Artificial-Intelligence-Driven Prediction of Load-Carrying Capacity of ECC-Strengthened Reinforced Concrete Beams Using Dense Learning Machine

Abstract

Ensuring the structural performance of reinforced concrete (RC) beams strengthened with engineered cementitious composites (ECC) demands a reliable prediction of their load-carrying capacity (LC). This task is complicated by the nonlinear interactions among material properties, geometry, and applied loads. This study introduces a rigorously validated dense neural network (DNN) model tailored to accurately predict the LC of ECC-strengthened RC beams, and provides a powerful data-driven tool for advanced structural design. A comprehensive database was assembled from published experimental programs and high-fidelity numerical simulations, which includes diverse beam geometries, reinforcement ratios, and ECC layer configurations. Among a suite of machine-learning techniques, the optimized DNN achieved superior predictive performance (R² = 0.975, MAE = 14.544, RMSE = 18.190), which outperforms linear, tree-based, and other nonlinear models. Sensitivity analysis revealed beam depth and ECC tensile strength as dominant drivers of LC, while ECC layer thickness exerted a comparatively minor influence. Importantly, the inherent strain-hardening capacity of ECC was shown to markedly enhance ductility, energy dissipation, and seismic resilience. These findings highlight the potential of artificial-intelligence-based approaches to restructure the design of ECC-strengthened RC beams, inform performance-based seismic design, and guide the next generation of robust, high-performance concrete infrastructure.