Constructing Industrial Knowledge Graph Through Ontology and Link Prediction

Given that the numerous data embedded in manufacturing processes and products are separated, it is challenging to tackle and integrate heterogeneous data in industrial scenarios. In this context, an industrial knowledge graph (iKG) has been developed as a promising semantic organisation to leverage the rich information from multiple resources. However, relations are usually missing and hidden in original iKGs, which results in the necessity for iKG completion. Given these two perspectives, a framework of iKG construction is proposed based on ontology and link prediction in this study. Firstly, an ontology design framework is deployed to generate domain-centric ontologies after extracting numerous data (e.g., entities and relations). Secondly, the missing relations between each couple of entities are discovered over existing knowledge to increase the number of edges that complete and refine iKGs. Thirdly, iKG visualisation is conducted by importing data into the generated ontology. The feasibility and effectiveness of the proposed framework are substantiated and demonstrated in a case study using real-world data.