Categories
Pages
-

DBIS

The Implementation of Federated Learning for a Data Space in the Plastic Packaging Industry

March 7th, 2023

Thesis Type
  • Master
Student
Lei Wang
Status
Finished
Presentation room
Seminar room I5 6202
Supervisor(s)
Stefan Decker
Christoph Quix
Advisor(s)
Lina Molinas Comet
Contact
lina.teresa.molinas.comet@fit.fraunhofer.de

Motivation and Thematic Background

The KIOptiPack project has more than 30 partners from industry and research institutes. The main goal of this project is to optimise the production and recycling of plastic packaging using AI. An essential element in the project is the data space infrastructure which provides the necessary data (in a federated fashion) for AI solutions. The basis for developing the data space is the Fraunhofer-developed concepts for International Data Spaces (IDS) and Gaia-X, particularly the Eclipse Dataspace Connector (EDC).

The master thesis goal is to develop a demonstrator for data space for supply/production chains in the plastic packaging industry.

The development and evaluation of a demonstrator for this data space are based on at least one concrete use case. The possible use cases are the following: 1. Federated Machine Learning (ML), i.e., distributed ML, without merging the necessary data in a central database, but by distributed execution of ML operations in different instances of the EDC. 2. A solution for the connection of external backend systems to the EDC, using common protocols from the field of Industry 4.0 (e.g. OPC UA, MQTT), ensuring that data streams can be processed efficiently. 3. The integration and transformation of data sets in different formats, where the interoperability of the data formats should be achieved in particular through the semantic description of the data structures.

What we offer

  • Write a thesis on an exciting and highly relevant topic, both industrially and academically, using the latest tools and technologies.
  • Guaranteed comprehensive support during your work.
  • The inter-institutional character of your work will give you a broad insight into various research topics at the RWTH and at FIT

How to apply

When applying, please briefly describe your prior experience, include information regarding relevant courses you took in this direction, and your transcript of records. If you are interested in this thesis or if you have any questions, please feel free to contact me.

Lina Molinas Comet, M.Sc.
lina.teresa.molinas.comet@fit.fraunhofer.de


Prerequisites:

Your Profile

  • You are motivated to conduct scientific and industry-oriented research.
  • You bring an independent way of working and critical thinking skills.
  • Interest in modern software architectures and web technologies
  • Practical experience with at least two of the following technologies:
    • Good programming skills in Java
    • Experience in Web Service development
    • Good knowledge of data management systems (e.g., SQL/NoSQL databases, Apache Kafka, Apache Spark)
    • Experience with data streaming systems (e.g., Kafka) or protocols in the IoT (e.g., OPC UA, MQTT)
  • A plus: knowledge of Semantic Web technologies and ontologies, good programming skills in Python