Predictive modeling of RC beam shear strength using machine learning across global design standards
摘要
Shear resistance in Reinforced Concrete (RC) beams is contributed by both shear reinforcement and concrete. This study examines the shear strength contribution of concrete in RC beams, comparing various national codes. Six RC beams of M35 grade concrete, with dimensions of 150 mm width, 300 mm depth, and 2000 mm length, were tested under three-point loading. Supports were placed at an effective span of 1700 mm, with varying longitudinal reinforcement: RCB1 to RCB3 were conventional mixes, while RCB4 to RCB6 consisted of 20% Metakaolin, 80% OPC, and 1.5% glass fiber. The experimental shear strengths were determined and compared with theoretical values from IS 456:2000, Eurocode 2, and ACI 318-19. The comparison revealed that the shear strength of concrete invades its shear resistance according to IS 456:2000. Codes like ACI 318-19 and Eurocode 2 supplied nearer approximations to experimental results, with Eurocode 2 being the most precise, observed via ACI 318-19 and IS 456:2000. Machine learning models, particularly Random Forest, effectively validated and predicted the enhanced shear performance of metakaolin and glass fiber-modified RC beams, aligning closely with experimental results. The study concludes that IS 456:2000 underestimates shear strength, suggesting the need for modifications to diminish reliance on shear reinforcement. These findings oblige structural engineers and researchers in deciding on appropriate shear strength values for secure and low-cost RC beam layout.