基于机器学习的受压混凝土氯离子传输预测模型  

Prediction model for chloride ion transport in compressed concrete based on machine learning

作　　者：郭冰冰 陈楠 李京钊 王艳 张永利[1] 牛荻涛[1,2] GUO Bingbing;CHEN Nan;LI Jingzhao;WANG Yan;ZHANG Yongli;NIU Ditao(School of Civil Engineering,Xi’an University of Architecture and Technology,Xi’an 710055,China;Key Lab of Structural Engineering and Earthquake Resistance of China Ministry of Education,Xi’anUniversity of Architecture and Technology,Xi’an 710055,China;School of Materials Science andEngineering,Xi’an University of Architecture and Technology,Xi’an 710055,China)

机构地区：[1]西安建筑科技大学土木工程学院,陕西西安710055 [2]西安建筑科技大学结构工程与抗震教育部重点实验室,陕西西安710055 [3]西安建筑科技大学材料科学与工程学院,陕西西安710055

出　　处：《建筑结构学报》2025年第4期198-209,共12页Journal of Building Structures

基　　金：国家自然科学基金青年基金项目(51908453);深圳市承接国家重大科技项目产业化应用研究项目(CJGJZD20220517141806015)。

摘　　要：机器学习预测模型已逐渐成为研究混凝土氯离子传输性能的重要方法,目前尚未考虑到荷载对氯离子传输的影响,为此建立了荷载作用下混凝土氯离子传输的机器学习预测模型。基于已有的试验研究,建立了包含2 458个样本的数据库。通过开展6种机器算法预测模型性能的评估,发现基于XGBoost算法的机器学习模型更适用于受压混凝土中氯离子传输的机器学习预测。进而,将SHAP方法与XGBoost算法结合,开展了受压混凝土中氯离子传输机器学习预测结果的可解释性分析。结果表明:应力荷载水平对于混凝土中氯盐侵蚀的影响仅次于暴露时间;氯离子侵蚀速率随着压应力水平的增大呈现先降低后升高的趋势;相较于Fick第二定律的解析模型,所建立的机器学习模型能够准确地呈现出氯离子浓度峰值的位置,可更好地预测混凝土中氯离子传输行为。Machine learning prediction model has gradually become an important method to study the chloride transport behavior in concrete,and existing machine learning prediction models have not considered the effect of load on the chloride transport.Thus,machine learning prediction model for chloride transport in concrete under compressive loading is developed.Based on the experimental research in the existing literature,a database containing 2458 samples has been established.After evaluating the performance of six machine learning prediction models with different machine algorithms,it is found that ML model based on XGBoost algorithm is most suitable for the chloride transport model in compressive concrete.Furthermore,SHAP method is combined into the XGBoost model to analyze the interpretability of machine learning prediction.The results show that apart from exposure time,stress level has the second most effect on the chloride transport.In addition,the transport rate of chloride in concrete decreases first and then increases with increasing compressive stress level.Compared with the analytical model of Fick’s second law,the established ML model can accurately predict the peak of the chloride concentration,and thereby it can predict the chloride transport behaviors in concrete better.

关 键 词：受压混凝土 氯离子传输 机器学习 应力水平 可解释性 

分 类 号：TU528[建筑科学—建筑技术科学]