An Automated Building Facade Renovation Design Framework Based on Explainable AI

Abstract

Facade retrofitting is a critical component of urban development, yet current practices suffer from inefficiency and strong stylistic subjectivity. To address these challenges, this study proposes and validates a novel intelligent design framework fully integrated into the SketchUp platform. This framework automates the identification and parametric reconstruction of facade geometric elements through Ruby scripting while innovatively integrating an explainable artificial intelligence (XAI)-based style recommendation engine. The core innovation lies in employing decision tree algorithms to analyze quantifiable architectural features-such as window-to-wall ratios, component types, and symmetry-thereby providing transparent, logic-based quantitative adaptation recommendations for diverse styles, including Neoclassical, Modern Minimalist, and New Chinese styles. The results demonstrate that this approach not only significantly enhances modeling efficiency but also improves the overall accuracy of style recommendations. Crucially, by translating AI’s “black box” decision-making process into clear “rules comprehensible to designers”, this research substantially strengthens human‒machine collaboration. Tansforms AI’s “black box” decision-making into clear, designer-understandable “if-then” rules, significantly enhancing human–machine collaboration. This not only provides innovative technical support for historic district preservation and sustainable urban renewal but also explores a trustworthy localized design solution that integrates “identification-analysis-recommendation-generation.” It holds broad application value in architectural design and engineering management.