混凝土的抗压强度；再生骨料混凝土的峰值强度和CFRP加固厚度对CFRP加固损伤再生骨料混凝土抗压强度的影响较大，CFRP断裂应变的影响最小。
Abstract: A comparative study was conducted on the predictive ability of four machine learning algorithms for the compressive strength of carbon fiber reinforced polymer (CFRP) reinforced damaged recycled aggregate concrete, including decision tree (DT), random forest (RF), gradient boosting regression tree (GBRT), and extreme gradient boosting (XGB) algorithms. DT, RF, GBRT, and XGB models were established. The SHAP method was used to interpret the GBRT model, analyzing the relationship between input features and predicted compressive strength. The results show that the GBRT model has relatively better performance, with determination coefficients R2 of 0.996 and 0.963 for its training and testing sets, respectively. The GBRT model can accurately predict the compressive strength of CFRP reinforced damaged recycled aggregate concrete. The peak strength of recycled aggregate concrete and the thickness of CFRP reinforcement have a significant impact on the compressive strength of CFRP reinforced damaged recycled aggregate concrete, while the impact of CFRP fracture strain is minimal.
结论
（1）考虑fco、εco、εh，rup、T、d和rFA因素的影响，DT、RF、GBRT和XGB模型均可用于预测CFRP加固损伤再生骨料混凝土的抗压强度，其中，GBRT模型的预测精度相对最高。
（2）SHAP分析结果表明，6个参数对CFRP加固损伤再生骨料混凝土的抗压强度影响程度大小顺序为：fco>T>d>εco> rFA >εh，rup，其中，fco、T的影响相对较大，εh，rup的影响最小。
参考文献
[1] XU J J,CHEN Z P,XIAO Y,et al.Recycled aggregate concrete in FRP-confined columns: A review of experimental results[J].Composite Structures,2017,174:277-291.
[2] 刘春阳,闫凯,李秀领,等.FRP约束再生混凝土构件研究进展[J].复合材料学报,2024,41(7):3490-3502.
[3] ZHAO X Y,CHEN J X,CHEN M G,et al.Prediction of ultimate condition of FRP-confined recycled aggregate concrete using a hybrid boosting model enriched with tabular generative adversarial networks[J].Thin-Walled Structures,2023,182:110318.
[4] WEI Y Y,WU Y F.Unified stress-strain model of concrete for FRP-confined columns[J].Construction and Building Materials,2012,26(1):381-392.
[5] 陈春红,俞江,刘荣桂,等.再生细骨料混凝土的性能研究进展[J].混凝土与水泥制品,2024(1):94-99.
[6] XIE T Y,GHOLAMPOUR A,OZBAKKALOGLU T.Toward the development of sustainable concretes with recycled concrete aggregates:Comprehensive review of studies on mechanical properties[J].Journal of Materials in Civil Engineering,2018,30 (9):04018211.
[7] MIRALDO S,LOPES S,PACHECO-TORGAL F,et al.Advantages and shortcomings of the utilization of recycled wastes as aggregates in structural concretes[J].Construction and Building Materials,2021,298:123729.
[8] JAYASURIYA A,SHIBATA E S,CHEN T,et al,Development and statistical database analysis of hardened concrete properties made with recycled concrete aggregates[J].Resources Conservation and Recycling,2021,164:105121.
[9] 曹玉贵,赵国旭,尹亚运.基于广义回归神经网络的纤维增强聚合物复合材料约束损伤混凝土强度预测[J].复合材料学报,2021,38(5):1623-1628.
[10] SALEHI H,BURGUENO R.Emerging artificial intelligence methods in structural engineering[J].Engineering Structures,2018,171:170-189.
[11] DARKO A,CHAN A P C,ADABRE M A,et al.Artificial intelligence in the AEC industry: Scientometric analysis and visualization of research activities[J].Automation in Construction,2020,112:103081.
[12] MASOOMEH M,HOSEIN N.Recent trends in prediction of concrete elements behavior using soft computing(2010-2020)[J].Archives of Computational Methods in Engineering,2021,28(4):3307-3327.
[13] FALCONE R,LIMA C,MARTINELLI E.Soft computing techniques in structural and earthquake engineering:A literature review[J].Engineering Structures,2020,207:110269.
[14] HUNT E B,MARIN J,STONE P J.Experiments in induction[M].Cambridge:Academic Press,1996.
[15] 吴贤国,王雷,陈虹宇,等.基于随机森林-NSGA Ⅱ高性能混凝土耐久性配合比的多目标优化研究[J].材料导报,2022,36(17):115-121.
[16] LIU K H,ZHENG J K,DONG S X,et al.Mixture optimization of mechanical, economical, and environmental objectives for sustainable recycled aggregate concrete based on machine learning and metaheuristic algorithms[J].Journal of Building Engineering,2023,63:105570.
[17] FRIEDMAN J H.Greedy function approximation: a gradient boosting machine[J].The Annals of Statistics,2001,29(5):1189-1232.
[18] CHEN T Q,GUESTRIN C.Xgboost:A scalable tree boosting system[C]//Association for Computing Machinery.Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining.New York:Association for Computing Machinery,2016:785-794.
[19] 黄振江.BFRP约束损伤混凝上的轴压力学性能与尺寸彭胜,刘广.基于多种机器学习算法的CFRP加固损伤再生骨料混凝土抗压强度预测[J].混凝土与水泥制品,2025(10):130-135,159.
PENG S,LIU G.Prediction of compressive strength of CFRP reinforced damaged recycled aggregate concrete based on multiple machine learning Algorithms[J].China Concrete and Cement Products,2025(10):130-135,159.
引用本文
彭胜,刘广.基于多种机器学习算法的CFRP加固损伤再生骨料混凝土抗压强度预测[J].混凝土与水泥制品,2025(10):130-135,159.
PENG S,LIU G.Prediction of compressive strength of CFRP reinforced damaged recycled aggregate concrete based on multiple machine learning Algorithms[J].China Concrete and Cement Products,2025(10):130-135,159.
下一个： 粗骨料超高性能混凝土（UHPC-CA）的设计方法、力学与收缩性能研究进展
相关文件