A Supercomputer Dreamed Up a Bird-Like Airplane Wing
Inspired by evolution in nature, Danish engineers used supercomputing to design a wing structure that resembles the interior of a bird’s wing or beak.
Evolution is the master engineer. Just look at the nearly hollow bones of a bird, impossibly strong and stiff for their feathery weight. They were forged over millions of years of natural selection with the random mutations of countless generations contributing piece by piece to the ever-improving design.
The human design process, by comparison, doesn’t take millions of years, but is also limited by our own creativity as well as practical production constraints. Most new cars, buildings, and bridges are largely based on what’s come before, because truly novel and risk-taking designs require multiple rounds of expensive prototyping and testing.
Over the past two decades, however, engineers have been trying to mimic the iterative process of evolutionary design — adapt, test, adapt, repeat — with the help of increasingly powerful computers. It’s called computational morphogenesis, a digital recreation of the ages-long biological process that gives all living things their distinct forms. The technology has mostly been used to design single structural components in larger systems so that they are stronger and lighter than before.
Now a team of Danish researchers has designed a surprisingly “organic” model for the inside of an airplane wing by harnessing the immense computing power of 8,000 CPUs. In a letter published in Nature, Niels Aage and colleagues from the University of Denmark showed off an intricately curved and fractal-like airfoil design that’s strikingly similar to the interior of a bird’s wing and beak.
“Nature has been optimizing itself using trial and error, meaning it has constructed a design and the species has been out flying and the next generation will improve on that,” said Aage in a Nature podcast. “That’s exactly what we’re doing. We come up with an initial design and then we improve on that iteratively. The result has many similarities to nature, but the process itself has equally as many similarities to nature’s own evolution.”
The process started with a 3D model of a standard airplane wing from a Boeing 777 with everything pulled out of it except the stiff exterior skin. Then Aage and his team divided the empty interior space into 1.1 billion individual voxels, which are like three-dimensional pixels or tiny Lego bricks.
“We then ask every point in space” — each of the 1.1 billion voxels — “if it should be material or void, such that we will end up with a structure that is as stiff as possible,” explained Aage, while using the least amount of material.
The process isn’t exactly trial and error. The computer is programmed with “rigorous mathematics,” said Aage, to apply strict design rules and variables to each point inside the wing.
“Over an iterative process, we repeatedly perform a structural analysis so we figure out where in the internal structure is the material heavily loaded,” explained Aage. “If it’s heavily loaded, there should be more material. If it’s not carrying any load, we can remove material. And through a couple of hundred iterations, we have a final design.”
The result was something that looked almost nothing like the interior of a conventional airplane wing, which is buttressed by box-shaped supports of intersecting straight lines. Instead, the evolutionary computing process spontaneously produced curved beams running the length of the wing and fan-like supports with a complex, open-lattice structure.
Writing in an accompanying commentary, Matthijs Langelaar of the Delft University of Technology in the Netherlands said that the same kind of complex computer modeling has been used by engineers to better understand fluid mechanics, thermodynamics, and acoustics, but never at this level of detail or resolution.
“[Aage and his team] produce designs comprising about 200 times more data than current state-of-the-art techniques,” wrote Langelaar, adding that the computational framework used by the Danish researchers allows for “six times finer detail in each spatial direction” than current methods.
The novel “giga-voxel” wing design achieved the stiffness and strength of a conventional wing while using up to 5 percent less material. Aage calculated that the decrease in weight alone would result in potential fuel savings of up to 200 tons per year, reducing both cost and greenhouse gas emissions.
Langelaar called the unprecedented resolution of the airplane wing “a leap forward in the capabilities of computational design,” but warned against predicting that supercomputers will soon compete with the engineering genius of evolution.
Aage’s computer models, for example, didn’t take into account the “remarkable fractural toughness” of natural materials, wrote Langelaar, which cannot yet be achieved with synthetic materials. Also, perhaps more importantly, we have no existing production technology that can produce the kind of wildly intricate designs dreamed up by the supercomputer.
In his podcast interview, Aage admitted that such components would have to be 3D printed and there are no such large, metal 3D printers on the market. But that doesn’t mean that the novel design is wholly impractical.
“What you can use it for right now, you can pick out some of the main trends in the design and use that in the next generation of aircraft.”
WATCH: What the Heck Is 3D Printing?