Deep learning-based modelling of polyvinyl chloride tube-confined concrete columns under different load eccentricities

This study presents a deep learning-based framework for predicting the load-carrying capacity of polyvinyl chloride (PVC) tube-confined concrete columns under various loading conditions. A comprehensive dataset of 200 samples was generated using finite element modeling, incorporating key parameters such as PVC tube thickness, concrete strength, and load eccentricity. Several machine learning algorithms, including Linear Regression (LR), Support Vector Regression (SVR), Decision Tree (DT), Random Forest (RF), Gradient Boosting (GB), and a novel hybrid Transformer-Convolutional Neural Network (Transformer-CNN) model, were employed for the prediction task. The results demonstrate that the proposed Transformer-CNN model outperforms traditional methods, achieving the lowest root mean squared error of 27.15 kN and the highest coefficient of determination value of 0.9875. The model's robustness was further validated using cross-validation techniques, ensuring its reliability for practical applications. To facilitate usability, a Python-based graphical user interface (GUI) was developed, enabling engineers to apply the model efficiently in real-world scenarios. This study highlights the potential of deep learning in advancing the design and analysis of PVC-confined concrete columns, offering a more accurate and efficient alternative to conventional methods.