A SHAP algorithm-based prediction of the interlaminar shear strength degradation of G/BFRP bars embedded in concrete exposed to marine environment

Interlaminar shear strength (ILSS) degradation of fiber-reinforced polymer (FRP) bars embedded in concrete reflects the service performance of FRP reinforced-concrete structures. However, predicting the ILSS of FRP bars, particularly under coupled conditions, remains a significant challenge. To address this issue, a dataset collected from published literature was compiled and SHAP algorithm-based method was employed to predict the ILSS retention (ILSSR). Firstly, the effects of FRP parameters (e.g., fiber/matrix type and fiber volume fraction), concrete properties (e.g., cover thickness and concrete alkalinity) and exposure environments on the ILSS degradation of G/BFRP bars embedded in concrete were systematically investigated using SHAP algorithm, further unveiling the dominant degradation mechanisms of FRP bars under marine environments. Based on SHAP analysis, an ILSSR prediction formula was developed, integrating all experimental data by considering both design and exposure factors. This formula offers superior accuracy (R 2 =0.89) than traditional regression methods (R 2 =0.64) while potentially reducing the need of repetitive testing for further predictions. Additionally, the proposed formula was used to evaluate the environmental reduction coefficient (ERC) of ILSS concerning bar diameter and concrete alkalinity, with the estimated ERC values ranging from 0.548 to 0.836. These findings provide insights for the selection of FRP composites and long-term durability evaluation of FRP bar-concrete structures in marine engineering. • Application of SHAP-based framework in predicting ILSS degradation of FRP bars. • Cross-validation of degradation mechanisms from a data-driven perspective. • Evaluation of environmental reduction coefficients concerning bar diameter and concrete alkalinity.