To synchronize designers early in the design process and help them understand the building forms and performance metrics needed to meet both qualitative and quantitative requirements is crucial for sustainable building designs. While the prevalent design exploration technique, which integrates self-organizing mapping with a Supervised Learning-based Agent Model (SOM-MLPNN), offers insights into design geometry and solar potential, it still has limitations in the range of the design exploration space, the efficiency and accuracy of data approximation, and the uncertainty in dealing with performance metrics related to more complex environmental factors. Using the solar potential of tower buildings as an example, this study proposes a method based on SOM-MLPNN by integrating a Generalized Tower Generator (GTG) and form description, testing its usability for the inclusion of complex surrounding environments, and improving its design exploration breadth as well as the efficiency and accuracy of data approximations. The results demonstrate improved efficiency and accuracy in data approximation across an expansive design space, facilitating visual and interactive exploration of design-performance relationships. The proposed method emphasizes the critical interplay between building geometry and environment-related performance, providing designers with a refined toolkit for discerning the correlation between design and performance, thereby enhancing their design decision-making abilities.
Read full abstract