Intelligent design of shear wall layout based on diffusion models

This study explores artificial intelligence (AI) for shear wall layout design, aiming to overcome challenges in data feature sparsity and the complexity of drawing representations in existing AI-based methods. We pioneer an innovative method leveraging the potential of diffusion models, establishing a suitable drawing representation, and examining the impact of various conditions. The proposed image-prompt diffusion model incorporating a mask tensor featuring tailored training methods demonstrates superior feature extraction and design effectiveness. A comparative study reveals the advanced capabilities of the Struct-Diffusion model in capturing engineering designs and optimizing performance metrics such as inter-story drift ratio (in elastic analysis), offering significant improvements over previous methods and paving the way for future innovations in intelligent designs.