The paper covers key innovations that enabled an accelerated process of design, fabrication, and installation of 23,000 individually unique panels with cold bending at a previously unprecedented scale and to an extreme depth. The first of these innovations is the use of 3-dimensional frames, as opposed to the more common approach of pushing flat frames out of plane. Fabrication of the panels in their final shape reduces stress in the structural silicone, avoids spring-back forces in the framing, and allows for rapid installation because no on-site deformation is required.
Due to the extreme degree of deformation, determining the flat shape of the glazing is not trivial. A special methodology based in Machine Learning was developed to reverse-engineer the flattened shape of the deformed panels, and was built on a database of 3,500 unique material simulations.
These innovations are underpinned by a data-centric approach to 3d modelling. Digital models are enhanced with metadata, which can then be parsed, queried, tabulated, or otherwise transformed to allow for rapid decision making. An interconnected network of data-rich models allows for production of information at any level of detail, including FEA simulations and high resolution fabrication models and documents. We discuss the system design innovations described above, and how the computational approach enabled them.