Phase Four

By Baselinemag  |  Posted 2002-11-01 Email Print this article Print

Milestones along General Motors' route to becoming the first all-digital automaker.

Phase Four
Months 6 to 16: Iterative Development
Data Types: Design and simulation models of all vehicle components, tools and dies required for manufacturing, crash and performance testing Storage Requirement: 900 GB

While building the concept car has created the "skeleton code" for the vehicle, it hasn't covered the real work of developing the design. Now development engineers must figure out how to build it, and how to make it work in the real world.

Executing a handoff from styling to development five years ago was no easy feat—even moving data from one designer to another usually meant dumping files to tape and hand- delivering them. But with the data from styling now stored in TeamCenter, the transfer is almost seamless—and so, in some respects, is the transition from styling to development.

Even with a full mock-up, there's lots of room for engineering changes in the product. Based on financial goals, the product managers look at the bill of materials and any trade-offs they may need to make on the design, in order to get it built under target cost. Cheaper parts or restyling lower the cost of manufacturing.

Gutmann's team finds low-hanging fruit here. Engineering knowledge was scattered across more than 150 Web sites within the GM intranet, and there was no effective way for engineers to access, update or add to it.

Those 150 sites have now been converted to a single "e-Product Development" portal. Engineers self-publish data. The site gives central access to all of GM's engineering applications. There's also an iMan portal—tied to TeamCenter—that gives access to all of GM's Unigraphics data through a browser.

Engineers also can check files in and out directly from Unigraphics, through TeamCenter. In effect, the TeamCenter database looks more like one big disk drive to the designer.

As the vehicle approaches its final configuration, GM puts the design through a series of tests. Now, suspension checks and even crash-testing are done by "math,'' using data on vehicle types and GM's farm of supercomputers.

This has been done with Silicon Graphics systems almost exclusively, but GM recently began replacing them with IBM's Regatta supercomputers. The reason is simple, according to GM's CIO for North American Vehicle Operations, Terry Kline. Regattas recover from crashes better.

"Supercomputing time is expensive," he says. "If you're halfway through a 20-hour simulation batch, and the system crashes and you lose everything, that's expensive." IBM's supercomputers, however, are equipped to recover from a crash close to the point of failure, Kline says, so the only lost time is what's required to bring the system back up.

Instead of building as many as a hundred different development-phase prototypes of a car or truck for testing, GM only needs as few as a third of that number now—to verify the results of tests run on the supercomputer.

For example, suspension systems are no longer put through their paces for the first time at GM's Milford, Mich., and Mesa, Ariz., proving grounds. Instead, they undergo virtual tests that let engineers tweak and tune a suspension multiple times before it is even built.

This also means that engineers can more rapidly step through iterations of design, as a car or truck gets built. Changes in materials planned for use in the manufacture of the vehicle can almost immediately be assessed for how they could affect vehicle performance and crash-worthiness—before the parts are ever stamped or milled to specs.

At this point, supplier involvement mushrooms rapidly, as GM sends out requests for bids on the various components of the design. All the engineering data to support those bids comes out of TeamCenter. In the past, getting engineering data to suppliers required reaching a secure File Transfer Protocol (FTP) site over a value-added network—possibly over a dial-up connection. Otherwise, the data had to be sent on tape, a process that was often faster than downloading from the FTP site.

But now, 17 of GM's top suppliers are connected directly to its TeamCenter system over its high-speed GigaMAN network. Engineering data needed to make bids is replicated across the wire to their servers.

Other suppliers can get their data by connecting to GM's iMan portal over a virtual private network, or ANX, an automotive transaction network operated by SAIC. In either case, the bandwidth available to pull the gigabytes of data that make up even a portion of a virtual vehicle is an order of magnitude larger than was available to even the largest of GM's suppliers five years ago.

One of those suppliers is Tesco Engineering, which designs and manufactures the equipment that make closures for GM cars and trucks—their doors, hoods and tailgates. Tesco gets a look at the design files for many GM vehicles right at styling freeze.

"We have the right of first refusal on equipment projects that GM's MFD (metal fabrication division) can't handle," says Ken Monville, a program lead at Tesco. Even in cases where GM builds its own dies and equipment, Tesco often designs the dies. This September, design work was about all that was coming out of GM, as it had pulled most of its manufacturing in-house to keep the metal division operating near capacity.

In the company's headquarters just a mile from the home court of the Detroit Pistons, Tesco engineers test the latest dies and robots being built for another customer, DaimlerChrysler. A pair of robots feeds sections of door into a die on an electric press, the press closes and opens, and the robotic arms pull out a complete "hemmed" door, the edges of the outer section neatly crimped over the inner.

"The cycles on those robots are still about 50% under what we need," says Jim Toenneskoetter, the president and COO of Tesco Engineering. The team working on the robots adjusts the set of directions that tells them which way, how fast and how much to move each gear, piston and actuator.

When Tesco was hooked into GM's GigaMAN network four years ago, it allowed them to pull down design files for the doors, hoods, trunks and tailgates almost immediately after the styling of the car was complete. It also ensured that the version of the design was the most current—a change that has saved Tesco and GM hundreds of thousands of otherwise-wasted man-hours.

"When we originally started working with GM, we would get math data on eight-millimeter tapes . . . in any way you could imagine," says Toenneskoetter. "You could never be quite sure of the release level of the files. Even when we got upgraded to a [frame relay system], it was slow—it could take eight hours to download a file."

Now, Tesco's engineers work directly with models pulled from iMan, and are able to use them in developing the models of manufacturing equipment. At the same time, GM's product and manufacturing engineers can pull down the latest work of Tesco's engineers whenever they want—and can correct problems without having to wait for the next design review cycle. That means fewer late engineering changes, and fewer surprises for both sides.

Before 1999, Tesco did most of its work in two-dimensional drawings. The move to the 3-D tools mandated by GM has driven other innovations at Tesco, including the use of simulation tools to program its robots. In the past, robots had to be completely programmed at installation. Now, by the time the robots themselves are born, the code that drives them is ready to load, and engineers only have to make adjustments.

Sixteen to twenty months into the design process, the project "book" for a new product like the SSR or CTS can include 100,000 files, including drawings for each of the vehicle's 13,000 parts and the dies and tools required to create them, crash and suspension test models, virtual reality representations of the vehicle, the bill of materials, and related project data. The total storage requirements for one car are closing in on a terabyte.

But all this is just a prelude to the real chore ahead: translating that data into a manufacturing plant assembly line (and the instructions and processes that will turn the bill of materials into a finished automobile), and the documentation that will help mechanics support them once they roll off the dealers' lots.


Submit a Comment

Loading Comments...
Manage your Newsletters: Login   Register My Newsletters