Ensemble Machine Learning for Predicting Flexural Capacity of FRCM- Strengthened RC Beams

Fabric reinforced cementitious matrix (FRCM) composites are progressively applied to strengthen reinforced concrete (RC) members because they offer improved compatibility, durability and fire resistance compared with epoxy-bonded fibre reinforced polymer (FRP) systems. However, predicting the ultimate flexural capacity of FRCM-strengthened beams remains challenging due to the complex interaction of material, geometric and loading parameters. Here, we develop and evaluate ensemble machine learning (ML) models to predict the ultimate load-carrying capacity of FRCM-strengthened RC beams. A curated database of 244 experimental tests from the literature was compiled, comprising 16 input features describing concrete compressive strength, reinforcement ratios, FRCM properties and specimen geometry. After systematic preprocessing and feature encoding, several ensemble regressors were benchmarked and three models Extra Trees, XGBoost and AdaBoost were analysed by 5 fold cross-validation with grid search for hyperparameter optimization. Among them, the Extra Trees regressor achieved the best performance, with a coefficient of determination R² = 0.898, root mean square error RMSE = 21.5 kN and mean absolute error MAE = 11.7 kN. Feature-importance analysis and SHAP-based explainability consistently identified concrete compressive strength, reinforcement ratio and cracking load as dominant predictors, in agreement with structural mechanics. These results demonstrate that ensemble ML models provide a reliable and interpretable tool for assessing the flexural capacity of FRCM-strengthened RC beams, also highlight the potential of data-driven methods to complement empirical and code-based design approaches in structural strengthening.