Virtual Design Takes off at P&G

When Tom Lange enters a supermarket or Big Box store, he sees waste. A lot of it.

"About 10 percent of the packaging in stores serves no useful purpose," he says. "It doesn't protect the product. It doesn't improve the customer experience. It doesn't do anything. It's only in there because no one engineered it out"

As senior director of modeling and simulation at Procter & Gamble, it's Lange's job to reduce waste not only of finished products but also in the process of designing, testing and creating finished products. Lange is also chief technologist for reliability engineering at P&G, and head of computer-aided engineering.

P&G has made a significant commitment to virtual computing in product design and development. During a conference call with investment analysts in 2003, P&G CEO A.G. Lafley said: "We are significantly expanding capabilities in computational modeling and computer-aided engineering, so we can do an increasing percentage of product and process design through virtual simulation."

In the 2008 annual report, Laffey further commented: "Virtualization is enabling P&G brands to co-design products with consumers. The same technologies allow us to show retailers virtual in-store displays for half the cost and less than half the time required for physical shelf designs. Computer modeling and simulation saved P&G about 17 years of design time in the last year alone."

Lange says the traditional paradigm of focusing solely on physical prototypes no longer makes sense. It's a very expensive and time-consuming process and isn't the best way to determine if a product is fit for use. "We don't build two or three bridges and then break them to see if they work," says Lange. "Why would we do that for products we're developing?"

Virtual designs

One of the big virtual success stories at Procter & Gamble was the development of the first plastic coffee canister.

The AromaSeal canister is a high-density polyethylene coffee container that replaced metal cans in use for 150 years. The new design is blow-molded with a proprietary six-layer barrier coextrusion that provides 12-month shelf life. The new plastic container is dent-resistant, lightweight and stackable. A built-in handle makes the can easier to hold.

The peelable seal includes a patented, one-way valve in the center, allowing freshly roasted coffee to off-gas in the container, eliminating back pressure and the potential for package explosion. Because of the seal, the canister can be filled and sealed immediately after it's roasted, instead of having to cool and naturally off-gas prior to being packaged. The seal also helps preserve freshness, keep air out and equalize pressure during shipping, which is important because the coffee is made in New Orleans and then shipped over the Rocky Mountains to the West Coast.  

"Without finite element analysis, we would not have been able to develop this canister," says Lange.

The coffee sold in the plastic canister boosted Folger's market share from 15 percent to 25 percent in three years. The Folger's brand is now owned by Smuckers.

In another example, P&G used FEA simulation to examine fitness of a large number of moving parts in a Braun electric shaver. Virtual simulation identified a single piece in an early design that couldn't pass a required drop test. The shaver was redesigned based on virtual examination.

P&G also does significant virtual testing of bottle strength when stacked in pellets during warehouse storage. "The bottle is a structural element in the warehouse," says Lange. Various load cases in bottles are tested. And that's increasingly important as P&G engineers take thickness and weight out of bottles to reduce solid waste, and cut resin costs.

Engineers also virtually simulate tendencies of metalized labels to peel. "It's all about materials' properties," says Lange. "You can answer those types of questions virtually. If you go ahead and make the stuff, it's an expensive proposition."

10,000 simulations

Lange estimates that P&G conducted 7,000 to 10,000 design simulations in 2008.  That work was carried out by a group of 10 highly skilled people.  "The only way you can do that much work is through automation of the analysis," he says.

"In a physical experiment, the test includes everything, even the things you don't know about" says Lange.  "In the virtual mode, you know everything that's included in the test. But you don't know what you don't know. Is one risk better than the other?"

He points out that virtual modeling, however, is based on testing conducted on real materials. P&G's simulation groups include scientists who conduct tests to build databases for the simulations. In an example outside of P&G, Moldlfow introduced simulation of plastic flows inside mold cavities in 1974. Today, simulations by Moldflow (now part of Autodesk) are based on tests performed on more than 8,000 standard plastics compounds, and more than 4,000 proprietary compounds developed for specific customers, such as General Motors.

Lange is quick to point out that the trend to virtual simulation is being driven in part by rapidly dropping costs for super computing. In 2001, a unit of computing power cost $1.50 in his estimation. Today, that same unit costs 15 cents. Within five years, he feels it will drop as low as one cent.

