Conceptual design of three-dimensional structural and sun-shading façades supported by machine-learning

Abstract

Façade design is a multi-disciplinary process that involves qualitative and quantitative aspects belonging to diverse disciplines – such as architecture, structural design and building physics. Since decisions made in the early design stage generally have a strong influence on the design outcome, the need for a holistic design strategy is evident. This paper presents an extension of a previous research work carried out by the authors regarding a holistic approach to the conceptual design of load-bearing and sun-shading façade systems. In this work, design aspects related to different disciplines are integrated into a holistic process thanks to the application of geometry-based methods (like graphic statics and solar geometry). At the same time, high-dimensional solution spaces are managed with the help of Self-Organizing Maps (SOMs). Particularly, the current paper focuses on the possibility to work with variable topologies. The proposed approach is applied in design explorations based on a multi-story building designed by the architect Christian Kerez and the structural engineer Joseph Schwartz. The case study shows how geometrical descriptions of the core design aspects supported by machine learning clustering algorithms support the designer in performing comprehensive design explorations and in making more conscious design choices.