|Brought to you by:
Members of the Gehry Technologies New York office outline a collaboration that spanned between four U.S. cities and Dubai
Editor’s Note: The following has been excerpted from Knowledge Engineering: The Capture and Reuse of Design and Fabrication Intelligence on the Burj Khalifa Office Ceiling by Neil Meredith and James Kotronis of Gehry Technologies:
Behind schedule and with the finish date of the building fast approaching, Gehry Technologies was presented with a unique problem. A complex, double-curved wood ceiling for one of the main entrances to the Burj Khalifa (then Burj Dubai) was under construction and it was becoming apparent that the proposed system would not work as designed. Instead of delaying the schedule or scrapping the design, an integrated team was quickly mobilized. This team was based in different design and fabrication domains, with all participants working toward the shared goals of redesign, fabrication, delivery, and installation of the new ceiling, all within a tight timeframe and construction site. Partners included Skidmore Owings and Merrill (SOM), Imperial Woodworking Company (IWA), ICON Integrated Construction and Gehry Technologies/New York (GT). Working through the design issues, the new team developed a strategy to strip the design back down to essential geometry and redesign the system from the ground up, satisfying design intent and a host of fabrication and constructability constraints through a shared parametric model.
The previous system—a stick-built plank system wrapped over a series of ribs—worked within the constraints of a small physical mock-up but did not scale up properly to the required geometries for the finished ceiling. A new system was developed using a pre-fabricated unitized panel approach. Although more risky in terms of on-site adaptability, building the panels offsite in IWA’s Chicago woodshop and then shipping them to Dubai in stages gave the added benefits of quality and speed. Now not only did the panels need to arrive on-time within a very aggressive schedule, but they needed to be built perfectly, arrive undamaged through shipping, and be pre-coordinated for installation and all surrounding building elements (structural, mechanical ductwork, interior finishes, lighting, etc.).
The first step in developing the drivers was mapping the physical constraints of the desired materials. The original design surface was then rationalized to maximize geometric simplicity while still meeting the design criteria and minimum bending radii of tested plank materials.
With a controlling surface in place, countless iterations of planking and panelization options were then tested. From the underside of the ceiling, the surface is a continuous field of planks. The panel joints are made to disappear on the outward facing surface. Visible from the back surface, rectangular panel divisions simply follow the same divisions as the planking, picking up lines as necessary. Having a parametric rig for the development of both the planking and the paneling was key as this was not a simple linear process, but instead an iterative back-and-forth conversation between maximum panel sizes, constructability, changing planking widths, and various design options for the planking itself. Prosaic constraints such as the size of the freight elevator available on site were worked along side more design-oriented issues such as the typical width of the plank.
Taking the wireframe planking curves and surfaces as inputs, a series of flexible components were designed to parametrically reconfigure themselves to adapt to various detail conditions. The first module to use this approach was the three-dimensional wood planks forming the exterior of the panel. Due to the desire for variegated veneers across the surface, individual planks with different veneers needed to be laid-up on the panels. Complicating this arrangement, many of the panels had a anticlastic surface topology, requiring individually cut and finished planks for many of the panels.
Another challenge with prefabricated units was predicting the distribution of wood grain across the entire ceiling surface. In a typical plank-based design such as a floor, veneers are sorted for aesthetic criteria and then selected by craftsman to ensure the even distribution across the surface, avoiding dark and light patches or clustered areas. This ceiling was built as a unitized system of panels, fabricated offsite and out of sequence. (Due to the schedule, some panels were still being built in Chicago while the initial grouping was being installed in the UAE.) Often called veneer “randomization” the eventual solution was hardly random, but instead sorted and mapped veneers to discreet planks in an automated system. The tool also allowing for multiple mappings to occur in an iterative digital design environment, giving the craftsman a tool to visually asses the veneer placement while still allowing for a new level of control during fabrication. In the end, every veneer for the entire project was tagged and managed across the entire ceiling surface through this method.
Remarkably, the entire panel fabrication was done with 2-D cutting profiles. That is not to say that the fabricator was not working in 3-D, but the understanding of 3D had more to do with assembly and fabrication of actual materials in space instead of digital models. GT produced a 3-D solid model, but the Document Template CAM outputs were completely 2D in nature. Instead of relying on robotic fabrication or multi-axis machining, a series of registration elements were output from the model for use during assembly. Looking at one piece of the assembly, the “waffle frame” is a well known and fairly easy-to-assemble construction given the prevalence of CNC fabrication, but to ensure the larger project goals such as continuity between adjacent panels, the edges of the frame were projected on to a single plane to create a “tabling” jig at 1:1 scale and CNC out of 1⁄4-inch MDF. In addition to the direct-to-fabrication cutting of the plank and frame pieces, various other jigs and templates aided panel continuity and assembly.
After gathering all the pieces together in to a “master model” details were then refined and expanded in the 3-D environment with all the appropriate links to the upstream driver geometry and the downstream panel fabrication models. While representing an investment in time, effort and infrastructure, the flexibility of the system allowed for rapid update cycles late in the design process. To take one example, 3D survey data of as-built conditions was continually cross-checked against the developing fabrication model. Fairly late in the design process it was discovered that some of the steel interfacing with the boat section was not within the assumed design tolerances and needed to be updated. Instead of remodeling these parts, the input surface geometry was slightly tweaked, producing a ripple effect of changes though the design surface. These panels were simply updated to the new surface and reissued just prior to the actual panel fabrication.