Intelligent super-fast camera prevents errors in 3D printing

The Vision-in-the-Loop AI camera installed on a 3D research printer from Flanders Make. The image on the right shows what the camera is analyzing.

Research centers imec and Flanders Make present a first proof-of-concept of a super-fast AI camera that automatically adjusts the 3D printing of metal parts. This is important for quality assurance of parts in fields such as aerospace, wind turbines or medical implants.

The technology behind it was developed within Vision-in-the-Loop (VIL), an imec.icon innovation project in collaboration with the Flemish companies Materialise, AdditiveLab, Dekimo and ESMA and research partners from Flanders Make (DecisionS & KU Leuven MaPS) and imec (UGent IPI & UA Vision Lab).

🎥   Watch IPI researcher Brian Booth and other project partners explain the innovations in this video report from Kanaal-Z (in Dutch).

📰   You can read the imec press release in Dutch. Below is the English version.

📰   Other press coverage:


[Translated from the Dutch language imec press release]

Intelligent super-fast camera prevents 3D printing errors

Artificial intelligence (AI) can adjust printing process every millisecond

Leuven, September 30, 2022 - Research centers imec and Flanders Make present a first proof-of-concept of a super-fast AI camera that automatically adjusts the 3D printing of metal parts. This is important for quality assurance of parts in fields such as aerospace, wind turbines or medical implants. The technology behind it was developed within Vision-in-the-Loop (VIL), an imec.icon innovation project in collaboration with Flanders Make and with the Flemish companies Materialise, AdditiveLab, Dekimo and ESMA.

To create customized metal parts in aviation, for wind turbines or for medical implants, laser-based 3D printers are often used. In this process, a laser melts away a metal powder layer by layer in a specific pattern. But any minute deviation from the melting temperature can cause defects during printing. Unlike mass production, quality control of each individual product is crucial, because each 3D print is essentially a new production process in itself. The annoying thing is that this control now has to be done after the fact, thus wasting valuable time, energy and resources every time an object has to be printed all over again.

Thanks to the VIL project, this checking can soon be done in-line and corrections can be made for impending errors during the 3D printing process. As soon as a superfast and intelligent camera detects that the laser is generating too much or too little heat at a certain location, the system immediately adjusts the laser's settings to prevent defects. The quality of each printed object is thus guaranteed, resulting in enormous time and cost savings. In addition, because defective objects do not have to be completely reprinted, no waste is left behind and energy can be saved.

The project partners optimized the AI camera that underlies the system so that it can move extremely fast with the laser, with no loss of quality in the captured images. Part of this innovation was the software that provides fast communication between laser and camera so they know and can track each other's position. The researchers also developed an AI learning model that, in the same span of 1.2 milliseconds, links images from the printing process to archival knowledge of what the printed object and any defects should look like. This archival knowledge consists of the analysis of CT scans of previously produced objects. To train the learning model, the researchers improved the way the CT scans are taken and succeeded in filtering out noise created by reflections from a CT scan on metallic objects. These results build on strategic basic research conducted within the Flemish AI research program.

The new method allows the AI camera to adjust the 3D printing process on a well-defined printer for one type of metal. The results so pleased the partners in the project that a follow-up project was immediately drawn up. This new imec.icon project, called MultipLICITY and started in September, will further market the proven technology and also extend it for multiple materials (plastics and metals) and for a printer with a multiple laser.

Brian Booth, researcher at IPI, an imec research lab at UGent: "The fastest way to bring innovations in image processing and artificial intelligence to the Flemish industry is to work directly with companies that want to increase their innovative power. That is exactly what we do within imec.icon projects. Our goal is to translate our scientific research results into concrete applications in industry, and these innovation projects are incredibly valuable for that."

"Companies are increasingly using 3D printing because of the unique advantages the technology offers. But when scaling up production to large volumes, new challenges arise. The ability to better monitor printing creates a more consistent and sustainable production process that gives companies the confidence to deploy the technology for volume production of essential industrial products," said Tom Craeghs, research manager at Materialise.

About imec.icon VIL

Imec.icon Vision-in-the-loop (VIL) focuses on high-speed image processing for real-time control of 3D printers. VIL is an imec.icon research project funded by imec and the Flemish Agency for Innovation and Entrepreneurship (Vlaio). Icon stands for Interdisciplinary Cooperative Research and consists of demand-driven R&D projects with industry. The industrial partners in the VIL project are AdditiveLab, Dekimo Products, ESMA NV, and Materialise. Participating as research partners were: Flanders Make (DecisionS & KU Leuven MaPS) and imec (UGent IPI & UA Vision Lab)

3D printing, also known as additive manufacturing, has been on the rise for more than a decade. Because it allows objects to be built layer by layer, it gives enormous design freedom. Moreover, it can also provide cost savings for products that are produced in small quantities and for which it is not cost-effective to make expensive molds. In industries such as aerospace or wind turbines and medical implants, for example, this is the case. What is important with 3D printing - and certainly in these sectors - is to keep the printing process well under control to guarantee the quality of the objects made. Thanks to the imec.icon VIL project, this will soon be possible in a much more efficient way.


Avatar
IPI
Image Processing and Interpretation

Related