AMEXCI’s previous Quarterly Review for Q2 was focused on the topic of Metal AM technologies, presenting past and recent developments, while putting new innovations into perspective. We continue our Quarterly Reviews by shedding some light on Additive Manufacturing printers and investigate if they have started to develop abilities to sense, think and learn.

By Akshatha Dyananda, Head of Innovation, AMEXCI.

The demands for process stability, quality assurance and repeatability increase as the industrial use of additive manufacturing matures. New solutions with the help AI for instance are developing in high speed. This has contributed to smarter 3D printers with somewhat human-like abilities.


Can AM printers sense?

The first questions to be lifted in this article is whether AM printers can sense. The answer to that is; yes, they can!
The use for sensors is common in 3D printing. Most of them are based on imaging techniques and the hardware are typically optical cameras, infrared (IR) cameras and X-ray imaging system. These sensors help to detect vapour plume, powder spatter ejection, melt pool characteristics, etc. A recent review paper, In-situ monitoring of sub-surface and internal defects in additive manufacturing mentions about some acoustic methods being used for defect detection. All these falls into the category of in-situ monitoring where the monitoring is happening while the build is in progress. The paper mentions another NDT (Non-destructive Testing) method called XCT. This method however is more used in ex-situ monitoring. What is worth mentioning is that there is a good corelation between the results from XCT and in-situ methods of defect detection.

Through the international project DREAM (Data dRiven process control in mEtal Additive Manufacturing), Germany’s Fraunhofer ILT and BCT GmbH, as well as Sweden’s AMEXCI and Interspectral, have joined forces to improve the process and quality control for Laser Beam Powder Bed Fusion or LPBF. Additive Manufacturing. The project not only will include automated defect detection in Additive Manufacturing, but also will work on synchronising the data from multiple sensor sources, combine them with information about the manufactured parts, and prepare them for further analysis. AMEXCI will apply its previously developed annotation and analysis platform to the pre-processed data to identify potential defects that may occur during the Additive Manufacturing process.

Using this data to adjust the build in real time is the next possibly logical step. For instance, ORNL intends to develop and test its own inspection techniques to identify new methodologies and approaches for quality assurance in Additive Manufacturing, using EOSTATE MeltPool Monitoring and EOSTATE Exposure OT (optical tomography).

Using Artificial Intelligence (AI) for Additive Manufacturing (AM) at AMEXCI.




Can AM printers think?

Yes, they can think! The data collected through sensors is collected and acted upon through real time action.
An €6.8 million EU funded, InShaPe project is working on just that- on flexible adaptation of the laser spot. The improved manufacturing process is based on a high-performance optical module with programmable intensity distribution and AI techniques to determine the optimal beam shape for the target object, determined for example by the material type and geometry. InShaPe also aims to develop an innovative process monitoring and control system for quality analysis that integrates multispectral imaging, i.e. simultaneous observation of light of different wavelengths in the area of Additive Manufacturing.

AddUp’s FormUp 350 3D printer comes with a suite of softwares;  AddUp Dashboard, Recoat Monitoring, and Meltpool Monitoring – committed to providing a fully closed-loop process. 

While the 3D printing machine manufactures have always led the efforts of build monitoring, machine independent, affordable and open architecture solutions are also available. Recently, Open Additive, LLC, Beavercreek, Ohio, USA, and Addiguru, LLC, Metairie, Louisiana, have announced an agreement to provide Addiguru’s Recoater, a Laser Beam Powder Bed Fusion (PBF-LB) analysis software, as a plugin to Open Additive’s AMSENSE® multi-sensor data collection and analysis platform. Addiguru’s software uses computer vision, artificial intelligence (AI), and machine learning (ML) methods to identify critical process errors and send real-time alerts.

Labelling process for annotating metal L-PBF data




Can AM printers learn?

Yes, they are learning! This means that the printers can adjust the build process without manual interference and make complex decisions through machine learning.
is a case from Massachussets Institute of Technology (MIT), Cambridge, USA, who have published the paper “Closed-Loop Control of Direct Ink Writing via Reinforcement Learning” discussing the use of machine learning (ML) to monitor and adjust the Additive Manufacturing process in real-time. To adopt machine learning, the researchers developed a system using two cameras pointed at the nozzle of the Additive Manufacturing machine. This system shines a light at material as it’s deposited and can calculate thickness based on the amount of visible light. The controller then processes these images and subsequently adjusted the feed rate and direction of the nozzle to counteract any errors.

Another example of applying machine learning is from Sigma Additive Solutions. Sigma Additive Solutions specialises in the development and commercialisation of real-time monitoring and analytics solutions for 3D metal and polymer advanced manufacturing technologies. Vendors such as TRUMPF, Essentium and Desktop Metal, are also collaborating on standards needed for the industry as a whole, with the aim of positively influencing technology directions. Sigma Labs has proven that machine learning can accurately predict where anomalies in 3D printed parts are likely to occur. Sigma found, though a series of experiments, that machine learning did predict anomalies in 3D additively printed parts better than any other single metric. They found that the models could be well trained to recognize their own anomaly types: lack of fusion, gas porosity, keyhole, and tungsten inclusions.

In conclusion, the projects and efforts mentioned above all aim, to ensure process stability, quality, repeatability, determine optimum build parameters, speed up the process for new material adoptions, increasing the ease of printing complex geometry. These techniques are most widely applied to the Metal Laser Powder Bed Fusion and are useful for research, development, and production scenarios.

For more information, please contact Akshatha Dyananda, Head of Innovation at AMEXCI:
akshatha.dyananda@amexci.com