Engineers Optimize Designs to Keep them Light and Strong

Weight optimization is a balancing act to juggle competing objectives.

The number of design variants involved in a lightweighting project is usually too numerous to be executed manually. Shown here is a typical automated workflow in ANSYS Workbench for optimizing composites.

ANSYS Workbench The number of design variants involved in a lightweighting project is usually too numerous to be executed manually. Shown here is a typical automated workflow in ANSYS Workbench for optimizing composites.

When giving talks at conferences, Pierre Thieffry, ANSYS’ product manager, often asks his audience, “Raise your hand! How many of you get your first simulation right?” As he expects, usually no hand goes up. That’s the point he’s trying to make—nobody does.

“When you do design evaluation, you do one, then another one, then another, and so on,” Thieffry says. “So rather than just doing one at a time until you stumble on the right one, why not automate the exploration process?”

That’s the principle behind software-driven weight optimization, or lightweighting, as the practice has come to be known among automotive and aerospace manufacturers. It’s now possible to delegate a few dozens to hundreds of design variations to a machine, capable of assessing them at a much faster speed than any individual human can.

In theory, you can automatically try out even thousands of design alternatives by varying the thickness, targeted region for trimming, and material. At that scale, however, you’d need to rely on a high-performance computing (HPC) system to be able to process the job in a reasonable timeframe. So in reality, the cost (measured in both time and money) of running these calculations, the need to keep the data manageable, regulatory requirements, and company-approved material choices are bound to exclude some design variants from the exploration pool. Lightweighting is not as simple as shaving off as much material as possible; it’s a balancing act.

Inevitable Conflicts

Bob Ryan, president of Red Cedar, part of CD-adapco, points out the competing objectives involved in vehicle lightweighting.

“Consumers want vehicles that have excellent crashworthiness, durability, handling and performance, but also provide good gas mileage,” he says. “This requires vehicles that have a good combination of stiffness in the right locations, combined with lower overall mass. It is a balancing act to find the best combination of materials, geometry and part thicknesses to yield the desired vehicle performance.”

Matteo Nicolich, enterprise solutions product manager for ESTECO, emphasizes a multidisciplinary approach to lighweighting, noting, “If each discipline is pursuing lightweighting on its own, they have to periodically reconvene with the others to discuss their designs. It’s a lot faster for all the disciplines to improve the product as a single system.”

Ryan agrees. “What carmakers typically experiment with during lightweighting is geometry, thickness and material,” he says. “Your intelligent software will try out various combinations of those variables. They’re not infinite; they’re confined to, for instance, the types of materials that meet your company standards.”

Ryan estimates that, in a typical lightweighting project, the combinations the software must try out “could be between 500 to 1,000.” Therefore, a manual approach is impractical. Engineers must rely on optimization software to automate the process.

ANSYS’ Thieffry similarly observes, “In lightweighting, you’re trying out different shapes, so you definitely need a way to automate the process using design parameters.” Adds ESTECO’s Nicolich, “You also need a way to mine and explore the data afterward.”

ESTECO’s modeFrontier comprises an environment to automate the simulation runs and a design space to gain insights from the simulation outcomes.

“In the second phase, you have multiple parties with multiple objectives,” Nicolich continues. “modeFrontier lets you analyze the tradeoffs and rank them. That’s where all interested parties from multiple disciplines can come together to discuss the options.”

Keeping the Data Manageable

The key to prevent a data explosion in optimization is to restrict the archive to only the most promising design variants and their associated parameters.

“Simulation software is now fast enough to recreate a scenario at any given time, so it makes more sense to save the input parameters and design configurations, not the whole simulation result file,” says Thieffry, noting it’s done usually in gigabyte scale for complex jobs. “You can always rerun the job with the same configuration—that’s the smartest way. If you’d like to save the result file, save the one for the design you decide to go with, not for every configuration you tried out. I’m sure the IT guy will like that strategy, too.”

But traditional product lifecycle management (PLM) packages may not be the best for optimization data archival.

“A PLM system won’t let you evaluate the performances of design variants, or iterate automatically,” Nicolich points out, noting that the intermediate steps you go through to get to a lightweight design are not product variants—most of them will never go beyond the computer screen. “They should be managed outside the product lifecycle. Otherwise, you’ll clog the PLM system with a huge amount of data.”

Lightweighting involves material properties, safety requirements and design parameters used in the simulation runs. The purpose of keeping them is to be able to recreate the same simulation scenario at any given time, and to be able to explain (or defend, as the case may be) a certain option chosen in the final phase. Both the data type and the treatment they require are significantly different from those managed in PLM, such as CAD file versions, approval records, change orders and supplier data.

Even though the optimization software is doing the heavy lifting in searching the space of possible designs, Red Cedar’s Ryan says engineers are not off the hook. The optimization software typically yields what Ryan describes as “a small number of design clusters or families, usually numbering less than a dozen, which show the greatest promise of delivering superior performance at lower cost.”

It normally takes additional human intelligence, expertise and experience to select the best option.

“There’s no one answer to an optimization question,” says ANSYS’ Thieffry. “You usually end up with a number of good designs that satisfy your criteria. You can’t just pick one answer based on the numbers. The optimal shape the computer proposes is numerically optimal, but it’s probably not something you can easily manufacture.”

Strength in Numbers

Red Cedar’s Ryan points out that modern optimization software is changing how lightweighting projects are performed.

“Traditionally, when engineers did design exploration, they’d simplify the simulation model, screen out seemingly unimportant variables, sample the model, build a response surface [an interactive map showing the design variable changes and the associated consequences], and optimize using that response surface,” he explains.

