The computer crunch: suppliers scramble to boost CAD/CAM/CAE "horsepower."

Charlie Martin pauses for a second as a wave of nostalgia washes over him. It's politically correct nowadays to sneer at engineering's old-fashioned tools of the trade: drafting boards, T-squares and blueprints. But just this second - as the weight of 10 years of "progress" hangs heavy in the air - it almost sounds like "the good old days" when he talks about the way things used to be.

Mr. Martin is director of engineering for Detroit Center Tool's Welding and Assembly Div., a major supplier of automotive production tooling. He oversees some of the most sophisticated computer-aided design/manufacturing/engineering (CAD/CAM/CAE) operations any supplier could hope to own. On a major project, DCT can design, build and simulate the operation of a major production system completely on computer, then send that computer model to an automaker customer over special telephone lines.

It's impressive, but it's also very complex and expensive. Training highly skilled engineers and technicians to run it adds another thick layer of cost. Multiply it all by the number of customers, three, four - or more - because each automaker has its own computer engineering system, add the inevitable start-up glitches, and it all equals something less than hacker heaven.

"Ten years ago, when we were still in a manual drafting mode, you could just run a copy (of a blueprint) to transfer information," Martin says wistfully. Other supplier sources join the chorus, pointing out that blueprints also were almost universally understood. No special translations were required, because, well, blueprints were blueprints.

Training wasn't a big issue, either, because most draftsmen started learning their trade in seventh or eighth grade, and their skills could last an entire career. Now skills can become obsolete in 18 months.

Cost? Well, the lead for those nerdy mechanical pencils did have to be replaced.

Has a backlash against computers begun?

Not likely. This computer crunch is all part of the massive change going on as North American automakers shift more design and engineering responsibility onto suppliers in an effort to cut costs and chop new product development times. The concept makes sense, and someday, when CAD/CAM/CAE systems are all standardized and can talk to one another, everyone will probably be glad it happened.

But right now the cost and complexity of taking over high-level computer design and engineering tasks on three or more redundant systems - while still delivering annual price cuts - is killing some suppliers. That's especially true at the second-and third-tier, where profit margins and resources are extremely limited.

Insiders say the problem is exacerbated further by first-tier suppliers trying to "dump" engineering work onto their subsuppliers in an effort to mitigate costs, particularly when there's a snafu with the original program.

What's more, the pace of change is quickening. OEM's in-house systems are becoming increasingly powerful as they move to complete digital mockups, or prototypes. They want to build and test vehicles and components completely on computer, with as few physical prototypes as possible. If suppliers want to keep winning new business, they have to keep up by adding more computing power - perhaps to the point of buying costly supercomputers.

For instance, earlier this year Chrysler Corp. and Dassault Systems announced a Digital Manufacturing Process System (DMAPS) described as a fully computerized end-to-end product and process management system that enables the automaker to design, construct and run a "virtual manufacturing facility."

DMAPS is built upon the CATIA computer-aided design system Chrysler first implemented in 1984. CATIA replaced engineering drawings and allowed for an integrated design. This led designers to better understand how their part related to others during the design of the vehicle. Now, Chrysler says, DMAPS allows engineers to build tooling and robotic assembly stations complete with the knowledge of how they fit into the process flow of all manufacturing operations.

"The aim of DMAPS is to simulate and validate everything we do in manufacturing - the tooling, manufacturing process, material handling, even the sub-system process layout," says Frank Ewasyshyn, vice president of advance manufacturing engineering. He expects the implementation of DMAPS to shorten Chrysler's manufacturing lead time by 20%. He also hopes it "allows us to more readily identify and eliminate process variation, improve quality and reduce tooling costs."

"DMAPS is more than a new product," adds Gus Olling, Chrysler's lead executive for CAD/CAM/CAE Research and Development. "It extends the use of CATIA into the process modeling domain. It introduces a new strategy to manage the manufacturing processes."

This is no problem for a supplier such as DCT, which already is geared up with high-level CATIA systems and can integrate its operations seamlessly with Chrysler. But vendors who deal primarily with General Motors Corp. or Ford Motor Co. and have EDS Unigraphics or PDGS systems will soon find themselves out of the loop at Chrysler unless they spend some big bucks on a CATIA setup.

"There's no question automotive and aerospace are moving toward total electronic description, and moving from two-dimensional to three-dimensional (computer) models," says Tony Affuso, vice president-product marketing and development at EDS Unigraphics, a division of General Motors Corp.'s Electronic Data Systems Corp. "I see more and more automotive and aerospace companies committing to total math or electronic data. As they outsource, there are cost and data translation issues," he says.

When physical prototypes must be produced, OEMS frequently are using sophisticated techniques such as stereolithography (SLA) that create plastic components directly from CAD data. This technique can shave weeks and sometimes months off prototype development times, so OEMS want their suppliers to get on the rapid prototyping bandwagon. The price of one SLA machine: $250,000 to $500,000.

