Structural Nonlinearity Extraction from Video Data for Damage Evaluation in Earthquake Events: Experimental Verification

Abstract

A development of a method of structural nonlinearity extraction is introduced for fast evaluation of structural damage conditions in post-earthquake events using the video data that is taken and shared in societies. The video data-based structural analysis has advanced rapidly in recent years due to advantages of non-contact data acquisitions, high spatial resolution in low-cost device, and so on. This study presents the experimental verification conducted by the shaking table tests, and to extract singularities spatial domains due to nonlinearity events using computer vision (CV)- based technology. In the shaking table test, two three-story aluminum frame models, whose modal frequencies are designed to become slightly different, are used to introduce some kinds of structural nonlinearities, such as hitting, boundary condition changes, and residual deformations in some of members. The present study solely concentrates on boundary condition nonlinearity, which is achieved by incorporating a controllable hinge member with a trigger magnet at the base of the columns to simulate the boundary constraint change from fixed end to hinge end during excitation. When the input excitation level is large enough, the magnets detach and cause the boundary constraint to change. A video camera is put in front of the model to capture the whole experiment, and the acquired video data are used for the verification. Optical flow method, as an effective video processing technique, can be used to estimate the real-world object motion between observer and scene by a dense field corresponding to the interframe displacement of each pixel. Farneback optical flow algorithm is selected here to extract the motion information of models; Then, the node strength index is introduced as the feature extraction to detect the nonlinearity events. The change in boundary conditions can significantly alter structural local motion information, leading to a clear mutation in node strength that can serve as an indicator of nonlinearity events. The results in this experimental study show the possibilities of video-based technology for the fast damage condition evaluation in the post-earthquake.