Abstract: Driven by the tripartite forces of enhanced hardware computing power, the innovation of the Transformer mechanism, and the explosive growth of data, the new generation of AI technologies (language, vision, multimodal) are deeply reshaping the creative expression paradigm of architectural design. Researching and applying AI technologies to implement applications in business scenarios has become a strategic choice for enterprises to build core competitive barriers and cultivate new forms of productive forces. This paper conducts a systematic study on the AIGC algorithms and workflows in architectural design, and proposes a set of enterprise localization deployment method system for the generation of architectural design schemes in the aviation industry. Based on the open-source large model Stable Diffusion, through fine-tuning training in the enterprise intranet environment, a scheme-aided design platform adapted to the architectural characteristics of the enterprise is quickly built to realize semantic inspiration stimulation and stylized creation of architectural schemes. Additionally, the large language model DeepSeek is deeply integrated with the enterprise knowledge base to accelerate the iteration and upgrading of the company's AI digital intelligence capabilities, providing a fast and reusable localized solution for building an enterprise design "brain".