Evaluation of crack self-healing performance in cement-based materials using deep learning and multidimensional characterization

Cracks are a persistent issue in cement-based materials, especially in corrosive environments, such as marine conditions, where aggressive ions exacerbate the problem. To address this challenge, a novel self-healing method, which combines layered double hydroxides (LDHs) with microbial mineralization, has been proposed, with its effectiveness comprehensively evaluated. A new core-shell healing agent was developed, incorporating calcium-based and aluminum-based inorganic minerals as the core along with microbial spores. The concept of deep learning was introduced to quantitatively assess the healing performance. A dense residual network (DRN) and a multidimensional crack evaluation method were applied to quantify the crack repair effects, revealing considerable improvements in all performance indices after a certain curing period. In addition, the principles of various crack characterization methods were analyzed and classified. Results indicate that these methods may reflect the surface repair effects, internal repair effects, or a combination of both. Furthermore, repair products were microscopically analyzed to identify their composition, morphology, elemental distribution, and pore size distribution. The self-healing mechanism was elucidated as the in situ LDH formation from the healing agents, which expanded to fill the cracks while immobilizing chloride ions, sulfate ions, and water molecules, thereby mitigating the damage from aggressive ions. The LDH formation also consumed a substantial amount of hydroxide ions, converting the solution into a mildly alkaline environment that favored microbial growth and mineralization, further enhancing and consolidating the healing effect. This study provides a scientific basis for crack repair in corrosive environments and for the multidimensional characterization of crack healing performance.