A Regression-Based Approach for Assessing the Buckling Coefficient of Stiffened and Unstiffened Elements

Abstract

Sustainable construction begins essentially by enabling the application of renewable and recyclable building materials. Cold-formed steel sections providing long-term sustainability and having a high strength-to-weight ratio are finding extensive application in the fields of construction and engineering. The Effective Width Method (EWM) and the Direct Strength Method (DSM) are the primary available analytical methods for the design of these sections. The design procedure for the buckling strength prediction of these analytical methods is inherently dependent on the critical buckling load, which in turn is influenced by the plate buckling coefficients. The plate buckling coefficient of either a stiffened or an unstiffened element is defined for varied boundary conditions and elemental stress ratios. Evaluating the design procedure for the critical buckling load, which is taken as the minimum of the elemental strengths, it was clearly evident that the analytical method yields highly conservative solutions. This conservative prediction is inherently caused by the application of individual plate buckling coefficients that fail to account for the inter-element interaction.The present research strives to propose a regression-based method for the determination of the critical buckling load. The symbolic regression analysis with the aid of machine learning is opted for to obtain more precise and reliable mathematical solutions for the buckling coefficients. The inter-element interaction and varied loading conditions are assessed for their influence and are accounted for in the expressions of the buckling coefficient. More reliable and accurate buckling strength prediction results with a broader conceptual approach are postulated.