摘要
桥梁在运营过程中面临着组合荷载的复杂环境,因此发展组合荷载下的损伤识别方法具有重要意义。本文提出了一种基于组合荷载响应特征融合的桥梁结构智能损伤识别方法,基于移动主成分分析对自重静载、温度准静态荷载、动态荷载下的结构响应数据分别进行特征挖掘,并将不同荷载下第一特征向量的组合作为机器学习模型的输入,建立结构的损伤识别方法。最后,以双跨连续梁的仿真模型进行了验证,研究结果表明,即使在大噪声水平下,以组合荷载特征向量进行损伤定位和定量的准确率分别可达91.65%和97.22%,比传统的单荷载下的准确率最高分别提升了32.40%和18.00%,表现出优异的损伤检测性能和抗噪性。
Bridges face a complex environment with combined loadings during operation,therefore it is important to develop damage identification methods under combined loadings.An intelligent damage identification method based on response feature fusion under combined loadings for bridge structures was proposed.The method began with the moving principal component analysis for the structural response data under static self-weight loading,quasi-static temperature loading,and dynamic loading,to obtain the first eigenvectors as damage-sensitive features.Subsequently,the combination of the first eigenvectors under different loadings wass used as the input of the machine learning model,so as to establish the intelligent damage identification method.Finally,a simulation model of a two-span continuous beam was used for validation.The results demonstrate that the accuracy of damage localization and quantification with the eigenvectors of combined loadings can reach 91.65%and 97.22%,respectively,even under a large noise level,which is up to 32.40%and 18.00%higher than that under the conventional single loading,showing excellent damage detection performance and noise immunity.
作者
钟玉琪
张红
张舸
周立成
刘泽佳
刘逸平
蒋震宇
杨宝
汤立群
ZHONG Yuqi;ZHANG Hong;ZHANG Ge;ZHOU Licheng;LIU Zejia;LIU Yiping;JIANG Zhenyu;YANG Bao;TANG Liqun(School of Civil Engineering and Transportation,South China University of Technology,Guangzhou 510641,Guangdong,China;State Key Laboratory of Subtropical Building Science,South China University of Technology,Guangzhou 510641,Guangdong,China;Guangdong Provincial Academy of Building Research Group Co.,Ltd.,Guangzhou 510599,Guangdong,China)
出处
《实验力学》
CSCD
北大核心
2023年第2期151-164,共14页
Journal of Experimental Mechanics
基金
国家自然科学基金(11972162)
中国博士后科学基金(2021M700886)资助。
关键词
结构健康监测
损伤识别
移动主成分分析
组合荷载
特征融合
机器学习
structural health monitoring
damage identification
moving principal component analysis
combined loadings
feature fusion
machine learning
