Innovations in Design and Fabrication Technology

Luc Bamberger, AIA

October 2018


The challenges of designing and constructing the United States Olympic Museum—a 65,000 SF, $45M project in Colorado Springs, Colorado—are numerous, including working with a large team comprised of the design architect (DS+R), the executive architect (AMD), exhibit designers (Gallagher and Barrie Projects), exhibit fabricators and content creators (CREO, Center Screen and BVI), as well as multiple layers of subcontractors and consultants. Innovations in technology, especially in 3D BIM and parametric modeling, and the ease with which these models are shared from one platform and one team to another, have made this collaborative process manageable, while also opening up possibilities for the expression of dynamic movement and novel form within the building design itself. 

Nowhere on the project is this more true than in the design, fabrication and installation of the skin system, a tessellated surface of nearly 10,000 unique aluminum diamond panels. Designed to emulate the taut, engineered suits worn by Olympic athletes, the skin is meant to express the tension and movement inherent in the twisting forms of the building itself. Using the latest software, fabrication and layout tools, the team was able to realize a complex design, manage multiple models and large data-sets, streamline fabrication and, ultimately, install thousands of panels over a curving substrate with extremely tight tolerances. 

The following is an overview of the process of taking the skin system from design through fabrication and installation, highlighting the technology harnessed to realize the design. It is important to note that this process was by no means linear. At each phase, feedback from other team members was essential to pushing the design closer to the finish line: programmatic changes influenced the skin geometry, skin geometry influenced structure, constructability changed panel design, as-built conditions in the field led to modifications in panel layout, etc. Only through this iterative process was the design of the skin system able to evolve to the point where it would ultimately be successful both aesthetically as well as functionally.


The overall exterior geometry of the building had its genesis in the 3D modeling software Rhino. Starting with conceptual free-hand sketches, the DS+R team moved quickly into Rhino to begin to formalize the exterior surfaces of the museum and café. The parti of twisting petals defined the control geometry at the edges of the four nested forms, defining the sweeping surfaces that would be clad in the aluminum diamond panels. The gaps between the pedals are zones where light is let into the galleries, ribbons of negative space that wrap around the central atrium and then diagonally above each top-floor gallery before turning down vertically at each corner of the museum. These voids would be detailed with glass and smooth aluminum composite panels that reach out to meet the exterior diamond panels.

The power of Rhino is two-fold: it allows the user to quickly model complex 3D geometry as well as harness the power of parametric modeling to subdivide surfaces based on a set of user-defined rules. In the case of the Olympic museum, the edges of the four petals were located, and then ruled surfaces (a twisting surface comprised of straight lines—think of taking a sheet of paper and twisting it) were developed to connect them, forming the rolling surfaces of the building skin.

In tandem with this effort, DS+R developed a script in Grasshopper (a Rhino plug-in) that parameterized the subdivision of the skin surface into a series of diamonds. The script interface starts with sliders that allow the user to input the number of equal divisions each edge should be divided, and then arrays U- and V-lines across the surface of the geometry created in Rhino, subdividing it into quadrilaterals of various sizes. The Grasshopper script then uses the intersections off the U- and V-lines to generate the diagonal lines that make up the edges of the diamond panels. DS+R was also able to input the parameters of a generic diamond panel (fold line, depth of return legs, etc.) into Grasshopper and then apply these parameters to each 2D diamond created by the intersections off the U- and V-lines to create the articulated 3D design surface.

Despite the power inherent in this parametric modeling, the process is not perfect. DS+R would review the geometry created by their Grasshopper script and re-model areas that were not aesthetically resolved. In general, about 80% of the panels created by the Grasshopper script were satisfactory while the remaining 20% required some amount of remodeling in Rhino.

