Cloud-Based Industrial Cyber–Physical System for Data-Driven Reasoning: A Review and Use Case on an Industry 4.0 Pilot Line

Alberto Villalonga,Gerardo Beruvides,Fernando Castao,Rodolfo E. Haber,Fernando Castano

DOI: https://doi.org/10.1109/tii.2020.2971057

IF: 12.3

2020-09-01

IEEE Transactions on Industrial Informatics

Abstract:Nowadays, reconfiguration and adaptation by means of optimal re-parameterization in Industrial Cyber–Physical Systems (ICPSs) is one of the bottlenecks for the digital transformation of the manufacturing industry. This article proposes a cloud-to-edge-based ICPS equipped with machine learning techniques. The proposed reasoning module includes a learning procedure based on two reinforcement learning techniques, running in parallel, for updating both the data-conditioning and processing strategy and the prediction model. The presented solution distributes computational resources and analytic engines in multiple layers and independent modules, increasing the smartness and the autonomy for monitoring and control the behavior at the shop floor level. The suitability of the proposed solution, evaluated in a pilot line, is endorsed by fast time response (i.e., 0.01 s at the edge level) and the appropriate setting of optimal operational parameters for guaranteeing the desired quality surface roughness during macro- and micro-milling operations.

automation & control systems,computer science, interdisciplinary applications,engineering, industrial