基于XGBoost模型集成学习的RC框架结构地震响应预测方法  

Seismic response prediction method of RC frame structure based on XGBoost model of ensemble learning

作　　者：赵煜东 许卫晓[1,2] 李静 杨伟松 赵继幸[1,2] 姜冠宇 ZHAO Yudong;XU Weixiao;LI Jing;YANG Weisong;ZHAO Jixing;JIANG Guanyu(School of Civil Engineering,Qingdao University of Technology,Qingdao 266525,China;Engineering Research Center of Concrete Technology under Marine Environment,Ministry of Education,Qingdao 266525,China;Qingdao Civil Air Defense Engineering Quality Supervision Station,Qingdao 266072,China)

机构地区：[1]青岛理工大学土木工程学院,青岛266525 [2]海洋环境混凝土技术教育部工程技术研究中心,青岛266525 [3]青岛市人民防空工程质量监督站,青岛266072

出　　处：《青岛理工大学学报》2024年第2期76-83,132,共9页Journal of Qingdao University of Technology

基　　金：山东省自然科学基金资助项目(ZR2020ME246;ZR2022ME029)。

摘　　要：为实现钢筋混凝土(RC)框架结构地震响应的快速预测,提出了基于集成学习的RC框架结构地震响应预测方法。设计低层、多层和小高层共3个RC框架结构作为研究算例,根据条件均值谱(CMS)选取地震动记录,通过弹塑性时程分析搭建样本数据库,以地震动强度信息和结构信息为输入预测结构响应,同时对模型进行特征重要性分析。研究结果表明,建立的XGBoost模型相比梯度提升回归树(GBRT)模型具有更好的泛化性能,特征参数中平均谱加速度(AvgS_(a))的相对重要性最大,提出的方法为快速预测RC框架结构地震响应提供了借鉴,具有较高的应用价值。In order to realize the rapid prediction of seismic response of reinforced concrete(RC)frame structure,a seismic response prediction method of RC frame structure based on ensemble learning is proposed.Three RC frame structures,namely low-rise,multi-story and small high-rise,were designed as research examples.The ground motion records were selected according to the conditional mean spectrum(CMS),and the sample database was built through elastoplastic time history analysis.The structural response was predicted with the input of ground motion intensity information and structural information,and the feature importance of the model was analyzed.The results show that the established XGBoost model has better generalization performance than the gradient boosting regression tree(GBRT)model,and the relative importance of the average spectral acceleration(AvgS_(a))in the characteristic parameters is the largest.The proposed method provides a reference for quickly predicting the seismic response of RC frame structures and has high application value.

关 键 词：RC框架结构 集成学习 地震响应 特征重要性分析 

分 类 号：TU375.4[建筑科学—结构工程]