Developing Machine Learning Models for Identifying the Failure Potential of Fire-Exposed FRP-Strengthened Concrete Beams

The resilience of fiber-reinforced polymer (FRP)-strengthened concrete beams under fire exposure is a critical aspect of structural engineering, with significant implications for safety and design. Despite extensive research on the fire resistance of concrete beams, there is a lack of studies on identifying the failure potential of FRP-strengthened variants when exposed to fires. This gap in the literature highlights the novelty and necessity of a comprehensive analysis that uses advanced machine learning techniques to predict failure potential accurately. The present study aims to address this gap by developing robust machine learning models capable of identifying the failure potential of fire-exposed FRP-strengthened concrete beams. Within this context, a comprehensive database comprising experimental findings on the fire resistance of such beams is compiled from existing literature. The research employs 54 different combinations of six different classification machine learning models, each integrated with nine distinct data preprocessing techniques, to define the most effective predictive model. The performance of these models is rigorously evaluated to identify the optimal combination. Furthermore, a feature importance analysis is conducted to describe the influence of various input parameters on the failure potential of the beams. This analysis provides valuable insights into which factors most significantly affect the structural integrity of FRP-strengthened concrete beams under fire conditions. The importance of this research lies in its potential to enhance the safety and design of FRP-strengthened concrete structures. By identifying key predictive features and developing accurate models, the study contributes to the optimization of design protocols and the advancement of fire safety standards in the construction industry.