A limited sample pool based on a reduced set of variables was the right approach in the 1990s, in Ryan’s view.

“At the time, the software wasn’t good enough to accurately regenerate designs that were too far away from the initial design,” he says. “We didn’t have ubiquitous computing power, and we didn’t have efficient design exploration algorithms that could work on any problem.

“But that’s no longer the case now,” he continues. “With software like Red Cedar’s HEEDS, you no longer need to simplify your simulation model or extensively screen out design variables a priori. You can explore your fully validated simulation model directly, without simplification.”

Ryan notes that it’s common for Red Cedar users to explore 400 to 500 design variables at a time in HEEDS. “If you want to do successful lightweighting today, you have to be able to look at a lot of variables, and a fairly large range of designs,” he says.

Made in Composites

The long history of traditional manufacturing materials, such as steel and aluminum, gives manufacturers the ability to accurately predict how they’ll perform. But in lightweighting, manufacturers turn to advanced materials made of a combination of multiple materials.

Roger Assaker, MSC Software’s chief material strategist, says, “Plastic reinforced with chopped glass fiber or continuous carbon fiber-reinforced polymers, for example, have different properties and characteristics than metals. The actual properties of the composites typically varies across the structures and in different directions.” Therefore, understanding the nature of composite materials may be the most critical area in the pursuit of lightweighting.

ANSYS’ Thieffry notes that if you “pick a standard material, there may be about four or five numeric values to represent its deformation rate, stretching behavior, and so on. But with composite materials, the number is multiplied. Adding the effects of the orientation of the materials [the direction and angle of the ply layout], for instance, is immensely complex mathematically.”

MSC Software Composite materials commonly used in lightweighting are designed from the ground up, including the direction and pattern of the fiber. Shown here is a close-up of composite materials in MSC Software.

Most simulation software comes preloaded with a database of standard manufacturing materials, but for composites, the user may have to rely on lab tests, public resources and material providers to obtain values representing the quantities of material.

“Currently, the material data available for composites is limited, so the Tier I manufacturers and original equipment manufacturers who are spearheading the lightweighting project would have to test the materials themselves to get accurate data, or work closely with a material supplier to get the data,” says Assaker.

Even though the nature of composites offers almost infinite possible varieties, the manufacturer’s own choices may be much more limited, Assaker points out.

“Qualifying new materials is kind of expensive, so you want to reuse preapproved materials if possible,” he adds.

The switch to advanced materials also requires thinking about other angles—literally. “With composites, the material’s properties and resistance to stresses and loads are not the same in all directions,” Assaker says, noting that the ply stack’s direction, angle and layout all affect its strength differently in each direction.

Treating composites like classic black metals, he says, can lead the engineer to build a part with more materials than necessary, thus compromising the project’s lightweighting goal.

“Material is one of the three pillars of MSC Software’s strategy,” says Assaker, noting that MSC Software’s acquisition of e-Xstream engineering, which he founded, is a result of that. “Our Digimat Software lets you model the composite material in detail from the ground up, starting from constituents (polymer/fibers) and underlying microstructure (fiber length, orientation).”

He recommends not just switching to composite throughout the whole design, but identifying the right composites for the chosen regions of the design.

“For example, in some areas of the design, you might want continuous carbon fiber; in others, perhaps a cheaper thermoplastic reinforced with chopped glass fiber is sufficient,” Assaker explains. “Digimat will allow you to accurately model all these types of composites, and help you choose the right material and converge to the optimal design.”

Designer-driven vs. Expert-driven

Simulation software makers have had good success in promoting their technology to the designers, thereby expanding the market beyond the specialists. Today, simple stress analysis and flow simulation have become an integrated part of CAD software and the conceptual design workflow. But what about lightweighting? Can average designers grasp its intricacies?

MSC’s Assaker says he believes they can at least play a role. “Going after the last kilograms or pounds should be the expert’s job, because there’re too many variables and complexities involved at that point. But topology optimization at the beginning of the design can be done by designers, if they have basic understanding of composite materials,” he says. “They can do the first 80% of lightweighting in topology.”

One company focusing on simplifying topology optimization is solidThinking, an Altair company. It recently released the latest version of Inspire, which allows designers to investigate structurally efficient concepts quickly. The company refers to Inspire 2014 as a concept development tool, as opposed to Altair OptiStruct, its structural analysis solver for design and optimization.

“With solidThinking Inspire 2014 we focused on enhancing the concept development process by proposing designs that can be rapidly iterated and easily exported to the user’s preferred CAD tool,” notes Andy Bartels, program manager for solidThinking Inspire in a company press release. “We put a strong emphasis on improving the usability of the software while adding new features like geometry simplification tools for easier model setup and analysis to help users verify their concepts, all directly in the Inspire interface. These new features will allow customers to apply Inspire to a much broader set of design problems.”

Automotive and aerospace manufacturers are currently the most aggressive in adopting lightweighting strategies. But there are signs suggesting other industries—consumer goods and electronics, for instance—have begun to experiment too. And the manufacturers’ balancing act is crucial in producing smartphones, large-screen TVs and vacuum cleaners that are light enough for us to carry or haul around, but strong enough to survive the inevitable drops and bumps.

More Info

Share This Article


About the Author

Kenneth Wong's avatar
Kenneth Wong

Kenneth Wong is Digital Engineering’s resident blogger and senior editor. Email him at [email protected] or share your thoughts on this article at digitaleng.news/facebook.

      Follow DE
#13048