Converting Relational Databases to Manufacturing Knowledge Graph for Product Quality Tracing in Additive Manufacturing

Abstract

Product quality tracing (PQT) in additive manufacturing (AM) is a significant and complex challenge crucial for fully understanding the processes, tracing quality issues, and optimizing process technologies. Typically, tracing relies on production data from historical relational databases, but these data lack semantic information and integration with the service context of manufacturing processes. A manufacturing knowledge graph (MKG) offers a solution by adding manufacturing semantic information to production data. This paper constructed a PQT model and an MKG database for AM. The PQT model and the schema layer of the MKG originate from the swimlane flowchart. Subsequently, a relational database was mapped to instantiate the MKG, and the natural language model was used to complete it. This integration constructs a PQT model and an MKG database centered on manufacturing resources, events, data, and states. The tracing subgraphs extracted from the MKG database demonstrate that this method effectively traces manufacturing process issues and anomalies. The proposed method provides a viable database and method for PQT in AM, enabling comprehensive tracing of manufacturing processes, identifying quality anomalies, and providing potential cause references for AM experts.