This study aims to explore the impacts of waste glass sand (WGS) replacement ratios and bond lengths on the bond performance and interfacial bond strain evolution of rebar-glass sand concrete, addressing the lack of accurate strain prediction in existing research to advance the engineering application of this material.
Specimens with WGS replacement ratios (0%, 10%, 30% and 50%) and bond lengths (3d = 36mm, 5d = 60mm) were tested. Distributed fiber optic sensors (DFOS) enabled high-precision strain monitoring, and a Time-Frequency Deformable Fusion Neural Network (FreqDeformNet) was developed to analyze nonlinear strain sequences.
The results indicate that under a bond length of 3 times the rebar diameter, a 30% glass sand replacement ratio and under a bond length of 5 times the rebar diameter, a 10% glass sand replacement ratio, exhibit the optimal bond performance. Compared to the 3d bond length, the strain variation in the 5d bond length demonstrates a more pronounced shear lag phenomenon. FreqDeformNet outperforms traditional models, with the mean absolute error reduced by 17.4%, 92% and 77%, respectively, and the highest average coefficient of determination.
The integration of DFOS and FreqDeformNet can reveal the bond performance of rebar-glass sand concrete under different conditions, address the issues of cross-variable dynamic interaction and long-term dependencies in strain data and provide a high-precision prediction method for bond strain.
