"Ephemeral Warehouse," Feature Article, September 2005

Today's complex 3-D engineering files definitely aren't your father's (or grandfather's) hand-drawn blueprints.

It used to be that anybody could look at blueprints at any time, as long as you didn't lose or destroy them. Digital designs, though, are created on software and computers that are outdated when they're delivered. Computer files can be hard to retrieve in as little as five years down the road. That's a big problem for the engineering community and, of course, for corporations, government agencies, and organizations that store information digitally—in short, for everyone.

Engineers of all stripes can forever debate the merits of two-dimensional versus three-dimensional design and speak of the vagaries of parametric, kinematic, or solid modeling. The point is academic if the files won't last.

No system exists to archive digital designs or anything else that is created on computer or housed on the World Wide Web. And that's a huge problem for society, experts say.

Unless a company maintains every system and every computer ever used to create data—a highly unlikely event—it'll be gibberish to computers of even the very near future.

U.S. Air Force engineers designed the B-52 bomber in 1952. It's still flying today and the military has plans to keep it around for another 45 years. So blueprints and digital files for all of the parts ever created have to be kept up to date and within easy reach for future engineers.

Yes, CAD images can be printed or scanned, but much information is lost when a three-dimensional file becomes 2-D and when attendant engineering information like manufacturing and product data management files aren't included, says William Regli. He's an associate professor in Drexel University's Department of Mathematics and Computer Science and in the Department

of Mechanical Engineering, and he's hard at work on
engineering an archival system for 3-D data.

The lack of such a system might sound like a pointless problem to some engineers. After all, engineers design for today.

"Let's suppose that you have the design for a bolt of
a bridge," said Maurice Smith. "Let's say you put that in an electronic file. That bridge will hopefully last up to 100 years.

"Now let's say the bolts turn to rust, or they want the original drawings to do maintenance," Smith added. "The system used to create them doesn't exist. What can you do?"

Smith is a staff engineer at The Kansas City Plant, a National Nuclear Security Administration facility managed by Honeywell Federal Manufacturing & Technologies. The plant produces components for nuclear devices for the U.S. government. He's working on a digital archival system for his company.

Most of those working like Smith on ways to archive digitally engineered data cite the B-52. The name is a coincidence, but the first flight of the B-52 bomber was in 1952. It still flies today. Third-generation designers rely constantly on first- and second-generation designs and blueprints. The early work is as relevant today as it was 53 years ago.

The Air Force plans to keep the B-52 flying for perhaps 45 more years, according to Smith. Engineers will need today's digital files at their fingertips for years to come.

Manufacturers of civilian aircraft face the same problem.
As Smith puts it, "How would you like to fly on a commercial aircraft, where the man who's made the replacement part for the wheel strut says, "Gee, that original CAD design is hard to read; I'll just guess on how to make it."

The Kansas City Plant has an obvious interest in archiving all designs. Smith heads an initiative to produce the NICE system—short for Networked Information Control Environment. The system attempts to archive CAD data and make it searchable. Engineers tested an early version of the system last year, but it's far from complete. The complex program must operate free of proprietary software, use Web standards, and be searchable using a number of different methods.

"If we could have bought one like that, we obviously would have," he said. "No one has anything like that."

His team has been working on NICE the past three years. "We've barely scratched the surface," Smith said. "So far, we've only really defined the problem."

The U.S. Department of Energy has been quite receptive to the plea for an archival system that Smith first put forth about five years ago. Officials know they need to save this information, and they understand the scope of the problem.

Of course, most information today—not just engineering data—is created and stored digitally on computer systems that become outdated sooner than bread gets stale.

Already aware that rich, historical information is lost every second because it's stored digitally, the nation's preeminent archive, the Library of Congress, joined the National Science Foundation this spring to grant 10 university teams a total of $3 million to look at managing long-term digital information accessibility.

Digital materials with research or cultural value are at great risk of loss because of factors like insecure storage methods and obsolete file formats, said William LeFurgy, digital initiatives project manager at the Library of Congress.
Millions of digital objects, such as Web sites documenting the early days of the Internet, have already been lost.

EBay's first home page may seem merely quaint now, but could make rich ground for scholars in the future. It's probably already gone for good.

A part can go from CAD to reality easily enough. But when that computer file is archived, much of it can be lost, including associated manufacturing planning, analysis, and tolerance information.

"For many years, technology to produce content has run far ahead of technology to archive content," LeFurgy said. "As a result, much, if not most, significant digital content is now at risk of loss. If we're going to keep this accumulated knowledge, we must push digital preservation technology forward, fast.

"Current tools are wholly inadequate for the task at hand," he added.

According to Regli, the Drexel professor, one issue crops up continually when it comes to CAD preservation: proprietary software. Data can't be separated from the software that created it.

As Smith puts it, and as anyone who still has floppy disks filled with personal documents created in WordPerfect can tell you: "If you can't open it, it's essentially gone."

Engineering software vendors don't have a stake in making sure the designs you create today can be retrieved from their systems tomorrow.

