An industrial analytics methodology and fog computing cyber-physical system for Industry 4.0 embedded machine learning applications

Thumbnail Image
thesis_hardbound.pdf(5.63 MB)
Full Text E-thesis
O'Donovan, Peter
Journal Title
Journal ISSN
Volume Title
University College Cork
Published Version
Research Projects
Organizational Units
Journal Issue
Industrial cyber-physical systems are the primary enabling technology for Industry 4.0, which combine legacy industrial and control engineering, with emerging technology paradigms (e.g. big data, internet-of-things, artificial intelligence, and machine learning), to derive self-aware and self-configuring factories capable of delivering major production innovations. However, the technologies and architectures needed to connect and extend physical factory operations to the cyber world have not been fully resolved. Although cloud computing and service-oriented architectures demonstrate strong adoption, such implementations are commonly produced using information technology perspectives, which can overlook engineering, control and Industry 4.0 design concerns relating to real-time performance, reliability or resilience. Hence, this research compares the latency and reliability performance of cyber-physical interfaces implemented using traditional cloud computing (i.e. centralised), and emerging fog computing (i.e. decentralised) paradigms, to deliver real-time embedded machine learning engineering applications for Industry 4.0. The findings highlight that despite the cloud’s highly scalable processing capacity, the fog’s decentralised, localised and autonomous topology may provide greater consistency, reliability, privacy and security for Industry 4.0 engineering applications, with the difference in observed maximum latency ranging from 67.7% to 99.4%. In addition, communication failures rates highlighted differences in both consistency and reliability, with the fog interface successfully responding to 900,000 communication requests (i.e. 0% failure rate), and the cloud interface recording failure rates of 0.11%, 1.42%, and 6.6% under varying levels of stress.
Industry 4.0 , Smart manufacturing , Machine learning , Cyber-physical systems
O'Donovan, P. 2018. An industrial analytics methodology and fog computing cyber-physical system for Industry 4.0 embedded machine learning applications. PhD Thesis, University College Cork.
Link to publisher’s version