Predicting the patch load resistance of stiffened plate girders using machine learning algorithms

In the incremental launching method employed for steel bridge construction, the girder is subjected to patch loading which occurs at the piers’ position. This loading significantly affects the girder resistance in the construction stage. Therefore, prediction of the girder resistance under this loading is important. This paper proposes a new approach for predicting the patch load resistance of stiffened plate girders using an extreme gradient boosting algorithm (XGBoost). A total of 170 experimental data on stiffened plate girders under patch loading collected from the literature serves as the training and testing data to build the predictive model. To demonstrate the efficiency of the proposed model, its predictions were compared with those obtained from other Machine Learning (ML) methods such as support vector machines (SVM), decision tree (DT), random forest (RF), adaptive boost (Adaboost), and deep learning (DL). The accuracy of the proposed model was validated against the existing equations taken from the design standards (EN-1993-1-5 and BS 5400) as well as existing formulae in the literature. The comparative results reveal that the proposed model provides better and more accurate predictions than the existing formulae.