A combination of speedy processors, larger hard disks, and less expensive
memory has enabled a dramatic leap in desktop-analysis capabilities. "Models
running in our software have doubled in size over the past two years," says Bob
Haubrock, director of simulation and test products at SDRC.
This performance boost has given engineers unprecedented access to real-time analysis.
"We are doing problems on desktop computers that just a few years ago we would have had to do on a mainframe," says David Dearth at Applied Analysis and Technology, a Huntington Beach, CA, engineering consulting firm. "Our analysis capabilities have gone up by orders of magnitude."
For example, engineers can now process models with several hundred thousand degrees of freedom on a mid-range workstation running MSC/NASTRAN, according to Ken Blakely at MacNeal-Schwendler Corp. "Five years ago, these were large models for mainframes."
Now, MacNeal-Schwendler has introduced MSC/NASTRAN for Windows, bringing the same functions of its supercomputer-class software to Pentium PCs. "We have reduced the barriers to doing analysis," he says.
"In 1987, it was considered by everybody to be almost magic that we could do 6,000 to 10,000 degrees of freedom on a 386 PC," Dearth recalls. "That problem would run for a couple of days, but we could do it."
Recently, Dearth performed static analysis on a pressure-bulkhead for an aerospace contractor using Algor FEA software on a Pentium PC. "It was almost 50,000 degrees of freedom, and it ran in less than an hour," Dearth says. "That's absolutely amazing." It's also an indication that Algor's PC-based FEA products are as powerful as its UNIX-based products, says Algor's Peter May.
Empowering an individual. At Burlington Northern (BN) Railroad, mechanical engineer Scott Landrum single-handedly developed a new freight car to carry more steel at less cost-all on a personal computer. After developing the design with AutoCAD, he performed the necessary structural analysis with Algor Linear Stress software.
"The really good thing about this was just one person was able to do the design and structural analysis," he says. "In the past, it took a team."
Landrum ran a number of analyses based on various load scenarios, constantly tinkering with the design to find an optimum weight. The result: A 47,200-lb car, compared to 65,400 lbs for the earlier model; and the new, lighter car can carry an extra 12 to 15 tons of steel. The entire project, from concept to the first car put into service, took nine months. BN won an initial order for 175 cars worth more than $8 million, and another 50 are being built. "It's pretty successful," Landrum says.
Wheel design. Innovation in Composites says its award-winning one-piece hollow bicycle-wheel design depended on current computing technology. "There's no way we could do this project without desktop computing and analysis," according to Doug Olsen, vice president of engineering.
The company developed an injection-molded plastic wheel, reinforced with carbon fibers, Olsen explains. The big question: Would it be strong enough? Engineers used Rasna software to run through numerous scenarios. "I can do ten times the work, and ten times the what-ifs, as I did five years ago," Olsen says. "You do so many what-ifs, you get a better part." The design won a Society of Plastic Industries (SPI) award for overall best part.
Slashing design time. Mitsubishi says it cut product-development time 40% for new circuit breakers with advanced software modeling. The software helped with key design challenges: placing contact points in a vacuum or heavy insulation, and making sure the circuit breaker didn't open so slowly that it melted from electricity "jumping" across the contacts.
In one case, engineers used SDRC I-DEAS software, running on Silicon Graphics workstations, to perform stress analysis on a tank used in gas-insulated circuit breakers. Mitsubishi engineers have long done such analysis, but it used to take an entire day to receive results from the company's computer center in Osaka, Japan. "Now that we can analyze with I-DEAS in real time, analysis results can have an immediate impact on the design," says Masao Narita, manager of the Circuit-Breaker Development Group. "We can perform design and analysis in parallel."
Replacing impractical tests. Computer analysis plays a crucial role in designing military spacecraft that can withstand radiation from a nuclear explosion. "Since underground nuclear testing was curtailed, and because above-ground simulator tests are expensive, accurate analysis techniques are critical," says Anthony Botting, who works as a composite structures design engineer at Mission Research Corp.
A team at Mission Research had one chance to gather data from an actual field test. First, they developed analysis models using COSMOS/M from Structural Research and Analysis Corp., then checked simulation data with actual test results. In one analysis, the company tested a space-based interceptor bridge structure that connects two propulsion tanks and provides support for thrusters, pumps, and pressurized lines.
The part, fabricated from an advanced metal matrix composite material of silicon-carbide/aluminum, forms a non-symmetric ring-like structure. It includes flanges, tapped holes, a number of pump and control line cut-outs, and thruster seats. Mission Research also tested four other rings similar in shape to the bridge, but of simpler construction-one bare, and the others having 3-ml plasma-sprayed alumina coatings. The researchers placed circumferential strain gages at various locations on the rings and bridge.
They first analyzed the uncoated ring to check calculated loads, silicon-carbide/aluminum material properties, and analysis procedures. Then they moved to the coated rings to generate information for analyzing the bridge itself.
"Our analysis followed a building-block scheme, wherein we generated a simple model of the solid ring and validated it against measured natural vibration frequencies," Botting explains. They used four-node shell elements at five degree-arc increments, with some mesh refinement near slot boundaries, and "tuned" the model against actual test data.
When finally analyzing the bridge, Botting says, "the correlations of data and calculated strains are quite good." The next step? Modeling larger structural assemblies-and ultimately complete spacecraft systems. "Continuing research will give spacecraft designers and analysts the capability to predict accurately possible optical and sensor misalignments, time delays, and jitter associated with radiation." And all the work was done on a 50-MHz, 486 PC that sells for about $1,000.
Spacelab equipment. Computer analysis also helped in designing health-monitoring equipment for European astronauts exposed to long periods of weightlessness. The European Space Agency commissioned Innovision, Odense, Denmark, to develop a 234-lb respiratory monitoring system, 49-lb washout gas-supply system to mount on a standard Spacelab rack, and 88-lb cycle ergometer for mounting on the center aisle of the European space shuttle.
The design team had to ensure that the equipment's fundamental eigen frequency exceeded 36 Hz to protect it from resonance caused by vibration during the space-shuttle launch. Using COSMOS/M software, they created a variety of models, ranging from 7,000 to 25,000 degrees of freedom, with boundary spring elements at shuttle-equipment interface joints supporting each model.
"PC-based COSMOS/M analysis was very competitive to mainframe FEA," says Gö-Jacobsen, head of analysis at Innovision. "In addition, the user interface is incredibly easy to use." Lab tests showed the analysis work agreed closely with actual results.
As processing speed continues to double every one to two years, industry observers say engineers can expect more analysis advances, such as semi-automated design optimization. That would continue the trend toward trying to more tightly integrate analysis with CAD, using the results to automatically refine some geometry within the CAD model.
Says Bruce Jenkins at Daratech, a Cambridge, MA, research and consulting firm: "The amount of raw power that's going to be available will make any engineering computation problem solvable at the desktop.