If you’ve ever tried printing an AI-generated 3D model, you know that the experience can come with some obstacles: the 3D model may have an undesirable “AI bumps” (small, irregular imperfections or protrusions), details that are not feasible for your 3D printer to achieve, and, critically, mechanical stability oversights.
There has been lots of buzz around creating AI-generated models that are viable for 3D printing. For instance, Backflip, an AI model generator from the creators of Markforged, is promising AI 3D model generation built for 3D printing. More recently, MeshyAI announced the launch of Creative Lab, a platform for generating 3D printable models. Now, researchers from the Massachusetts Institute of Technology (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL), along with researchers from Google, Stability AI, and Northeastern University, have developed the “MechStyle” system. The purpose of this system is to make real-world objects with AI, creating functional items, while also exhibiting the user’s intended appearance and texture.
Parts adapted with the MechStyle system (Image credit: Faruqi et al.)
How Does MechStyle Work?
It works a little differently from other AI 3D model generators. Instead of beginning with a text, image, or video prompt, the MechStyle process begins with a 3D model, either one that users upload themselves or a preset asset of objects like vases and hooks. Then, they can prompt the tool with text or images to create a personalized version. The geometry will therefore be modified by a generative AI model, while MechStyle simulates how those changes will impact particular parts. In this way, MechStyle ensures vulnerable areas remain structurally sound. Essentially, it creates an AI-enhanced blueprint that users can 3D print and use in the real world.
In an article about MechStyle published by MIT, they gave the example of 3D printing a wall hook. A user could upload a 3D model of a wall hook, state the material they’ll be printing with, and prompt the system to make a custom version, with directions like, “generate a cactus-like hook.” The AI model will work in tandem with the simulation module and generate a 3D model resembling a cactus while also having the structural properties of a hook. So, it’s about two components working together: the stylization process (which works based on its understanding of the text prompt) and the feedback received from the simulation module.
“We want to use AI to create models that you can actually fabricate and use in the real world,” MIT Department of Electrical Engineering and Computer Science (EECS) PhD student and CSAIL engineer Faraz Faruqi said. “So MechStyle actually simulates how GenAI-based changes will impact a structure. Our system allows you to personalize the tactile experience for your item, incorporating your personal style into it while ensuring the object can sustain everyday use.”
Strength through Physics: The Finite Element Analysis
A vital component to the MechStyle project is the group’s use of finite element analysis (FEA). This is a physics simulation method creates a sort of heat map indicating which regions are structurally viable under a realistic amount of weight, and which ones aren’t. Then, as AI refines the model, the simulation identifies the parts of the model that are weakening and prevents further changes.
However, running the FEA simulations every time dramatically slows down the AI process, so MechStyle is designed to know when and where to do additional structural analyses. “MechStyle’s adaptive scheduling strategy keeps track of what changes are happening in specific points in the model,” Faruqi added. “When the genAI system makes tweaks that endanger certain regions of the model, our approach simulates the physics of the design again. MechStyle will make subsequent modifications to make sure the model doesn’t break after fabrication.”
The MechStyle iterative workflow (Credit: Faruqi et al.)
Through FEA and adaptive scheduling, MechStyle could generate objects that were as high as 100 percent structurally viable. The team tested 30 models with styles resembling bricks, stones and cacti, and discovered that the most efficient way to create structurally viable objects was to dynamically identify weak regions and tweak the generative AI process to mitigate its effect. The researchers realized they could stop stylization completely when a particular stress threshold was met, or gradually make refinements to prevent at-risk areas from reaching that point.
MechStyle: Where Can It Be Improved?
The CSAIL researchers explained that the MechStyle system can make sure that a user’s model remains structurally sound, however, it cannot yet improve 3D models that were not viable initially. If someone uploads a model that’s not viable to MechStyle, they’ll receive an error message. In the future, the team wants MechStyle to be able to improve the durability of those faulty models.
Remember how MechStyle also starts with a 3D model uploaded by a user, or a selected preset? That’s also something the researchers hope to improve. They want the platform to use generative AI to create the 3D models, instead of relying on pre-made designs. To learn more about MechStyle and see five example applications, find the paper published about the project here.
What do you think of MechStyle? Let us know in a comment below or on our LinkedIn or Facebook pages! Don’t forget to sign up for our free weekly Newsletter here, the latest 3D printing news straight to your inbox! You can also find all our videos on our YouTube channel.
*Cover Image Credit: MIT News