Introduction

Performance-driven design in architecture increasingly relies on tools like generative design and building simulations, which, despite their benefits, pose challenges such as high computational Performance-driven design in architecture increasingly relies on tools like generative design and building simulations, which, despite their benefits, pose challenges such as high computational demands and complex design spaces. Machine learning-based surrogate models offer a faster alternative but remain inaccessible to non-experts. This research introduces D-Predict, a prototype integrating surrogate modeling, generative design, and interactive visualizations to streamline design exploration. By enabling intuitive collaboration between architects and artificial intelligence, the framework empowers architects to efficiently evaluate and enhance building performance, bridging the gap between technical complexity and practical usability.
This research was made possible through a collaborative effort between Simon Fraser University (SFU), Perkins&Will, and Mitacs. We are grateful for SFU’s academic resources, Perkins&Will’s industry expertise, and Mitacs’ funding support, which together advanced innovation in computational design.