Neural networks-based formulation for predicting ultimate strength of bolted shear connectors in composite cold-formed steel beams

Abstract

In recent years, the use of artificial intelligence-based methods in engineering problems has been expanded. In the current study, the method of artificial neural networks (ANN) has been employed to predict the ultimate strength of bolted shear connectors in cold-formed steel (CFS) composite beams. For this purpose, multilayer perceptron (MLP) networks with a hidden layer were used. Three parameters affecting the performance of these networks, including the training algorithm, the activation function in the hidden layer, and the number of neurons in the hidden layer, were examined and the most accurate network was selected. The input and target data for training the network were provided by conducting an extensive numerical study on the behavior of bolted shear connectors in CFS composite beams. Consequently, using ABAQUS software, finite element (FE) models validated with experimental results were first developed. Then, 216 models with different characteristics were analyzed and a reliable database was provided for the development of neural networks. Moreover, in order to prove the high accuracy of the ANN method, the stepwise regression (SR) method was also developed as one of the powerful regression-based methods, and the performances of these two methods were compared. Finally, the most important purpose of this study is to propose an accurate ANN-based formulation in order to predict the ultimate strength of bolted shear connectors in CFS composite beams. Due to the fact that so far no relationship has been proposed to predict the resistance of shear connectors in CFS composite beams, the formula presented in this paper can be helpful in the design process of this type of beams.