This study includes an interpretable machine learning (ML) framework for predicting the shear contribution of externally bonded fiber-reinforced polymer (FRP) composites in reinforced concrete beams. A database including total 313 experimental specimens was collected from previous experimental research. The data screening process has been conducted using the Isolation Forest algorithm, resulting in 268 cleaned specimens. The cleaned database was divided into a training subset containing 214 specimens and an independent test set containing 54 specimens. The trained subset was enlarged into 5204 synthetic data using two advanced generative models including Wasserstein generative adversarial network and conditional Variational autoencoder (CVAE). Separate constrained monotonic neural network (CMNN) models were then trained on both datasets and WGAN-based CMNN achieved R2=0.9524 for the synthetic training dataset and R2=0.9120 for the independent test set, whereas the CVAE-based CMNN achieved corresponding values of 0.9632 and 0.9011. To improve practical applicability, response functions were extracted from WGAN-based CMNN and fitted with analytical expressions to derive a closed-form prediction equation. The proposed equation was independently validated using separate unseen test specimens, which were not used in CMNN training and achieved R2 = 0.79, RMSE = 24.98 kN, MAE = 19.65 kN, MAPE = 21.72%, VAF = 79.35%, U95 = ±54.94 kN, SI = 3.04, and PI = 0.11. Compared with ACI 440.2R-17, CSA-S806.12, CNR-DT200 R1.2013, TR-55, and JSCE, the proposed equation showed superior accuracy while maintaining a transparent and design-oriented format.
An interpretable and uncertainty-aware machine-learning framework for estimating the shear capacity of FRCM-strengthened beams enables accurate, transparent, and uncertainty-aware assessment of shear capacity in FRCM-strengthened concrete beams.
Xiangsheng Liu, G. Figueredo, G. Gordon et al.· Journal of composites for co...· 0 citations
A comprehensive data-driven framework integrating ensemble machine learning models with systematic hyperparameter sensitivity analysis and explainable artificial intelligence techniques is proposed, demonstrating that the XGB model significantly outperforms the other approaches, achieving superior accuracy and robust generalization.
Qaim Shah, Waheed Ali Khoso, Fawad Iqbal et al.· Discover Artificial Intellig...· 0 citations
The compressive behavior of carbon fiber-reinforced polymer (CFRP)-confined ultra-high-performance concrete (UHPC) short columns involves nonlinear interactions among concrete properties, confinement characteristics, and specimen geometry. In this study, a database of 144 circular specimens was compiled using nine input variables: specimen diameter (D), height (H), CFRP thickness (t), number of CFRP layers (Layers), unconfined concrete compressive strength (fco′), concrete ultimate strain (εco), CFRP tensile strength (ff), CFRP ultimate strain (εf), and CFRP elastic modulus (Ef). Eight regression algorithms were evaluated using an 80:20 training–test split, with ten-fold cross-validation conducted within the training subset for hyperparameter selection. ANN achieved the best test performance, with an R2 of 0.96, an MAE of 8.37 MPa, and an RMSE of 11.32 MPa, while SVM and CatBoost also showed competitive predictive accuracy. The selected ML models exhibited substantially lower prediction errors than seven existing empirical equations. SHAP analysis further identified CFRP layer number, unconfined concrete strength, specimen height, and CFRP thickness as influential predictors and revealed interactions among confinement, matrix deformability, and specimen geometry. The proposed framework provides an accurate and interpretable supplementary approach for assessing the compressive strength of CFRP-confined UHPC within the parameter range represented by the database.
Large rupture strain fiber-reinforced polymer (LRS-FRP)-confined concretes are increasingly used in safety–critical infrastructure due to their high ductility and load-carrying capacity; however, accurate prediction of compressive strength (CS) in non-circular sections remains challenging due to non-uniform confinement induced by geometric irregularities, which limits the reliability of existing empirical models and design codes developed mainly for circular sections. To address this limitation, this study develops a reliability-oriented, data-driven framework that combines Bayesian-optimized ensemble machine learning, model interpretability, and uncertainty quantification. Six algorithms including random forest (RF), extremely randomized trees (ERT), extreme gradient boosting (XGBoost), histogram based gradient boosting (HistGBM), light gradient boosting machine (LightGBM) and categorical boosting (CatBoost) were trained using an experimental database of 174 non-circular LRS-FRP-confined concrete specimens. Model interpretability was achieved using Shapley additive explanations (SHAP), while predictive reliability was systematically evaluated through uncertainty-aware performance assessment. All models demonstrated strong generalization, with testing coefficients of determination (R
2
) ranging from approximately 0.96–0.99, and boosting-based methods consistently outperforming bagging approaches. CatBoost (testing R
2
≈ 0.985) exhibited the best overall performance, the lowest prediction errors, and the most reliable uncertainty estimates. Accordingly, the overall performance ranking was identified as CatBoost > HistGBM > XGBoost > ERT > LightGBM > RF. The results clearly indicate that high predictive accuracy alone is insufficient for reliable modeling of non-circular LRS-FRP-confined concrete and that uncertainty-aware evaluation is essential. SHAP-based analysis yielded physically consistent insights, identifying LRS-FRP thickness, unconfined concrete strength, and section corner radius as the dominant contributors to CS, while highlighting the critical role of post-transition LRS-FRP stiffness in sustaining effective confinement. Overall, the proposed framework offers an interpretable and reliability-aware alternative to conventional models and provides a robust predictive tool for engineering design and assessment of non-circular LRS-FRP-confined concrete.
This study proposes an experiment-driven machine learning (ML) framework to address experimental limitations in estimating compressive strength within a controlled experimental domain of locally sourced fiber-reinforced concrete (FRC). Concrete mixes incorporating 1, 2, and 3% lathe scrap steel fibers were prepared using a standard 1:2:4 mix with brick chips as coarse aggregate. Due to practical constraints, intermediate fiber contents (e.g., 1.5% and 2.5%) were not experimentally cast. To overcome this limitation, supervised ML models were employed to interpolate compressive strength at fractional fiber dosages, thereby reducing experimental cost and time. A total of 48 experimental observations obtained from laboratory-tested cylindrical specimens were used to develop and evaluate four ML algorithms—Gaussian Process Regression (GPR), Artificial Neural Network (ANN), Support Vector Machine (SVM) and Random Forest (RF)—under a five-fold cross-validation framework. Among these, GPR exhibited the best performance among the evaluated models (R² = 0.884, RMSE = 1.647 MPa), effectively capturing the nonlinear relationship between curing age, fiber content, and compressive strength. SHAP-based interpretability analysis indicated that curing age is the primary governing factor in strength development, while fiber content contributes through secondary nonlinear effects. The results further indicate that moderate fiber inclusion improves compressive strength, whereas excessive fiber content may reduce performance due to workability loss, fiber clustering, and increased matrix heterogeneity. These findings are consistent with established concrete mechanics, where fiber reinforcement plays a more dominant role in tensile behavior than in compressive strength. The proposed framework demonstrates the capability of probabilistic ML models for accurate interpolation within controlled experimental conditions using high-quality laboratory data. This study provides a practical, data-driven approach for strength estimation and mix optimization of low-cost, locally sourced FRC, supporting sustainable construction practices and efficient utilization of industrial waste materials.