Once the skin geometry was developed, it was relatively easy to modify as the building evolved from schematic design all the way through the construction document phase. Initially, the Rhino design surface (without the diamond subdivisions) was exported as an .sat file that AMD brought into the master Revit model as a Revit family. This 2D surface (in the sense that it had no depth) was then used to create roof and wall assemblies of predetermined buildup by using the “model by face” tool in Revit. This Revit geometry was then shared with the structural engineer so they could design the structural frame to meet these assemblies. As the design evolved, models were shared between the various team members to refine their models and get them into alignment. For example, if a programmatic change resulted in a change to the structural model, the model would be converted to a 3D .dwg and brought into Rhino to allow DS+R to modify their skin as needed. This updated surface was then used to update the respective exterior assemblies in the Revit model.

This iterative process worked well throughout the design process (which continued after CDs for a six-month collaborative design assist phase with the skin subcontractor) and ultimately resulted in the final skin geometry file, complete with panel edges and 3D articulation, being issued to the contractor as part of the contract documents.


Once the final design surface was issued by AMD, the skin subcontractor, MG McGrath, took this Rhino geometry and began their work required to fabricate and install the metal panels. An early step was to mock-up a typical aluminum diamond panel to see how material limitations affected folds, returns, connections, etc. MGM went through between twenty to thirty iterations before finally settling on a typical panel design and connection detail that they and the design team were satisfied with.

The next step was to get the design surface into Catia and define the control geometry common to all the panels. Similar to Grasshopper, Catia has a parametric wire-frame engine that allowed MGM to define this low-resolution geometry and use it to begin generating the basic panel layout. Next, more complex parameters derived from the physical panel mock-up, and tolerances defined by the design team—including the 3/16" panel joint dimension—were added to the Catia model to generate a first-pass at all the diamond panels. Unique conditions across the exterior skin like expansion joints and transitions required additional modeling of the diamond panels in Catia. In addition, all the ACM return panels on the project were unique and required individual modeling in Catia.

Along with the panel geometry contained in the Catia model, a master fabrication template was developed that takes this geometry and applies a set of preassigned parameters to automate the fabrication process. This template can be applied to all the panels on the project to develop unfold drawings; one-to-one CAD files for laser cutting, etching of part number and fold lines; 11 x 17 assembly drawings; as well as Excel files of all the data to track parts, production, control geometry, layout, etc. By utilizing this master template approach and integrating it with the Catia model, MGM has found they can reduce the teams on this front-end portion of the project to only one or two people who manage the model and the resulting data.

After the panelized design surface has been developed in Catia, the as-built conditions of the framing in the field are surveyed to ensure the design surface is correct. Predetermined points on the light gauge exterior framing are scanned to verify the framing is within a 3/8" of tolerance in three dimensions (the amount of variance the z-girts can make up). In addition, point cloud scanning is utilized to get a sense of the overall geometry and if there are any unforeseen deviations. While the light gauge framing is being installed, this surveying is done daily, and a variance log is provided to the framers to correct issues as they go. Once all the framing is in place, a survey is provided to MGM to do final checks with the Catia model. As a last check, a scan of the sheathing is provided to make sure there are no anomalies on the top surface receiving the z-girts and the panels. All of this data is fed back into the Catia model and any updates to the geometry are made.

By using the Catia model shared on the cloud as the single-source of truth for the fabrication and installation, the field team can install panels on one elevation of the building, the assembly team can proceed on a second elevation, while the detailing team can finish on the third. All relevant, current information is located in one place and there is never issues about version control and design changes happening in a silo.

Once an elevation has been released for fabrication, the cut files and assembly drawings are sent to the shop for fabrication. The first step is to lay out the unfolded panels on the 48" x 120" sheets of anodized aluminum to both maximize yield and to keep a consistent grain direction across all panels. The panels are then laser cut and etched to include both the panel number and fold lines. The next step is to fold the panels which is all done manually on press brakes, with each panel requiring twelve to thirteen folds. Pins, stiffeners and sealant are then applied to finish each panel. MGM has developed an in-house software solution that allows them to track each panel from fabrication to shipping and to install. Each panel is given a bar code after fabrication has begun and is scanned at six stages of the process: cutting, bending, crating, shipping, receiving and installation. The user is prompted at each stage to conduct various QC operations and the data collected along the way helps MGM track efficiency and fine-tune the process as they go.


