Geometric characteristics identification for concrete interfacial transition zone using efficient group enhanced residual UNet

The interfacial transition zone (ITZ), as the weakest region in concrete, requires precise characterization for understanding both the mechanical properties and aggressive particle transport behaviour of concrete materials. The ITZ in concrete exhibits significant heterogeneity, which leads conventional small-scale testing methods such as microhardness measurement inadequate for accurately characterizing its geometric properties. This study proposes the Efficient Group Enhanced Residual UNet (EGER-UNet) to identify ITZ geometric characteristics, using 3D scanned images. The proposed EGER-UNet integrates Global Mixing Module (GMM), Group Aggregation Bridge module (GAB), and Residual Block (RB). The GMM is employed to extract the striped characteristics of ITZ. The GAB enhances ITZ identification accuracy by grouping and fusing multi-scale information, including high-level features, low-level features, and prediction masks. The RB increases network depth while ensuring rapid convergence of the proposed model. The identification results of ITZ were consistent with those obtained by the microhardness method. The recognition speed of proposed method was up to 2000 times faster than that of the microhardness method, promoting the research on mechanical property and durability of concrete.