Can AI Catch 3D Printing Errors?

3D printing errors


As much as we love 3D printing, we do have to admit that it still isn’t a 100% reliable process and there will inevitably be 3D printing errors. Things sometimes go wrong with even the best designed print job. Objects can slide off the print bed, sections of the print can collapse, layers can delaminate or the filament can break. Most of these failures are pretty obvious – but considering how long it takes to print the average job, how many of us sit there watching the whole process? Most people that have printed a few projects have come back to find that something has gone horribly wrong and the printer has merrily deposited half a spool of expensive filament on top of the mess.

Now a California tech company is running a kickstarter for a device that, it says, can detect problems and stop the printer automatically, to avoid wasting more filament on a job that’s already failed. The Autonomous Intelligent Management System, by startup Delcos Systems, should work with any printer that has a filament runout sensor and if Delcos raises the money to put it into production it should cost around $120.

Meet The AIMS

The AIMS is a small, simple-looking device. It’s powered by a USB cable, has a connector that mounts to the printer’s runout sensor, and its most visible feature is a wide-angle camera. All you have to do is connect it up, position it so the camera is observing the print bed, and start the job.

On the inside, the AIMS isn’t as simple as it looks. Its electronics are based around an ARM Cortex A53 processor, and Delcos say it uses artificial intelligence to spot errors in the print job. The camera monitors the position and movements of the print head, and analyses it using a neural network that’s pre-trained to tell the difference between a print job that’s going smoothly and one that’s run into problems.

According to Delcos the AIMS can recognise a range of 3D printing errors, including stringing of the molten filament, a head that’s skipping or sticking, or a project that’s slipped out of position. When it detects an error it sends a signal through the runout sensor, tricking the printer into thinking the filament has run out; this should stop the printer within 15 to 20 seconds. If the filament actually does run out the AIMS can detect that too, and send the same signal. A nice touch is that the neural network is all internal; the AIMS doesn’t need an internet connection, so it can work completely offline.

Will It Happen?

So far AIMS is a prototype, and Delcos are trying to crowdfund final development and setting up for production. A $94 pledge on Kickstarter will get you one, if they manage to get the device to market. Even if they don’t, though, this isn’t an idea that’s just going to fade away. Expect more error-detecting devices to appear, and the same technology to be built into printers themselves. Hopefully it won’t take long; we’re huge fans of anything that makes 3D printing more user-friendly.

Leave a comment