AI-Based Design: Prediction of Load-Carrying Capacity in Cold-Formed Members

Cold-formed steel Class 4 profiles are extensively used in modern structural engineering because of their low self-weight and high load-carrying capacity. Despite these advantages, thin-walled elements are highly susceptible to local and distortional buckling, which curtails their ultimate yielding strength. Consequently, structural design requires computing effective section properties. Eurocode 3 Part 1–5 addresses this through the Effective Width Method, an iterative procedure that penalizes plate dimensions according to slenderness and stress profiles. However, this method is computationally inefficient and time-consuming, especially when dealing with complex configurations like perforated plates. This study proposes a Machine-Learning alternative to bypass these iterative calculations. An Artificial Neural Network (ANN) was developed as a high-speed surrogate model to predict effective area and section modulus ratios. The optimized network proved reliable performance, with errors below 10% for over 98% of the analyzed cases, thereby offering an efficient tool for practical structural design.