This paper introduces a deep learning-driven workflow and toolkit, the Structural Embodiment Toolkit on the McNeel Grasshopper platform, that supports the conceptual structural design process by streamlining form-finding, solid geometry generation, and AI-powered visualisation within a single CAD environment. The toolkit equips users with adaptors for incorporating results from established form-finding tools such as Combinatorial Equilibrium Modelling (CEM) and Kangaroo Physics, facilitating the transformation of the structural skeleton into solid geometries with appropriate cross-section based on internal forces. Leveraging the transformative potential of deep learning, it allows user-friendly access to deep learning-based image generation within Grasshopper with concurrent multiple control options, including line, depth and semantic segmentation. The toolkit also offers components to enhance the conventional render-based visualisation pipeline with the ability to mass-produce for model training, embodying a self-reinforcing ecosystem. This streamlined process and toolkit, exemplified through a detailed case study to demonstrate improvements in design ideation, communication, and collaboration, underscoring the distinctive potential of deep learning in structural design.
«
This paper introduces a deep learning-driven workflow and toolkit, the Structural Embodiment Toolkit on the McNeel Grasshopper platform, that supports the conceptual structural design process by streamlining form-finding, solid geometry generation, and AI-powered visualisation within a single CAD environment. The toolkit equips users with adaptors for incorporating results from established form-finding tools such as Combinatorial Equilibrium Modelling (CEM) and Kangaroo Physics, facilitating the...
»