| 卷积神经网络在结构损伤诊断中的应用 |
| |
| 引用本文： | 李书进,赵 源,孔 凡,张远进.卷积神经网络在结构损伤诊断中的应用[J].建筑科学与工程学报,2020,0(6):29-37. |
| |
| 作者姓名： | 李书进 赵 源 孔 凡 张远进 |
| |
| 作者单位： | 1. 武汉理工大学 土木工程与建筑学院,湖北 武汉 430070;
2. 武汉理工大学 安全科学与应急管理学院,湖北 武汉 430070 |
| |
| 摘 要： | 对卷积神经网络(CNN)在工程结构损伤诊断中的应用进行了深入探讨; 以多层框架结构节点损伤位置的识别问题为研究对象,构建了可以直接从结构动力反应信号中进行学习并完成分类诊断的基于原始信号和傅里叶频域信息的一维卷积神经网络模型和基于小波变换数据的二维卷积神经网络模型; 从输入数据样本类别、训练时间、预测准确率、浅层与深层卷积神经网络以及不同损伤程度的影响等多方面进行了研究。结果表明:卷积神经网络能从结构动力反应信息中有效提取结构的损伤特征,且具有很高的识别精度; 相比直接用加速度反应样本,使用傅里叶变换后的频域数据作为训练样本能使CNN的收敛速度更快、更稳定,并且深层CNN的性能要好于浅层CNN; 将卷积神经网络用于工程结构损伤诊断具有可行性,特别是在大数据处理和解决复杂问题能力方面与其他传统诊断方法相比有很大优势,应用前景广阔。
|
| 关 键 词： | 损伤识别 卷积神经网络 深度学习 框架结构 小波变换 |
Application of Convolutional Neural Network in Structural
Damage Identification |
| |
| LI Shu-jin,ZHAO Yuan,KONG Fan,ZHANG Yuan-jin.Application of Convolutional Neural Network in Structural
Damage Identification[J].Journal of Architecture and Civil Engineering,2020,0(6):29-37. |
| |
| Authors: | LI Shu-jin ZHAO Yuan KONG Fan ZHANG Yuan-jin |
| |
| Affiliation: | 1. School of Civil Engineering and Architecture, Wuhan University of Technology, Wuhan 430070,
Hubei, China; 2. School of Safety Science and Emergency Management, Wuhan University of
Technology, Wuhan 430070, Hubei, China |
| |
| Abstract: | The application of convolutional neural network in damage identification of engineering structure was discussed in depth. Taking the identification of damage location of multi-layer frame structure nodes as the research object, a one-dimensional convolutional neural network model based on original signal and Fourier frequency domain information and a two-dimensional convolutional neural network model based on wavelet transform data were constructed, which could learn directly from the structural dynamic response signals and complete the classification diagnosis. The type of input data sample, training time, prediction accuracy, shallow and deep convolutional neural network and the influence of different damage degree on damage identification were studied. The results show that the convolutional neural network can effectively extract the damage features from the dynamic response information of the structure, and has a high recognition accuracy. Compared with using acceleration response sample directly, using frequency domain data after Fourier transform as training samples can make the CNN convergence faster and more stable, and the performance of deep CNN is better than that of shallow CNN. The convolutional neural network can be used in the damage diagnosis of engineering structures, especially in the big data processing and the ability to solve complex problems, and has great advantages and wide application prospects compared with the other traditional diagnosis methods. |
| |
| Keywords: | damage identification convolutional neural network deep learning frame structure wavelet transform |
|
| 点击此处可从《建筑科学与工程学报》浏览原始摘要信息 |
|
点击此处可从《建筑科学与工程学报》下载全文 |
|
