Prediction of ultimate load carrying capacity of short cold-formed steel built-up lipped channel columns using machine learning approach

This study presents the prediction of the ultimate load carrying capacity of cold formed steel (CFS) built-up back-to-back channel columns having fixed boundary conditions under axial compressive load. There were 60 non-linear finite element models developed in ABAQUS, 12 of which were validated using experimental data while the remaining 48 models were validated based on AISI specification design standards. The finite element analysis and experimental results were also compared to the ultimate strength from the AISI specification. A parametric study was carried out using the validated finite element model in addition to the use of machine learning models to predict the ultimate load of CFS sections. Here, the machine learning models such as Artificial Neural Network (ANN), Gradient Tree Boosting (GTB) and Multivariate Adaptive Regression Splines (MARS) were developed for comparative evaluation of model predictions. Based on the performance evaluation using several statistical indices, MARS and GTB models were found to provide relatively accurate predictions of the ultimate load of CFS sections.