ELAMU Successfully Completed: Three Years of Research on Digital Material Certificates

After a three-year project duration, the ELAMU research project was successfully completed at the end of September 2026. On September 17, the project partners met for a joint wrap-up meeting to reflect on the past three years, the results achieved, and their collaboration on the project.

Processing Material Certificates with AI

Material certificates contain important information about the properties and chemical composition of materials. However, in companies, these documents are often available in various, poorly standardized formats. ELAMU’s goal was therefore to increase the level of digitization in the processing of material certificates, reduce the likelihood of errors and the manual effort involved in processing them, and make the data they contain more readily available for downstream processes. NeurologIQ Engineering GmbH, Robert Josef Wolf GmbH & Co. KG, AVENTUM GmbH, and the University of Siegen collaborated on this project.

Based on various user studies, a prototype was developed during the project’s duration that automatically extracts data from digital material certificates and presents the values in a structured format. To achieve this, various AI-based methods are combined, including Optical Character Recognition (OCR), Intelligent Character Recognition (ICR), and, in particular, Large Language Models (LLMs).

However, since material certificates are safety-critical documents, it is not sufficient to simply extract the information automatically. Users must be able to verify and confirm whether the recognized values are correct. A key focus of our research was therefore on how end users want to interact with such an application and what control and validation mechanisms are needed to ensure that AI-supported processing can meet the high demands of industrial practice.

Combining Research and Teaching

As part of the project, we were also able to explore these questions in greater depth through several project and thesis papers in collaboration with students from the University of Siegen. For example, in his master’s thesis, one student developed a method for determining the uncertainty of recognized values and subsequently evaluated it together with participants from the project. Another thesis addressed the question of how AI users would like to provide feedback on recognized information. This was prompted by a recurring finding from our evaluations: AI support is considered useful in the long term, especially when the system can learn from incorrect recognitions and the corrections made by users.

Other studies examined the processing of material certificates within a broader work context. For example, a bachelor’s thesis investigated how a system can support the verification of delivered material against purchase orders and material certificates and how this changes existing work practices. Another project—followed by a bachelor’s thesis—focused on the adoption of new systems and the question of what design requirements arise for a context-aware annotation and communication system that supports employees in the implementation and use of such software.

With the end of the project term on September 30, ELAMU officially came to a close. We would like to thank all project partners, students, and participants in our studies for their collaboration and the many valuable insights provided over the past three years!