From Medieval Ornament to Computational Design: AI-Based Generative Frameworks For Contemporary Architectural Decoration

Abstract

Traditional architectural ornamentation, particularly in medieval Islamic architecture, constitutes a structured system of cultural self-expression, aesthetic logic, and symbolic meaning. However, in contemporary design practice, the integration of such ornamentation faces challenges stemming from industrialization, standardization, and the increasingly widespread use of data-driven design tools.

This study explores the reinterpretation of Azerbaijani ornamental systems specifically, the Nakhchivan architectural school using a hybrid computational design model. The proposed approach combines rule-based methods (grammar of forms), crystallographic symmetry analysis, and generative models using artificial intelligence (diffusion-based), which allows for the control of variations while preserving structural and cultural integrity.

A multi-level validation system is applied, including both qualitative expert assessment (n = 15) and quantitative metrics, such as symmetry preservation, pattern complexity, and structural similarity. The results show that rule-based systems provide higher structural accuracy, while AI-based models increase formal diversity. A hybrid approach yields the most balanced results across structural, aesthetic, and cultural criteria.

This study contributes to research in the field of architectural design by formalizing ornamentation as a multi-level computational system and positioning artificial intelligence as a constrained design tool that accounts for cultural characteristics. The proposed methodology offers practical applications in parametric facade systems, interior surfaces, and digital fabrication workflows.

Keywords: Islamic geometric patterns, generative design, grammar of forms, diffusion models, parametric design, architectural heritage, computational ornament