Performance in generation: An automatic generalizable generative-design-based performance optimization framework for sustainable building design

Abstract

Performance-based design has become increasingly important for designing environmentally friendly buildings. It is an interdisciplinary and time-consuming process that requires architects to possess not only knowledge of design, performance simulation and optimization, but also the ability to make decisions that effectively align with environmental goals while exploring a range of design schemes. To promote such kind of design, this study proposes an innovative generalizable generative-design-based performance optimization framework (GGPOF). The framework combines generative design with performance-based design techniques to enhance various environmental aspects of buildings, including thermal comfort, daylighting and solar radiation. To facilitate performance-based design and optimization, the GGPOF automatically generates building variants and parametric simulation models, quick iteration with simulation feedback, sensitivity analysis and multi-objective optimization. A case study with two design scenarios demonstrates the high efficiency of GGPOF in green building design. More specifically, GGPOF can automatically generate every optimized design solution in an average time of 1.72 min. The optimal designs also achieve remarkable improvements in the predicted mean vote index for thermal comfort and the daylight factor compliance rate with increases of up to 86.57% and 72.73%, respectively, compared to the original user-defined design. The innovations of this study are as follows: first, it proposes a systematic and comprehensive solution for optimizing building performance from the architects' design perspective; second, it develops an automatic generalizable performance optimization framework at the early design stage.