卷积神经网络在结构损伤识别中的应用研究

Research on the Application of Convolutional Neural Networks in Structural Damage Identification

为了提高结构健康监测系统损伤诊断的自动化程度和识别精度,将卷积神经网络引入到结构损伤识别中。首先介绍了卷积神经网络的相关理论,对输入层、卷积层、池化层和全连接层以及输出层的相关理论和参数设置进行了简要介绍;建立了用于结构损伤识别的一维卷积神经网络模型,采用卡塔尔大学看台试验数据集验证了该模型用于结构损伤识别的可行性及可靠性。识别结果表明:所建立的模型的损伤识别平均准确率大于90%,满足工程结构健康监测精度要求。

In order to improve the automation and identification accuracy of damage diagnosis in structural health monitoring system,convolutional neural network is introduced into structural damage identification.Firstly,a brief introduction was given to the relevant theories and parameter settings of convolutional neural networks,including input layer,convolutional layer,pooling layer,fully connected layer,and output layer;A one-dimensional convolutional neural network model for structural damage identification was established,and the feasibility and reliability of this model for structural damage identification were verified using the Qatar University stand test dataset.The identification results show that the average accuracy of damage identification of the established model is more than 90%,which meets the accuracy requirements of engineering structural health monitoring.