Effect of Member Grouping and Pool Size of Discrete Cross-Sections on the Optimal Design of a Large-Scale 3D Steel Frame

This paper examines the effect of member grouping and selected pool size of discrete cross-sections on the optimal design of a large-scale steel frame structure. Member grouping is commonly applied to structural optimization problems to significantly reduce the solution space of candidate designs and improve the efficiency and robustness of the search to achieve near-optimum solutions. At the same time, discrete sets of cross-sectional steel shapes (profile lists) assigned to a particular member group are often limited to a subset of cross-sections to further limit the search space and ensure practicality in the final design. However, for large-scale structural systems (e.g., tall buildings), limiting the number of member groups coupled with the exclusion of lighter cross-sectional shapes may significantly impact the quality of the optimal solution (i.e., weight of the frame). This work presents a case study of a large-scale three-dimensional (3D) steel frame to evaluate the sensitivity of the optimal design to practical combinations of member groups and steel profile lists used in multi-story steel frames. A recently developed metaheuristic search algorithm was applied to the steel frame and independent optimization experiments were performed for 16 unique member grouping and profile list combinations (cases) to quantify their influence on the variability and quality of the optimal design. Results reveal a 33% difference in the optimal (final) weight between member group and profile list combinations producing the smallest and largest solution spaces. Yet, the lightest design (corresponding to the largest solution space) produced a greater number of unique steel shapes and required significantly higher computational cost.