9 Ways AI Could Affect 3D Printing

AI is upon us, but what might it affect in the 3D printing universe?

Many of us are now dabbling in the world of AI tools. Once it was discovered that large language models (LLMs) were able to perform astounding tasks with digital data, the floodgates have been opened with a cascade of startups and fascinating AI applications to perform all manner of tasks previously done by humans.

There are some uses of AI in 3D printing already, with perhaps the most notable being the ability to detect print failures via optical camera monitoring (e.g. “spaghetti detection”). However, as time passes there will be many more areas in 3D printing that might be optimized through the use of advanced AI tools.

But what could they be? I pondered this question and came up with the following list of possibilities. There is no doubt that at least some of these will already be under development in secret underground labs, and we could see them emerge in the near future.

Advanced Slicing

Preparing a job for 3D printing isn’t as simple as hitting the “slice” button. For many objects, particularly complex ones, there could be many considerations. For example, perhaps a different wall thickness might be required in one region, or extra support is required in one spot, or a hole requires extra reinforcement, or a less-stressed zone might withstand less dense infill.

Decisions about such slicing factors are done by expert operators that “know” what to do by examining the object in question. But what if that knowledge was embodied in an AI? Could a super-intelligent AI-powered slicer be developed that could automatically recognize differing regions of a FFF print job and instantly generate a series of modifiers to optimize the print?

If such a thing existed, I believe everyone would want to use it.

Resin Job Preparation

When preparing a resin 3D print job, the key issue is almost always the support structure. Without proper supports the print job will probably detach during printing and the job will fail, and possibly even damage the resin 3D printer.

You might think that the answer is to simply make a lot more supports and make them thick and strong. While that wastes material, it also makes a mess of the surface quality and could make post processing removal of the supports problematic. The result is that most resin 3D printer operators spend considerable time tweaking their support structures. Personally, I can spend up to an hour doing this on a complex 3D print job.

What if this could be done automatically and reliably with an AI that could understand the geometry, weights and other factors? That would be highly beneficial, and should be integrated into slicing systems.

Part Design

Need a part? Design it yourself. Oh, you don’t have CAD skills? Find it in a repository, and good luck. This is the issue that creates a barrier for many using 3D printing.

Recent research has been exploring the concept of “Text to 3D”, where a text prompt generates a 3D model automatically. The technology is very rough right now, but one could see this rapidly developing into something useful that would, for many people, replace the need for CAD tools and even trigger a whole new wave of creators.

Repository Shifts

Following on from the previous point, what happens to online printable 3D model repositories if Text to 3D tools become powerful? Would we even need repositories when we can just ask for anything we want?

My thought is that basic items might disappear from repositories. Once someone trains an AI on the McMaster-Carr catalog, for example, every basic part could be simply generated. On the other hand, true creatives would still have a place for unique “ideas” for 3D models. Who would have thought of a 3D model of, say, a zombie riding a sentient pickle with dragon wings? You’d go to a repository to find items you couldn’t imagine on your own.

Print Profiles

A big issue today is print profiles, where a machine is matched with a specific branded material. The resulting profile more or less guarantees good print results. But these are tediously developed by hand: there are zillions of printer-material pairs to create.

What if an AI was trained on a large number of known good profiles? Would it be possible for it to “recognize” an arbitrary machine-material pair and instantly generate a likely useful profile from scratch?

That would be extremely interesting, and something that might even be integrated into the machines themselves to dynamically adjust to the material loaded.

Material Development

There are more 3D printer materials than ever, but the number of them is vastly less than the known materials developed by chemical companies. Each company has a massive portfolio of chemical mixes that contains, somewhere, mixes that are suitable for printing.

How do they find them? Which materials might be suitable for printing applications? I think an AI system might be able to help here.

Material and Process Selection

A typical 3D print scenario goes like this: “How do I print this?” Then an expert must determine which 3D printing process to use, which materials, and so on. This happens even before figuring out the print parameters and slicing process.

It’s a process that I believe could be implemented in an AI tool. Examine an object and its intended usage, and then determine the correct style of printing and recommended materials.

Buying Guides

Another common scenario is this: “Which 3D printer should I buy?” I get this constantly, and I end up asking a bunch of questions about budget, skill level, frequency, part types, usage, etc., in order to get enough data to make an assessment and prepare a recommendation. This is a tedious mental process.

Now imagine an AI trained in a similar way, but somehow hooked up to a comprehensive database of all available current machines on the market. I think you could get very good product recommendations, and quickly. However, once something like that is set up, I expect manufacturers to game the system. But then, maybe, you just need more training to overcome that.

CAD Conversions

There are several tools that can examine a vast library of CAD designs to determine which items are 3D printable. This is very useful for companies that don’t want to hold warehouses of old parts and instead want to print them on demand.

But what about all the parts that did not make the cut? The parts that exist and cannot be 3D printed?

What if we had an AI that would examine the CAD design and intended function and re-designed the part in such a way that it could be 3D printed? This could enable companies to print nearly everything they’ve designed.

This could be so powerful that it could unleash unprecedented manufacturing demand for printing. Instead of finding five percent of the parts can be printed, 100% can be 3D printed. I expect a 3D printer manufacturer to be working on this concept simply because it would supercharge demand for their products and materials.

An AI -3D Future

And there you have it: nine ways 3D printing could be utterly changed with AI tools in the future. Some of these will undoubtedly come to pass, while others are more speculative.

Either way, we’ll find out soon.