Machine learning based prediction for maximum base shear, top displacement, and vibration period for SCBF under nonlinear response history analysis

Machine learning tools have been used in this research to predict the response of a special concentrically braced frame (SCBF) to earthquake using non-linear response history analysis. The target features were the first two modes of vibration (T1 and T2), maximum base shear, and maximum top displacement. A detailed model for three different configurations was modeled in Opens espy to generate the training and testing data. The model captures the nonlinearity of both the material and geometric properties used in the model. A total of 4500 different cases were analyzed in Opens espy (1500 for each configuration). Three machine learning algorithms, Random Forest, XGBoost, and Adaboost, were used in this research; each algorithm was trained to predict the target features mentioned above. Cross-validation technique with 20 folds was used to split the data for training and testing. The input features were different for each target feature to get the highest accuracy of the output. The prediction of the maximum top displacement was performed after the prediction of T1 and T2 because T1 and T2 increase the accuracy of the maximum top displacement prediction. The last prediction is the prediction of the maximum base shear because it depends on the maximum base shear and T1 and T2. A graphical user interface (GUI) was created depending on the trained models.