GM Plans Digital Turnaround

Once upon a time,General Motors sold three out of every five cars and trucks in North America. Now, it's less than half that. For GM, passing the competition now is its job 1. So, instead of clay, metal and wires, it now builds new cars out of math. Whether GM's effort to digitize everything it does will add up to gains in the marketplace remains to be seen.

In a brick-lined courtyard on the campus of General Motors' technical center in Warren, Mich., a team of engineers is taking a look at a newly designed truck. This is no ordinary pickup. It has the rounded fenders and arched hood of a '50s Chevy roadster, and a roof that folds into the back of the cabin. It's a convertible—the Chevrolet SSR, due in dealer showrooms this December.

This is also no ordinary setting to review a prototype. The truck rotates slowly on a turntable. Its shiny, cherry-red finish reflects the summer sunlight and the trees that border the courtyard.

When the lead designer doesn't like what he sees, he changes it. In an instant, the truck is repainted in magenta, and then black. The truck drives itself around the courtyard, so the lead engineer can see how it looks from different angles.

All this is possible because the truck and the courtyard don't really exist. This is GM's alternate universe—the GM Virtual Reality Center. When the lights come up and the engineers' 3-D eyeglasses come off, the truck—and the courtyard—fade away. The truck prototype, the paint changes, and the sunny summer day exist only on the disk drives of a Silicon Graphics supercomputer. The whole presentation has been projected on the glass walls of GM's version of a Star Trek "holodeck."

At GM, bits and bytes have replaced the clay and metal that once was used to build prototypes of new products. The SSR will be fully built in software before a single piece of sheet metal is cut.

A few years ago, reviewing a new pickup truck like the SSR would have required building multiple life-sized prototypes to get a feel for what they would look like. Changing the basic color would take three days and 100 man-hours of work. Now, using a remote control, anyone can change the color in a second.

The SSR isn't a one-hit digital wonder. To regain market share the company has lost steadily over the last 30 years, GM is trying to use technology to meet the challenge set by Robert Lutz, GM's vice chairman for product development. His mandate: produce cars "that people gotta have—not ones they're willing to settle for."

Click to download a chart detailing General Motors' financial and technological progress. (Portable Document Format)

To restore the thrill, GM has created a design process more akin to creating video games than automobiles. In fact, the software used by GM's designers to create this simulation—Silicon Graphics' Alias/Wavefront AutoStudio for three-dimensional surface modeling and product simulation —is cut from the same code as that used to create cars for games like Sony PlayStation's "Gran Turismo." (See our Dossier on Silicon Graphics.)

But this isn't an arcade; it's serious business. GM is digitizing every aspect it can of the manufacture of a vehicle, from the design of the factory floor right down to assisting the implant of the last fastener.

Every aspect of the SSR's design and engineering was executed "in math"—with computer modeling and design software. The AutoStudio prototype of the SSR was turned into an electronic blueprint for building the car in EDS' Unigraphics design software, which guided the engineering of the vehicle. On-screen drawings were shared electronically with suppliers to ensure that their parts fit the design. Supercomputers used the mathematical models to do virtual crash testing.

The SSR comes the closest yet to a goal stated in 1998 by Jay Wetzel, then vice president in charge of GM's engineering operations, of designing, viewing and simulating the performance of a "virtual vehicle" without actually having to build a single physical prototype. Throughout the SSR's entire development process, only 60 prototypes were built—and the very first of them was a working concept car for the 1999 Detroit Auto Show. "It went straight from math to fabrication for the show," says Don Siefkes, virtual reality manager at GM's Warren facility.

GM is now moving toward the endgame for reaching Wetzel's goal: carrying out manufacturing in an all-digital system as well. That requires some serious computing power. Warren now is home to the 15th largest supercomputing center in the world. But the logic behind this effort isn't rocket science. By eliminating bottlenecks and errors as well as avoiding the "re-inputting" of data, Chief Technology Officer Tony Scott says GM hopes to cut years of time—and billions of dollars—from its product development cycle.

The effort to re-engineer GM's product development process began in 1995, with the launch of its global Synchronous Math-Based Process program, focused on integrating all of GM's engineering computer tools around the world with one core design and manufacturing system, EDS's Unigraphics. The efforts accelerated over the next two years as GM's new CIO, Ralph Szygenda, put together a team to consolidate the company's engineering efforts.

The point man for this effort is Kirk Gutmann, hired in 1997 by Szygenda as vice president of product development information systems.

