AI has the potential to allow engineers to design products faster than before while meeting design specifications, sometimes in new and unique ways.
Jean Thilmany, Senior Editor
Artificial intelligence could be said to be the new hot buzzword, as it seems to be making inroads into all types of software.
“We don’t see a lot of AI yet in a CAD environment, but it’s coming,” says Andreas Vlahinos, chief technology officer Advanced Engineering Solutions, a research and design firm in Castle Rock, Colo.
AI is a broad field focused on using computers to do things that require human-level intelligence.
But how CAD will make future use of AI is still up for debate, he adds.
While some CAD makers are delving into AI functionality, the marriage of AI and design software is in the early stages, says Jon Hirschtick, chief executive officer of Onshape, which makes cloud-based CAD software. “AI has great potential, but so far no one has illustrated how it will unfold,” he says.
AI doesn’t have a one-size-fits-all definition within any industry yet, says said Gian Paolo Bassi, chief executive officer at Dassault Systèmes. “Today, there’s a huge debate about what AI is. People say AI is machine learning, or they say it’s related to the neural network or to neuroscience. Definitions vary.”
The machine learning that AI depends on is actually already present to a certain degree in the CAD systems that include topology optimization and generative design capabilities. “The primary functions of these features within CAD is to automate the analytical steps of design, Vlahinos says. The computer generates designs from an engineer’s preliminary directions.”
The key focus of AI in CAD right now is design optimization achieved through the creation of more intelligent designs which are lighter, stronger and more economical. And, in some cases, more artistic, continues Vlahinos.
Typically, designers create their design step by step, analyzing certain junctions to get critical feedback about performance. They tweak the design if it doesn’t meet performance needs or customer specifications. The incorporation of AI, as it stands now, allows the designers to skip these time-consuming steps allowing the task to get over quickly and effectively.
Last year, for example, Autodesk released generative design to subscribers of its Fusion 360 Ultimate product development software. The design concept allows engineers to define design parameters such as material, size, weight, strength, manufacturing methods, and cost constraints–before they begin to design. Then, using artificial-intelligence-based algorithms, the software presents an array of design options that meet the predetermined criteria, says Ravi Akella, who headed the product management team for Autodesk’s generative manufacturing solutions before moving last year to become director of product development at Roblox. The feature focuses on helping designers define the problem they’re trying to solve, he says.
“The software asks the user preliminary questions. ‘What sorts of materials would you consider for your design? Where does it connect with other things as part of an assembly? What are the loads? What are the pieces of geometry?’” Akella says.
After a short period of time, the software then presents designers and engineers with an array of design options that best meet their requirements. Designers choose the best design. Or, if none of the options meet their needs, they can begin the generative process again, this time offering slightly different inputs.
Like other big-name CAD makers, SolidWorks also includes topology optimization capabilities within its CAD software.
“We expect the computing platform to anticipate your design goals,” says Bassi.
But, Vlahinos adds, the AI in those systems is used for simulation rather than for design. It’s by continual simulation that the designs are found. The tools allow the engineer to skip all the step-by-step analyzing. The human is still involved in the process and must validate the simulation the CAD system returns.
“The generative process could get you plus or minus 15 percent of the real answer but with 2 percent of the effort,” he says. “So, you know how to make the heat exchanger this way or that way – you’ve isolated the design alternatives and you can find them right away and validate.”
Even though the tools function through simulation, “You get amazing design insights and design innovation so you can see how something can be done,” he says.
But Vlahinos cautions against relying fully on the current AI-enabled CAD optimization tools.
“They are not simulation replacement,” he says. “Don’t let the vendors oversell them. They are design guidance, like a spell checker for you design concept. But they do give you more amazing results.”
Still, by helping engineers more quickly meet their prescribed design specifications, AI also frees up time engineers can spend focusing on other aspects of the piece—like its inherent shape or its artistic merits.
And – with proper validation in place — these tools can help ensure parts will meet manufacturing specifications and allow for much quicker design than traditional methods. It also can create unorthodox, sometimes never-before-seen-shapes that can be manufactured through 3D printing.
The engineer as artist
Design for manufacturability, as it’s called, is of course an important—some might say bedrock—necessity. AI techniques have a role to play in other aspects of design. And some of its uses may not have been conceived of yet, as CAD makers focus on these first AI implementations, Vlahinos says.
“Right now, it’s ‘Please tell me what the optimal shape is to achieve my engineering goal,” Vlahinos says. “We could see AI answering other questions in the future.”
Though his career has focused on rapid product development — Vlahinos recognizes that AI could help engineers design products faster than before — at the same time it offers engineering company customers new and unique ways to meet needs they may not even know they have.
For instance, broaden the view beyond the focus on manufacturability and AI can also lend artistic value to an engineered piece or product, he says.
“We’ve never properly valued the artistry of the design. But we could,” he says.
A product’s design artistry is, of course, subjective, so putting a monetary value to that number — as opposed to function — has always been elusive. Likewise, the capability to add never-before-seen geometries that create swirls and whorls in new and unexpected ways to pieces can bring a great deal of satisfaction, or headaches, for designers, depending on their liking to bring creativity to their engineering work.
If CAD can evolve, in the not-too-distant future, everyday objects like your blender, electric toothbrush or even the engine within your automobile, will take the shape of nothing you’ve ever seen before, said Hod Lipson, a mechanical engineering professor Columbia University and director of the school’s Creative Machines Lab. He is a roboticist who works in the areas of AI and digital manufacturing.
Most 3D printers take their printing instructions from 3D CAD files. Because the 3D printer receives its instructions from CAD files, the printers are limited in the shapes that those CAD systems generate, Lipson says. CAD software only allows for designers to work with recognized geometries: circles and ovals, squares and rectangles, and so on, he says.
That’s changing as topology optimization and generative design capabilities make their way into design tools, Vlahinos adds.
So the day of the twisted blender may be upon us sooner than we think.
Feature and character recognition, which have been part of AI for many years, are part of the SolidWorks system. In fact, they’re so standard that many users may not recognize the AI component of those features—until, for instance, they begin to type a misspelled word they use frequently and see that word corrected automatically, Bassi says.
And AI has a role in CAM as well. For instance, SolidWorks CAM automatically generates a part’s manufacturing toolpath after design. CAM software uses the CAD models to generate the toolpaths that drive computer numerically controlled manufacturing machines. Engineers and designers who use CAM can evaluate designs earlier in the design process to ensure that they can be manufactured, Bassi says.
“The toolpath captures design strategies and recognizes features and types of materials, so you can have a CAM solution that’s almost completely automated,” Bassi says. AI drives the way the toolpath is automatically created.
“You can create a toolpath in a couple of clicks. You don’t need a lot of details for intelligent manufacturing,” Bassi said.
One thing is certain, Vlahinos says. AI will never take the human engineer or designer out of the equation.
Even intelligent machines need guidance. That means engineers will always be vital to the design process, he adds. A human will always be needed to view shapes and designs in the same way other humans will. To translate a part’s use, — its form, and its function — with an eye toward other human users.
Filed Under: 3D CAD World, 3D printing • additive manufacturing • stereolithography, AI • machine learning