基于一维密集卷积网络的悬臂梁断裂损伤识别  

Fracture Damage Identification of a Cantilever Beam Based on One⁃Dimensional Densely Connected Convolutional Network

作　　者：沙春 SHA Chun(School of Civil Engineering,Chongqing Jiaotong University,Chongqing 400074,China)

机构地区：[1]重庆交通大学土木工程学院,重庆400074

出　　处：《东莞理工学院学报》2023年第3期101-107,共7页Journal of Dongguan University of Technology

摘　　要：基于振动的损伤识别是结构健康检测的重要任务,提出了一种基于加速度时程响应的悬臂梁断裂损伤识别方法。使用有限元分析模拟悬臂梁作为研究对象,通过分离裂缝模型施加裂缝模拟损伤状态,施加瞬态荷载获得损伤状态对应的加速度时程响应数据,利用偏移采样处理后的数据建立一维密集卷积网络回归模型,并与标准一维卷积神经网络模型和残差网络对比,最后在原始数据中添加白噪声模拟真实环境检验模型的实际应用效果。结果表明建立模型的识别精度以及效率均远远好于其他神经网络,并且在20分贝的噪声环境下效果也比较显著。证明了使用一维密集卷积网络对梁裂缝问题进行损伤识别的优越性和可行性。Vibration⁃based damage identification is an important task of structural health monitoring,and a fracture damage i⁃dentification method for cantilever beam based on acceleration time⁃course response is proposed.ANSYS is used to simulate a canti⁃lever beam as the research object.Discrete crack model is used to create the cracks to simulate the state of damage;applying a tran⁃sient load obtains acceleration time⁃course response data corresponding to the damage state;a one⁃dimensional intensive convolution⁃al network regression model is built by using the offset sampling processed data,and is in contrast with the standard one⁃dimensional convolutional neural network model and the residual network.Finally,white noise is added to the original data to simulate the real application effect of the real environment test model.The results show that the recognition accuracy and efficiency of the model are far better than those of other neural network,and the effect is also significant in the noise environment of 20db.It proves the superi⁃ority and feasibility of using one⁃dimensional densely connected convolutional network for damage identification of beam cracks.

关 键 词：损伤识别 加速度时程响应 悬臂梁 裂缝 一维密集卷积网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TU317[自动化与计算机技术—控制科学与工程]