A graph-enabled parametric modeling approach for façade layout generative design

Façade layout is a crucial design element that has significantly benefited from the advancements in generative artificial intelligence. However, current generative methods such as procedural modeling and parametric modeling exhibit two primary limitations. Firstly, they rely on drawings or images instead of more flexible and structured data inputs. Secondly, they often neglect the spatial sequence of the façade in the optimization process. This paper introduces a graph-enabled parametric modeling approach (GEPMA) for façade layout generation. The approach utilizes graphs to represent structural data inputs and incorporates spatial sequence-aware optimization for the output. The process begins by decoding a hierarchical attributed graph (HAG) to extract essential façade information and define key modeling parameters. Facades are then generated and simulated based on two predefined objectives: energy use intensity (EUI) and spatial daylight autonomy (sDA). Subsequently, the façade variants are optimized under four different spatial sequences using a non-dominated sorting genetic algorithm II (NSGA-II). GEPMA was evaluated through a case study of a residential building in China. The optimal solutions improved sDA by an average of 7.71 % (up to 12.02 %) and reduced EUI by an average of 0.40 % (up to 0.62 %) compared to the baseline façade. This study contributes to the field of generative design by presenting a flexible and editable methodology and providing practical insights for façade design. • A flexible and editable graph-enabled parametric modeling method for façade layout. • A spatial-sequence aware optimization method for multiple façade elements. • The EUI and sDA have been improved by up to 0.62 % and 12.02 %, respectively. • Practical design strategies integrated with design modules are summarized.