Prediction of the Compressive Strength of Concrete Circular Columns Confined with FRP Using Neural Networks

This study aims to predict the compressive strength of circular concrete columns confined with Fiber Reinforced Polymer (FRP), particularly for normal and high-strength concrete under axial loading. Existing predictive models have limitations, such as restricted applicability to specific ranges of concrete strength and the inability to account for FRP variations. To address these challenges, this research employs Neural Networks (NN) to enhance prediction accuracy and efficiency. A dataset of ۵۷۴ data points was compiled from prior studies, encompassing various FRP types and concrete strengths. The NN models were trained using Levenberg-Marquardt (LM) and Bayesian Regularization (BR) methods, with different configurations tested to optimize performance. K-fold cross-validation was performed to ensure robustness. The models were validated and compared with existing approaches using  and MSE as performance metrics. Results showed that the NN models achieved up to an ۱۱.۶۷% improvement in  and an ۸۴.۲۴% reduction in MSE, significantly outperforming traditional methods. This study highlights the potential of NN-based approaches to provide reliable and accurate predictions for FRP-confined concrete columns. These findings offer valuable insights for engineers and designers, paving the way for safer and more efficient structural design practices in the construction industry.

کلیدواژه ها:

Fiber-reinforced polymer (FRP) ، Neural Network ، Confinement ، High strength ، Concrete columns