A machine learning-based method for analyzing the axial load capacity of rectangular reinforced concrete columns.

Resumo

The axial load capacity of columns is a critical variable for structural systems, as the failure of a single element can lead to the collapse of the entire structure. Therefore, an accurate prediction of this variable is essential to ensure the safety of the system. In this context, machine learning methods have gained prominence in modeling the behavior of reinforced concrete structures. Accordingly, this study aimed to investigate the axial load capacity of rectangular reinforced concrete columns using Artificial Neural Networks (ANNs) and Tree/Forest-based algorithms. Ten input parameters related to the geometric and mechanical properties of this type of structural element were selected in order to capture their behavior under different loading conditions. Prior to training, data preprocessing was carried out to identify the importance of each input variable with respect to the output of interest, as well as to detect the presence of outliers in the dataset.Model training was performed using supervised learning, with the Python programming language and the Scikit-Learn library. To assess model performance, error metrics such as the Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and the coefficient of determination (R²) were used.The results indicated that both models (ANNs and tree/forest algorithms) achieved satisfactory performance in predicting the ultimate load capacity of the columns. However, the tree/forest-based model stood out, which can be attributed to its robustness against potential outliers and its efficiency during the training process.