There is a tight relationship between innovation and inspection. The more innovation in a new product or process, the more dimensional measurement is needed to ensure that nothing goes (or went) wrong. Common sense tells us that any change increases the opportunity for errors.

Moreover the innovation / inspection relationship is probably not a linear one. Even a modest increase in innovation—more advanced products, for example, or bigger modifications to processes—can lead to a doubling or tripling of the measurements needed. Great examples of this are two Wisconsin companies—J.P. Pattern and Waukesha Kramer Inc. (formerly known as Waukesha Foundry).

J.P. Pattern's (JPP) and Waukesha Kramer's skills in handling prototypes—and willingness to innovate—won them a crucial place in America’s push to someday end its energy dependence on fossil fuels. They won a contract to produce the prototype for 18 castings for the National Compact Stellarator Experiment (NCSX). This is a program of the Princeton Plasma Physics Laboratory (PPPL), a unit of the U.S. Department of Energy managed by Princeton University in New Jersey.

JPP and Waukesha relied heavily on a 3000i portable coordinate measuring machine (PCMM) from ROMER Inc., Farmington Hills, Michigan. ROMER provides pre-sales and post-sales support, application assistance, training and on-going telephone support for the complete CMM system

In their work on the NCSX prototype, JPP and the foundry amply demonstrated the principle that Innovation Multiplies Inspection. “We measured every dimension and every surface we could reach, and that was all of them,” said John Puhl, president of JPP in the Milwaukee suburb of Butler, Wisconsin.

“This was an R&D project for us and the customer so we checked everything on the mold and casting we could think of,” said Bill Norris, vice president of product development for Waukesha Kramer.

Puhl and Norris agreed that, fundamentally, the 3000i was the cornerstone of their efforts to manage the risks of dimensional variance that always accompany innovation. Their inspections and verifications went far beyond the already extensive requirements of the Princeton contra.

For Waukesha, the NCSX casting—poured in June 2004—was also a quantum leap into foundry research and development. Patternless casting was an innovation all by itself.

A time- and money-saver for large castings produced only once or twice, chemically bonded (“No-bake”) sand is allowed to “set up” or harden into blocks for the top (cope) and bottom (drag) halves of the mold. The reverse of the shape of the part to be cast is cut into the hard, smooth and consistent sand, usually with a five-axis CNC router.

In conventional casting, the sand sets up around cope and drag patterns. These replicate halves of the part and allow for shrinkage as the metal cools. By eliminating conventional mahogany or plastic resin patterns for large castings, several weeks can be cut from lead times and costs cut by as much as $200,000.

And there was the physical size. “The NCSX prototype was huge for us,” said Norris. “The mold measured 128 by 108 by 70 inches. It was so big that we had to have a rigger come in twice and move the mold under the router’s spindle,” he said. The XYZ axis travel of the router is 96 by 144 by 36 inches.

The mold weighed about 60,000 pounds and the raw casting 7,500. After cleaning and rough machining, the finished casting weighed about 5,500 pounds. Without the portable 3000i, there would have been no way to inspect something that big other than “hand” methods such as measuring tapes, templates and height gauges. They probably would have been unacceptable to the NCSX program managers.

The metal is a proprietary nonmagnetic stainless steel similar to AISI 316.

CHALLENGE: When Nothing is The Same

The biggest dimensional issue confronting JPP and the foundry was that no one had ever produced a large steel casting shaped anything like this before, new process notwithstanding. The casting is an open ring with an inner diameter of more than four feet. Two large, irregularly shaped “wings” required three molds, a drag and two copes rather than one of each. The mold therefore had two parting lines instead of the usual one.

A T-shaped feature “roller-coasters” around the inner diameter, and two flanges run around the exterior of the part. Eight cores were required to mold these features.

The engineers at Princeton in partnership with Oak Ridge National Laboratory (ORNL) in Tennessee designed this casting as part of a toroidal plasma-containment structure with a helical twist. The 18 castings—the production run that JPP and Waukesha prototyped—will hold 18 cryogenically cooled electromagnets. Their magnetic fields will keep the NCSX fusion plasma in place. A fusion plasma has an operating temperature of around 100 million degrees Centigrade, similar to that of the stars, which these fusion energy experiments mimic. No known material can withstand such temperatures.

A high demand for dimensional data was built into the structure of the project. There are a lot of players in the NCSX program, nearly all of them top engineers at leading laboratories and large companies and physicists. They have voracious appetites for data, dimensional and every other type that were plugged into the contract.

In addition, the prototype was to verify that the castings could be made, Puhl and Norris pointed out, and to demonstrate that feasibility to foundries bidding on the 18 production castings. For PPPL, the prototype helped establish a risk-management framework so bidders wouldn’t feel compelled to add large bid premiums to cover unknowns. For the NCSX, it was a rite of passage.

Before pouring, the casting’s three molds and eight cores were exhaustively measured during and after machining. “First we measured all the molds and cores against the against the CAD model which had an allowance for shrinkage,” said Puhl. Shrinkage is always a crucial variable in foundry operations.

“As the metal in a casting of this size cools, it shrinks about two and a half inches overall,” he pointed out. No one can ever exactly calculate the shrinkage because there are so many variables.” The biggest cause of variability is the geometry of the casting’s thickest and thinnest sections.

“We measured the actual casting with the ROMER to find out the actual shrinkage,” Puhl added,

For PowerINSPECT, this inspection point-to-CAD comparison is a basic function; it is done as the points are gathered. The main dimensional-measurement tool for both the mold and casting was three-spheres / three points analysis,” Puhl noted.

To help with measuring the casting, JPP put dozens of one-inch diameter hemispheres into the mold surfaces at critical locations. They projected about three-eighths of an inch out from the metal surfaces as reference points for measuring. Since this was not finish work, the as-cast locations and dimensions were close enough for JPP’s purposes.

