Holographic Dynamic Displacement Monitoring And Damage Detection Of Bridges By Fusing Sparse-Point Acceleration Measurements

The Structural Health Monitoring System(SHMS)provides increasingly crucial technological and data support for the safe operation and scientific management of bridges.However,methods for collecting dynamic holographic data from bridges and the theory for diagnosing their safety status urgently need enrichment and improvement.In this study,with the support of the National Natural Science Foundation of China,as well as the Shenzhen Science and Technology Program,the research of structural responses monitoring,and damage detection based on data fusion of computer vision measurements and sparse accelerations has been studied.The main research tasks and fruits are as follows:1.Theoretical expressions for the image correlation transfer rules based on Zero Mean Normalized Cross-Correlation(ZNCC)have been derived.A down sampling template matching algorithm was proposed to expedite target tracking calculations by analyzing the distribution pattern of upper and lower bounds on local correlation coefficients in images.Furthermore,a subpixel refinement method,based on Gaussian surface fitting of correlation coefficients,has been introduced to augment the precision of the target tracking algorithm.The sources of measurement errors in vision displacement measurement methods,encompassing camera motion,camera calibration,and hardware limitations,have been thoroughly investigated.Corresponding mathematical models and calculation formulas have been established,alongside proposals for mitigating measurement errors.Displacement monitoring tests have been conducted on a suspension bridge model and the Baishatuo Yangtze River Bridge,applying the proposed method to monitor both dynamic and static displacements and identify low-order frequencies.The research outcomes illustrate that the proposed down sampling algorithm reduces computation time by approximately 80% compared to traditional exhaustive search methods.Moreover,the dynamic displacement measurement error for the model bridge remains within 2%,signifying an algorithmic enhancement in displacement measurement efficiency and accuracy.2.A data fusion method that utilizes complementary filtering of vision-based displacements and accelerations was proposed to reconstruct accurate displacements by leveraging the correlation mechanism of the two signals in both the time and frequency domains.Specifically,the fusion displacements comprise quasi-static and dynamic components provided by the vision-based displacements and accelerations,respectively.The parameters such as target accuracy,time-window length,and sampling frequency were theoretically analyzed to understand their impacts on the accuracy of the fusion displacements.The visionbased displacements and accelerations within the same frequency range are employed to automatically determine the scaling factor.Moreover,the change rule of scaling factors in the image coordinates were investigated for full-field scaling factor calculations,which overcomes the complex operation problem of precise measurement of structural dimensions and distance measurement in existing calibration methods.The measurement error expressions resulting from translational vibrations of the camera was derived and mitigated by measuring the camera’s accelerations.The research demonstrates that the proposed data fusion method surpasses conventional computer vision approaches,delivering high-precision,wide-bandwidth,and low-noise dynamic displacements at measurement points.3.The dynamic mechanics model of the multi-span beams with arbitrary boundary conditions under a moving load were established,and the structural responses and their physical meaning were derived and analyzed,revealing that the forced vibration response components of displacement modal coordinates contain the corresponding mode shapes.Then a method was proposed to extract multi-order spatial high-resolution mode shapes by separating the forced vibration components from the limited fusion displacements.The spatial resolution can be calculated by the sampling frequency of accelerations and the speed of the moving load,while the orders depend on the number of measurement points.Additionally,the analytical solutions of the mode shape functions for beams with arbitrary boundary conditions were derived and a method to solve unknown parameters in the functions by fitting visionbased displacements and accelerations was proposed.After obtaining high-resolution spatial mode shapes,the full-field dynamic displacements were reconstructed using modal superposition.Research results demonstrate that the proposed method can extract spatially dense mode shapes even if only displacements at the mid-span are measured,and the reconstructed full-field displacement error decreases as the number of fused displacement measurement points increases and the moving load velocity decreases.4.According to the principles of modal superposition and modal perturbation,the expressions for modal shapes and moving load-induced displacements after reduction in local stiffness of arbitrary boundary beams are derived.The associated mechanisms between the reduction in local stiffness of structures and their dynamic characteristics and dynamic responses are revealed.The analysis indicates that multi-order high-resolution modal shapes extracted using the response separation method can be utilized to characterize the true state of structures and identify structural damage.Based on the characteristics of the full-field modal shapes,a damage identification index(WEADR)based on the weighted fusion of area differences of multi-order modal curvatures before and after damage is proposed.Moreover,a damage identification index(RCC)based on changes in the cross-correlation function of fullfield displacement responses before and after damage is proposed.Numerical analysis results demonstrate that in the case of a limited number of measurement points,the peak coordinates of the two types of damage identification indices can accurately locate structural damages.5.Experiments were conducted on a simply supported beam model bridge and a scaled model of a suspension bridge in the laboratory,a 2×40 m continuous composite beam bridge and the Dongshuimen Yangtze River Bridge in Chongqing,China.The experiments involved the use of external cameras and onboard accelerometers to collect structural responses.Analytical studies were carried out on the proposed methods for displacement monitoring based on data fusion,mode shape extraction,and full-field displacement reconstruction.The feasibility and effectiveness of the proposed methods in acquiring the full-field dynamic displacement responses of structures under different types and scales of bridges and different loading conditions were verified.For the laboratory model beam,dynamic data collection was conducted under multiple damage conditions,and damage detection of the model beam was achieved based on two types of damage identification indices proposed herein.