By 1997, GM had already seen a 13% improvement in engineering productivity, and Wetzel said at the time that he expected to see another 30% improvement by 2000.

To achieve those goals, Gutmann and Wetzel proposed making GM's design process totally digital. That would mean consolidation and standardization of GM's engineering systems, and a major investment in high-speed networks and data center capacity.

The main component of the plan: a single global system for storing, tracking and managing every aspect of the information associated with each of GM's products, from concept sketch to engineering drawings of every component that goes into a finished vehicle.

Doing the Math

Digitization of design and manufacturing is no small challenge for GM and its integrator, EDS. The design of a single medium-duty truck can incorporate 86,000 data files, according to product development engineers Steve Powell and Rosemary Hamill.

But there's a big payoff in overcoming that obstacle—the digital route to vehicle design dramatically reduces the cost of developing each new vehicle, Gutmann says. In the mid-'90s, it took GM 48 months to move from "style freeze" to first product. By 1999, the digitization effort cut that to 18 months.

"The same engineering team that did one product can [now] do five in the same time," says David Cole, president of the Center for Automotive Research in Ann Arbor, Mich.

The reduced cost also means GM can sell fewer cars and still make a profit on a product line. "It lets you drop the volume of products, so you're much better able to change directions if a product doesn't work out," says Gutmann. That means GM can design bolder vehicles like the SSR. "Risk aversion drives you toward mediocrity," Cole says.

Now this plan is being rolled out worldwide across GM's 14 engineering centers, and incorporated into the design process for GM brands such as Saab, Vauxhall and Opel in Europe, Holden in Australia, and localized versions of GM products built and sold in Asia and South America. The company's goal now is to become "the number-one benchmark company in engineering," says Gutmann.

Turning the Corner

In the 1950s, General Motors sold nearly 60% of all automobiles in North America. Now, after decades of attack from Asian and European rivals, stodgy designs and slow reactions to changing demands of customers, GM's share is down to 28%, little more than one in four vehicles sold.

But the company in 2001 reversed its market-share decline for the first time in decades. And it's beginning to lick its chops. "Every indicator we track—market share, profitability, quality, customer satisfaction, etc.—says we're headed in right direction for the first time in years," says CTO Scott.

Scott claims that GM is now number three in quality, behind Toyota and Honda, and ahead of DaimlerChrysler, Nissan and Volkswagen. With its "common-math model,'' as Scott terms it, the big automaker uses video and data technology to deliver desirable vehicles faster than rivals.

But there may not be a payoff. Japanese carmakers have succeeded without a big high-tech push. "When you look at Honda or Toyota, they don't have anywhere near the technical sophistication of GM," says Dennis Virag, president of the Automotive Consulting Group in Ann Arbor, Mich. "But they are masters at product development."

"Toyota and Honda don't have 50 people sign off on a design,'' Cole says.

As a result, Honda's development cycles now average around 14 months. At an auto industry conference this August, Honda President and CEO Hiroyuko Yoshino said the company now planned to reduce that to an average of 12 months.

"At GM, in the past, they could have a lot of stuff languish for years waiting for approval," says Cole. "And a lot of concepts, by the time they got to the approval process, were too expensive to do and wouldn't work."

Time Trials

To change that, GM now insists on a single way of doing design. Five years ago, every business unit within GM did things differently. There were four different design packages and more than 560 different engineering and simulation applications across the company. Engineering knowledge was scattered across 150 internal Web sites. Every business unit tracked its product data and project planning differently. Globally, only Opel's engineering center in Germany was linked in any way to engineering operations in the U.S.

Since 1997, GM has consolidated its engineering operations, reducing the number of servers by 80%. Four design systems have been replaced by one mandated standard for every GM engineer and first-tier supplier—EDS' Unigraphics. The 150 Web sites have collapsed to a single engineer- ing portal.

Every new GM vehicle now is managed like a software project, with its essential files—everything from the initial styling through crash-test simulations—tracked from the earliest requirements to final field modifications. This product "source code" is managed by Team Center Engineering Edition, a product-data-management applica- tion developed by EDS (with heavy input from GM). TeamCenter puts the components of every car that GM makes—and all of the tools that go into making them—into a 150-million terabyte storehouse of product information that all of GM's divisions and business partners can draw on with a click of a mouse.

Copyright © 2004 Ziff Davis Media Inc. All Rights Reserved. Originally appearing in Baseline.