- Topology optimization designs by mimicking natural selection
- Giga-voxel resolution allows unprecedented design detail
- New aircraft wing design could result in significant fuel savings
Ever since Leonardo da Vinci sketched experimental flying machines in his Codex on the Flight of Birds in 1505, humans have looked to the avian world for insight into getting off the ground.
500 years later, a supercomputer followed da Vinci’s lead and has reimagined a commercial aircraft wing using a technique that mimics the organic structure of a bird’s anatomy.
This technique, called topology optimization, works like natural selection, only hundreds of times faster. Engineers designate a block of material and a set of constraints, such as dimensions or maximum weight, and then the computer takes over the process.
The optimizer then redistributes material from the block to minimize or maximize objectives within the design domain,ultimately determining both the shape and size of each feature.
Scientists then analyze the output and identify regions which can benefit from additional material and generate a new design. “We repeat the cycle until the designs stop changing,” says Niels Aage, professor of mechanical engineering at the Technical University of Denmark (DTU).
Each design cycle is performed using rigorous mathematics such that the new design always outperforms the previous. In the case of the aircraft wing, the end result is a structure that is very curved at the root and almost straight at the tip – like a bird’s wing.
“The fact that the optimization framework produces designs that resemble existing aircraft means that 120 years of previous research hasn’t been wasted,” says Aage. “But at the same time, many exciting new features have appeared.”
Survival of the strongest
Mother Nature has had eons to perfect her creations. But to develop from raw materials to finished concept in just a few months instead of over several millennia takes a lot of computing power.
Just as the quality and of the photos you snap with your phone are limited by the number of pixels displayed, topology optimization is limited by the number of voxels (3D pixels) the design system can handle — typically only a few million.
Because of this constraint, most topology optimization is used only on simple structures. But if a significant boost in resolution capability becomes possible, it means an opportunity to redesign and improve larger, more complex structures.
Aage first tested his new optimization code on the JESS cluster at DTU. But when the team needed more resources, they applied through the Partnership for Advanced Computing in Europe (PRACE) and received time on the Curie supercomputer in France.
With a peak performance of 2 petaFLOPS, Curie’s 90,000 cores produced designs with giga-voxel resolution — more than 200 times what has been previously reported — allowing the researchers to model a full-scale aircraft wing for the first time.
Light as a feather
Topology optimization is particularly useful for minimizing structural weight subject to constraints on stresses — just what is needed for a 70 ton aircraft screaming through the sky at 600 mph.
By extracting the novel spar and rib configuration and re-analyzing these features alone, Aage’s team estimate that the wing would be 2-5 percent lighter than existing aircraft wings, resulting in significant fuel savings.
Aage admits that the full optimized design is too intricate to be manufactured using current methods, but rapid advances in 3D printing and additive manufacturing may soon overcome that obstacle.
First introduced in 1988, topology optimization remains a relatively new method for solving engineering problems, and many of the possibilities it offers have yet to be explored.
Recently published in Nature, Aage’s paper has highlighted only two features that could be implemented in aircraft design today, but their experiments revealed the opportunity for many more.
The same methods that produced the novel aircraft wing design could be used for many other applications, such as an improved aircraft fuselage, automotive design, or even high rise buildings.
Perhaps some future inventor may use high-resolution optimization to finally realize the dream of da Vinci’s personal flying machines. If not, we still have lighter planes and stronger buildings to look forward to.
