Generative AI-Driven Structural and Safety Design across Multiple-Domain Engineering

Generative AI-driven structural and safety design is a crucial element of smart construction. Traditional human-centric methods are often inefficient and overly dependent on individual expertise, frequently falling short of engineering requirements. Targeting the intelligent design demand of multiple-domain engineering (e.g., buildings, MEPs, and infrastructures), this study presents a high-quality, efficient design method driven by generative AI. Based on generative AI models such as diffusion models, generative adversarial networks, and graph neural networks, an intelligent design algorithm is proposed based on "data-knowledge-mechanics" fusion learning. This method effectively addresses the challenges of limited training data and quality inconsistency while ensuring compliance, safety, and cost-efficiency. Case studies involving building structural design, fire sprinkler design, and rock-filled dam design demonstrate that AI-generated design outcomes achieve expert-level quality, meeting design code requirements with significantly improved efficiency (approximately 10 times faster than traditional workflows). This research advances intelligent construction by creating a self-evolving framework integrating domain knowledge with data-driven learning.