Neural networks for predicting shear strength of CFS channels with slotted webs

Abstract

Cold-formed steel channels are made with staggered courses of slots for reduced thermal conductivity and improved energy efficiency of cold-formed steel buildings. The reduced shear strength of the slotted channels must be accurately evaluated in the design to ensure the safety of the buildings with such members. This paper proposes artificial neural networks for predicting the elastic shear buckling loads and the ultimate shear strengths of the channels with slotted webs. The neural networks were trained using a large dataset consisting of 3512 finite element simulation results and a ten-fold cross-validation method. Hyperparameter tuning with a grid search was performed to determine the optimal hyperparameters of the models. The effects of the channel properties on the elastic shear buckling loads and the ultimate shear strengths of the channels were evaluated using the SHAP method. The selected neural networks with optimal hyperparameters showed excellent agreements with the finite element simulation results and exceeded the accuracy of the available design equations.