| Bridge bearings are important member for connecting superstructure and substructure,which running condition is related to the safety of bridge.Laminated rubber bearings are widely used in small and medium span highway bridges with the advantages of simple structure,easy manufacture,less steel consumption,low cost and convenient installation,etc.However,the bearing damage is also quite common because of environmental erosion,material aging and post operation and other factors.Once bearings voided,the stress of structure will be changed,further causing girder damage and reducing the service life of the bridge.The width span ratio of many new bridges are large,some even close to 1.So a large number of low order two-dimensional modes appear.When detecting the bearing damage,if only considering the longitudinal axis of the beam,it will inevitably lead to a larger error,or even a miscarriage.Therefore,the traditional damage detection method has some limitations for the detection of the bridge bearings.It is difficult to accurately identify the location and extent of the damage of bearings.The paper uses two-dimensional model to solve the problem of bearing damage of highway bridge.Large horizontal connection stiffness beam is simplified as a plate.From frequnencies and mode shapes of the plate,emphasizing variation rates of natural frequency and Gauss curvature mode correlation coefficients.After comparative analysis,Gauss curvature mode correlation coefficient is determined as the damage identification index for the location of damaged bearings.On the basement,neural network is used to identify the damage severity.Thus,a method based on Gauss curvature mode correlation coefficient and BP neural network is proposed to detect the bearing damage of highway bridge.Firstly,the indoor experiment was carried out,where using the thin plate to simulate simply supported bridge.Then the finite element model was established.Weather the method was suitable for the damage detection of bridge bearings,which was explored by comparatively analysing theoretical and experimental data.Eventally,the results of testing data were proved that the method was applicable.Secondly,taking the actual project as an example,the finite element model of continuous bridge was established for numerical analysis,and the modal test was carried out,which is beneficial to the further research on the quantitative analysis of the damage of the continuous bridge.Both theoretical and experimental results indicate that the method is feasible and effective for bridge bearing damage detection.Gauss curvature mode correlation coefficients are obviously different between intact and damaged bearings,further achieving the purpose of damage location.And the Gauss curvature mode correlation coefficient decreases with increasing damage severity.BP neural network is used to identify the damage severity of bearings effectively. |
