SOFTWARE AS PROTOTYPES
Back about 40 years ago, according to Richard Smith, director of CAD/CAM products and services for The Boeing Commercial Airplanes, whatever software Boeing had for aircraft design and simulation was all internally developed. Comments John Givens, director of engineering math process for General Motors Powertrain (Pontiac, MI), “Because of the expense to develop aircraft, [the aircraft manufacturers] have developed computer-aided tools for a lot longer time than we have in automotive.”

Boeing developed FlyThru, a high-performance CAD visualization system. This program was first used to “preassemble” the Boeing 777, thereby helping engineers determine the geometric relationships and mechanical interferences between parts. This program was the basis of the digital mockup. Boeing also developed Easy5 (now owned by MSC.software; Bellevue, WA), a set of engineering analysis software tools used to model, simulate, and analyze dynamic systems containing hydraulic, pneumatic, mechanical, thermal, electrical, and digital subsystems. Other Boeing software performed noise protection modeling. Hydraulics analysis was done using commercially available software.

The company’s Voxmap PointShell evaluates part proximity and part interferences. This software program detected over 10,000 part interferences in the initial computer modeling. Boeing also had two programs for ergonomic analysis, which helped determine whether assemblers had enough space when building an aircraft. The same software helped assess the ability of people to perform aircraft maintenance.

Over time, commercial software products have reached and exceeded the capability of the software Boeing had developed. The result is that, one by one, internally developed software programs have been taken out of service. For instance, Boeing’s structural analysis programs have been replaced by Elfini (finite element analysis system from Dassault Systemes) and Nastran. Moreover, estimates Smith, nearly two-thirds of the software used for airplane design is Catia 5 from Dassault Systemes. Boeing enhanced Catia with other applications, such as BCSLIB-EXT, a utility for solving very large problems that cannot fit in the central memory of a computer. This let engineers view up to 500 parts at a time and electronically assemble major airplane sections. Boeing’s EPIC (electronic preassembly integration on CATIA) tests the fit of components.

The company currently simulates the manufacturing process “to varying degrees,” according to Smith. For instance, it simulates the placement of airplanes in the factory so that aircraft don’t bump into each other during production. Some internally developed software exists for simulating assembly.

Boeing’s migration from internally developed design and simulation software to commercially available software is based on a simple, basic criterion: When the features in the externally created software meet or exceed the internally developed software, when the accuracy from the externally developed software provides the answers Boeing demands, then it’s time to switch. When the two types of software have reached parity, there’s an assessment and, adds Smith, “probably quite a bit of turmoil to make a decision. It’s hard to turn the tide. People get attached to [software] for one reason or another.”

The three main sections of the longer-range 777 jetliner are being connected at the Boeing final assembly plant in Everett, WA. The forward section of the airplane, manufactured by Boeing in Wichita, KS, was loaded into a tool and then joined with two other large body sections that were manufactured in Japan. The design and engineering of this massive structure were largely conducted on computers, not with physical prototypes, as has historically been the case for aircraft development.

Sourcing computer-aided tools is similar to automotive engineering/design. GM’s Givens guesses that 30% of GM’s current software tools are internal developments; the rest are commercially available. Internally developed software, he continues, is “another source of structural cost—the cost to maintain internally developed tools.” For that reason, GM’s engineering departments are trying to use commercial software more; however, as with Boeing, continual internal software development is basically driven by the lack of commercial software availability. There is also an “awkward” situation in the move toward commercial software, explains Givens. “Sometimes the proponents of the internally developed stuff are the authors.”

SOME PROTOTYPING REQUIRED
In the past, new aircraft models required a long series of experiments to quantify and validate all the elements of a new design. Typically, the experiments began with wind tunnel tests on several models of varying size and complexity. These were followed by a flight test program involving building a full-scale aircraft—and flying it.

Nowadays, aerodynamics analysis at Boeing is almost all done from internally developed programs. Virtually all of the company’s aerodynamic modeling is done without a wind tunnel. “Physical models have gone down in significance,” continues Smith. “We produce a number of aerodynamic test cycles digitally, and one cycle physically. The physical model [measuring three to four feet in wingspan] validates the simulation, to see if the simulation is giving us the right answers.” That’s a huge difference compared to 30 years ago. Back then, Boeing engineers used the wind tunnel model to create new and working designs.

