Cost optimisation of steel I-girder cross-sections using genetic algorithms

• Published in: Structures (Oxford) 2023-09, Vol.55, p.379-388
• Main Authors: Martins, J.P., Correia, J., Ljubinković, F., Simões da Silva, L.
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Cost optimisation of steel structures; Genetic Algorithms; Steel I-girder optimal design
• ISSN: 2352-0124; 2352-0124
• DOI: 10.1016/j.istruc.2023.06.030

In this study, the economical use of steel I-girders subject to bending with built-up cross sections is explored. The cross-sections are optimized in terms of cost while respecting the design rules of Eurocode 3, namely, bending and shear cross-section resistance, M−V interaction, and flange-induced buckling, whereas lateral torsional buckling is neglected. The optimization is carried out by employing a Genetic Algorithm allowing for one or more bending moment and shear force combinations to be applied to the cross-section. Finally, a large parametric study is carried out for different bending moment and shear force combinations, hence covering a wide range of applications of I-girders, and an extensive database of optimized built-up sections is generated. These results, which are made available, allow for a quick selection of an optimal cross-section, hence presenting a valuable source for design practitioners. Lastly, using these results, a cost comparison is carried out between the optimized built-up cross-sections and commercial IPE and HL hot-rolled profiles, selected to comply with the safety requirements for the same pair of bending moment and shear force. This comparative study shows that despite the welding and cutting costs, the use of built-up cross-sections in large-span girders results in a more sustainable solution, as substantial material savings of up to 30–35% can be attained. This is due to their flexibility in geometry, which nowadays may be intelligently controlled and adjusted for virtually any particular case using sophisticated optimization algorithms.