Architectural constraints in a generative design system: interpreting energy consumption levels

This paper investigates the possibility of encoding architectural design intentions into a generative design system, using as a test bed the School of Architecture at Oporto [Portugal], designed by Álvaro Siza. Based on language constraints derived from Siza’s original design, the generative system [GS], consisting of a genetic algorithm and the DOE-2.1E building simulation program, created facade solutions resulting in lower annual energy consumption, while acting simultaneously as a diagnosis mechanism for problems occurring in the existing building. Results suggest this GS may be a useful tool to architects during the design process, by identifying potentially problematic areas and suggesting ways to approach them. Experiments were also performed for other geographical locations apart from Oporto, to test the algorithm’s capability to adapt solutions to different climates within the same language constraints.