“For us, this was a proof of concept for the machining required for patternless casting.” Despite “Pattern” in the name, Puhl’s company does far more short-run, fast-turnaround machining and mold making than pattern making.

The basic goal in all the measuring and inspection was to make sure the part can be machined out of the casting, that there is enough metal on every surface whether molded or cored. “Actually, the shell tolerance was not all that tight,” Puhl said, “plus or minus a quarter of an inch.”

Machining stock is metal poured in excess of the dimensional specifications. Intended to be machined away, it accommodates any slight shift in core placements, or mold movement at the parting line and warpage from heat-treating—not just shrinkage.

SOLUTION: Leave Nothing Unmeasured

“This was a VERY complicated project and the most complicated mold we ever made,” said Norris. “All 11 pieces had to go together precisely in a one-off job with not one but two parting lines. Also, the mold was so big that it pushed the foundry’s physical capabilities for size and weight.”

From the outset, his measurements linked inspection with innovation. Inspection began as soon as the casting had cooled enough to clean and work on. Its dimensions were compared with those of the mold, just as the mold’s dimensions had been compared to the CAD files. “We used J.P.’s ROMER arm to look for any possible errors we might have had made,” Norris continued, “in the tool paths for the 5-axis CNC router [which he programmed], in the calibration of the machine tool, or a CNC program hiccup.

“All these are possibilities for error and to minimize their impact we need independent hardware and software verification,” he said. The 3000i's role in all this was “huge.”

Added Puhl, “We measured every major surface. Where there was a hole, we measured its inside surface as if it were a geometric feature and not just a location point.” Some as-cast holes served two purposes, as locators and to accommodate rods and bushings during assembly, he noted.

“The full set of measurements took nearly a full day, six to seven hours,” Puhl said. “That included three or four critical cores plus time for setup and cleanup. The only delays we ran into a few dimensional anomalies caused by a core shift of, say, a sixteenth of an inch. The arm played a major role in communicating that kind of process verification,” he added.

“We were not so much worried about the fine points of inspection as simply whether or not there was enough metal to machine,” Puhl continued. “For that, the red-green-blue indications of PowerINSPECT ‘weather map’ were ideal for our purposes. What we most needed we got right away from PowerINSPECT. That was to check whether profiles of these surfaces were in or out of tolerance” as shown by the original CAD files.

After casting, Norris said, “we measured it for any distortion that occurred during cooling and to see if any of the cores had moved during the pour. We focused on the tolerances in the critical areas such as that T-shaped cross section around the inside of the casting.

“Shrinkage was not an issue for us,” he added. “We were leaving half an inch of stock for machining, so we knew there was enough metal there.”

Norris was also looking for anything out of specification caused by:
• Removing the raw casting’s gates, risers and other mold “plumbing.” These are removed by snag grinders and welding torches, and both can impart stress to the casting.
• Weld repairs, if necessary, as their heat imparts stress.
• Sand blasting for cleaning, which introduces stress.
• Heat treating, whose warpage was checked with measurements right before and right after.
• Anything uncovered by the 100% X-ray or 100% die penetrant tests.

“During measurement of the mold and the casting,” said Puhl, “for the ROMER users, there was a lot of twisting and turning and going back and forth from one side to the other. There was a lot of stretching out to surfaces while standing in the middle of the molds and, later, in the casting. “That would have been a major nuisance to do without the ROMER arm’s infinite rotation and portability.”


BENEFIT: NCSX Design Review Team Says ‘Go’

JPP’s and Waukesha Foundry’s plans and inspection proposals were blessed in the NCSX Final Design Review, a formal step in all such government programs before money start to flow to contractors. The design review report said, in part, “Prototype R&D activities, including those at JPP and Waukesha, have shown that the manufacturing plan is sound and that the risks associated with sector fabrication [the 18 production castings] can be managed.

“Inspection / quality assurance plans are in-place and were successfully exercised on the prototype hardware,” the report continued. The Stellarator design review was chaired by Carl Strawbridge, deputy director of Spallation Neutron Source Project at ORNL.

Summarizing, Bill Norris observed, “We may buy our own portable CMM. This project and the success we had with patternless casting supplied a big part of the rationale for it. Dimensional inspection of molds is usually done with templates, and they are really only good for looking for large errors such as core shifts and mold shifts,” he pointed out.

“Templates obviously cannot reveal tiny increments in tolerances; they are just not that precise. Going patternless,” Norris continued, “was our decision, not J.P.’s or Princeton’s. We chose to bid it that way. No other foundry did.” Asked why he took the risk, Norris cited

• The fact that the NCSX job was a prototype, and there was a high possibility the geometry might be changed. Making those changes in patterns, molds and cores would be much more time consuming that modifying the CNC program for the router. “The beauty of patternless is revealed by engineering changes,” Norris said.

• The high cost of pattern equipment for large castings, an estimated $150,000 to $200,000 for this job.

• Lack of space at the foundry to store all the patterns and core equipment.

“Patternless casting is still new in the industry” Norris said. “This success of this project may have put Waukesha in a uniquely strong competitive position.”

NCSX has a budget approaching $90 million and several alternative approaches are under development, funded at similar levels. All of America’s government-run nuclear labs—including Oak Ridge, Los Alamos, Princeton Plasma Physics and Livermore—are working on fusion. The combined projects account for hundreds of millions in the DoE budget every year.

JPP is well known to the fusion experimenters. Counting NCSX, it has worked on components for three fusion projects. “We are bidding on a fourth,” said Puhl, “and it will be even more complex than this one, if that’s possible. With the ROMER 3000i, we know we have the dimensional-measurement capability to handle anything they might throw at us.”