Hackernoon logoHow AI Could Save the 3D Printing Industry and the Future of Machines by@himanshu_ragtah

How AI Could Save the 3D Printing Industry and the Future of Machines

Author profile picture

@himanshu_ragtahHimanshu Ragtah

Ex - Tesla. Kairos Society Fellow. SpaceX OpenLoop. TEDx speaker.

3D printing is a billion-dollar market with a variety of use cases- from healthcare, replicas to architecture, airplane parts.

Example use cases by market share-

However, the unsolved billion-dollar problem with the 3D printing industry is currently shape distortions. This problem plagues all industries and a majority of print jobs that are done have to be discarded due to unwanted shape distortions.
3D printers are super unpredictable when it comes to what will actually get printed out. If you have ever printed something from a 3D printer (FDM, SLA, etc.) you have likely ended up discarding various print jobs before the print came out the way you wanted it to.
Examples:
Pile up enough print job errors and you are left with a box of them-
‘Box of Shame’ src: myminifactory

Reasons for Shape Distortions

Shape distortions can happen due to various reasons-
  • Material expansion during printing
  • Material shrinkage during printing
  • Input values in the printer software (step/mm, print speed, layer height, temperature)
  • Mechanical issues with printer motor/belt
  • The print bed is not level
  • The print bed is not heated
  • The order in which layers are being printed
Example:
How the temperature of the plate can distort your print
How different layers at different temperatures can lead to shape distortions
Material shrinkage leads to warping and distortions
I have personally faced these issues when printing various kinds of hardware - from fixtures, enclosures, connectors to parts for aerial and ground robots.
(Typical parts of a FDM printer)

Latest development

Researchers from USC are working on this problem of shrinkage and expansion and have demonstrated that they can increase the print accuracy by 50% using Convolutional Modeling.
The model has been trained on data from previous print jobs to have the same accuracy across different materials and applications.
For example, if you are printing using PLA material and using the MakerBot 3-D printer, you can use the model if there is previous data on print jobs using the same printer type and material.
MakerBot 3-D printer: With over a 100,000 such printers worldwide, it is one of the most popular 3D printer.
The software being used: PrintFixer (imgsrc: Nathan D)
The software that this intelligence will be packaged in is called PrintFixer. As you can see, it is checking for shape distortions (expansions (red), shrinkage(blue)) on the CAD model based on data from previous print jobs.
50% accuracy is certainly progress. Even higher accuracies can make a huge difference in the healthcare industry which relies heavily on 3D printers. 
Dental implants require high accuracy and preventing shape distortions during 3D printing will certainly be a game-changer.

Tags

The Noonification banner

Subscribe to get your daily round-up of top tech stories!