Research On Damage Identification Of Steel-Concrete Composite Beams Based On Abnormal Deformation

The research group are conducting the NSFC Project named Research on the method of obtaining the bridge deformation and deducing the structural state based on the differential analysis of image contour(51778094).Our team carried out a series of static load experiments on steel-concrete composite beam with initial damage.This paper is dependent on the NSFC Project and is targeted to deducing structure damage identification.Our major work include:1.We applied step load on a seven meter –span steel truss-concrete composite test beam with multiple initial damage.Meanwhile we gathered the deformation data by traditional displacement gauge,fixed spot camera and vehicular camera which can obtain holographic image information should be emphasize2.To get rid of the limit of camera angle,we imported image-mosaic method to establish overall model with a certain sum of local HR beam images form vehicular camera.We used new technology which combined by two image-mosaic methods based on the gray matrix and improved SIFT algorithm and it is proved well-work in intergrating local image into overall beam photograph.3.Referring to the related content of mathematical morphology and the image informatics,we promoted the edge-profile superposition algorithm of hologram image.This algorithm was used to analyze the holographic profile of the test beam under different load conditions.And then the holographic displacement curves of the test beam were obtained.Comparing the displacement data deduced from the holographic displacement curves and measured directly from deformation gauge,the validity of new data acquisition technology can be assessed.4.The artificial neural network is used to identify the structure damage.The beam abnormal deformation is defined as the load displacement under damaged state,then we determine the holographic displacement curve.Next the deformation characteristics of the nondestructive structure and the structural holographic deformation curve under the damage conditions are characterized as damage feature.Finally we turn the problem of inverse analysis into a neural network mapping discriminant.5.Training neural network with the specimens of the deformation curves of the test beam under static load with no initial damage and multiple damage,the relatively-reliable neural network mapping discriminant can be achieved with the network advantage of generalization ability with itself.Then we input non-training samples displacement curve into the neural network,and the damage location and damage degree can be concluded at the network output end.