The supplier community also is abuzz with another rumor that has some pulling out their hair: Ford is contemplating phasing out its in-house PDGS mechanical design software and replacing it with a system developed by Structural Dynamics Research Corp, (SDRC). To suppliers, that means adapting and training for yet another system while still retaining PDGS capabilities during a lengthy phase-out period. A Ford official only will say "We are in discussions with (SDRC) on a strategic partnership."

DCT's Mr. Martin won't detail how much his company has invested in its CAD/CAM/CAE, operation, but the Automotive Industry Action Group (AIAG) a not-for-profit automotive trade organization, estimates a typical CAE workstation or "seat" costs $45,000 to $90,000 - for starters. That includes hardware, software updates and initial training - but no operator.

A typical CAE supplier operation needs two seats per customer. Jerry Widmer associate director for AIAG, and an executive on loan from Eaton Corp., says his company has 12 seats for six different high-level CAE systems, representing an investment of about $750,000, not including maintenance, periodic software updates, operators or additional training.

Compared to GM, Ford and Chrysler (who have 8,000, 5,400 and 2,300 seats, respectively in North America doing all manner of CAD/CAM/CAE work) this computing power shift is small potatoes. But makers of high-end computing equipment and software have noticed the change. So have independent engineering houses and others who are helping suppliers bridge their computing gaps.

"We're seeing it. Large auto manufacturers are putting off more engineering responsibility to suppliers, and suppliers are looking more to computer simulation and becoming more sophisticated in their use of it. That type of analysis is growing dramatically," says Greg Clifford, director for third-party supercomputer applications at Cray Research Inc. Cray says half of the automotive orders for its new low-cost ($250,000) air-cooled J90 compact supercomputer are from suppliers.

Silicon Graphics Inc., another major producer of engineering computers and supercomputers, says it's also selling more of its high-end systems to auto suppliers, although it doesn't have exact numbers.

Independent engineering houses that traditionally have handled overflow" engineering work from the OEMs also are noticing a shift. Michael J. Presser, vice president, West Side Operations, for APX International says 90 to 95% of APX's customer base used to be OEMs. That has dipped to about 70% now and could soon drop to 50% as supplier work floods in, he says.

Automakers are swiftly winnowing the ranks of their first tier suppliers and demanding that the ones left have global - as well as high-powered - engineering capabilities. Many suppliers aren't equipped to do that, at least not at the level of the OEMs, so they look outside for help, Mr. Presser says.

Hriday R. Prasad, vice president strategic planning, Research & Development at Modern Engineering, Inc. tells a similar story. Modern also is providing high-powered global engineering capabilities to an increasingly number of suppliers who don't have the resources to do it on their own - and it recently beefed up its already advanced computer simulation capabilities.

Computer simulation speeds product development by - among other things - turning complex mathematical data into easy-to-understand visual images. For example, it's difficult to solve a door-fit problem by only looking at dimensional data, Mr. Prasad explains. But with computer simulation, the data is turned into a life-like three-dimensional door that can be opened and closed. With that kind of visualization, the problem can be spotted immediately, he says.

Both Modern and APX also report they are doing a fair amount of business simply translating CAD/CAM/CAE work for second-and third-tier suppliers who can't afford to have multiple engineering systems compatible with all their customers. Most OEMs will not accept or translate engineering work done on incompatible systems.

But it's not all bad news.

The cost of the most powerful software and computing equipment is dropping dramatically. While Cray's J90 series still isn't cheap, these systems are as powerful as the multi-million-dollar supercomputers it built only a few years ago. Furthermore, Cray says these new systems are completely compatible with the really big Crays used by almost every automaker OEM in the world.

In addition to adding more and more powerful computers to its lineup, SGI is offering an increasing array of less-expensive systems "the size of pizza boxes" as one user says.

And even expensive SLA rapid prototyping systems are coming more within reach of smaller suppliers. Richard P. Fedchenko, vice president of strategy and market development for 3D Systems, Inc., a major vendor of SLA systems, says "Service Bureaus" which lease time on SLA equipment are ordering machines faster than he can deliver them. The Detroit area has one of the highest densities of SLA service bureaus in the U.S., he adds.

Furthermore, he promises that 3D will introduce a new, dramatically less expensive rapid prototyping machine in November 1995. It will use an entirely new type of technology, but - like other RP technologies - it will allow CAD designers to turn their computer data directly into physical prototypes. It won't compete with SLAs, but it will "show amazing detail" he says.

EDS Unigraphics has an innovative program where a company can rent a turnkey CAD/CAM/CAE system, complete with hardware, software and on-site training for two or three years. A supplier eager to bid on GM business but reluctant to commit up front with something as ambitious as a 10-seat operation could build the cost of the "rental" into its bid. If approved "we could get there within a week or two," Mr. Affuso says.

