Reconstruction of meso-scale FEM based on XCT images for 2D composites

For establishing the three-dimension finite element model (FEM) of plain woven composites with real meso-structures, the histogram of oriented gradient (HOG) was used to extract the features of XCT sub-images with crossed or uncrossed warp and weft yarns for training support vector machine (SVM). Then the generated weight file and sliding window located the cross position of two different types of yarn in the entire XCT image. On this basis, the corresponding line segments of the warp and weft yarns was extracted by Hough transform respectively. According to the principle of straight line intersection, the coordinates of the four intersection vertexes of the warp and weft were calculated. Combined with the spatial distribution characteristics of two kinds of yarn of plain woven structure, the meso-structure identification was realized with fewer user involvement.