How many people does it take to design a car? Not just shape the exterior surface, draw the interior, or establish the architecture, but to do all of these things from start to finish. How many people does it take? If you ask Dave McLellan, consultant and former Corvette engineering director, this question, take a seat first.
"Ten, to do a car from beginning to end in one year. That's all you need." McLellan's concept is not the latest step to increase shareholder value at the expense of employees. Rather, it's a culture change brought about through the marriage of the virtual world of computer-aided design (CAD) and expert systems, and their placement in the hands of experienced designers and engineers. This, he says, will compress the time and personnel necessary to bring a vehicle from clean sheet to assembly line, and let a company increase its coverage of the market. And those "extra" employees? Read on. He's got them handled too.
"There's just so much stuff that already can be automated in the computation process," McLellan observes, "that if you think about not jumping out of the virtual world into the real world more than you absolutely have to, large chunks of time–wasted time–can be cut out of the process. And with this come reductions in cost."
McLellan began his career at General Motors in 1959, and has seen the evolution in vehicle design first hand. "When I started out, engineers didn't want to sit at drawing boards drawing the car ugly line by ugly line," he says. "That wasn't any fun. So the car companies hired high school graduates, taught them how to cut sections, and they were the people who actually drew the cars with the engineers supervising the process." Shifting to CAD hasn't fundamentally changed the system, he says. "Non-degree engineers are designing cars on CAD tubes now.
"So, if you want to design a car with 10 people, some of them are going to have to have fairly serious science skills," he mentions. "And it will take automating all of the design processes that can be automated. The grunge work, so to speak. These 10 people will have to have different skill sets than today's stylist, and they will have to be located together in order to take the vehicle all of the way to releasing the tools to build it."
But the auto industry depends on a "belt and braces" approach because of the costs involved. "I understand the culture and sentiment behind that thinking," he says matter-of-factly, "and you should put the process together on paper first, pro forma, to see what the opportunity is. Then, as a systems engineer, you can think about going through each of the elements and each of the interfaces to understand how they interact, and why you have to go out and build a prototype. You probably won't get there in the first iteration, and I'd never commit a mainstream, high-volume product as a first step." Still, McLellan says there is precedent for this change in culture and concept.
"Boeing started out by maintaining a parallel process–old way and new–when it started down this road," he says. "By the time it did the 777, it was all in CAD, and the first airplane was sold when they were done with it. It wasn't at all like your typical automotive prototype." And, McLellan adds, if you aren't sure you're asking all of the right questions while in the virtual world, stop the process and build a prototype to see if it works. "Resort to that only when you have to. The idea is to get all of that stuff into the virtual world," he says.
One of the challenges facing the creation of expert design systems is visualization of test results conducted onscreen. "You can do all of the packaging, all of the interferences without ever having to go out and make parts," he says, "because you already have the CAD file." Yet, often the process stops so a dimensionally representative mockup can be made. Why? "Because you didn't feel capable of defining the load case or the important attributes of the part well enough to test it in the virtual world," McLellan responds.
Still, McLellan says automakers are trying to understand the foundation elements that make up a vehicle, which is a major step toward the creation of a seamless expert design system. "They are trying to codify some of the things you'd imbed in an expert system. It would operate at the level where, after styling defines the outer shape of the door, the computer would determine the structure necessary to make it work," he says. "All of the grunge work in a door like hinges, locks, beams, glass, structures and the like would be semi-automated in the CAD system."
Not only should this speed up the process, making it easier for automakers to produce vehicles that are closer to the buyers' current desires, it should reduce design and engineering costs as well. "If you remember the old shadow diagram," says McLellan, "one stylist affects 10 engineers, and 10 engineers affect 100 people doing the tooling. The ugly costs that bankrupt you occur at the end of the process, but it isn't the guy grinding out the tools that makes it that way. It's the decisions made in styling and engineering that add cost. An expert system would eliminate a lot of the rework that takes place in styling and engineering."
One area ripe for exploitation, according to McLellan, is computational fluid dynamics (CFD). "Today you still jump out of the styling studio and make a mockup of the instrument panel and its ductwork and nozzles to see if it will work," he says. "You can't be sure how it will work until you try it, and any arguments you have with the stylists are useless until you can prove whether or not the design works. It takes months to make and test the pieces, and then more time to settle the argument. And those are months wasted along the critical path."
McLellan suggests a robust CFD package, first and foremost, should be easy to use. A stripped-down version would take the design software's CAD data, model the ductwork and nozzles, and determine if the airflow was sufficient or if a redesign was necessary. The resulting design would then move to a more robust system where a heat load module would correlate the window area with sun load under various conditions (think Arizona in the summer), and compare this to a comfort factor model to establish optimal climate control ramp-up times. "We have the information to solve this in the virtual world," he says, "and eliminate the need to build hardware, but we don't."
