Deep Learning-Assisted Analysis of GO-Reinforcing Effects on the Interfacial Transition Zone of CWRB

Understanding the enhancing mechanisms of graphene oxide (GO) on the pore structure characteristics in the interfacial transition zone (ITZ) plays a crucial role in cemented waste rock backfill (CWRB) nanoreinforcement. In the present work, an innovative method based on metal intrusion techniques, backscattered electron (BSE) images, and deep learning is proposed to analyze the micro/nanoscale characteristics of microstructures in the GO-enhanced ITZ. The results showed that the addition of GO reduced the interpore connectivity and the porosity at different pore throats by 53.5-53.8%. GO promotes hydration reaction in the ITZ region; reduces pore circularity, solidity, and aspect ratio; enhances the mechanical strength of CWRB; and reduces transport performance to form a dense microstructure in the ITZ. Deep learning-based analyses were then proposed to classify and recognize BSE image features, with a high average recognition accuracy of 95.8%. After that, the deep Taylor decomposition (DTD) algorithm successfully located the enhanced features of graphene oxide modification in the ITZ. The calculation and verification of the typical pore optimization area of the location show that the optimization efficiency reaches 9.6-9.8%. This study not only demonstrated the deepening of the enhancement effect of GO on the pore structure in cement composites and provided new insights for the structural modification application of GO but also revealed the application prospect of GO in the strengthening of CWRB composites and solid waste recycling.