Generative Design

What Generative Design Could Bring to AEC

Computer-designed parts remove humans from the equation to make alien-looking components that are lightweight and super strong.

According to NASA, it costs approximately $10,000 to get a pound of cargo into space. That’s $625 per ounce or $22 per gram. Suffice it to say, saving even a little bit of weight can save a lot of money.

Increasingly, for engineering challenges where weight is the enemy, performance is critical and physical constraints are complex, companies from Airbus to Mercedes Benz to General Motors are turning to a new technology to help meet their needs: generative design.

Generative design architecture uses advances in computing power and additive manufacturing—also known as 3D printing—to eschew traditional design processes in favor of a workflow that lets software handle most of the decision-making.

The process proceeds as follows:

A user inputs a list of design constraints—these could be attachment points, required strength, size, length, width, type of material or other inputs. The computer then searches through all possible options in 3D space to find designs that meet the list of parameters. The designs can then be virtually tested to see which ones work best.

Usually a group of designs, each with its own pros and cons, is then fed back to the human user, who can pick which one best suits their design needs. At its core, the technology aims to let pure physics drive the design process.

“From a structural standpoint, you can say I need this amount of resistance to torsion; this amount of  horizontal support, and I need this amount of lateral stiffness; I need this amount of axial strength,” said Steve Pilz, an additive manufacturing engineer at ANSYS Inc., an engineering simulation software company based in Canonsburg, Pennsylvania. “You can design that in.”

Because the computer isn’t limited by any constraints other than what you program in, it often returns designs that look almost alien in nature: Material is only used where it’s most necessary and parts often wind up containing lattice structures devoid of any straight lines and riddled with empty space.

Benefits for engineers

Generative design, when implemented correctly, has many benefits for engineers. It allows them to manufacture bespoke parts that can handle complex loads and forces coming from multiple directions.

“When the designs are 2D and all the loads are in the same plane, like building a truss, humans are pretty damn good at coming up with the optimal design,” said Dr. Andreas Vlahinos, principal of Advanced Engineering Solutions LLC, a computer aided engineering services firm based in Castle Rock, Colorado.

But when structures are twisted and compressed with complex 3D forces, the human brain begins to struggle. “It becomes so physically complex—this loading and resisting these moments—that the human brain can’t come up with these designs on its own,” Pilz said.

Generative design also excels when performance is more important than cost and when parts don’t need to be mass produced. Weight reductions equate to massive cost saving for the aerospace industry and worthwhile performance and efficiency benefits for the automotive industry.

Companies like Stanley Black and Decker are using the technology to make tools lighter, stronger and simpler to manufacture by combining multiple pieces into one. When combined with MRI and CT scan technologies, generative design is helping doctors and surgeons make bone implants and other medical devices built for the precise topology of each individual patient’s own body.

Generative design limitations

As amazing as the possibilities for generative design architecture is, the technology is still limited by cost and manufacturing constraints. If the performance of the part isn’t important, the low cost of traditional manufacturing techniques are always going to win out over additive.

Likewise, if a part needs to be mass produced, additive becomes too expensive. You’re not going to see generative designed cupholders in your Ford F-150 any time soon, but you can already see examples creeping into Formula One racing cars. Engineers are also limited by practical matters and existing infrastructure.

“You wouldn’t additively manufacture a steel bridge,” Pilz said. “Yes, you could create these wildly intricate shapes, but from a cost standpoint with the infrastructure we have, it’s cheaper to get a bunch of I-beams, cut them, rivet them together and throw them on a concrete pier.”

Moreover, while specific material constraints can be input into generative design software, it’s often still easier for an engineer to look at the available materials and parts and come up with a realistic design solution.

“[The computer] will come up with weird designs but are they manufacturable?” Vlahinos asked. “Nobody has a nice complete set of manufacturing constraints, but we’re getting there.”

As mentioned previously, additive manufacturing is vital for getting the most out of generative design. While it’s possible to include design constraints in the software that will force the computer to design parts that are buildable using other manufacturing techniques, generative really shines when paired with additive because of the freedom that it gives.

Whether you’re working with plastic or metal, printing components layer by layer greatly expands the library of potential shapes. “For instance, if I say, ‘give me a perfect, spherical hole inside a solid piece of steel,’ you can’t make that [with traditional manufacturing techniques] because you can’t get a cutting tool in there to remove that material on the inside—you’d need some kind of access hole,” Pilz said. “With additive manufacturing, that’s easy to do—it’s better to do—because you don’t have to print that material at all.”

Expanding possibilities

Even though additive is still in its relative infancy and is still beleaguered by numerous technical and quality control challenges, the technology continues to expand its range of materials and geometries. Combined with improvements in software—especially the ability to design, model, test and print parts in the digital realm—and increases in computing power, the realm of what’s possible with generative design architecture is only poised to expand in the future.

Whether humans will one day completely turn over control of the design of everyday objects to computers remains to be seen, but Pilz seems to think it’s possible that we might come close.

In the meantime, engineers and manufacturers will need to be extremely strategic about where they choose to employ this technology.

The human element is still critical for weighing the pros and cons, calculating the cost benefit analysis and choosing the best design from the pool of potential solutions offered up by the computer. Engineers are also necessary for getting all the different software to communicate and for understanding the limits of additive manufacturing and how to program the generative tools in a way that works within those constraints.

For now, “the human has to be in the loop,” Vlahinos said. “There’s no way around it.

Here’s how to build a bridge underwater.