Predictive insights into the key determinants of interlayer bond strength in self‐consolidating concrete: A focus on thixotropy, surface roughness, and packing density

In this study, the factors influencing the bond strength of self‐consolidating concrete (SCC) mixtures were investigated using different aggregate gradations, water‐to‐cement ratios (W/C), and cement replacement materials (zeolite and pumice powders). The bond strength of concrete samples was evaluated at delay times of 15, 30, 60, and 100 min using two methods: shear strength test and a pull‐off test (bond strength). The effects of rheological parameters, workability, paste volume, roughness, and packing density on bond strength were examined. The results indicate that the average reduction in strength in the pull‐off test is 6.5% higher than that in the shear strength test. Additionally, the results show that at shorter casting delays (15 and 30 min), thixotropy exhibited the strongest correlation (inverse) with interlayer bond strength among all evaluated parameters. The R 2 values for the 15‐min delay were 0.81 (pull‐off test) and 0.84 (shear test), while for the 30 min delay, they were 0.64 and 0.85, respectively. At longer casting delays (60 and 100 min), surface roughness demonstrated the strongest direct correlation with bond strength. For the 60 min delay, the R 2 values were 0.93 (shear test) and 0.86 (pull‐off test), while at the 100 min delay, they were 0.87 and 0.81, respectively. At shorter delay times (15 and 30 min), a relatively strong correlation was observed between paste volume and bond strength. The R 2 values at a 15‐min delay were 0.71 (pull‐off test) and 0.53 (shear test). However, the influence of paste volume significantly decreased at delay times greater than 30 min. Moreover, viscosity is closely related to thixotropy and roughness, and its impact is reflected in these two parameters, hence not considered independently in bond strength calculations. Finally, time‐dependent multiple linear regression models were developed using surface roughness, paste volume, and thixotropy as predictors. These models achieved high predictive accuracy, with R 2 values ranging from 0.96 to 0.99.