Prediction of Moment Capacity of Flush End-Plate Connection: A Machine Learning Approach

This study focuses on the importance of flush end-plate connections in steel structures and the potential for failure if bolts are damaged, leading to shear, bending, or fatigue failure. To avoid failure, it is critical to predict the connection’s moment capacity against anticipated loads and design criteria. Simulated results from experiments and numerical simulations were compared with actual results. Current finite element models and experiments have shown limitations in making accurate predictions. This paper developed improved machine learning techniques to accurately forecast moment capacity, using various parameters such as bolt diameter, end-plate width, and nominal yield stress. To create our models, a wide range of works with data spanning 25 years were examined. The analysis illustrated that among all the ML models that were examined in this study, the eXtreme Gradient Boosting (XGBoost) model demonstrated the best prediction performance which is also confirmed by comparing its predictions with those of the existing models. This study highlights the potential of machine learning techniques in accurately predicting the moment capacity of flush end-plate connections.