New design software transforms a concept into a multitude of configurations

deepSPACE is not a futuristic movie, a new video game, or the next season of a classic TV series. In fact, new design software developed by an aerospace engineer at the University of Illinois at Urbana-Champaign isn’t about space at all. This new tool takes your concept and requirements and quickly generates conventional or unusual design configurations, including a 3D CAD model and performance evaluations.

“We wanted to do for engineering and design what big AI language models did for text,” Jordan Smart said. “Right now, when you open engineering design software, you’re greeted with a blank screen. With deepSPACE, you tell it your needs and it generates 100 to 1,000 feasible concepts in the time it would take a human to examine them one or two. This gives you a much better picture of the larger design space.

And Smart said deepSPACE isn’t limited to just physics-related questions: “It’s trained on a mix of historical and simulation data, but can use standard cost estimation tools and get at least this level of return for a cost analysis.”

To demonstrate its flexibility, Smart and research partner Emilio Botero used deepSPACE to generate physical system designs on girders, wheels and planes, but also on operational logistics networks. They have created partnerships with aircraft and automobile manufacturers to ensure that deepSPACE is useful to researchers and industry professionals.

The research is published in the AIAA AND ASCEND AVIATION FORUM 2024.

“We learned that while individuals want deepSPACE fully loaded, companies prefer to create custom models linked to their own data and knowledge. On the back end, we can create our own models to use for research or design , but it can also be used from scratch. It’s a teachable platform,” Smart explained.

According to Smart, deepSPACE is more efficient than older optimization algorithms. “Where others have reported running 20,000 simulations to begin setting up their design space, we were able to achieve similar results with only around 250 samples. So with about 100x fewer data points, you can get a real sense of the tradeoffs. in design space.

“When you’re designing an airplane and you want to know what effect changing the wing, adding an engine, or increasing the payload might have on the design, those kinds of sensitivities and tradeoffs are complicated Traditional methods may require thousands of different design points before they can reasonably interpolate between them. Because deepSPACE builds a full generative model, it is able to interpolate much more successfully on fewer data points. “We’re able to do the same level of prediction with the same level of accuracy, faster and more cost-effectively.”

The lower cost makes deepSPACE particularly valuable in aerospace applications. “We rely on simulation because building airplanes is expensive. But we are studying how it can be used in other industries,” Smart added.

The fact that deepSPACE provides a 3D CAD file is an additional feature. Smart said that the output of other image generation programs cannot be opened and used with other design software with all of its layers and effects still intact.

“With deepSPACE, you get the exact same type of raw file as if a human had created it. So any types of edits or changes you would want to make are there and available. It fits right into your workflow work as if you had subcontracted the work to another company and it was part of their deliverables.

Smart said deepSPACE can create a unique design conversation with the human engineers who train it. Smart explained: “One of the designs generated by deepSPACE seemed absurd to us. We said, “Clearly something is wrong. It was designed according to a set of requirements, but nothing like that was in the training data. But when we looked at the results, the actual simulation results for what it generated looked reasonable and met the requirements. »

The aircraft in question had relatively short wings with offset control surfaces at the rear to provide balance and stability. Smart said he wasn’t mining the simulation or making something that couldn’t be built, so they started looking at it more closely and realized they’d seen something like that somewhere. Eventually, they discovered that it resembled a real plane built and flown by a leading aircraft manufacturer.

“I had set up the training data, the simulation and the actual learning algorithm. We gave deepSPACE a training set from three conventional tube-and-wing aircraft, the Concorde and a mixed wing body concept From there he started generating his own concepts and comparing them to simulation and learning Sometimes it generated something non-physical, but from there. , he was learning where the edges are.

“Without a human telling us ‘don’t think about this or that,’ he was able to conduct his own experiment, like brainstorming, and come up with something we didn’t expect. My personal bias would have been to throw it away,” Smart said. .

Smart said deepSPACE was able to show him the simulation results and how the design met his requirements. He found a viable solution to the problem, exactly as he was designed to do.

“We’ve given him a set of tabulated historical data, from which he improves his understanding and begins to explore and experiment. I can build a basic model to get the results, but then I can treat it like a playground or a sandbox I can run a new simulation that is not in the historical data, see how this adds to my knowledge database.

“For years I felt like we had incredible analysis capability, but the bottleneck had become us. We have simulations, but a human simply can’t run thousands of simulations over and over again, reject the bad ones and find the good ones and build that kind of intuition. deepSPACE is the first generation of systems designed to be like an engineer in your pocket You can define the problem and come back later to find a multitude. different options, and go further with much more information about the capabilities you already have.

Although created with academic and industry professionals in mind, Smart has other ideas: “My goal is to get middle school students to use something from deepSPACE. They may not know physics or have all the skills needed to make a CAD drawing, but if they have an idea for a car, train, spaceship or something like that, they can talk to deepSPACE and run it, then they can make their own changes and see what happens next.