LEE- Mimari Tasarımda Bilişim-Yüksek Lisans

Bu koleksiyon için kalıcı URI

http://hdl.handle.net/11527/19395

Gözat

Konu "architectural representation" ile LEE- Mimari Tasarımda Bilişim-Yüksek Lisans'a göz atma
  • Öge
    An artificial intelligence based methodology in architectural design process: Integrated diffusion model-aided design (DMAD)
    (Graduate School, 2024-10-15) Müezzinoğlu, Can ; Yazıcı, Sevil ; 523211007 ; Architectural Design Computing
    The thesis focuses on the integration of Diffusion Models (DMs) that synthesize text, images, or videos based on written or visual prompts into architectural design and how architectural design process can shift. The thesis aims to develop a methodology that integrates traditional design and computational design with DMs. The design process in this thesis is separated into three distinct phases: (1) preliminary design, (2) design development, and (3) presentation of finalized design. Each phase utilizes different capabilities of DMs. The designer can use Ideative DMs in the Preliminary Design Phase to generate new ideas to contribute to concept generation. Variative DMs can be used in the Design Development Phase to generate image alternatives to the design produced by the designer in the preliminary design phase. These alternatives can be related to mass model decisions, solid-void experiments, material selection and facade decisions. In the Presentation of Finalized Design phase, the designer works with Post Productive DMs to edit the visuals which were produced with Variative DMs, for presentation. The integrated DM-aided design framework was first tested through a preliminary study, followed by a case study involving two experts who were architects undertaking their PhD degrees in architecture. In the case study, data was obtained through questionnaires and through the collection of inputs used in the design process and outputs produced. The data was analyzed under two main categories: (1) phase-based evaluation of each DM separately and (2) the overall evaluation of the design process. In phase-based evaluation, the study examines the use of various DMs across three design phases. During the preliminary design phase, DM can enhance exploration and evaluation of alternatives, providing valuable new ideas for designers. In the later phases, DMs facilitate the generation of alternatives and the creation of visuals for presentations. The effectiveness of different detection methods for image analysis is assessed using image difference method. As seen in the case study, the outputs from the various environments used throughout the design process are used as inputs in the next design environment, allowing reflection in the process. Integrating DMs into architectural design creates a dialogue between the designer and the tool, especially in the early stages. As the design process advances, the designer becomes a decision-maker. The integration of multimedia and multiscale approaches enriches the design process by expanding the search space and addressing specific design challenges efficiently. Additionally, DMs expand the design search space by generating alternative outputs throughout the process. The framework also allows for transitions between different design stages and scales. The proposed methodology has the potential to develop new strategies for architectural design.