While the panels are being fabricated in the shop, the field team is working on-site to install the z-girts from which the panels will be hung. A series of control z-girts with the intersection point of four panels pre-etched into them are shot in with a robotic total station marking precise locations every two to three rows of panels so the installation team can ensure they are going up correctly. Standard z-girts are then filled in around the control girts, after which the entire assembly is water-tested one more time to ensure that there are no issues with unsealed penetrations.

Once the water testing of the substrate is signed off on, the installation of insulation and panels can begin. Work proceeds from the bottom up, with each panel being screwed to the z-girt and the next panel being connected to the next with a series of pins and receivers. Along the way, the panel tracking system will prompt the installers to conduct various QC checks every few panels, including checks of the water proofing, layout, panel quality, etc.

The final coordination of as-built conditions happens at the interface of the diamond panels and the ACM return panels, a lesson learned while constructing the mock-up. Once the diamond panels are laid out on an elevation, the edge conditions are scanned and surveyed (both points and point-cloud) and this information is fed back into the Catia model to finalize the geometry of the ACM return panels. The return legs of the ACM panels must align precisely with the edge formed by of the adjacent diamond panels which runs from straight to curved several times along its length, all while maintaining constant joint dimensions both between adjacent ACM panels and between ACM and diamond panels. ACM panels are then fabricated in the shop, a much more time-intensive process than fabricating a single diamond panel. Where the ACM turns from the vertical to horizontal at the top of the building, the ACM panels are semi-conical in shape and require hand-braking of the metal at half-degree increments every 1/16" of an inch, taking nearly twelve hours of fabrication time per panel. Once complete, the ACM panels are shipped to the site to finish the skin installation.


By no means has this process been perfect. At the time of writing, no panels have actually been installed on the building. Z-girts are being installed and panels are being fabricated for the first of the four elevations. The mock-up—after multiple panel installs, modifications and revisions—has been approved by the design team and met with favorable response, and the design and construction teams are optimistic that the project will ultimately be successful. It is beneficial, though, to point out a couple of lessons-learned to improve the process both moving forward on this project and also on future projects.

The benefits of having one point of truth on a project—a federated model that brings together all the 3D building information of the various team members—are numerous as we have seen. However, not all subcontractors rely on this 3D model and the sophisticated layout tools required for executing the design. There is still reluctance amongst certain trades to commit to this technology—be it resistance to change, cost for adopting technology, difficulty in training employees, etc. There have been occasions on this project where a subcontractor has not utilized the coordination model for their scope and proceeded with an attitude of “we’ll make it work.” Not to say that they have not been successful: their work meets project specifications and exhibits the necessary craft. However, it may not match what has been developed in the coordination model, and the general contractor and design team must take this into account as they move forward so they are not taken by surprise.

Another issue that requires careful coordination is the initial location of the project base point: the physical point in space which should be common to each model (despite the best efforts to have one federated model, there will often be multiple models early in a project with multiple model-specific base points). Early mis-coordination of this base point between subs resulted in errors in the construction of the mock-up, with the structural steel being out of alignment with the subsequent framing for the skin and curtain wall systems. Luckily, this is what mock-ups are for: to work through all the issues you will likely encounter during construction and to ensure there are processes in place to avoid these mistakes moving forward. In the end, these issues were resolved and the team is confident they will not resurface on the main building.

Ultimately, the design and execution of the skin system on the USOM would not have been possible without the use of the latest 3D BIM and parametric modeling software available, at least not within the budget and time-frame required. Through an iterative process of collaboration between the various team members, the design of the skin system evolved to a point where it could be efficiently fabricated and installed, despite there being thousands of unique panels arrayed over a complex surface. Moving forward, the architecture profession will continue to use the latest digital tools to push the limits of what is possible, but it is only with skilled designers and craftspeople behind the tools that these designs will ultimately be successful and stand the test of time.

Next post