Smarter Motors, Fewer Failures: Edge AI for Integrated Control and Predictive Health

Session details:

This paper introduces a novel edge AI architecture for predictive motor maintenance, combining real-time diagnostics and control on a unified embedded platform. This novel system performs high-resolution signal acquisition and efficient machine learning inference directly at the edge—eliminating the need for cloud connectivity. The solution employs a hybrid pipeline of FFT, autoencoders for dimensionality reduction, and one-class SVM (Support Vector Machines) to detect early-stage motor anomalies, such as bearing faults, achieving 99% precision and 85% recall.
Designed for harsh industrial environments, the architecture supports sealed, fanless enclosures and operates within tight power and thermal constraints. It reduces the bill of materials while enhancing diagnostic accuracy, making it ideal for scalable deployment in Industry 4.0 and 5.0 settings. The session will walk through the complete edge AI lifecycle—from data acquisition and feature extraction to model deployment and inference—highlighting key trade-offs in accuracy, compute efficiency, and deployment readiness.
Attendees will gain actionable insights into how embedded AI can transform motor-driven systems by reducing unplanned downtime, extending asset life, and enabling intelligent, autonomous maintenance strategies in modern factories.