<span lang="EN-GB">Shear Capacity Evaluation of Beams Made of Cold-Formed Steel with Web Openings Using a Hybrid Neural Network-Genetic Algorithm Method

This study investigates the ultimate shear capacity of cold-formed steel (CFS) I-beams with web openings, which are essential for utilities but detrimental to structural strength. Finite element models of ۱۰۰۰ mm long beams under three-point bending were analyzed, considering two openings with varied spacing, shape (rectangle, circle, square), dimensions, section thickness, lip stiffeners, and distance from supports. A comprehensive parametric study was conducted. Results confirmed that increased opening spacing significantly enhances ultimate strength, while the presence of lip stiffeners also provided a measurable improvement. For optimization, a hybrid approach combining a neural network (NN) and a multi-objective genetic algorithm (NSGA-II) was employed. The NN was trained using FEA output data to predict performance, and its predictions served as objective functions within the NSGA-II to maximize load capacity while minimizing beam mass. This process generated Pareto-optimal design sets for each opening shape, effectively illustrating the critical trade-off between structural strength and material weight for engineers. Finally, the optimal designs proposed by the hybrid AI model were validated against independent, high-fidelity finite element analyses. This validation confirmed the model's high predictive accuracy and robustness, demonstrating its practical utility. The developed framework provides a reliable and efficient computational tool for designing optimized, lightweight, and code-compliant perforated CFS beams in modern construction projects.

کلیدواژه ها:

cold-formed steel ، optimization ، Genetic Algorithm ، Artificial Neural Network