Bend, don’t break: new tool enables economical glass design

Curved glass facades can look amazingly beautiful, but traditional building methods are extremely expensive. Glass panes are typically made with “hot bending,” where the glass is heated and formed using a mold or specialized machinery, an energy-intensive process that generates excess waste in the form of individual molds.

Cold-bent glass is a cheaper alternative in which flat panes are bent and attached to the frames on the job site. However, given the fragility of the material, finding a shape that is both aesthetic and manufacturable is extremely difficult. Now, an interactive, data-driven design tool allows architects to do just that.

Created by a team of scientists from IST Austria, You come, UJRC, and KAUST, the software allows users to interactively manipulate a facade design and receive immediate feedback on the manufacturability and aesthetics of the paneling, a very convenient way to navigate various realizations of the designer’s intentions.

The software is based on a deep neural network trained on special physical simulations to predict the shapes and manufacturability of glass panels. In addition to allowing users to interactively adapt a planned design, it can automatically optimize a given design and can be easily integrated into an architect’s usual workflow. The software and research results were presented at SIGGRAPH Asia 2020.

Hot curved glass and cold curved glass

Hot-curved glass has been in use since the 19th century, but it wasn’t until the 1990s that it became generally available. However, the process remains prohibitively expensive and the logistics of transporting the curved glass are complicated. An alternative, cold curved glass, was developed around ten years ago. It was cheap to make, easy to transport, and the geometric and visual quality was better than hot curved glass. The technique also allowed architects to use special types of glass and to accurately estimate the deformation stress on the panels.

The problem was that the design of cold curved glass facades is a huge computational problem. Ruslan Guseinov, IST Austria post-doctoral fellow and co-first author, explains: “Although it is possible to calculate when an individual panel will break or provide a safety margin for additional loads, working with the entire facade – which often includes thousands of panels – is just too complex for conventional design tools. In addition, using a computer with traditional calculation methods to obtain constraints and shapes each time a change was made would take too long to be usable.

Enable new technology

Thus, the team’s goal was to create software that would allow a (non-expert) user to interactively edit a surface while receiving real-time information on the curved shape and associated constraints for each panel. individual. They opted for a data-driven approach: the team performed over a million simulations to create a database of possible curved glass shapes, represented in a conventional computer-aided design (CAD) format in architecture. Then, a deep neural network (DNN) was trained on this data. This DNN accurately predicts one or two possible glass panel shapes for a given quadrangular limit frame; these can then be used in a facade sketched by an architect.

The fact that the DNN predicted multiple forms was “one of the most surprising aspects of DNN”, adds Konstantinos Gavriil, co-first author and researcher at TU Wien. “We knew that a given limit did not uniquely define the panel, but we did not anticipate that the DNN would be able to find multiple solutions, even though it had never seen two alternative panels for one. limit. From the set of solutions, the program selects the glass geometry that best matches the facade design, taking into account characteristics such as soft frames and reflections.

The user can then adapt their model to reduce stress and otherwise improve the overall appearance. If this proves too difficult, the user can automatically optimize the design at any time, resulting in a “best fit” solution that significantly reduces the number of infeasible panels. Ultimately, either all panels can be built safely or the user can choose to heat bend a few of them. Once the user is satisfied with the form, the program exports the shapes of the flat panels and the frame geometries required for the construction of the facade.

Precision and efficiency

To test the accuracy of the simulations, the team fabricated frames and glass panels, including panels under extremely high stress. In the worst case, they observed minimal deviation from the planned shapes (less than the panel thickness) and all panels were workable as planned. The team further verified that the data-driven model faithfully (and efficiently) reproduced the result of the simulations.

“We believe we have created a practical new system that combines geometric design and responsive design with manufacturing and enables designers to effectively balance economic, aesthetic and technical criteria,” concludes Bernd Bickel, professor at IST Austria. . In the future, the program could be expanded to include additional functionality for practical architectural design, or be used to explore different materials and more complex mechanical models.

Publication

K. Gavriil, R. Guseinov, J. Pérez, D. Pellis, P. Henderson, F. Rist, H. Pottmann, B. Bickel. 2020. Computer design of cold-bent glass facades. ACM transactions on charts. DO I: https://doi.org/10.1145/3414685.3417843

Additional information

Project page: http://visualcomputing.ist.ac.at/publications/2020/CDoCBGF/

Funding Information

This project has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement n ° 675789 – Algebraic representations in computer-aided design for complex shapes (ARCADES), from European Research Council (ERC) in the framework of grant agreement n ° 715767 – MATERIALISABLE: intelligent computational design and modeling oriented towards manufacturing, and SFB-Transregio “Discretization in geometry and dynamics” thanks to grant I 2978 of the Austrian Science Fund (FWF). F. Rist and K. Gavriil were partially supported by core funding from KAUST.