McLellan's thoughts on exterior airflow are similarly to the point. "There's no good reason you can't get within 5%–10% of the actual values with a good CFD program that uses your existing CAD data. When you're doing a race car and the last mile per hour is important," he cautions, "you use wind tunnel testing to reach your goals. But passenger cars aren't as aero-critical as race cars. At its most basic, you don't want to the vehicle to lift or have weird crosswind characteristics, and you don't want the windows or mirrors whistling at you. These are things you could sort out in a good aero program well ahead of time. However, even the best systems out there can't do spinning wheels or moving ground planes, are very fickle to use, and it can take months to do two or three variations on a vehicle." McLellan believes that these files should be so robust that the results from these computerized tests can be shipped directly to the toolmakers while any validation tests are run in the wind tunnel.
Beyond the creation of interweaved CAD programs, a coterie of expert engineers and designers, and the need to modify current corporate culture, the biggest barrier to this wholesale change is the cost of building vehicles to the current paradigm. McLellan suggests one way of breaking through this obstruction is to replace the current stamped-and-welded chassis with a simple structure made of honeycomb composite sheets. These would be toy-tabbed together and bonded, with L-shaped channels added to reduce the strain in the bond area.
"There's not a lot of car that's part of the foundation at-surface," McLellan states. "Usually the stuff that is at-surface is there for somebody's convenience. That's why it's part of the primary structure. It doesn't have to be, and it obviously doesn't service very well." A simple, "folded box" chassis of the type he describes would replace complex chassis panels with a much simpler structure bridged by filler panels, and radically change the industry's tooling requirements. It also would take a major leap of faith for an OEM to embrace this construction method, although it can buy and dissect a vehicle built this way in order to get comfortable with the idea today. (See box.)
"It's the steel stampings in multiple progressive die sets that are the primary barrier cost for OEMs," says McLellan. "So if you are looking at low volume, quick turnaround, or being able to build specialty vehicles, there's an incentive to look at other methods." As McLellan observes, the switch from wood to welded steel body construction was no less traumatic or scorned. "The world changes," he says.
Which, finally, brings us back to those "extra" employees. "The typical industry response is to lay these folks off, but that's no answer at all," McLellan says with a hint of exasperation. "Why wouldn't you clone the process and use it to produce 10 different vehicles, introducing variation into the marketplace as a strategic marketing strategy? If you could double the number of unique-look automobiles and market them as Cadillacs, for example, you now have a strategy to completely change the dynamics of the market. The main component sets (powertrain, brakes, steering, etc.) stay the same, but each is tuned differently. As long as you've got those systems in place, you can build as many variants as you can imagine." Imagine that.
Proof of Concept
Rod Trenne is the prototype for Dave McLellan's new breed of designer/engineer. An employee of UGS since 1996, Trenne has a varied automotive background that includes time at GM building complete vehicle mockups, creating jigs and fixtures for Saturn, and doing packaging for Cadillac. "This gave me a real in-depth overview of what goes into a vehicle," he says, "and after 22 concept car build programs, you get a real feel for what works and what doesn't."
Trenne's next stop was Mosler Automotive (Riviera Beach, FL; www.moslerauto.com). Together they approached UGS about participating in the project to create an American-made supercar totally in cyberspace, and the company responded by putting Trenne on its payroll. Mosler, for his part, was looking for more than just a nice shape to house the Corvette component set he had in mind. He wanted an American road car that could double as a race car, and sell for less than $175,000.
"When I went down to Mosler, I thought I was doing the styling with the door seals, hood and door openings, and all of that stuff," Trenne says. "I thought he was going to provide the engineers to do the chassis for me, but it turned out that was part of the design brief too." It didn't slow him down much, though. The MT900 went from screen to street in six months. In that time, Trenne created a 90 lb. "folded-box structure" from FAA-approved honeycomb composite panels supplied by Teklam (Corona, CA; www.teklam.com). "The idea for using these flat panels came from my days building assembly jigs and fixtures," he says. "I was looking for a way to build a car that would be as accurate as a jig. The panels are accurate to 0.003 in., and the CNC router that trims them is accurate to another couple of thousandths. When you put it all together, it's just as accurate as the fixtures used in production assembly, and each one is the same as the one before."
The interior panels are glued to the chassis, in effect creating a tub inside a tub. "It slides inside the box you created from the composite panels, and is clamped and glued in place," says Trenne. "Then you drop the body on top and glue that on. By the time everything goes together, it's all supported and fixtured by the box you made from these super-accurate panels. Plus it's very light." The first MT900s weigh 2,600 lb. with air conditioning, stereo, and leather trim. Trenne believes the road car will come closer to 2,400 lb. once the build process is optimized. "It's way overbuilt right now," he says.
If you believe this construction process is limited to niche vehicles, think again. "I'm sure it would work for any type of vehicle, especially SUVs. You could get rid of the frame rails, lower the floor by as much as seven inches, and save close to $300-million in tooling," he says. So why hasn't anyone done it? "It's funny," he sighs. "When you show these things to closed-minded people, they just never seem to go anywhere."