INTERNAL DAMAGE DETECTION OF PREFABRICATED RC BEAMS UPON CONVOLUTION NEURAL NETWORK AND DIGITAL IMAGE CORRELATION TECHNIQUE

This paper combined convolution neural network with DIC technique to address the problem of identifying internal damage of prefabricated RC beams. A convolutional neural network (CNN) model is proposed to detect the internal damage, and the displacement cloud image dataset of the damaged concrete beam is generated from ABAQUS secondary development. The displacement contour image is employed as the input of the model, and the excellent image processing ability of CNN is utilized to retrieve the internal damage index of the structure. A 3D IOU algorithm is designed to improve the network performance. The DIC technology was used to capture the displacement contour image of the concrete damaged samples, and the method proposed was tested and verified, which realized the damage identification of RC structures. The results shown that present method can achieve high precision with robustness in damage identification of RC structures, which provides a new idea for the quality inspection of prefabricated components.