摘要
钢筋混凝土柱在侧向地震力作用下具有弯曲、剪切和弯剪三种失效模式。不同的失效模式下钢筋混凝土柱具有不同的地震损伤特征。因此,准确地判别钢筋混凝土柱的失效模式对于准确评估钢筋混凝土结构的抗震性能具有重要意义。利用已有的钢筋混凝土柱滞回加载试验数据,采用机器学习方法,提出了一种钢筋混凝土柱失效模式两阶段判别方法。其中,第一阶段以钢筋混凝土柱的基本设计参数为输入变量,采用机器学习中的回归算法,建立钢筋混凝土柱的受弯承载力、受剪承载力、弯曲变形和剪切变形预测模型。第二阶段以钢筋混凝土柱的受弯承载力、受剪承载力、弯曲变形和剪切变形作为输入变量,采用机器学习中的分类算法,对钢筋混凝土柱的失效模式进行自动判别,实现了准确判别钢筋混凝土柱失效模式的目的。研究结果表明:极端随机树、AdaBoost、随机森林和梯度提升算法分别对受弯承载力、受剪承载力、弯曲变形和剪切变形的预测效果最佳;极端随机树、梯度提升算法和最近邻居法分别对弯曲失效、剪切失效和弯剪失效具有最佳的分类效果;相比已有的钢筋混凝土柱失效模式分类方法,提出的两阶段分类方法具有与真实失效模式最为接近的分类结果,分类精度可以达到96%。
There are three main failure modes, i.e., flexural-critical, shear-critical, and flexural-shear modes, of reinforced concrete(RC) columns under lateral loadings. Different failure modes commonly lead to various damage characteristics of RC columns under earthquakes. Therefore, an accurate classification of failure modes of RC columns is important for seismic performance evaluation of RC structures. A two-step classification approach for failure modes of RC columns was proposed in this study. It incorporates machine learning methods based on the available test data on RC columns under cycling loads. At the first step, regression algorithms of machine learning methods are adopted to predict the flexural capacity, shear capacity, flexural deformation, and shear deformation, using the design parameters of column as input variables. At the second step, classification algorithms of machine learning methods are employed to identify the failure modes of RC columns, with the flexural capacity, shear capacity, flexural deformation and shear deformation as the input variables. Based on the proposed approach, an accurate identification of the failure modes can be realized for RC columns. The results show that the extremely randomized trees, AdaBoost, random forest, and gradient boosting algorithms show the best performance on predicting the flexural capacity, shear capacity, flexural deformation, and shear deformation, respectively. The extremely randomized trees, gradient boosting and K-nearest neighbor algorithms exhibit the best performance on classifying the flexural-critical, shear-critical and flexural-shear failure modes of RC columns, respectively. Compared to the existing methods, the proposed two-step classification approach has more similar classification results to the real failure mode. The classification accuracy can reach as high as 96%.
作者
于晓辉
王猛
宁超列
YU Xiaohui;WANG Meng;NING Chaolie(College of Civil Engineering and Architecture,Guilin University of Technology,Guilin 541004,China;School of Civil Engineering,Harbin Institute of Technology University,Harbin 150090,China;Shanghai Institute of Disaster Prevention and Relief,Tongji University,Shanghai 200092,China)
出处
《建筑结构学报》
EI
CAS
CSCD
北大核心
2022年第8期220-231,共12页
Journal of Building Structures
基金
国家自然科学基金项目(51778198,51808397)
黑龙江省自然科学基金资助项目(YQ2020E023)。
关键词
钢筋混凝土柱
机器学习
回归算法
分类算法
失效模式
reinforced concrete column
machine learning
regression algorithm
classification algorithm
failure mode
