The GovDesignHub recently had the opportunity to attend some of the sessions and panel discussions from last year’s Autodesk University (AU). This annual conference, held in mid-November in Las Vegas, Nevada, brings together innovators in architecture, engineering, construction, product design, and manufacturing to share ideas, advance industry practices, and discuss the potential of exciting new technologies.
One of these future technologies is generative design. Generative design is the next frontier in computer-aided design (CAD) for engineers in virtually all manufacturing industries. It harnesses the power of AI to develop new high-performance design concepts that help solve complex challenges, reduce manufacturing costs, increase customization, and optimize performance.
During one of the sessions at this year’s Autodesk University, Ryan McClelland, Research Engineer for the Instrument Systems and Technology Division at NASA Goddard Space Flight Center, gave his insights into how NASA uses generative design and AI to produce exciting new components for spaceflight structures.
During his presentation, McClelland explained how generative AI has revolutionized the process in which NASA designs and manufactures parts. This has enabled the agency to take a “small to mid-sized metallic structure” from requirements to fabrication in just a few days, and do so in a mostly automated fashion.
McClelland…turned to generative design to…create a bracket solution that would meet all requirements while still being manufacturable. The AI produced 31 iterations of two different designs in one hour.
With generative design helping to conceptualize parts and components, and advanced manufacturing solutions helping to bring them to life, NASA has seen benefits that extend behind time savings. The agency has gained the ability to optimize the characteristics of components – including their strength and weight – while drastically decreasing the cost of component fabrication.
According to McClelland, the use of generative design and digital manufacturing has enabled NASA to, “…improve the mass stiffness and strength [of components] by a factor of two to four, while also reducing development costs by a factor of ten.” This technology has delivered incredible results for NASA, expediting projects and improving mission-critical components.
“[We have realized] some massive, great gains,” McClelland claimed. “[NASA] is taking [the design and fabrication of a component] that usually takes a month and having it done in a few days – having a structure that weighs 50 kilograms and making it weigh 20 kilograms.”
Generative design is also changing the role of the engineer, enabling them to own the process from design to fabrication. It’s bringing new, novel component designs to the table that may not have been considered previously. Its inherent speed and relatively low cost are also enabling engineers to iterate on designs more quickly and efficiently – making them able to try things and fail fast with far fewer consequences than with traditional approaches to design and manufacturing.
During his presentation, McClelland shared how he and another colleague created an ultra-futuristic and ultra-optimized component called the “EXCITE Tip/Tilt Bracket” with the help of generative design. The story of how this component was designed and manufactured with AI and advanced manufacturing is a testament to its potential for those who manufacture things for the government.
Not just any bracket
NASA needed to design a bracket to attach a tilt mirror to the back of a telescope for a balloon-borne platform whose mission was to analyze the atmospheres of exoplanets. However, even though the bracket itself was a simple design, the requirements would prove to be challenging. The bracket had to be very stiff, needed a form factor that would not interfere with the general assembly, and had to be light; about 0.2 kilograms.
McClelland then began a “human design” alongside another senior designer. After a few days of working on a solution, they had iterated four different designs. However, none of the four designs met all of the criteria for the bracket that was needed.
McClelland explained how the initial iterations of their design were either too heavy, not strong enough, or simply impossible to manufacture. “The first iteration was way too heavy. The second iteration, we just pocketed it out to remove mass and make it lighter, but it was not nearly stiff enough,” said McClelland, “Then, we got a little bit better in the third iteration, but still couldn’t meet the requirements since it wasn’t very manufacturable. And then [my associate at NASA] finally produced a radical design that actually met the requirements, but wasn’t really manufacturable from either CNC machining or [3D] printing.”
Generative AI to the rescue
McClelland and his colleague then turned to generative design to help them create a bracket solution that would meet all requirements while still being manufacturable.
“[We have realized] some massive, great gains. [NASA] is taking [the design and fabrication of a component] that usually takes a month and having it done in a few days – having a structure that weighs 50 kilograms and making it weigh 20 kilograms.” – Ryan McClelland
The AI produced 31 iterations of two different designs in one hour. These designs met all the requirements for the bracket design; with at least some of the iterations greatly exceeding the baseline requirements.
One area where the generative AI was able to deliver measurable benefits was in bracket stiffness, which saw a three-times improvement over human-designed concepts. McClelland explained why this is important, “The stiffer [the bracket] is, the less likely it is [the be impacted by] the vibration of the rocket,” said McClelland, “So it’s a big advantage for us.”
Lastly, and maybe most importantly, the part was produced quickly and at a low cost. McClelland shared that, with the help of the generative AI, the CAD was easy to fabricate via automated CNC, only cost about one thousand dollars to manufacture, and only took three days to produce.
These results were so significant that NASA has since used generative AI for about 40 more different applications since this project.