Hybrid XGBoost–LSTM Model for Structural Health Monitoring of Reinforced Concrete Beams

Structural Health Monitoring (SHM) of reinforced concrete (RC) beams is critical for ensuring the safety and longevity of civil infrastructure. Conventional SHM approaches often rely on manual inspection or standalone machine learning and deep learning models, which are limited in capturing nonlinear damage characteristics and temporal degradation patterns under varying loading conditions. To address these limitations, this paper proposes a hybrid XGBoost– LSTM model that integrates gradient boosting–based feature learning with long short-term memory–based temporal sequence modeling for effective damage detection and severity assessment of RC beams. Initially, damage-sensitive features are extracted from sensor-based structural response data in both time and frequency domains. XGBoost is employed to perform nonlinear feature selection and preliminary damage estimation, enabling the identification of the most influential structural parameters. The selected feature sequences are then fed into an LSTM network to model the time-dependent evolution of structural damage. The proposed hybrid framework is evaluated using multiple performance metrics and compared against conventional machine learning and deep learning models, including support vector machines, random forest, standalone XGBoost, and LSTM. Experimental results demonstrate that the hybrid XGBoost–LSTM model achieves superior accuracy, robustness under noisy conditions, and improved damage severity prediction, with performance gains of up to 10–15% over baseline models. The findings confirm that the proposed approach provides a reliable and scalable solution for intelligent SHM of RC beams, supporting the development of data-driven, real-time infrastructure monitoring systems.