Knowledge Graph Integration to Improve Reliability and Quality of Maintenance Data in the Oil and Gas Processing Industry

Abstract

Data quality and reliability are fundamental challenges in asset maintenance management in the oil and gas processing industry, where equipment data is scattered across dozens of heterogeneous operational tables, including SAP notifications, work orders, pipe inspections, corrosion monitoring results, and metering data. This data fragmentation makes it difficult to trace relationships between entities, identify anomalies, and perform consistent cross-validation. This study proposes a Knowledge Graph (KG) integration approach as a data representation layer that connects equipment entities as central nodes with child nodes from monitoring tables through typed relationships. Implementation is carried out through the development of a Graph Manager component based on Flask and D3.js that synchronizes data from a PostgreSQL relational database to a Neo4j graph database on a scheduled basis. Evaluation is carried out on three aspects of data reliability, namely referential consistency, attribute completeness, and timeliness of data updates, before and after the implementation of the KG. Test results in a production environment show significant improvements in equipment tag inconsistency detection, decreased equipment history search time, and increased semantic search accuracy when the KG is combined with embedding vector (pgvector)-based search. This study concludes that the graph approach provides a real contribution to data governance and operational reliability of maintenance systems.