Adaptive Edge AI Orchestrator for Real Time Data Analysis and Process Control in Manufacturing Lines

The Industry 4.0 has exposed the manufacturing and transportation lines to large volumes of sensor and machine data, which needs to be processed real time. Nevertheless, the legacy manufacturing models continue to be plagued by the major issues of latency, bandwidth limitations, and the dangers of transferring sensitive manufacturing information to centralized clouds. These restrictions inhibit fast detection of anomalies, responsive actions to the process variations and resource optimization within the factory floor. As such, manufacturers suffer unjustified downtimes, energy wastages, and high cost of operation. To overcome these issues, this paper presents the Adaptive Edge-AI Orchestrator (AEO)-a distributed, edge-intelligent platform that is intended to be used in smart manufacturing to do real-time analytics and dynamic process control. Each edge node presents AEO as modular and containerized microservices with which low-latency inference is made possible, as well as workflow optimization with reinforcement learning, and the deployment of AI modules with no interruptions to production. Most prominently, AEO presents a peer-to-peer collaborative learning framework that is decentralized, in which edge nodes share learned knowledge instead of raw data, which guarantees privacy of data, generalization of models and continuous learning. AEO can be used to coordinate nearly in real-time through 5G-enabled mesh connectivity, thus decreasing the time response to anomalies, improving the product quality, and decreasing energy usage. It is also designed in a modular manner that enables it to integrate easily into the existing manufacturing infrastructures. In general, it can be stated that AEO is a paradigm shift in intelligent manufacturing-adaptive control, scalability, and real-time decision-making are being made available at the edge.

Ҳали таржима қилинмаган