Reverse engineering, CFD analysis push U.S. Bobsled Team down high-tech design path
The push-start is crucial for bobsledding; it and gravity are all the power allowed in competition. It’s ironic then that the sport’s sled, which requires no motor and has a deceptively simple aerodynamic design, would attract the attention of former NASCAR driver Geoffrey Bodine and chassis and suspension design specialist Bob Cuneo.
That’s where the modern success of the U.S. Bobsled Team begins. Due in part to improvements made by Bodine and Cuneo’s non-profit Bo-Dyn Bobsled Project, the U.S. team earned a gold in the woman’s two-person event, and silver and bronze in the men’s four-man event in the 2002 Winter Olympics in Salt Lake City, Utah.
But Cuneo wants to further that success with new sled designs based on computational fluid dynamics (CFD) studies of the two- and four-person sleds. He sought out the expertise of Capture 3D Inc., Geomagic, and the Massachusetts Institute of Technology (MIT) to help create a 3D computer model and perform the CFD analysis.
Sled design on a shoestring
By Olympic standards, bobsledding in the United States is managed on the smallest of budgets. From the 1960s to the 1990s, the team was forced to work with sleds formerly used by the European teams. Now the U.S. Bobsled Team designs, tests and manufactures its own sleds for less than a tenth of the budget for many European teams.
Cuneo has been involved in designing sleds for the U.S. team since 1992, when Bodine became interested in helping the U.S. team with its equipment and training. The two worked together for years on Bodine’s NASCAR vehicles. The Bo-Dyn Bobsled Project is a combination of Bodine’s name and Cuneo’s company, Chassis Dynamics.
The pair was armed with decades of racing knowledge they quickly adapted to bobsled racing, but they needed to go beyond old-fashioned hand methods – creating shapes based on their expertise and experience, building a prototype out of foam and plastic, then making the final sled.
“In this age of computer-aided design (CAD),” says Cuneo, “this was the wrong way to go about it. The process was extremely labor-intensive, but it’s the best we could do.”
Cuneo knew the next logical step was creating a 3D computer model that could be used for CFD study. Since no native CAD data for the sled existed, he first needed to capture the sleds and athletes in 3D, create a 3D model, and find the right CFD solution.
The solution began taking shape at a round-table discussion sponsored by the U.S. Olympic Committee that brought together equipment designers from a number of sports and experts from several engineering fields. There Cuneo was introduced to Kim Blair, the head of sports technology at MIT, where new software to analyze low-speed aerodynamics was under development. Blair wanted to make the software available to the U.S. Bobsled Team.
To get the project underway, Cuneo began searching for a company to scan the two- and four-man sleds with the athletes inside. He had no takers. The main drawback was including the athletes in the scan; most scanners would require that the athletes remain still for hours in a difficult position. Simply breathing during that time span would repeatedly move them more than the 50-micron accuracy they needed from the scan.
Cuneo thought of designing a sled in a CAD program, using another program to generate and manipulate models of the athletes, then combining the two for CFD testing. But an engineer wouldn’t be an athlete, so he or she wouldn’t be able to precisely position how the athletes sit, ride and brace during races; all of the locations would be best guesses.
“The athletes’ participation can’t be duplicated,” says Cuneo. “You can design a superior sled on the computer, but once you put the athletes in, it’s junk because the estimations are wrong.”
According to Cuneo, this is because a bobsled’s aerodynamics more closely match those of a motorcycle than a car. With cars, the vehicle’s body is the main aerodynamic unit. With motorcycles and bobsleds, people are part of the unit since they are exposed to the air. Scanning the athletes in the actual sleds was the only way to create an accurate 3D model and study the entire mechanism.
Optical triangulation-based scanning
Cuneo found the 3D data capture answer at Capture 3D, a scanning solution provider in Novi, Mich. Capture 3D was willing to take on the challenge, and the company knew it had the right equipment to accurately capture the athletes and sleds.
The Bo-Dyn team traveled to the Capture 3D facility with one two-man sled, one four-man sled, and four U.S. Bobsled Team athletes. There the sleds were scanned and the data processed into polygonized meshes over the course of three days.
Scanning preparation began with sleds being lifted onto sawhorses so the geometry could be captured from all sides, including the under carriage. The sleds were then covered with a powder to reduce surface reflection. The final step was to apply barcoded identifying markers and flat white circular targets to the sled. The markers are part of the Capture 3D TRITOP digital photogrammetry system.
The TRITOP system is based on the theory of triangulation, using two known points to determine the location of a third. The markers and a scalebar are recorded by a high-resolution reflex camera from pre-determined views. These views are bundled together in TRITOP’s software, which automatically recognizes and calculates the xyz coordinates of the markers and applies scale to the locations.
Capture3D imported the xyz coordinates collected by the TRITOP system into the ATOS IIe bright-light scanner software. As the scanner was positioned around the sled, it used the predefined coordinates to determine where to scan the object and how to orient the data it collected in the sled’s global reference frame. Each point-cloud patch from the scan data had a unique identification and placement, enabling the data to automatically align in the ATOS software.
“A benefit of combining the TRITOP system and the ATOS scanner is increased accuracy on larger objects,” says Steve Albrecht with Capture 3D. “The marker coordinates communicate that there’s only one place for the scan data to go in the final model. Alternative laser scanning technologies would have required us to orient and align a great number of scans by hand – a huge undertaking for a large volume scan like this.”
When the athletes were scanned, markers were applied to their helmets, which also had a powder coating to reduce shine. The athletes positioned themselves in racing formation for seven seconds while 3D data and marker center point locations from their helmets and visible torsos were recorded. Each helmet was then scanned alone to get full definition and tighter tolerances.
Because the coordinates on the helmets were assigned while the athletes were in competition position, each separately scanned helmet was automatically aligned on the final point-cloud model as if the athletes and helmets were still in the sled. This process is called digital assembly. There was enough data around the athletes to generate a complete model even though no makers were applied to them.
The combination of the TRITOP system and the ATOS IIe scanner gave Albrecht an accuracy of 50 microns (0.002 inches). The total process required 103 scans, a number that would have been considerably higher had the ATOS scanner not been able to scan such large areas.
NURBS surfaces in hours, not weeks
Once scan data was collected and aligned from each sled, the athletes and the helmets, Albrecht created final stereolithography (STL) files, or polygonized meshes, in the ATOS software. The files were then sent to Geomagic to create final surfaced 3D models.
Srdjan Urosev, senior applications engineer with Geomagic in Detroit, Mich., imported the STL files into Geomagic Studio. Geomagic Studio automatically generates an accurate digital model from any physical part based on data input from 3D scanners.
Urosev did a quick preview for each sled using Geomagic Studio’s push-button surfacing function. He then cleaned up the data where needed, filling in holes and closing gaps to complete the models.
Urosev worked on the scan data for the sleds and athletes separately, then merged the scans to create the final models. All of the data aligned instantly using the target registration from the scanning session. He also extracted planes and axes that ran through the base of each sled to use for alignment within the CAD package.
“Typically an application like this would take weeks to reverse engineer,” says Urosev. “With Geomagic Studio, I only spent two to three hours on each model.”
In the end, there were two models of the two-man sled and two of the four-man sled. Urosev compared the final NURBS surface models with the original STL file to ensure tolerance zones were met. The final files were exported as NURBS surface IGES files.
MIT is now using the files for CFD analysis, which can take a few months – time that Cuneo is willing to wait.
“We’re really excited to take this next step,” Cuneo says. “It’s really ambitious, but it would be nice to use the results for a new design in time for the 2006 Winter Olympics in Torino, Italy.”