Subscription Freedom: Automatic Industrial Data Subscription Based on Recommendation System

In smart factory, the number of sensors and industrial applications is increasing exponentially. The data transfer relationship between sensor data and applications has also become intricate. Message queuing telemetry transport (MQTT) has become the de-facto standard in the IoT field due to its decoupling of data transfer in time and space. However, even though existing MQTT message servers have been able to reach 100 million connections, the way data is subscribed still relies entirely on manual work. Based on a recommender system, this work proposes an innovative automated industrial data subscription approach to improve the level of automation to better cope with the large-scale dynamic scenarios of smart factory data transfer. This work deploys MQTT broker to the cloud for larger datasets and computing power, and combines OPC UA semantic model to make MQTT publish/subscribe more standardized and automatic. The architecture and algorithm for auto-subscription are proposed, and the corresponding datasets are constructed and obtained on the MQTT broker. The similarity between different subscribers are mined through natural language processing (NLP) model and user-based collaborative filtering. Then the usefulness of the topics is calculated to generate the auto-subscription list. The simulation proves that this work can realize automatic industrial data subscription effectively.