Autonomous Edge Intelligence for Smart Manufacturing Environments

Abstract

Industries with legacy manufacturing systems have recently become intelligent, connected, and highly automated production environments due to the rapid evolution of Industry 4.0 technologies from traditional paradigms [1]. Modern manufacturing centers produce massive quantities of data from both industrial Internet of Things (IIoT) devices and sensors, as well as robotics and programmable logic controllers, security systems, and cyber-physical systems. This data has been processed and analyzed in the cloud for most of its life; however, more demanding needs around real-time decision-making, ultra-low-latency requirements, as well as privacy preservation compliance and operational reliability are stressing the capabilities of centralized architectures. An emerging and promising paradigm known as autonomous Edge Intelligence (AEI) has also emerged to facilitate both edge computing and Artificial intelligence directly in manufacturing environments, allowing for localized analytics, i.e., A Comprehensive Framework for Autonomous Edge Intelligence in Smart Manufacturing Environments. A proposed framework that incorporates the interconnectedness of distributed edge nodes, machine learning models and industrial IoT infrastructures with cloud-assisted orchestration mechanisms to enable predictive maintenance; quality inspection; process optimization pertaining to material flow path generation and task execution by each node; energy management from embedded sensors providing pressure and temperature data coming off machines contributing towards reducing carbon footprint of operations as well managing final output delivery adopting technologies like autonomous production control. By moving intelligence closer to the manufacturing equipment, it greatly reduces communication latencies and bandwidth usage while improving operational resiliency & overall system scalability. This study examines previous contributions in edge-enabled manufacturing systems and recognizes key issues related to latency-sensitive industrial applications. This paper proposes a multilayer framework, which comprises a sensing layer, edge intelligence layer, autonomous decision layer, manufacturing execution deep learning module (DLMM), and cloud coordination. Edge devices are equipped with deep learning, reinforcement learning, and federated learning to support advanced machine learning on the edge device to support localized model training and inference. Additionally, the framework integrates dynamic resource allocation and cooperative edge-cloud cooperation mechanisms for improved computational efficiency. Experimental evaluation shows that using remote computing capacity in this way brings significant improvements in response time, production efficiency, fault detection accuracy, and network utilization while being more energy efficient than traditional architectures based on cloud-based manufacturing. The experiments expose a processing latency reduction of more than 70%, an equipment fault prediction accuracy over 95%, and network traffic around the edge reduced by approximately 60%. Additionally, our framework enables scalable deployment of the learning method across various manufacturing environments while preserving data privacy and operational reliability. The results prove that autonomous edge intelligence can transform self-optimizing manufacturing ecosystems to dynamically respond and adapt to real-time changes in their operating condition. It advances the progress towards future intelligent automation, distributed decision-making in next-generation smart factories that yield significant productivity and sustainable industrial operations. Advancements in edge AI hardware, industrial communication technologies, and autonomous orchestration mechanisms are expected to augment usage of autonomous edge intelligence over a composite portfolio of manufacturing sectors.