Civil Engineering Departmental Seminar - Research Advances in Generative AI-Driven Engineering Design

Research Advances in Generative AI-Driven Engineering Design

Generative AI-driven intelligent engineering design constitutes a critical component of intelligent construction. To address the pain points of conventional engineering design, including heavy reliance on manual experience, high cross-disciplinary collaboration costs, and low efficiency of design iteration, this prestation will systematically introduce generative AI-enabled intelligent design methodologies and state-of-the-art applications across multiple engineering scenarios, including building structures, earth-rock dams, fire protection systems, bridges, and photovoltaic supports.

Based on diffusion models, generative adversarial networks (GANs), graph neural networks (GNNs), and multimodal large language models, the research proposes a data-knowledge-mechanics integrated learning framework. The framework embeds implicit empirical knowledge extracted from engineering drawings, design code constraints, and structural mechanical performance indicators into the generative workflow, enabling a closed-loop design pipeline that integrates schematic generation, performance verification, and optimization iteration.

For residential building structures, the system can automatically complete component layout, section sizing, and model generation. For earth-rock dams, it realizes dam zoning layout and stability evaluation. Engineering case studies demonstrate that AI-generated design schemes are comparable to expert-level designs and meet the requirements of preliminary design. The proposed approach significantly improves design efficiency, code compliance, and structural safety.