Shear strength assessment of FRP-reinforced concrete beams without shear reinforcement: database-driven approach and prediction models Available to Purchase

Fibre-reinforced polymer (FRP) reinforcement is increasingly used as a corrosion-resistant alternative to traditional steel in concrete structures. However, accurately predicting the shear strength of FRP-reinforced concrete (FRP-RC) beams, particularly those without shear reinforcement, remains challenging owing to the involvement of multiple design variables. Existing models often rely on semi-empirical approaches, which can produce inconsistent results. An approach based on machine learning (ML) to enhance the accuracy of shear strength prediction for slender FRP-RC beams is proposed. Eight ML models were evaluated – five ensemble methods (AdaBoost, XGBoost, CatBoost, LightGBM and random forest) and three single models (K-nearest neighbours, linear regression and ridge regression). An extensive database of FRP-RC beam specimens without shear reinforcement from the literature was used for training and validation. Model inputs were based on parameters used in existing design methods. Among the models, CatBoost showed the best performance, achieving a mean value of 1.00, a coefficient of variation of 0.079, a coefficient of determination of 0.994 and root mean square error of 0.0031. A parametric study using CatBoost further explored the influence of key variables. Based on the findings, a new multiplicative design equation was developed to provide a practical and reliable tool for predicting the shear strength of FRP-RC beams in structural design.