Bridge Damage Detection Based On Moving Principal Component Analysis Combining With Transfer Entropy

The bridge structure is usually affected by various factors in the environment for a long time,which will lead to aging of the bridge materials,and then the declines of structural performance and the bearing capacity.The normal operation of the traffic will also be affected,therefore the health monitoring of bridge structure is particularly important.At present,some achievements has been made in structural health monitoring(SHM),but an urgent challenge that the high cost caused by a large number of sensors installed in the SHM system should be solved.Addressing this problem,a damage detection method based on moving principal component analysis(MPCA)combining with transfer entropy is proposed in this thesis.The main works of this thesis are as follows:(1)The significance of SHM in civil engineering is elaborated.The identification methods of structural damage detection are summarized,as well as the research condition and the current challenges of bridge health monitoring.(2)The theory of vehicle-bridge coupling vibration is described.A jointed identification method of bridge damage based on principal component analysis and transfer entropy is proposed to deal with the problem of massive data.In this thesis,a simple supported beam bridge subjected to a moving load is taken as an example.Firstly,the installed sensors are divided into two equal parts considering the midspan of the bridge as the boundary position,forming two data matrixs from the two part datasets.A moving window is defined to cut the two data matrixes synchronously,and the datasets in the window are analyzed by the method of PCA respectively.Moving the window along the time axis,two time series of the first eigenvalue are obtained.Then the transfer entropy of the two first eigenvalue sequences is calculated by the windowed transfer entropy method,and a new damage index,bridge damage detection based on principal component analysis combining with transfer entropy is proposed,When the moving vehicle passes through the damage location,the damage index will have mutation,which is used to locate the damage.At last,an theory of unilateral upper limit threshold is introduced to enhance the robustness of the method.(3)Numerical simulation: using ANSYS software,a simple beam bridge model is established with the element type of Plane 42.The dynamic responses of the beam bridge with different damage position and different vehicle speed are calculated using the Transient Analysis module of the software.The acceleration signal of each observation point on the beam bridge is collected and the damage index is calculated.The results show that this method can accurately locate the damage.(4)Experimental verification: a simply supported hollow rectangular steel beam model with the length of a 6m is used for experimental verification.The results show that the method can effectively locate the damage of the bridge model at different vehicle speeds,which verifies the effectiveness and reliability of the method.