AIPDORCS: Artificially Intelligent Preliminary Design of Reinforced Concrete Structures

Abstract: Leveraging 106 engineers’ expert assessments of preliminary structural design in 48 reinforced concrete building models, we compiled two experimental Graph Neural Network (GNN) tools to demonstrate feasibility for automated classification of structural schematic layouts, a key step toward building generative artificial intelligence (AI) tools for design. Contributions include a robust project database, a model-to-graph conversion tool, and a structural design scoring application. Acknowledging limitations related to modelling assumptions and a relatively small dataset, this research clarifies the opportunity and the obstacles to AI-driven advancements in preliminary structural design.