基于机器学习的钢筋混凝土矩形柱破坏模式预测研究  

Research on failure mode prediction of rectangular reinforced concrete columns based on machine learning

作　　者：张海[1,2] 马小平[1,2] 苏三庆[3] 王威[3] 蔡玉军[1,2] ZHANG Hai;MA Xiaoping;SU Sanqing;WANG Wei;CAI Yujun(State Key Laboratory of Intelligent Construction and Maintenance for Extreme Geotechnical and Tunnel Engineering,China Railway First Survey and Design Institute Group Co.,Ltd.,Xi an 710043,Shaanxi,China;Research Institute of Architectural&Planning Design/Research Center of Transit Oriented Development,China Railway First Survey and Design Institute Group Co.,Ltd.,Xi an 710043,Shaanxi,China;School of Civil Engineering,Xi an University of Architecture&Technology,Xi an 710055,Shaanxi,China)

机构地区：[1]中铁第一勘察设计院集团有限公司,极端环境岩土和隧道工程智能建养全国重点实验室,陕西西安710043 [2]中铁第一勘察设计院集团有限公司,建筑与规划设计研究院/TOD研发中心,陕西西安710043 [3]西安建筑科技大学土木工程学院,陕西西安710055

出　　处：《建筑科学与工程学报》2025年第2期48-57,共10页Journal of Architecture and Civil Engineering

基　　金：中铁第一勘察设计院集团有限公司科研开发项目(2022KY51ZD(ZNJC)-03,2022KY54ZD(ZNXT)-05);中国施工企业管理协会青年创新项目(2023-B-009);中国铁路西安局集团有限公司科学技术研究开发计划项目(Y2023034)。

摘　　要：针对传统分析方法识别效果差、数据依赖性强等问题,以既有试验数据为基础,建立矩形截面钢筋混凝土柱的数据库,应用K邻近、随机森林、支持向量机、梯度提升决策树、深度神经网络等机器学习算法,实现矩形柱破坏模式的有效识别与预测。借助机器学习强大的自学习、自适应能力,精准预测钢筋混凝土矩形柱的破坏模式,并为震后结构的维修加固与损伤评估提供依据。结果表明:机器学习技术对弯曲破坏均有良好的识别效果,随机森林和梯度提升决策树算法的准确率和回归率均达到100%,可用于矩形柱弯曲破坏模式的精准预测;机器学习技术对于剪切破坏的识别效果差别不大,准确率均达66.67%,K邻近、支持向量机、梯度提升决策树的回归率最高,达到100%;对于弯剪破坏模式,随机森林和梯度提升决策树的准确率最高,达到83.33%,支持向量机的预测效果较差。Aiming at the problems of poor recognition effect and strong data dependence in traditional analysis methods,a database of rectangular reinforced concrete columns was established based on existing test data.The machine learning algorithms such as K-nearest neighbor,random forest,support vector machine,gradient boosting decision tree and deep neural network were applied to realize the effective recognition and prediction of failure modes for rectangular columns.By leveraging the powerful self-learning and self-adaptive ability of machine learning,the failure mode of rectangular reinforced concrete columns was accurately predicted,providing a basis for for the maintenance,reinforcement and damage assessment of post-earthquake structures.The results show that machine learning has a good recognition effect on bending failure.The accuracy and regression rate of random forest and gradient boosting decision tree are both up to 100%,and they can be used to accurately predict the bending failure mode of rectangular columns.The recognition effect of machine learning technology on shear failure is not significantly different,with an accuracy rate of 66.67%.The regression rates of the K-nearest neighbor,support vector machine,and gradient boosting decision tree are the highest,reaching 100%.For bending-shear failure mode,the accuracy of random forest and gradient boosting decision tree is the highest,reaching 83.33%,while the prediction effect of the support vector machine is poor.

关 键 词：钢筋混凝土 矩形截面柱 机器学习 破坏模式预测 试验数据 

分 类 号：TU393[建筑科学—结构工程]