Automated Process Knowledge Graph Construction from BPMN Models

Abstract

Enterprise knowledge graphs are increasingly adopted in industrial settings to integrate heterogeneous systems and data landscapes. Manufacturing systems can benefit from knowledge graphs as they contribute towards implementing visions of interconnected, decentralized and flexible smart manufacturing systems. Process knowledge is a key perspective which has so far attracted limited attention in this context, despite its usefulness for capturing the context in which data are generated. Such knowledge is commonly expressed in diagrammatic languages and the resulting models can not readily be used in knowledge graph construction. We propose BPMN2KG to address this problem. BPMN2KG is a transformation tool from BPMN2.0 process models into knowledge graphs. Thereby BPMN2KG creates a frame for process-centric data integration and analysis with this transformation. We motivate and evaluate our transformation tool with a real-world industrial use case focused on quality management in plastic injection molding for the automotive sector. We use BPMN2KG for process-centric integration of dispersed production systems data that results in an integrated knowledge graph that can be queried using SPARQL, a standardized graph-pattern based query language. By means of several example queries, we illustrate how this knowledge graph benefits data contextualization and integrated analysis. In a broader context, we contribute towards the vision of a process-centric enterprise Knowledge Graph (KG). BPMN2KG is available at https://short.wu.ac.at/BPMN2KG, and the sample queries and results at https://short.wu.ac.at/DEXA2022.