Machine learning algorithms for axial strength prediction of fully FRP-confined circular concrete columns

This study investigates the application of advanced machine learning (ML) algorithms to predict the peak axial strength of circular concrete columns fully confined with fiber-reinforced polymer (FRP). Unlike existing ML models that rely on limited experimental data, this research utilizes a comprehensive database comprising 1517 samples of FRP-confined circular concrete columns. Four state-of-the-art ML models — Support Vector Regression (SVR), Multi-Layer Perceptron (MLP), Gradient Boosting Regressor (GBR), and Extreme Gradient Boosting (XGB) — were employed to leverage this extensive dataset. To prevent overfitting, a robust 10-fold cross-validation strategy was implemented. Optimal hyperparameters for each model were determined through a genetic algorithm-based optimization process to maximize performance. The predictive accuracy of the models was assessed using four established error metrics, demonstrating significant improvements over traditional design formulations. Among the ML approaches, the XGB algorithm achieved the highest predictive accuracy with minimal computational cost. This study underscores the advantages of using a large, diverse dataset to enhance the reliability of axial strength predictions, paving the way for more robust, data-driven design methodologies in structural engineering.