This is where the AlloySort project comes in, building on the findings of its predecessor, the MetallClass project. While MetallClass focused on the classification of metal alloys, AlloySort takes it a step further: The ongoing project aims to integrate non-destructive, real-time analysis of heterogeneous metal recycling materials directly into industrial production facilities. This will enable targeted sorting directly on the conveyor belt at industrial companies in the future.
From 20 hours to 1 second: A powerful combination of PGNAA and AI
The process is currently being tested in a demonstrator: It is located in the pilot plant of project partner AiNT GmbH and is being supervised there by the researchers. The facility is equipped with a conveyor belt on which various test samples are transported. These include material streams or batches that are sometimes complex and are analyzed using prompt gamma neutron activation analysis (PGNAA). The resulting measurement data is high-resolution but highly noisy. Until now, the analysis took around 20 hours.
Significantly more efficient results are now being achieved through the use of modern AI algorithms such as neural networks and convolutional neural networks: They reliably determine material compositions within a second.
In early December 2025, the research team met at AiNT GmbH in Stolberg to discuss the project’s status and coordinate next steps. inIT presented new results on the classification and regression of PGNAA data. Prof. Dr. Lange-Hegermann draws a positive interim conclusion: “The results so far show that PGNAA, in combination with data-driven methods, enables reliable analysis of alloy compositions in a very short time. AlloySort makes a promising contribution to more efficient and sustainable industrial recycling.”
The coming months will focus on further developing the demonstration plant, expanding the data sets, and continuing the validation of the mathematical models. Additionally, the goal is not only to identify the alloy type but also to determine the mixing ratios.
The project is funded by:
European Union, co-financed by the Ministry for the Environment, Nature Conservation, and Transport of North Rhine-Westphalia. A measure of the Environmental Economy Strategy.
Funding lines: ERDF/JTF Program NRW 2021-2027: GreenEconomy.IN.NRW
Funding code: EFRE-20800105