The prediction of the ultimate base shear of BRB frames under push-over using ensemble methods and artificial neural networks

This study aims to develop machine learning (ML) models to predict the base shear of buckling restrained braced frames (BRBF). Four machine learning (ML) algorithms [random forest, artificial neural network (ANN), XGBoost, and Adaboost] were used to conduct this task. The training data were generated by conducting pushover analysis in OpenSeesPy. The BRBF model in OpenSeesPy considered both geometric and material nonlinearity. Six different configurations were used in this study. Each generated data point has unique frame properties (column, beam, BRB, boundary condition, leaning column and dead load, number of bays and stories, bay width, and story height). The learning and testing processes were conducted for each BRB configuration individually and for the combined data of all the configurations. Several statistical analyses were done to evaluate the prediction model and to study the importance of the BRBF properties based on their influence on the prediction. For that matter, the number of stories had the highest effect on the prediction values, and it shows that the higher the number of stories, the lower the maximum base shear that the frame can provide. The second most important feature is the core area of the BRB, where increasing the core area increases the base shear and vice versa. Furthermore, XGboost showed the best-predicted results, followed by Adaboost, Random Forest, and, Finally, artificial neural network (ANN). Finally, a graphical user interface based on the models was developed for the preliminary estimation of the base shear of buckling restrained braced frames. © 2023, The Author(s), under exclusive licence to Springer Nature Switzerland AG.