Knowledge Graph Supported Machine Parameterization for the Injection Moulding Industry

Abstract

Plastic injection moulding requires careful management of machine parameters to achieve consistently high product quality. To avoid quality issues and minimize productivity losses, initial setup as well as continuous adjustment of these parameters during production are critical. Stakeholders involved in the parameterization rely on experience, extensive documentation in guidelines and Failure Mode and Effects Analysis (FMEA) documents, as well as a wealth of sensor data to inform their decisions. This disparate, heterogeneous, and largely unstructured collection of information sources is difficult to manage across systems and stakeholders, and results in tedious processes. This limits the potential for knowledge transfer, reuse, and automated learning. To address this challenge, we introduce a knowledge graph that supports injection technicians in complex setup and adjustment tasks. We motivate and validate our approach with a machine parameter recommendation use case provided by a leading supplier in the automotive industry. To support this use case, we created ontologies for the representation of parameter adjustment protocols and FMEAs, and developed extraction components using these ontologies to populate the knowledge graph from documents. The artifacts created are part of a process-aware information system that will be deployed within a European project at multiple use case partners. Our ontologies are available at https://short.wu.ac.at/FMEA-AP, and the software at https://short.wu.ac.at/KGSWC2022.

This research has received funding from the Teaming.AI project, which is part of the European Union’s Horizon 2020 research and innovation program under grant agreement No. 957402.