基于卷积神经网络的脉冲型地震动破坏力的分级评估  

Evaluating the Destructive Power Grade of Pulse-like Ground Motions Based on Convolutional Neural Network

作　　者：曾春梅 蒲武川[1] ZENG Chun-mei;PU Wu-chuan(School of Civil Engineering and Architecture,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学土木工程与建筑学院,武汉430070

出　　处：《武汉理工大学学报》2022年第8期61-68,共8页Journal of Wuhan University of Technology

基　　金：国家自然科学基金(52178504)。

摘　　要：为了对脉冲型地震动的破坏力进行定量分级评估,基于地震动小波时频图,提出了一种基于卷积神经网络的脉冲型地震动破坏力等级分类方法。选取230组脉冲型地震动,基于4种多层钢筋混凝土结构模型,分别开展单向和双向地震输入下地震响应分析,根据结构层间位移角划分地震动破坏力等级,利用VGG16深度卷积神经网络进行迁移学习,建立地震动小波时频图与破坏力等级之间的映射模型。利用该模型可以实现对脉冲型地震动的破坏力预测,应用于震后结构损伤评估等领域。In order to quantitatively evaluate the destructive power of impulsive ground motion,based on the wavelet time-frequency diagram of ground motion,a classification method of the destructive power of impulsive ground motion based on a convolution neural network was proposed.230 impulse ground motions were selected,and the seismic response analysis was carried out based on four kinds of multi-story reinforced concrete structure models.The destructive power was graded according to the story drift.The VGG16 deep convolution neural network was employed to establish the mapping model between the wavelet time-frequency diagram and the destructive power grade after transfer learning.The model can predict the destructive force of impulsive ground motion and can be applied to structural damage assessment after an earthquake.

关 键 词：脉冲型地震动 地震动破坏力 卷积神经网络 地震动强度参数 双向地震动 

分 类 号：P315.3[天文地球—地震学]