Edge-Cloud Synergy in Industry 4.0: Optimizing Real-Time Latency for Smart Manufacturing Facilities M. Aminul Haque, Md. Rasel-Ul-Alam

Modern Industry 4.0 smart manufacturing environments heavily rely on Industrial Internet of Things (IIoT) sensors and machine learning analytics to drive predictive maintenance (PdM). However, conventional
cloud-centric architectures suffer from high network latency, bandwidth congestion, and single-point-of-failure risks, making them ill-suited for mission-critical, time-sensitive asset health monitoring. This
paper proposes a dynamic Edge-Cloud Synergistic Architecture (ECSA) designed to optimize end-to-end processing latency, balance computational workloads, and deliver real-time predictive maintenance
across smart factory floors. A multi-tiered architectural testbed integrating vibration, acoustic, thermal, and current IIoT sensors was evaluated across simulated high-speed production workloads. The system
uses lightweight on-device anomaly detection models at the edge for immediate inference, while offloading complex fleet-level retraining, deep temporal modeling, and historical trend analysis to cloud
infrastructure. Empirical evaluation demonstrates that the proposed hybrid edge-cloud framework reduces response latency by 78.5% (from 312 ms in cloud-only setups to 67 ms at the edge) and decreases
outbound network bandwidth usage by 64.2% via local data aggregation and model quantization. Furthermore, the multi-tiered inference strategy achieved a fault detection accuracy of 97.4%, reducing Mean
Time to Repair (MTTR) by 31.5% and increasing Overall Equipment Effectiveness (OEE) by 8.3%