Digitalizing physical processes holds great promise in various fields ranging from engineering to healthcare. High value assets such as manufacturing factories and oil platforms drive the development of Digital Twin (DT) concept to optimize the performance of such entities even further.
Knowledge graphs represent a natural mean of structuring highly heterogenous data that is common in domains where Digital Twins are considered. Complementary to this, the real-time aspect and analytics of Digital Twins data are equally important.
Many promising technologies exist that can facilitate the use of knowledge graphs for data representation as well as handling of real time aspects and data analytics in this context. Technology examples used in the implementation of a generic Digital Twins software framework in the context of the manufacturing domain (see UI screenshot below, depicting various aspects of a manufacturing line, such as machines and flow of product parts between machines) include FastAPI, Neo4j, Kafka, Influxdata that are interconnected using Python libraries. The main challenges in combining various such technologies for implementing Digital Twins are scalability and generalization to various domains in which Digital Twins need to be developed.
The thesis aims to build upon the existing prototype (see screenshot of the current UI above) and explore various techniques and tools for combining knowledge graphs representations of Digital Twins with real-time and analytics aspects, focusing on cognition, scalability and genericity of the solution. Examples of concrete tasks could include (but are not limited to):
- Identify and implement at least one feature for a Digital Twin for example time series analysis, optimizing resources in a manufacturing factory or prediction of machine failure.
- Identity and implement a generic solution for integrating domain expert knowledge into the Digital Twin to support cognitive capabilities, for example anomaly detection or process optimization in the production.
- Identify and implement at least one feature to improve the quality of the existing prototype such as unit testing, continuous integration and deployment or logging.
- Evaluate implemented feature according to previously defined benchmarks.
Required knowledge and skills: Good programming skills in Python or Java languages; Good knowledge in data analytics; Good knowledge in modelling languages such as UML; Great advantage with knowledge about knowledge graph and semantic web technologies.
References:
- SINDIT: https://github.com/SINTEF-9012/SINDIT
- Detail of the AAS: https://www.plattform-i40.de/IP/Redaktion/EN/Downloads/Publikation/Details_of_the_Asset_Administration_Shell_Part1_V3.html
- Abburu, S., Berre, A. J., Jacoby, M., Roman, D., Stojanovic, L., & Stojanovic, N. (2020, October). Cognitive Digital Twins for the Process Industry. In Proceedings of The Twelfth International Conference on Advanced Cognitive Technologies and Applications (COGNITIVE 2020), Nice, France (pp. 25-29).