Home >

news ヘルプ

論文・著書情報


タイトル
和文:Nonlinear seismic response and residual drift determination combining ground acceleration data with recorded videos 
英文:Nonlinear seismic response and residual drift determination combining ground acceleration data with recorded videos 
著者
和文: BOROSCHEK Ruben, YEOW Trevor Zhiqing, 楠浩一.  
英文: Ruben Boroschek, Trevor Zhiqing Yeow, Koichi Kusunoki.  
言語 English 
掲載誌/書名
和文:Bulletin of Earthquake EngineeringBulletin of Earthquake Engineering 
英文: 
巻, 号, ページ        
出版年月 2025年3月28日 
出版者
和文: 
英文: 
会議名称
和文: 
英文: 
開催地
和文: 
英文: 
DOI https://doi.org/10.1007/s10518-025-02142-9
アブストラクト While we can only instrument a small number of structures with traditional sensors, cities have become surrounded by video cameras that record response of whatever is in their field of view. In this paper we have a proof of concept that if we combine the seismic response data derived from video cameras with input obtained from accelerometers installed at ground level, the linear and nonlinear response of the structure can be estimated. We present an application and validation in a laboratory environment on a plane steel frame under varying shaking amplitudes. Linear and nonlinear responses are monitored using standard acceleration and displacement sensors and later compared with the displacements derived from videos using standard computer vision techniques. The comparison of computer vision derived displacements with traditional sensor data gives excellent results with differences in maximum values in relative to base displacement time histories of less than 10%. Later, the derived computer vision relative to ground displacements are used in parametric input-output system identification to estimate the evolution of the modal parameters as a function of response amplitude. For this case, the synchronized input from an inertial accelerometer is used. The result from this identification process is nearly identical to the ones obtained using traditional acceleration sensors. To extend its use, double differentiation of the computer vision derived displacement is used to estimate accelerations in a reduced frequency band with practically no difference with acceleration records on the selected band. There are several advantages detected from combining standard sensors and computer vision techniques, like full space definition of possible monitoring points and limiting the distortion of acceleration records due to structural rotations and the possibility to obtain residual displacements. The methodology to obtain reliable displacement is presented together with the determination of varying modal property as a function of shaking intensity and damage level.

©2007 Institute of Science Tokyo All rights reserved.