Physical rule-guided generative adversarial network for automated structural layout design of steel frame-brace structures

Due to the complexity and repetitiveness of building structural design, the application of artificial intelligence (AI) to assist engineers has become a hot research topic in recent years. However, in the field of steel frame-brace structures, the performance of the AI generative models remains to be improved, particularly in incorporating essential physical design rules into the design process. To address this gap, a physical design rule-guided generative adversarial network, namely FrameGAN v2, is proposed. The primary goal of FrameGAN v2 is to synthesize high-quality steel frame-brace structural drawings while ensuring adherence to the specified design rules. To validate the effectiveness of the proposed model, a comprehensive analysis and comparison are conducted between FrameGAN v2, the original FrameGAN, and expert-designed structures. The results reveal that FrameGAN v2 achieves better performance in terms of both visual and physical properties, which indicates the high potential in automated structural layout design of steel frame-brace structures.