Enhancing the manufacturing processes and productivity in Industry 5.0 with the integration of machine learning and cyber-physical systems

Abstract

This paper aims to examine the synergistic integration of Cyber-Physical Systems (CPS) and Machine Learning (ML) as a foundational enabler for Industry 5.0, focusing on creating human-centric, sustainable and resilient manufacturing ecosystems. The study utilizes a synthesis and review approach, analyzing recent advances in ML-driven CPS applications such as workflow optimization and predictive maintenance, alongside enabling technologies like 5G and digital twins. The findings show that the integration of ML in multi-tier Edge-Fog-Cloud CPS brings significant operational advantages for the shop floor, such as sub-millisecond real-time control, more efficient human–robot interaction and a 10–20% reduction in energy used by the shop floor. But, the benefits of such physical manufacturing solutions rely on the need to work through a number of operational challenges. These include formally verifying non-deterministic ML policies, reducing IT/OT cybersecurity threats, including data poisoning and signal spoofing, reducing data heterogeneity in Federated Learning (FL) and lowering high deployment costs for small and medium-sized enterprises (SMEs). Future implementation requires addressing the need for Explainable AI (XAI) for transparency, FL for privacy and reinforcement learning for human-in-the-loop control. The proposed paradigm is applicable in smart factories, autonomous production lines and supply chain optimization, helping manufacturers maximize asset output and reduce environmental impact. This paper highlights the essential shift from “technology-driven” (Industry 4.0) to “value-driven” (Industry 5.0) manufacturing, identifying the CPS-ML convergence as the critical engine for this transition.