Development of Simplified and Machine Learning-Based Predictive Models for Elastic Distortional Buckling of C-Shaped Cold-Formed Steel Sections

The primary objective of this study is to develop simplified analytical expressions and machine learning models for predicting the elastic distortional buckling of C-shaped cold-formed steel (CFS) sections with and without standard punchouts under compression, major-axis flexure, and minor-axis flexure. A comprehensive database of elastic distortional buckling loads for SFIA standard C-section members was established through finite strip analysis using CUFSM. Based on the database results, for C members, simplified analytical formulas were developed using the methods described in AISI S100-07. Combined with the analytical formulations, one load case (axial compression with holes) was used to train and evaluate an artificial neural network, which demonstrated good accuracy. Extension to the other load cases is now underway. In this case, there is a strong correlation between the CUFSM-generated values and the projected results. When finished, ML models will provide a reliable supplement to the analytical expressions by enabling quick calculation of buckling loads without the need for in-depth finite strip analysis. The data-driven model and the streamlined analytical expressions together offer a framework for forecasting elastic distortional buckling behavior, which is meant to assist researchers and engineers by increasing design efficiency and serving as a potential foundation for upcoming revisions to the AISI S100 Specification.