基于多输入卷积神经网络隔震支座沉降识别  

Seismic isolation bearing settlement recognition based on multi-input convolutional neural network

作　　者：赵丽洁[1,2] 李纯 沈金生 王昊[3] ZHAO Lijie;LI Chun;SHEN Jinsheng;WANG Hao(School of Civil Engineering,Hebei University of Engineering,Handan 056038,China;School of Water Conservancy Engineering,Tianjin Agricultural University,Tianjin 300392,China;School of Civil Engineering,Tianjin Chengjian University,Tianjin 300384,China)

机构地区：[1]河北工程大学土木工程学院,河北邯郸056038 [2]天津农学院水利工程学院,天津300392 [3]天津城建大学土木工程学院,天津300384

出　　处：《地震工程与工程振动》2024年第4期62-69,共8页Earthquake Engineering and Engineering Dynamics

基　　金：国家自然科学基金项目(52208193)。

摘　　要：为了避免地基不均匀沉降导致隔震支座沉降以及对上部结构造成的隐性损伤,针对隔震支座沉降识别方法进行研究,提出一种基于多输入卷积神经网络(multi-input convolutional neural network,MI-CNN)的隔震支座振动信号识别模型。首先,采集隔震支座水平方向加速度和位移信号,采用归一化预处理和数据增强方法扩充样本;然后,将样本输入到所建立的网络模型中并进行训练;最后,利用完成训练的网络模型进行沉降识别。结果表明:相较于传统单输入卷积神经网络(Convolutional neural network,CNN)模型,MI-CNN模型易于训练,可最大程度地发挥CNN对沉降信号特征的提取能力,且具有更好的沉降位置识别准确率和更小的沉降程度识别误差,以及针对不均衡数据集更稳定的识别效果。研究结果可为隔震支座沉降识别提供新思路。In order to avoid the settlement of seismic isolation bearings caused by uneven foundation settlement and the hidden damage to the superstructure,a vibration signal identification model based on multi-input convolutional neural network(MI-CNN)is proposed to identify the settlement of seismic isolation bearings.First,the horizontal acceleration and displacement signals of seismic isolation bearings are collected,and the samples are expanded using normalised pre-processing and data enhancement methods.Then,the samples are fed into the established network model and trained.Finally,the settlement identification is performed using the trained network model.The results show that compared with the traditional single-input CNN model,the MI-CNN model is easier to train and can maximise the ability of CNN to extract features from the settlement signals,and it has a better accuracy in identifying the settlement location,a smaller error in identifying the settlement degree,and a more stable identification effect for the unbalanced data set.The results of this study can provide new ideas for the settlement identification of seismic isolation bearings.

关 键 词：卷积神经网络 隔震支座 不均衡数据集 沉降识别 

分 类 号：TU375.4[建筑科学—结构工程] TP183[自动化与计算机技术—控制理论与控制工程]