For the 777, computations regarding aerodynamics were double checked by “conventional” computational fluid dynamics (CFD) and a number of wind tunnel tests to determine whether the results were reasonable, says Jean Jacques Chattot, professor of mechanical and aeronautical engineering, and director of the Center for CFD at the University of California (Davis, CA). “Indeed, they were quite good.”

Fact is, CFD and wind tunnel tests are complementary, and they don’t necessarily produce the same results. When discrepancies do occur, explains Smith, “there would be quite a bit of assessment to determine what should be the right answer, why is one giving a different answer than the other. I wouldn’t say that either one is automatically accepted.”

Up to the Boeing 777, maybe 25% of an aircraft design was produced with CAD—drafting for the most part, no 3D. The 3D solids modeling was used for special studies. For instance, Boeing used solids modeling to figure out a problem with the landing gear in the mockup for the 767. The 3D model showed that the tool holding the landing gear was twisting a “tiny bit.”

For the 777, some systems and subsystems require special-ized modeling—and mockups. One mockup was for the 777 nose to check critical wiring. A second mockup, an “iron-bird,” was a complete working physical prototype of the aircraft’s internal dynamic systems. This proved out the integration of electronics and hydraulics, which were simulated separately though not integrated with the 3D solid modeling tools.

FIRST FLIGHT, AND TEST
Designing and simulating the 777 consumed about 2,200 workstations linked to a four-IBM mainframe cluster in Puget Sound, plus four more mainframes in other locations, and 3 terabytes of data. This work validated the tooling and assembly plans for approximately two million parts, a validation that usually occurs once assembly begins in the plant. The work paid off for the Boeing 777, resulting in these benefits:

• Elimination of more than 3,000 assembly interfaces, without any physical prototyping
• 90% reduction in engineering change requests (6,000 to 600)
• 50% reduction in cycle time for engineering change request
• 90% reduction in material rework
• 50x improvement in assembly tolerances for the fuselage

Regarding that last item, the fuselage is 200-ft long. Alignment was off by 0.023 in.—about the thickness of a playing card—while for most other airplane parts in previous aircraft the alignment is to within a half inch of each other. Likewise, the wing tip on that very first Boeing 777 out of production was off by 0.001 in. By comparison, the wing tip on the Boeing 747 was off by 4.0 in.

That first Boeing 777 out of production flew. It went through flight test. It was then refurbished and delivered to a customer. It was the next five or so airplanes out of production that went through non-destructive tests, such as jet engine checks for particular aircraft models under test. Also, there were structural tests that bent and flexed the wings and cabin, and pressure tests that compressed the cabin—numerous fatigue tests that run through an enormous number of cycles to both simulate actual performance in service, such as takeoffs and landings, and to show the state of the aircraft after 20-plus years of use. Boeing’s modeling methods have been so good that the FAA has accepted them in lieu of certain physical destructive structural tests on finished aircraft.

AIRPLANES AND AUTOMOBILES
“A number of differences exist in the nature of cars and airplanes,” concludes Smith. Physical size is one. The shape of the product and its function is another. The operating environment is third. Think about this: If your car’s engine stops while you’re driving, you can pull over to the side of the road and call for help. “That doesn’t really work in an airplane,” says Smith. “Airplane products are designed and built to pretty incredible standards—standards that would not be cost effective in the auto industry.”

Agreeing with Smith is Kevin Mixer, research director for automotive and heavy equipment industries at AMR Research (Boston, MA). Mixer points out three other differences between the two industries. First, a plane is not mass produced like an automobile, so a different set of manufacturing constraints apply, as well as simulating those constraints. For instance, aircraft rework tends to occur at the end of production, not in the midst of it.

Second, the automotive engineering/design culture is different than that in the aircraft industry. Take the reuse of parts and interchangeable parts, for example. That, continues Mixer, has been “a challenge for a number of vehicle manufacturers to drive through engineering because people like to design things new.”

Last, Mixer points out, while some of the automotive companies are trying to move closer to the Boeing aircraft-design/simulation model, they can never get to that point strictly speaking. Airplanes, Mixer says, “are much more tool, more utilitarian, than cars.” Cars have those elusive qualities of fashion and image. “You can’t necessarily simulate how people are going to react.”