Explainable ensemble learning framework for bond strength prediction in 3D printed concrete structures

Abstract

Accurate prediction of bond strength between reinforcement and concrete is critical for ensuring the structural reliability and durability of reinforced elements, particularly in emerging construction technologies such as three-dimensional concrete printing (3DCP). Traditional empirical and semi-empirical bond models are often limited by simplified assumptions and insufficient ability to capture complex nonlinear interactions among geometric, material, and reinforcement-related parameters. To address these limitations, this study proposes a comprehensive data-driven framework integrating ensemble machine learning models with systematic hyperparameter sensitivity analysis and explainable artificial intelligence techniques. An extensive experimental database comprising 550 samples and nine influential input variables was compiled and analysed. Random forest (RF), extreme gradient boosting (XGB), and AdaBoost (ADB) models were developed and rigorously optimized using multiple performance metrics, including RMSE, MAE, R², and CVRMSE. The results demonstrate that the XGB model significantly outperforms the other approaches, achieving superior accuracy and robust generalization. Residual diagnostics confirm unbiased predictions and stable error distributions. Furthermore, SHAP and partial dependence analyses provide transparent insights into the dominant influence of geometric ratios and reinforcement characteristics on bond strength. Finally, the optimal model was embedded into a user-friendly graphical interface to support practical engineering decision-making. The proposed framework offers an accurate, interpretable, and deployable solution for bond strength prediction in modern concrete construction.