Building a Knowledge Graph from Deviation Documentation for Problem-Solving on the Shop Floor

Abstract

The description of deviations on the shop floor includes information about the deviation itself, possible causes and countermeasures. This information about current and already processed deviations and problems is a valuable source for future activities in the context of problem-solving and deviation management. However, extracting information from unstructured textual data is challenging. Furthermore, the relationships among the heterogeneous data are hard to represent. This paper proposes a framework to extract the knowledge contained in the deviation documentation and store it in a knowledge graph as triples. The proposed knowledge graph can then be used for the decision support system in production and will support more application scenarios in shop floor management in the future.