There's also good news on the data transmission front. For years automak-ers have used "electronic CAD interchange" technology to exchange computer-aided design files with their suppliers. The Big Three automakers, for example, maintain large mainframes with electronic "mailboxes" that suppliers use to send and receive CAD files.

Until now, first- and second-tier suppliers looking for similar benefits have avoided implementing mainframe-based mailbox systems because of their big installation and maintenance costs. But a new smaller network system developed by CTI Communications offers a similar system that is much more affordable.

While mainframe mailbox systems can cost up to $1 million to install and require expensive, dedicated telephone lines, a typical CTI "FastSync Hub" configuration costs between $30,000 and $50,000 and uses standard, analog or digital dial-up telephone lines.

But perhaps the most encouraging news of all comes from the AIAG: It's well on the way to making CAD/CAM/CAE systems that can easily communicate back and forth.

Called the Standard for the Exchange of Product model data (STEP) it is an international standard for sharing computerized product design data. It makes it possible for otherwise incompatible CAD/CAM/CAE systems to communicate accurately, without expensive and troublesome electronic translators.

Jack Pokrzywa, a Program Manager at AIAG, emphasizes STEP is no pipe dream. Most of the world's automakers already have endorsed the program, and the U.S. Big Three are working with major suppliers such as Eaton Corp., Dana Corp., TRW Inc., Allied-Signal Corp., Delphi and others on several pilot programs. Dana Corp. is exchanging drive shaft data with GM Truck for instance.

The first phase is expected to be completed in the first quarter of 1996. The second phase will last "about a year" and the third phase should wrap up in mid-1998. But some impressive demonstrations are expected as early as next year.

Some industry experts are skeptical about how fast it will take off because STEP still can't handle some of the most sophisticated aspects of CAE. "And the technology isn't standing still," one reminds. But almost everyone is rooting for STEP's success.

"The Big Three are pushing it here at AIAG. They wanted to initiate a program or pilot that would migrate this technology down the supply chain," says AIAG's mr. Widmer. "The suppliers just want to satisfy their (OEM) customers. Internally, they are really looking (for STEP) to save them a bunch of money. I hear once a week about a supplier complaining about having to buy a certain CAD package to satisfy the requirements of a customer."

But even all this technological wizardry still won't solve what DCT's Mr. Martin calls the biggest bottleneck of all in the CAD/CAM/CAE world: The cost and complexity of training people to use the systems. "But that's another story," he says. Indeed it is.

`Domain decomposition'

The latest breakthrough in computer-aided engineereng

The pace of change in computer-aided design/manufacturing/engineering (CAD/CAM/CAE) has been staggering in the past 10 years, but it continues to accelerate. The latest breakthrough sounds like it belongs with forensic scientists rather than automotive engineers, but proponents say "domain decomposition" will allow engineers to analyze solids and structures that change shape because of interaction with a fluid - a missing link in many high-level CAE tasks that has long challenged engineers.

Unveiled earlier this year, this new method is part of the first series of results from a five-year research project at the Computational Mechanics Laboratory at Carnegie Mellon University in Pittsburgh, funded by a $1 million grant from Algor Inc., a major developer of mechanical engineering analysis and design optimization software.

"The applications for domain decomposition are staggering," says Michael Bussler, president of Pittsburgh-based Algor. "For example, automotive engineers can predict the strength of a car antenna bending in the wind."

The new method has numerous applications in bio- and civil-engineering as well, Mr. Busler says, such as studying the flow of blood through flexible vessels and determining how wind-induced vibration of bridge cables affects the strength of the bridge.

Analysis of objects that change shape because of contact with a flowing substance has always been difficult because the behaviors of the fluid and the solid are very different and have highly complex. inter-relationships.

"Imagine the flow of air around a flexible airplane wing," says Dr. Omar Ghattas, associate director of the laboratory. "The airflow creates pressures, which change the shape of the wing. The changed shape of the wing, however, affects the flow, which further changes the pressures on the wing. We really can't separate the two behaviors."

Domain decomposition builds on the principles of finite element analysis (FEA), long used by automotive engineers to analyze how computer models of vehicle structures and components withstand physical stresses and strains. FEA breaks a computer model down into smaller pieces called finite elements.

The computer analyzes the model by making computer analyzes the model by making computations on each of the finite elements to locate potential problem areas of the design. With domain decomposition, the FEA is carried a step further by breaking the model down into "subdomain" parts.

"The computer performs FEA on each of these separate parts and then `glues' the individual solutions together where the subdomains meet," says Dr. Ghattas. "When gluing the subdomains together, we use equations that represent the effects that the subdomains have on each other. We have applied this technique to fluid-solid interaction problems where the fluid and solid represent two adjacent subdomains."

Like most ultra-sophisticated software, however, domain decomposition needs lots of computer "horsepower" to run. It is expected to be very useful when "next-generation: computers capable of performing multiple calculations at the same time are introduced - they will drastically reduce the time it takes to run the analysis.