Experimental and Regression Modeling of Short-Term Flexural Behavior of Steel- and GFRP-Reinforced Early-Age Concrete Beams

To address the problem of corrosion, glass fiber-reinforced polymer (GFRP) bars have been introduced as a viable alternative to conventional steel reinforcement in concrete structures. While extensive research has been conducted on the flexural behavior of RC beams reinforced with steel and GFRP bars over both normal-term and long-term periods, studies focusing on fresh concrete beams are almost non-existent. Consequently, this research investigates the impact of steel and GFRP longitudinal reinforcement, as well as the influence of varying concrete compressive strengths, on the flexural behavior of RC beams. The study employs 3-point bending experiments and machine learning (ML) predictive analyses. Specifically, the short-term (fresh) concrete reinforcement compatibility and the effects of steel and GFRP bar reinforcements on beam flexural behavior were examined across three concrete compressive strength categories: low (C25), moderate (C35), and high (C50). A notable contribution of this research is the application of different ML regression models, utilizing Python’s library, for deflection prediction of RC beams. The failure mechanisms of the beams under static loading conditions were analyzed, revealing that composite bar RC beams failed through flexural cracking and demonstrated ductile behavior, whereas steel bar RC beams exhibited brittle failure characterized by shear cracks and sudden failure modes. The ML regression models successfully predicted the deflection values of RC beams under ultimate loads, achieving an average accuracy of 91.3%, which was deemed highly satisfactory. Among the 18 beams tested, the highest ultimate load was obtained for the SC50-1 beam at 87.46 kN. In contrast, while the steel-reinforced beams exhibited higher load-bearing capacities, it was observed that the GFRP-reinforced beams showed greater deflection and ductility, particularly in beams with low and medium concrete strengths. Based on these findings, it is recommended that the Gradient Boosting Regressor, an AI regression model, be utilized to guide researchers in evaluating the load-carrying and bending capacity of structural beam elements.