Prediction of Bearing Strength for Cold-formed Steel Bolted Connection Using Artificial Neural Network (original) (raw)

Cold-formed Steel (CFS) is a high-quality material known for several construction advantages. Still, its complex non-linear stress and strain behavior made it difficult to establish its full-strength capacity. Due to the lack of comprehensive studies on CFS, especially its joint connections, most of the specifications on current structural codes utilize experimental results from hot-rolled steel. The present study conducted an experiment on a bolted lap joint specimen using CFS members and developed a design model using an Artificial Neural Network (ANN) to analyze and predict the bearing failure of the connection. The results of the study showed that the parameters, including CFS thickness, bolt size, bolt spacing, and bolt configuration, have a significant influence on the connection's bearing capacity. The most influential parameter is bolt size, with a relative importance of 43.67%, followed by bolt spacing, bolt configuration, and CFS member thickness, with 22.03%, 18.22%, and 16.08%, respectively. The existing CFS bearing equations provided by NSCP 2015 failed to consider parameters like bolt spacing and bolt configuration, and the calculated bearing capacity values underestimated the actual connection's bearing strength. The 4-5-1 neural network model emerged as the most suitable ANN model with regression values of 0.9813, 0.9632, 0.9084, and 0.96736 for training, validation, testing, and overall regression value of the model. The generated ANN prediction values have successfully modeled the behavior of the CFS bearing strength.