Transforming Architectural Practice Through Computational Design and Machine Learning: A Decision-Support Framework for Energy and Daylight Optimization

Abstract

Architectural design requires balancing aesthetic goals, functional needs, and environmental performance, often involving complex trade-offs. This research integrates machine learning and computational design to optimize building enclosure design, focusing on energy efficiency and daylight performance in Jordan's climatic conditions. Among the tested models, Artificial Neural Networks (ANN) proved the most effective, excelling in identifying critical design features and uncovering hidden influences among variables such as material properties and glazing systems. The findings demonstrate that machine learning can support architects in exploring design possibilities and understanding trade-offs, while ensuring they retain the final decision-making authority. By highlighting interactions that conventional methods might overlook, this approach allows architects to tailor materials and structures dynamically, optimizing performance without compromising design goals. Though cost analysis was not directly included, the framework sets the stage for its integration in future studies, enabling even more comprehensive decision-making. The results emphasize that design solutions should be adaptive, allowing different walls or façades to have unique material and structural configurations. This flexibility helps architects achieve efficient, context-specific designs that align with sustainability goals. By leveraging machine learning, this research bridges the gap between creative design and performance-driven optimization, offering a practical framework for innovative architectural practices.