Scalable data fabric for smart manufacturing: A connector-based approach to industrial data integration

The growing adoption of Industrial Internet of Things (IIoT) technologies enables new opportunities for data-driven manufacturing. However, large-scale deployment remains hindered by heterogeneous device protocols, inconsistent semantics, and the continued use of legacy systems. Existing approaches often lack a unified method for achieving technical, syntactic, and semantic interoperability across diverse industrial devices. We propose a Data Fabric architecture tailored for IIoT environments. It defines four core functionalities essential for interoperable data integration: uniform messaging, semantic integration, automated metadata contribution, and Connector service templates. These functionalities are implemented through reusable Connector services, IIoT gateways that combine protocol translation, ontology-based metadata generation, and automated contribution to a central knowledge graph. We validate the proposed approach on a distributed fleet of 15 devices covering three representative IIoT device types: OPC UA (Open Platform Communications Unified Architecture)-enabled controllers, including a real production machine, simulated industrial robots, and a physical industrial camera. The evaluation demonstrates real-time synchronization of system state with the knowledge graph, semantic asset discovery, and subscription-cycle latencies around 50 ms under isolated conditions, scaling near-linearly up to 50 concurrent clients. Our results confirm that the proposed architecture offers a practical and extensible realization of the Data Fabric paradigm, addressing core interoperability challenges in IIoT data management with soft real-time feasibility demonstrated at lab scale.