ASPECTS OF THE DEVELOPMENT OF FEDERATION LEARNING SYSTEMS IN A HUMAN-ORIENTED INDUSTRIAL ENTERPRISE
DOI:
https://doi.org/10.31891/2219-9365-2024-80-11Keywords:
Industry 4.0, Industry 5.0, human orientation, protection, modelsAbstract
In the transition from Industry 4.0 to Industry 5.0, the focus of modern industry shifts from people-oriented to people-centric. Federated learning (FL) is the most effective innovative concept of machine learning in the smart industry. This publication analyzes approaches to the development of human-centric FL systems implemented in industrial enterprises using controlled robotic platforms, the Internet of personel Internet of Things, and the Industrial Internet of Things. The schemes of locally-centralized and globally-centralized FL are analyzed. A step-by-step implementation of locally-centralized and globally-centralized FL has been developed for an industrial enterprise.
It was found that locally-centralized FL provides a high level of privacy protection, since data is processed and stored directly on the device. Using a centralized server reduces data processing delays. However, the generality of the model is limited, since the aggregated data may not take into account the operating conditions of the entire enterprise.
A wide range of conditions can be taken into account and the model can be enriched with the help of global-local-centralized FL. The model proved to be useful in the development of systems for industrial companies using the Internet of personel Internet of Things and the Industrial Internet of Things. Such an architecture increases the risk of loss of privacy due to centralized data exchange. The transfer of large volumes of updated models is critical in the case of resource-intensive processes.