Machine Learning‐Based Prediction of Structural Behaviour of Optimised Cold‐Formed Steel Elements

Abstract

Cold‐formed steel (CFS) elements are increasingly used in modern construction; however, their complex buckling behaviour often makes the design and optimisation process computationally expensive and prone to inaccuracies. This study introduces a novel integrated design framework for the complex structural optimisation of CFS elements susceptible to different buckling modes. To this end, the cross‐sectional dimensions of CFS columns are first optimised for maximum load‐bearing capacity in accordance with Eurocode 3 provisions, considering a wide range of practical and design constraints. The resulting optimised dataset is then used to train several regression‐based machine learning (ML) algorithms, enabling rapid prediction of optimal cross‐sectional dimensions and corresponding load capacity for given material properties and loading conditions. The results demonstrate that the trained ML models achieve very high predictive accuracy, with coefficients of determination reaching R2 = 0.99, while significantly reducing the computational effort and manual iteration associated with the Eurocode 3 effective width method. The proposed ML‐based optimisation framework provides an efficient tool for system‐level design of CFS structures comprising multiple members with varying load demands.