摘要
为解决震后大量钢筋混凝土框架通信机楼损伤评定需求,基于卷积神经网络研究从构件层次至整体结构的损伤评定方法。首先对汶川地震、鲁甸地震、芦山地震等多次地震后大量钢筋混凝土框架结构损伤调查图片筛选处理,建立了钢筋混凝土框架梁、柱损伤评定数据集。然后通过对3个关键问题的研究建立了钢筋混凝土框架基于卷积神经网络的震损评定方法:训练和建立YOLOv5网络模型完成从结构震害照片中检测识别出梁、柱构件的任务,并改进优化了YOLOv5网络模型的检测性能;优选比较3种网络模型(ResNet50、MobileNetV2和AlexNet模型)对梁、柱构件损伤水平评定的精确性,最终建立了基于ResNet50的梁、柱构件损伤评定模型;给出了从构件层次到整体结构的损伤水平确定方法,并通过对一栋实际震损框架进行损伤评定验证了文中方法的可用性。结果表明,文中方法与专家的损伤评定结论一致性高,优化后的卷积神经网络模型精确度和稳定性好,对震后钢筋混凝土框架结构损伤评定具有良好的适用性。
In order to solve the demand for damage assessment of a large number of reinforced concrete(RC)frame communication buildings after earthquakes,this paper studied the damage assessment methods from the component level to the overall structure based on convolutional neural networks(CNN).Firstly,a large number of damage survey pictures of RC frame structures after earthquakes such as the Wenchuan earthquake,Ludian earthquake,and Lushan earthquake were screened and processed,and a damage assessment dataset of RC frame beams and columns was established.Secondly,a damage assessment method for RC frames based on CNN was established through the study of 3 key issues:The task of detecting and recognizing the components of beams and columns from the photos of structural damage was completed by training and establishing the YOLOv5 network model;the detection performance of the YOLOv5 network model was improved and optimized;3 network models(ResNet50,MobileNetV2,and AlexNet model)were selected and compared for the accuracy of damage level assessment of beams and columns.Finally,a damage assessment model of beams and columns based on ResNet50 was established.The method of determining the damage level from the component level to the overall structure was given,and the availability of the method in the paper was verified by damage assessment of an actual damaged frame,and the results show that the method in the paper has high consistency with the damage assessment conclusion of experts,and the optimized CNN model has good accuracy and stability,and has good applicability to the damage assessment of post-earthquake RC frame structures.
作者
毛晨曦
郭永超
张昊宇
张亮泉
MAO Chenxi;GUO Yongchao;ZHANG Haoyu;ZHANG Liangquan(Key Laboratory of Earthquake Engineering and Engineering Vibration,Institute of Engineering Mechanics,China Earthquake Administration,Harbin 150080,China;Key Laboratory of Earthquake Disaster Mitigation,Ministry of Emergency Management,Harbin 150080,China;School of Civil Engineering,Northeast Forestry University,Harbin 150040,China)
出处
《自然灾害学报》
CSCD
北大核心
2024年第5期157-167,共11页
Journal of Natural Disasters
基金
国家自然科学基金面上项目(52178513)。
关键词
钢筋混凝土框架
通信机楼
卷积神经网络
震害调查
损伤评定
RC frame structure
communication building
convolutional neural networks
post-earthquake survey
damage assessment
