Vibration Measurement And Damage Identification Of Civil Structure By Computer Vision Technology

In recent years,with the vigorous development of social economy,science and technology in China,the structural systems of building structures and bridges are becoming more and more complicated.Due to the damage of some key elements,some structures may collapse.This will cause a negative impact on the national economy and life safety of some people.Therefore,the importance of structural health monitoring and structural damage identification is highlighted.The development of computer vision technology provides a new idea for the dynamic response measurement of building structures.Compared with traditional measuring instruments,the displacement measurement by computer vision technology does not require the installation and maintenance of expensive sensors.The theory and implementation procedures of this new technology in the dynamic response measurement have been successfully studied in these years,however its application in the quantitative detection on the location and intensity of structural damage of building structures are less involved.Therefore,the study of vibration measurement and damage identification of civil structure by computer vision technology is proposed in this thesis by combining computer vision technology with modal parameter identification,damage location and damage quantification.The main contents of this thesis are as following :(1)The framework of displacement measurement by computer vision technology is first introduced systematically introduced.Based on this framework,a computer vision-based dynamic displacement detection algorithm is proposed by combining computational vision technology with Harris angle detection algorithm,KLT feature tracking algorithm and MLESAC algorithm.In order to verify the effectiveness of the proposed methods,the displacement measurement of a steel truss model with a span of 5.6 meter was conducted.The experimental results show that the method can obtain the structural vibration response with very accurate precision.(2)Two modal parameter identification methods,which are data-driven SSI/data and Nex T/ERA methods,are compared and analyzed for the measurement displacement data of the steel truss.The differences in the procedure to obtain the system matrix A and the observation matrix C are compared between these two algorithms.By the implementation of the two algorithms,the series of video data for measured displacement of the steel truss are adopted to identify its dynamic modal parameters.The identification results show that the data-driven SSI/data method is superior to the Nex T/ERA algorithm in identifying the natural frequency and vibration modal shape.(3)Combining the data-driven SSI/data algorithm with the Stochastic Damage Localization Vectors(SDLV)method,the procedure for the detection of damage location of building structure by computer vision technology is proposed in this thesis.In order to evaluate the effectiveness of the proposed method,the series video data of dynamic displacement for the 5.6 m steel truss model are taken as an example to investigate the damage location of structural elements.The analyzed results show that the method can effectively identify the damage location of the structure at a reasonable accuracy level.(4)Based on the information of identified structural damage location by the SDLV method,two optimization objective functions are proposed to identify the intensity of damaged structure with one or multiple damaged elements.Taking the previous steel truss as an example,the estimation of the extent in the damage element(s)under various working cases are investigated.The identification results show that the damage-intensity detection method based on the two optimization objective functions can effectively identify the damage extent of damage elements.However the identified results based on the second optimized objective function(with the combined information of the natural frequencies and modal shapes)are more accurate than those results based on the first optimized objective function(only with the information of the natural frequencies).