Towards a semantic edge processing of sensor data in a smart factory

Tracking #: 2184-3397

This paper is currently under review
Paula Zalhan
Gheorghe Cosmin Silaghi
Robert Andrei Buchmann

Responsible editor: 
Guest Editors SemWeb of Things for Industry 4.0 - 2019

Submission type: 
Full Paper
The global world is experiencing a digital revolution with Industry 4.0 and Internet of Things (IoT) which enable enter-prises to optimize their production processes using intelligent, interconnected equipment and sensing devices. By harness-ing the data generated from sensor networks, smart factories can remotely monitor their assets, and can make decentral-ized decisions. However, handling massive amounts of data that are continuously generated with a fast rate within the industrial environment is a demanding task. Semantic technologies have shown great promise in achieving machine-interpretability of IoT data. This paper proposes a semantic sensor stream processing pipeline that employs Apache Kafka to annotate in a scalable way the sensor data using the Semantic Sensor Network ontology, then store the annotated output in a RDF triplestore for reasoning. We are investigating whether it is preferable to do this on the consumer or on the pro-ducer side, considering the publish-subscribe model of Apache Kafka. The Design Science approach was followed, moti-vated by a Smart Factory application scenario that involves sensors for heartbeat, proximity and location. In this setup, we are investigating where it is preferable to execute the semantic annotation in a distributed, Kafka-based configuration. The experimental evaluations show that the "semantic edge" approach fulfills the low-latency processing requirement.
Full PDF Version: 
Under Review