The assignment: Reverse engineer a railroad tamper casting based on 40-year-old drawings that don’t come close to matching the actual example. It would have been impossible to capture all the surfaces with a scanner, and you’ll need to invent a couple of new approaches to create your digital model and verify its accuracy. The final product – a complete set of accurate digital patterns that can be used to make new castings, including cores and core boxes – must be delivered in two weeks.
This is a true story, and the assignment described here was the request was made to Advanced Simulation Technology (www.simtec.com), a design engineering group that promotes its ability to leverage technology to deliver what customers need when they need it.
AST supplied the ingenuity for the project, while the technology came from Geomagic, which made it possible to generate an accurate digital model from an existing tamper within the two-week deadline.
A tamper compacts gravel under a newly laid railroad track. Vibrating heads work like a jackhammer, pounding the gravel to provide a level track bed on straight sections of track, and at the proper elevation on curves.
HUB Corp., Roanoke, VA, was contracted to machine new tamper casting patterns by Southern Cast Inc., Charlotte. HUB had worked with Chip Potter, president of AST, on previous projects and recognized his ability to deploy available technologies.
“Chip certainly has the technological knowledge, but the key is that he listens to what the customer needs,” says Don Cloeter, who with his brother John, owns HUB.
AST specializes in analytical modeling for design and manufacturing, including two areas key to the HUB project: interfacing solid modeling with CNC machining, and scanning and surfacing large, complex parts. The tamper casting assignment started with just two resources: the 40-year old drawings and an existing casting.
The drawings provided only basic dimensions. There was no information on draft angles that enable the pattern to be removed from the sand molds without damage. Internal features, different floor depths, and blend radii that define smooth edges were among the data not defined by the drawings.
AST started by using a GOM ATOS IIe scanner from Capture3D to collect data for as many surfaces as possible on the original casting. The ATOS scanner uses white light to project fringe patterns onto an object’s surface. The patterns are captured by cameras positioned on each side of the scanner’s sensor head.
It took AST about a day and a half to complete the scanning. The process was made more difficult than usual by the need to move the 300-lb casting to capture as much surface data as possible. The casting had numerous surface imperfections picked up by the scanner, and internal surfaces under the flanges and inner holes could not be captured.
Polygonal data from the ATOS scanner was brought into Geomagic Studio, software used by manufacturers to transform scan data into accurate models for downstream CAD, manufacturing, and engineering analysis.
AST cleaned up the model in Geomagic Studio, removing holes, smoothing surfaces, and automatically filling in sections not captured by the scanner. Normally, the next step would be to generate a NURBS surface model from the polygon model. But, the job requirements led AST to a different approach.
Instead of generating surfaces from polygons, AST used Geomagic Studio to automatically create a symmetry plane based on an educated guess of where it might be located. Then, AST cut sections from planes that were created parallel or perpendicular to the symmetry plane.
“The ability of Geomagic Studio to provide accurate symmetry planes using intervals that we defined based on offsets was critical to making our process work,” says Steve Lelinski, AST’s research engineer.
The coarsely cut sections were brought into Pro/Engineer for alignment and to create dimensions and features, and the original drawings were used as a conceptual reference. As the model was refined, AST shuttled back and forth between Pro/Engineer and Geomagic Studio. Geomagic Studio was used to cut the model into ever-smaller orthogonal slices, starting with 2-in. sections and whittling down to sections just 0.125-in. thick. As the model was refined, details were checked and revised in Pro/Engineer.
“The most important reason for this approach,” says Lelinski, “was that we needed a parametric model that could be edited. Also, the part had to be a combination of the scan data and the original drawing.”
A finished surface model is difficult to change, according to Lelinski, and AST needed to be able to revise features on the part easily and quickly. In fact, once the model was finished and HUB presented it to the customer, a feature needed to be moved. Because AST had created a parametric, section-based model, the change was easy to make.
“If we were working with a surfaced model,” says Lelinski, “we would have had either to start at the beginning by moving the polygons and resurfacing, or to break apart the surfaces, make changes, and attempt to merge them back together. Neither of these options is quick, nor easy.”
AST’s sectioning approach lent itself to a model that needs to reflect both the intent of the drawing, and the actual details of the physical part. Arriving at this kind of combination model would have been difficult using the typical scan-to-surface procedure, and changes would have been more time-consuming, according to Lelinski.
“By creating the model in Pro/E using the sections from Geomagic Studio as our guide, we were able to put features in their proper location, with the correct dimensions and thicknesses from the beginning,” he says.
The flexibility of the sectioned model also made it easy to determine draft angles for the casting model. AST used the scan data to obtain measurements at the top and bottom of surfaces, then mathematically calculated the angles. Then, the angles were adjusted when cross-sections were laid into the model.
The final bit of AST ingenuity came during verification of the finished model. This was done through a process that Chip Potter calls “reverse inspection.” Instead of the typical process of inspecting an as-built part against the CAD model, AST used Geomagic Qualify to inspect the CAD model against data from the tamper casting.
Color-mapped visual results automatically created by Geomagic Qualify made it simple to see deviations between the CAD model and the scan model. Overall, the CAD model was accurate, except for problems with an off-center bore and variable wall thicknesses. After about a half-dozen iterations correcting problems, fine-tuning draft angles, determining casting shrinkage, and re-inspecting, AST arrived at an accurate model. The entire process – scanning to processing to verification – was accomplished within two weeks.
There is no good way of telling how much time was saved by the AST approach, simply because there was no other way to reproduce the casting accurately within a reasonable period of time. It would take hundreds of hand measurements to match the equivalent of the one-eighth-inch sections created in Geomagic Studio, and the curved surfaces of the casting would have made accurate measurements impossible. A coordinate measuring machine (CMM) wasn’t suitable for the job either, because the probe wouldn’t be able to account for shifting, shrinkage, and warpage in the physical casting.
“There was no other way to do this cost-effectively, since the drawings did not represent the casting as built,” says Potter. “We were given this project after others had tried and failed. Using our methods, in two weeks we were able to build a complete parametric model that captured design intent that was lost 40 years ago.”
The completed model was delivered to HUB, which imported it into SurfCAM software to make modifications such as pulling out cores, creating split lines for core boxes, and generating patterns to fit the existing casting. Tool paths were created and the model sent to HUB’s CNC machines for cutting. The finished casting patterns were cleaned up and lacquered, then delivered to Southern Cast before the 30-day deadline passed.
For HUB’s Don Cloeter, the project proves that skill and technology can provide cost-effective answers to problems that might normally be considered impractical to solve.
“I was pleased with the quality of AST’s work, and my customer was satisfied with our work. I think this shows that skilled professionals using new technologies can solve difficult problems at a reasonable cost if they are given the opportunity. Our methods might appear more expensive than offshore outsourcing at first, but in the end you’ll pay about the same or less for a better solution.”