摘要
随着人工智能在工程设计领域应用的逐渐深入,AI开始介入设计、优化、建造运维等工程全生命周期,尤其是生成式AI,其在设计中的应用场景逐步被挖掘,带来了设计范式的新变革,引发了设计行业的广泛讨论与应用探索。与此同时,生成式AI在工程设计中的应用难点也逐步显现:垂直领域模型训练量不足、缺乏对参数之间交互影响的认识、未构建精准提示词工程等问题制约了其在行业的深度应用和价值呈现。本研究以Stable Diffusion为例分析其在建筑设计中的应用难点及成因,针对性提出优化方法,并通过项目实践验证了其有效性,从而实现生成式AI在多种建筑设计场景下的高质量可控输出。
With the deepening application of Artificial Intelligence in the field of engineering design,AI continues to intervene in the design,optimisation,construction,operation and maintenance of the whole life cycle of projects.The application scenarios of Generative Artificial Intelligence in design have been gradually explored,bringing changes in the design paradigm and triggering extensive discussion and application exploration in the design industry.However,the application difficulties of Generative AI in engineering design have gradually emerged:insufficient training of vertical domain models,lack of understanding of the interaction effects between parameters,and failure to construct accurate prompt engineering have constrained its deep application and value presentation in the industry.This study takes Stable Diffusion as an example to analyse its application difficulties and causes in architectural design,proposes targeted optimisation methods,and verifies its effectiveness through project practice,enabling high-quality controllable outputs of Generative AI in various architectural design scenarios.
作者
胡旭明
季新亮
周振阳
HU Xuming;JI Xinliang;ZHOU Zhenyang
出处
《世界建筑》
2025年第6期83-89,共7页
World Architecture
