基于时空差分图卷积神经网络的古代石拱桥损伤识别研究  

Research on damage detection of ancient stone arch bridges based on spatio-temporal difference graph convolutional neural network

在线阅读下载全文

作  者:张承文 淳庆[1,2] 马宇坤[3] 林怡婕 ZHANG Chengwen;CHUN Qing;MA Yukun;LIN Yijie(School of Architecture,Southeast University,Nanjing 210096,China;Key Laboratory of Urban and Architectural Heritage Conservation of Ministry of Education,Southeast University,Nanjing 210096,China;School of Architecture and Design,Harbin Institute of Technology,Harbin 150081,China)

机构地区:[1]东南大学建筑学院,南京210096 [2]东南大学城市与建筑遗产保护教育部重点实验室,南京210096 [3]哈尔滨工业大学建筑与设计学院,哈尔滨150081

出  处:《东南大学学报(自然科学版)》2025年第2期370-379,共10页Journal of Southeast University:Natural Science Edition

基  金:国家重点研发计划资助项目(2023YFF0906100);江苏省自然科学基金资助项目(BK20241347)。

摘  要:基于长期结构健康监测数据,采用时空差分图卷积神经网络,实现了古代石拱桥损伤的自动识别。首先,提出了一种改进的有向图矩阵构建方法,以避免有向图矩阵构建时的环境影响问题。其次,针对古代石拱桥,采用时空差分图卷积神经网络,构建训练模型。然后,以全国重点文物保护单位北京卢沟桥为算例,进行了为期32个月的结构健康监测。最后,根据监测结果训练并测试模型性能。研究结果表明,所提损伤识别方法的完全准确预测率可达92.26%,标签准确率为99%,精确率均值为93.5%,召回率均值为90.4%,误报率均值为3.0%,F_(1)得分均值为0.917。研究成果可为桥梁文物的预防性保护提供科学支撑。Based on long⁃term structural health monitoring data,a spatio⁃temporal difference graph convolu⁃tional neural network was used to achieve automatic damage detection for ancient stone arch bridges.Firstly,an improved directed graph matrix construction method was proposed to avoid environmental impact.Sec⁃ondly,for ancient stone arch bridges,a spatio⁃temporal difference graph convolutional neural network was used to build a training model.Then,taking the Beijing Lugou Bridge,a national key cultural relics protec⁃tion unit,as an example,a 32⁃month structural health monitoring was conducted.Finally,the model was trained and tested based on the monitoring results.The research results show that the proposed damage detec⁃tion method has a global accuracy rate of up to 92.26%,a label accuracy rate of up to 99%,an average preci⁃sion rate of 93.5%,an average recall rate of 90.4%,an average false alarm rate of 3.0%,and an average F_(1) score of 0.917.The research results can provide scientific support for the preventive conservation of bridge cultural relics.

关 键 词:石拱桥 建筑遗产 损伤识别 图神经网络 结构健康监测 

分 类 号:TU363[建筑科学—结构工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象