Curved glass facades can be incredibly beautiful, but conventional construction methods are very expensive. Panels are usually made with a “hot fold” where the glass is heated and formed by specialized molds or machines, an energy-intensive process that generates excess waste in the form of individual molds. Cold-folded glass is a cheaper alternative, with flat glass bent and fixed to construction frames. However, given the fragility of the material, it is very difficult to find a shape that is aesthetically pleasing and manufacturable. Now, with an interactive data-based design tool, architects can do just that.
The software, created by a team of scientists from IST Austria, TU Vienna, UJRC and KAUST, allows users to interactively manipulate the design of the façade and receive instant feedback on the manufacturability and aesthetics of the paneling – a very convenient way to carry out various implementations. the intentions of the designer. The software relies on a deep network of neurons trained in special physical simulations to predict the shapes and manufacturability of glass panels. In addition to allowing users to adapt the planned design in an interactive way, it can automatically optimize a particular design and easily integrate it into the regular workflow of the architect. The results of the software and research were presented at SIGGRAPH Asia 2020.
Cold folded and cold folded glass
Folded glass has been used since the 19th century, although it was not generally available until the 1990s. However, the process remains expensive and the logistics for transporting the folded glass are complicated. The alternative, cold-folded glass, was developed about ten years ago. It was cheap to make, easy to transport, and the geometric and visual quality were better than hot quality folded glass. The technique allowed the architects to use special types of glass and accurately calculate the deformation stress of the panels.
The point was that the design of cold-folded glass facades is a huge computational problem. Ruslan Guseinov, PhD and first author of IST Austria, explains: “It is possible to calculate when an individual panel will break or provide a safety margin for additional loads, working with the entire façade – often thousands of panels – It is too complex for a regular designer for tools. “In addition, using a computer with conventional computational methods to obtain stresses and shapes each time a change was made would take too long to be available.
Enabling a new technology
Thus, the goal of the group was to create software that would allow a user (non-expert) to edit a surface interactively while receiving real-time information about the curved shape and stresses associated with each panel. They decided on a data-driven approach: the team conducted more than a million simulations to build a database of possible curved glass shapes, represented in a format common in computer-aided design (CAD) architecture. The deep neural network (DNN) was then trained in these data. This DNN accurately predicts one or two possible shapes of glass panels for a given quadrangular boundary frame; these can then be used on a façade sketched by an architect.
Konstantinos Gavriil, the first author and researcher at TU Vienna, adds that DNN predicted that it was “one of the most amazing aspects of DNN”. “We knew that a particular boundary did not define the panel in a special way, but we did not anticipate that DNN could find multiple solutions, even though it had never seen two alternative panels for a single boundary.” In the set of solutions, the program selects the geometry of the panel that best suits the design of the facade, taking into account features such as smoothness of the frames and reflections.
The user can adjust his model to reduce stress and improve the overall appearance. If this is too difficult, the user can automatically optimize the design at any time, which provides the “best fit,” which significantly reduces the number of non-viable panels. Eventually, all the panels can be built safely or the user can bend some of them hot. When the user is satisfied with the form, the program exports the flat panel shapes and frame geometries required to build the facade.
Accuracy and efficiency
To test accuracy Among the simulations, the team manufactured frames and glass panels, including high-stress panels. In the worst case, they saw a small deviation from the planned shapes (smaller than the thickness of the panel), and all the panels were manufactured as expected. The team verified that the data-based model faithfully (and efficiently) reproduced the output of the simulations.
“We believe we have created a new and practical system that matches geometric design and manufacturing and allows designers to effectively find a balance between economic, aesthetic and engineering criteria,” concludes IST Austria professor Bernd Bickel. In the future, the program could be extended to include additional features for practical architectural design or to be used to study different materials and more complex mechanical models.
Reference: Konstantinos Gavriil, Ruslan Guseinov, Jesús Pérez, Davide Pellis, Paul Henderson, Florian Rist, Helmut Pottmann and Bernd Bickelen “Computational design Design of cold folded glass facades”, November 2020, ACM Transactions in Charts.
DOI: 10.1145 / 3414685.3417843
Funding: European Union Horizon 2020 research and innovation program, Algebraic Delegations for ComplEx Forms in Computer Aided Design (ARCADES), European Research Council (ERC)