The beam-column joint is one of the important structural elements that control the structural behaviour during seismic excitation. Moreover, failure of any of these components may lead to the partial or overall collapse of the entire structure. In view of the safety concern, there is a need to evaluate the response properties, such as failure modes and ultimate shear capacity of the beam-column joints. Conventional methods to evaluate the shear capacity of beam-column joints are usually time-consuming. Therefore, in this study, an attempt has been made to evaluate the joint shear capacity and mode of failure of an exterior beam-column joint using ensembled machine learning (ML) approaches like Decision Tree, Random Forest, AdaBoost, CatBoost, and LightGBM. The present study highlights the potential use of the CatBoost model as a useful tool that can assist in predicting the shear strength and failure mode. The results of the study reveal that this model has performed well in predicting the load-carrying capacity and shear strength with high correlation coefficients of 0.9836 and 0.9978, respectively. Moreover, it has been observed that the prediction model provided by the CatBoost algorithm exhibits the highest accuracy for classification in the failure mode of beam-column joints.
Read full abstract