A large-scale topology optimization method integrating multiple subdomain volume and minimum length scale constraints for steel frame structure design

Conventional element-based topology optimization methods support a single global volume and minimum length scale constraint, sufficient for 3D components with fewer than a few million elements. However, for large-scale structures with over 100 million degrees of freedom, these constraints fall short. For instance, column areas may require larger dimensions for load robustness, while beam regions need smaller sizes for even load distribution. Moreover, the single global volume constraint may lead to inadequate material distribution in key areas, requiring manual adjustments that may lower design efficiency and structural performance. This study proposes an element-based approach that integrates multiple local volume constraints and multiple local minimum length scale constraints under various loading conditions, within a computationally efficient framework for large-scale topology optimization. Capable of handling up to 151 million degrees of freedom on a 24 GB GPU, this method produces high-resolution designs and offers precise control for complex large-scale engineering projects.