Ultimate drift ratio prediction of steel plate shear wall systems: a machine learning approach

Abstract

Predicting the ultimate drift ratio of steel plate shear wall (SPSW) systems is important for ensuring the structural integrity and performance of these systems under lateral loads. In this study, machine learning models are developed for predicting the ultimate drift ratio based on various design parameters using data from previous research on SPSW systems. These design parameters include the panel aspect ratio, column flexibility parameter, axial load ratio, web plate thickness and stiffness of horizontal and vertical boundary elements. A range of machine learning models is considered, including Random Forest, Lasso, Gradient Boosting, XGBoost, Adaboost, Artificial Neural Network and a stacked regressor. The models are trained and evaluated using data from 292 different SPSW models, and their performance is compared based on the R-squared value, root mean squared error (RMSE), and evaluation time. The results of this study demonstrate the effectiveness of machine learning techniques for predicting the ultimate drift ratio of SPSW systems. The results of this study show that machine learning techniques effectively predict the ultimate drift ratio of SPSW systems. Among the models considered, the ANN model achieved the highest R2 value, with a value of 0.94.