摘要
针对量大面广的砌体结构开展抗震性能评估,是我国防震减灾工作的重要内容。鉴于传统震害预测方法存在的局限性,以及考虑二次地震影响的结构震害预测方法研究的相对匮乏,该文收集了云南、四川和广西等地的83栋遭受二次地震作用的砌体结构震例,提出了基于支持向量回归(support vector regression,SVR)理论的砌体结构二次地震震害预测方法。详细阐述了该方法的基本原理及实施步骤,确定了砌体结构的震害影响因子及量化值,建立了震害样本数据库及预测模型,并将预测结果分别与BP神经网络、随机森林等方法的预测结果进行对比分析。结果表明,在小样本、少震害因子的情况下,基于SVR的震害预测结果可靠,适用于快速预测砌体结构的二次地震震害,可为考虑二次地震作用的结构抗震性能评估提供新思路。
Evaluating the seismic performance of large-scale masonry structures is a crucial aspect of earthquake prevention and disaster mitigation efforts in China.Given the limitations of traditional seismic damage prediction methods and the relative lack of research on methods that consider the impact of secondary earthquakes,this paper collected seismic damage data from 83 masonry structures affected by secondary earthquakes in Yunnan,Sichuan,and Guangxi.A seismic damage prediction method for masonry structures based on support vector regression(SVR)theory is proposed.The fundamental principles and implementation steps of this method are elaborated in detail.Seismic damage influence factors and their corresponding quantitative values for masonry structures are identified,leading to the establishment of a seismic damage sample database and a predictive model.The predictive performance is subsequently compared and analyzed against those of other methods,such as the BP neural network and random forest.The results indicate that,in cases of small sample sizes and limited seismic damage factors,the SVR-based prediction results are reliable and suitable for the rapid prediction of seismic damage to masonry structures.This approach provides a new perspective for seismic performance evaluation that considers the effects of secondary earthquakes.
作者
周强
陈天振
周杰
赵文洋
ZHOU Qiang;CHEN Tianzhen;ZHOU Jie;ZHAO Wenyang(School of Infrastructure Engineering,Nanchang University,Nanchang 330031,China)
出处
《自然灾害学报》
北大核心
2025年第5期141-150,共10页
Journal of Natural Disasters
基金
国家自然科学基金项目(51968047,51608249)。
关键词
支持向量回归
砌体结构
二次地震
抗震性能评估
support vector regression
masonry structures
secondary earthquakes
seismic performance evaluation
