Looking to improve aeroplanes, computer designs bird wing
Using big data and supercomputing results in potentially much more efficient aircraft components, Michael Lucy reports.
Slice open a pelican bone and you’ll find the hard shell is supported inside by little more than an airy web of bone filaments. Danish engineers using supercomputers to find an optimal structure for an aeroplane wing have arrived at a surprisingly similar result.
The design could make a wing 2-5% lighter than existing designs, write Niels Aage and colleagues at the Technical University of Denmark in a paper published in Nature this week. For an aircraft the size of a Boeing 777, this could mean a saving of between 40 and 200 tonnes of fuel each year.
The real innovation of the research is the unprecedented detail of the computation, which took 8000 processors several days to complete: the model is composed of more than a billion solid elements, or voxels. (Earlier attempts, using a mere five million or so voxels, produced unsatisfactory results.)
“It’s an interesting study in terms of the large scale,” says Gareth Vio, an aerospace engineer at the University of Sydney. “It’s common to do this in fluid dynamics, but not for structure.”
The approach used by the Danish team is known as “computational morphogenesis”. It aims to overcome the limits of designs based on human intuitions by mathematically describing a problem, such as making the lightest possible wing of a certain shape and material that can withstand certain forces, and letting the computer generate a form that solves it.
It’s a gradual process that approaches a solution over hundreds of iterations, each a little better than the last. Aage says it’s as if the computer is programmed to always walk uphill – but “it’s walking in the Himalayas in a heavy fog only carrying a small flashlight able to shine one metre ahead”.
The similarity of the final product to bird bones is not entirely surprising, since they too are the result of a long, iterative process – evolution – that addressed essentially the same problem of strength versus lightness.
“Birds have been doing this for a lot longer than we have,” says Vio, “and they’re pretty good at it.”
Vio also points to the work of Mike Xie and Grant Steven, who developed similar techniques at the University of Sydney in the 1990s (albeit with much less computing power). They also found themselves repeating the discoveries of evolution: they calculated that the optimal shape for an object hanging in air, for instance, is practically identical to a cherry.
There’s one major problem with the wing design: “It would be very hard to construct with current technologies,” says Aage. Nonetheless, he says the giga-voxel approach is already being applied in other areas of engineering, such as bridge design. And he hopes that some parts of the wing model, such as curved and diagonal ribs, may inspire new avenues of inquiry for structural engineers.
Vio doubts how novel some of these features are. “Boeing already has a patent for curved spars,” he says. Still, he adds, “There will come a time when you’ll be able to extrude something like this from a giant 3D printer. It’s the evolution of technology.”