Damage Identification of Corroded Reinforced Concrete Beams Based on SSA-ELM

Accurately quantifying corrosion damage in reinforced concrete (RC) beams is a significant challenge for structural health monitoring. This study introduces a novel damage identification method that integrates the Sparrow Search Algorithm (SSA)-optimized Extreme Learning Machine (ELM) to address this issue. By utilizing dynamic characteristics, including natural frequencies and mode shapes, as input features, the model predicts three critical damage indicators: the mass corrosion ratio (ηs), flexural capacity reduction factor (α), and flexural stiffness reduction factor (β). Validation through ABAQUS finite element simulations demonstrated the superior performance of the SSA-ELM approach compared to conventional ELM, achieving a 60–70% reduction in mean square error (MSE). Specifically, the MSE for ηs decreased from 2.1062 to 0.3174. The experimental validation conducted on seven RC beams with corrosion levels ranging from 0% to 14.1% confirmed the method’s reliability, with prediction errors for α and β ranging from 5 to 10%. This represents a 50% improvement in accuracy compared to conventional ELM, which exhibited errors in the range of 9–20%. SSA-ELM is a novel and more effective solution to the challenges (e.g., early convergence and convergence speed) faced by existing optimized ELM methods (especially GWO-ELM and GA-ELM). Furthermore, the practical implementation of the proposed framework includes a MATLAB R2024a-based graphical user interface (GUI) with Docker containerization, enabling efficient field deployment for structural assessment. Overall, this study establishes SSA-ELM as a promising tool for post-corrosion safety evaluation of RC structures.