AI to reduce rejects and reworking

Currently, additive production with recycled raw materials often generates waste or high post-processing costs. A new Fraunhofer IAPT project is laying the foundations for the profitable use of recycled thermoplastics in additive manufacturing and the expansion of production environments into industrial printer farms. Intelligent in-process control (IPC), digital twins and other applications of artificial intelligence (AI) are intended to stabilize 3D printing with recycled raw materials and ensure cost-effectiveness.

Recycled polymers offer cost savings and more sustainable production. However, the recycled materials differ from batch to batch, thereby increasing two challenges of additive production. Until now, most 3D printers have followed a static G-code. After starting a job, the machine executes a defined toolpath and a series of fixed parameters without taking into account what is happening in the printing process. At the same time, the flow behavior, moisture content and purity of the recycled raw materials vary - and therefore also the print result.

In addition, there are differences due to the individual wear and tear of the machines. According to Fraunhofer IAPT, nozzle wear, contamination and other machine-specific properties lead to unstable extrusion and dimensional deviations. The lack of consideration for the variance of recycled materials and the differences between individual machines increases reject rates. When scaling the production environment, the negative effect of static processes on efficiency is further exacerbated.

Digital twins make 3D printers capable of learning

In the new AI project on sustainability and profitability in industrial 3D printing, the experts at Fraunhofer IAPT are combining expertise in virtualization, digital twins and industrial AI. The aim is to change from an open loop without taking the printing process into account to a closed loop.

In this closed loop, observations on material quality or wear-related deviations in the behavior of the machines flow into the printing process in real time. Closed-loop printing uses the recorded data and adapts the process parameters within a print layer to the quality of the recycled material.

On the way to "closed-loop printing", researchers at the Fraunhofer IAPT are equipping 3D printers with sensors and computer vision. The systems monitor the print in real time and record layer height, bead width, vibration and extrusion behavior, among other things. AI algorithms analyze the data during production and adjust parameters such as extrusion rate, speed, temperature or laser power. In closed-loop printing, the 3D printer processes analysis results during the ongoing production process and continuously compensates for deviations - such as worn nozzles, environmental influences or material fluctuations in recycled raw materials.

Another project goal is to develop the 3D printers into learning systems. Digital twins of machines or machine parts are to identify the optimum parameter combinations for certain geometries, different material qualities and machine conditions during operation. Intelligent data management links the process data, geometry information, slicing parameters and quality metrics collected in the digital twin.

Unlike isolated data in separate files, such as .stl, G-code or logs, intelligent bookkeeping transforms every print - whether successful or not - into training data. The learning systems thus go beyond the direct reaction to deviations in "closed-loop printing", building up knowledge and using it for future construction projects.

Recycled materials as a building block for sustainable AM production

With a view to long-term profitability and seamless scaling, the Fraunhofer IAPT team is designing the project's control strategies and data framework directly for use in large printer farms. Edge devices on each machine take over local monitoring and control, while a central platform aggregates the data from all systems. This allows, for example, the knowledge gained by a 3D printer about a specific recycling material to be transferred to numerous other machines. Continuous optimization should therefore not only be possible for individual 3D printers, but also at fleet level.

According to the Fraunhofer IAPT, the concept of the current AI project provides companies with entry points for the use of recycled materials and more sustainable production, regardless of the size of their production environment. The combination of real-time control, virtual process understanding and structured data is intended to establish recycled materials as an integral part of industrial additive manufacturing across the entire AM chain.

Dr. Matthias Brück comments on the starting point of the project: "Today, recycling in additive manufacturing does not fail due to material availability, but due to process uncertainty. With adaptive, data-based control, we are transforming previous uncertainties into controllable variables. Sustainable 3D printing becomes plannable, certifiable and economically viable."