Genetic evolutionary deep learning for fire resistance analysis in fibre-reinforced polymers strengthened reinforced concrete beams

Created by W.Langdon from gp-bibliography.bib Revision:1.8414

@Article{Wang:2025:engfailanal,

• author = "Songbo Wang and Yanchen Fu and Sifan Ban and Zhuo Duan and Jun Su",
• title = "Genetic evolutionary deep learning for fire resistance analysis in fibre-reinforced polymers strengthened reinforced concrete beams",
• journal = "Engineering Failure Analysis",
• year = "2025",
• volume = "169",
• pages = "109149",
• keywords = "genetic algorithms, genetic programming, Fire resistance, Strengthened reinforced concrete beams, Light Gradient-Boosting Machine, SHapley Additive exPlanations, ANN",
• ISSN = "1350-6307",
• URL = "https://www.sciencedirect.com/science/article/pii/S1350630724011956",
• DOI = "doi:10.1016/j.engfailanal.2024.109149",
• abstract = "Fibre Reinforced Polymers (FRP) have become increasingly popular for strengthening reinforced concrete (RC) structures due to their structural benefits. However, a major concern with FRP-strengthened concrete members is their poor fire resistance. This study introduces a genetic evolutionary deep learning (DL) approach that uses the Light Gradient-Boosting Machine (LightGBM) algorithm, enhanced with a Genetic Algorithm for hyperparameter optimisation, alongside Genetic Programming (GP) to assess the fire resistance performance of strengthened RC beams. A substantial dataset comprising 20,000 data points, derived from numerically modelled results validated through experimental studies, underpins the data-driven DL analyses. The LightGBM model demonstrates high predictive accuracy for fire resistance time and deflection at failure of the FRP-strengthened RC beams, with coefficient of determination (R2) values of 0.923 and 0.789, respectively. Although the GP model shows lower accuracy (R2 values of 0.642 and 0.643), it provides explicit equations that facilitate a deeper understanding of the DL model and ease of application. A graphical user interface software, incorporating these two DL models, has been developed to enable engineers to apply these insights in practice without requiring coding skills. Furthermore, an assessment of feature influences was conducted, visually depicting their impact on the output results, thus enhancing interpretability for engineering applications",

}

Genetic Programming entries for Songbo Wang Yanchen Fu Sifan Ban Zhuo Duan Jun Su

Citations