As a result, simulations are much more fully fleshed compared to point estimates made in past years. An example in plastics is development of stress-strain curves that show performance of compounds at a variety of temperature and pressure points, not just the single-point information on supplier data sheets. As the quality of simulation improves, so does the capacity to capture more information and test more of what you didn't before. Lange describes the simulation process in part as automation of activities done by experts.

Started with Fortran

Lange has been a big fan of computing power since he first punched code on Fortran cards in a college class in 1974. "Computing has changed engineering as much as aviation changed travel," he says. Lange received a BSChE degree from the University of Missouri in 1978. He joined P&G that year as a Product Technical Engineer. His group is part of corporate R&D at P&G, which receives $2 billion in annual funding-more than the GDP of some African nations.

P&G's work in simulation dates to the 1980s when the company began work on reliability engineering - which is basically the study of why systems fail. Lange calls that "pathology work" and it's the low end of potential for virtual study. Reliability engineering makes broken systems work faster, rather than designing optimal systems from the beginning, in Lange's view.

P&G engineers studied systems used at Los Alamos, and then developed models that could predict systems' performance. The approach was first used on a retooling of a production line for Pampers diapers. The tools were then used to improve product designs, avoiding $80 million in capital costs in the 1990s. 

Lange is careful to distinguish simulation work from the work done by product design teams. "They're worried about, shape, equity and artwork. Lange defines equity as  locomotion that gets people excited. We work on the simple things: Can we pack it? Will it break? Does the lid fit? Are we making the most economical use of materials?"

Physical prototypes become the confirmatory experiment later. "So instead of the prototype being a ‘Let's see what happens'; it's ‘We expect this to work.'"

For touch-and-feel prototypes, P&G makes widespread use of rapid prototyping equipment. "Those aren't what I'm talking about," says Lange. "I'm talking about the ones where someone says, ‘Make me three pallets' worth I can run on the packing lines."

Modeling and simulation employees are embedded in the businesses at P&G where designs are made. The core group works on tool sets that are used by the deployed employees.

As the power of simulation has grown, so has its deployment within Procter & Gamble. It has evolved beyond product development into process development. In what Lange describes as a virtual race track, engineers start with a CAD file and simulate the progression of a bottle on a packaging line. Simulations show the tendency of some bottle designs to bump and fall, clogging the line. Expensive work-arounds are avoided by the line simulation. Sometimes a simple change in container design can solve the problem.

Role of pathology

Lange says many organizations use their modeling simulation groups to perform design pathology work. That is, to determine why designs did not work after the fact. It's a virtual trial-and-error system," says Lange. "I only use pathology work to build the credibility of our organization. The real goal is to conduct analysis-led discovery. I want to develop 128 different versions to make sure that we are optimal when we first go into production."  Lange describes analysis-led discovery as determination of the optimum space where design engineers should be working.  "When you're operating in this space, you're really cooking."

As a final step, Lange says it's important that engineers involved in virtual simulation formally quantify their savings with involvement of financial staff for credibility. Savings at P&G are broken into four buckets: capital avoidance, materials savings, innovation savings, and new business creation, such as the Folgers can. Lange says his group saves about five times its costs on average based on data confirmed by P&G finance officers.

"When I go to management ask for $1 million for a new computer, they are taken back," says Lange. "But I ask, how much would it cost for 153 mass spectrometers? How much would it have cost for the molds to make those prototypes? How many people would it take to build and test those prototypes? That computer cost is actually pretty cheap."

The P&G modeling and simulation group operates in a 3,500 core processing environment that includes work for computational chemistry. Lange's group likes to work on big models. Million-element problems are not uncommon. "We are trying to use models that are predictive, not just relatively correct," he says. That's a sea change from the way finite element analysis has traditionally been used, he says.  

There are some caveats in the outlook for modeling and simulation of engineering work.

One is the lack of adequately trained candidates coming out of engineering schools. Lange feels that schools are still training engineers much as they were 30 years ago and not providing adequate capabilities in computer skills.  P&G maintains relationships with 65 different universities around the world, in addition to government-funded research centers, such as Los Alamos.

Another is the lack of engagement of smaller companies. "The Fortune 50 really gets it," says Lange. "If you have 50 engineers in your company, at least one or two of them should be doing simulation and modeling work," he says. "At some point, you have to make a decision to do this." Some small companies, however, are successfully using engineering service providers for modeling and simulation work.

Read up on all the candidates for Design News' 2009 Engineer of the Year!