Capture of Crack Evolution for Evaluation of Concrete Properties Using Dynamic Mode Decomposition

Cracks on concrete surface provide the earliest indication in structural damage diagnosis and durability evaluation. Currently, most image-based crack detection methods focus on static image, which is not optimal for tracking the evolution of cracks. To address this shortcoming, a novel crack detection method based on dynamic mode decomposition (DMD) of video images is developed. The DMD spectrum, based on two-dimensional matrix formed by a sequence of video frames, is employed to identify cracks from each video frame according to the distribution of eigenvalues. The method is demonstrated using the flexural test of ultrahigh-performance concrete (UHPC) in the laboratory. The surface cracks identified agree well with those of the original image. The total crack area and its projections on two major directions were computed. The relationship between crack development and force-displacement state was investigated. The crack propagation of UHPC and its effect on mechanical characteristics were revealed from analysis of the captured images. In this crack detection procedure, the cracks of concrete surface do not require any special marks and the video can be recorded using a smartphone. The simplicity of the proposed method and its ability to produce good results make it attractive for practical applications.