Edge based machine learning cuts latency and bandwidth for industrial sensors
Seamless Deployment of Machine Learning Inference at the Far Edge: An Empirical AWS IoT Greengrass Architecture for Industrial Predictive Maintenance
Distributed, Parallel, and Cluster ComputingArtificial Intelligence
Summary
Industrial machines produce lots of sensor data that needs quick analysis to predict problems. Sending all this data to the cloud is slow and uses a lot of internet bandwidth. The authors built and tested a system that analyzes data right where it’s collected, on the edge, sending only important information to the cloud. This edge system made decisions much faster and used much less bandwidth, while keeping accuracy high for detecting anomalies. They also checked how the system handles losing cloud connection but left a deeper study for the future.
What this means in practice
- •For industrial maintenance teams: Enable factories to run predictive maintenance with local AI that reduces data sent to the cloud and gives faster alerts of machine issues.
- •For smart building operators: Implement AI-driven monitoring at building sensors to reduce cloud dependency, speeding up fault detection while saving network costs.
Authors
Prince Kwabena Appiah Boadu, Jeffery Baafi Kyei, Enoch Kofi Koranteng, Griffth Selorm Klogo
Abstract
Industrial IoT sensors produce continuous, high-volume data streams that predictive maintenance needs to analyse in real time. Cloud-centric architectures send all raw telemetry to central servers, which costs latency, bandwidth and resilience. This paper presents a hybrid cloud-edge architecture for manufacturing predictive maintenance built on AWS IoT Core and AWS IoT Greengrass v2, with local machine learning inference at the edge and selective cloud synchronization. We deploy two configurations on real AWS infrastructure and compare them: a cloud-only baseline (Configuration A) and the proposed hybrid edge architecture (Configuration B). We measure end-to-end inference latency, bandwidth consumption and inference accuracy. Configuration B cuts cloud-bound bandwidth by 86.8% (63.6 KB to 7.2 KB per 200-reading window) and decision latency by 256x (23.1 ms cloud round trip to 0.09 ms local inference). At the edge, with no cloud involvement, it keeps an anomaly-class F1 score of 0.84 and 91% recall. The evaluation also instruments system uptime and Stream Manager queuing under cloud disconnection; the quantitative disconnection experiment is left as planned work. Our review of the literature points to four gaps that this work addresses: the lack of empirically validated Greengrass v2 deployments for predictive maintenance, unquantified bandwidth savings from selective edge synchronization, unmeasured resilience under cloud disconnection, and the open MLOps feedback loop at the edge.