"To a degree, the CAD and PLM vendors' customer base sticks with them because it's hard to move from one system to another," Regli said. "They own your data, in some sense. It's hard to see how archiving would benefit traditional CAD companies."

But even an engineering company that upgraded every single legacy design each time it upgraded CAD software couldn't be sure those designs would last.

"No software company on Earth was around 75 years ago and few were even around 25 years ago," Regli said. "You can't assume your CAD system will be around in 25 years."

Perhaps more surprising, other factors like the need to save associated bills of material and manufacturing instructions make for a tangle of problems in creating a digital-engineering-data preservation system.

For the past 15 years, Regli has worked on ways to preserve what he calls design repositories. By that, he means CAD designs and their attendant information, like the manufacturing data and simulations.

He and Ali Shokoufandeh, an associate professor in the Drexel University computer department, received one of the Library of Congress grants, for nearly $500,000.

For the 3-D archival system, the pair seeks to store CAD data in a nonproprietary language—in this case, STEP, the standard for exchange of product-model data. All CAD files can be translated into this standard operating language, regardless of the software in which they were created.

"You bring all things to a common ground, get rid of predefined platforms, and try to create a set of mechanisms that can do the things the end user wants to do," Shokoufandeh said.

Still, STEP probably can't keep pace for long with today's advanced CAD systems. Even if all CAD models could be stored within the STEP format, only the models themselves would still exist. The rationale behind the design—the reason engineers chose a certain option—would be lost, as would necessary information like tooling specifications.

"You need to have a way to extract the knowledge you want to preserve down the road, so you'll understand the context of the design, why decisions were made," Regli said.

Tying the information to a search engine has proved a problem. No method exists for searching 3-D data. Regli envisions a search engine that would retrieve drawings and documents in a number of ways.

"So a manufacturing engineer could search on manufacturing properties, people in cost estimation could search on cost properties, and design engineers could search on design properties," he said.

Smith works with Regli and others to create this overall digital archival system. He'd like to see a search engine to which an engineer could describe, roughly, the type of bolt he or she sought. The computer might return design images of all bolts that fit the broad description. The engineer could narrow the search appropriately, based on that returned information.

Most companies either store original blueprints, or scan and store 3-D or 2-D data, said David Wilson. He owns Open Archive Systems Inc., a Windham, N.H., firm that scans documents, helps companies manage their 2-D and CAD designs, and aids in raster-to-vector conversion. Converting 2-D images to 3-D CAD files can help update documents and make them searchable, but, again, information is lost in the translation, Wilson said.

Companies may also store blueprints or CAD documents as portable document files (PDFs) or as tagged image files (TIFs). These are 3-D digital files that can be accessed fairly universally from any computer. Again, much is lost, including geometry, when smooshing a 3-D file as flat as a pancake, as Smith said.

For engineers of the future, those TIFs, PDFs, or blueprints might not contain vital manufacturing information. And that might mean the part can't be made, even with the stored blueprint.

"Let's say I want to manufacture and machine a part," Regli said. "You often need to design special fixtures to mount the part at the machining station. You might need to create intermediary features in the design for mounting that you later remove."

Some members of the engineering community are striving to create a nonproprietary method that will archive and search for CAD data. Ten universities have received a total of $3 million from the Library of Congress and the National Science Foundation to fund research.

But if a future engineer looks at a scanned print of the design that doesn't include the mountings, he can't make the part."Basically, people forget how to make things," Regli said. "Yes, they have the final object, but even if they have the blueprints, they don't remember how it was made."

Engineers could reverse-engineer a part to see how it was made. But, for his part, Smith and his colleagues tried a little experiment with a British team of academics who are also working on the preservation project. The Honeywell team made a part from reverse-engineered data gathered from the British scientists. Like the childhood game of telephone where much is lost in translation, the reverse-engineered part came back slightly altered from the original.

To create their large-scale archival system, Regli and Shokoufandeh want to describe all information mathematically, following standards set by the Web community and by the National Institute of Standards and Technology.

"If I write down the logic of the system, someone 20 years from now will understand it because it's math," Regli said.

Such a system has the ability to supersede or replace a proprietary CAD system. Regli and Shokoufandeh are also coming up with search methods. They've created some of those methods themselves, while some are based on pattern recognition, data mining, or computational vision methods.

Which brings Regli and Smith to their point: For complex digital archival storage and retrieval systems like theirs to ever work, a number of communities need to come together.

"This isn't something engineers or corporations or computer scientists will solve alone," Regli said.

He collaborates with NIST, with Smith's group—and thereby with the Department of Energy—and with a team at the University of Edinburgh that has a background in digital scientific databases.

Mechanical engineering is only one piece of the three-dimensional digital-archival problem. It reaches across nearly all disciplines.

"We're focusing on the mechanical, but what about the electrical?" Regli asked. "There are plenty of people building dams and bridges and we have no idea how that gets done. We can only look at one little slice of the problem."

According to Smith: "The problem can't be solved by one person, one group of scientists, or even one country. It almost spans our species. Any place that's adopted the efficiencies of using computers faces this problem."