Prediction of shear strength of infilled reinforced concrete frames using efficient hybrid BR-ANN model

Reinforced concrete (RC) frames with infills have been widely used in conventional low-rise buildings. Due to the presence of infill walls, the shear failure as well as lateral bearing capacity of the structural system will be changed significantly compared to the bare RC frames. The purpose of this study is to predict the shear strength of masonry-infilled RC frames using hybrid neural network models, which are combined based on the Bayesian regularization (BR) algorithm and Artificial neural network (ANN). A database containing 153 test results is gathered from the literature to construct the machine learning models. The shear strength predicted by BR-ANN in this study is then compared with the conventional ANN using the Levenberg-Marquardt algorithm. Four statistical metrics, including the goodness of fit R2, root-mean-squared error RMSE, mean average error MAE, and a20-index are calculated to evaluate the prediction performance of the ANN models. The comparison emphasizes that the BR-ANN model accurately predicts the shear strength of infilled RC frames with a high R2 of 0.92, a small RMSE of 12 kN, and a20-index of 0.7. Moreover, the influence of input design parameters on the shear strength is assessed. Finally, a graphical user interface tool is developed for practically calculating the shear strength of infilled RC frames. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2024.