Abstract

The study presented an AI-accelerated embedded system platform that supported predictive maintenance within industrial automation environments. The research addressed the increasing demand for intelligent monitoring systems that ensured reliable operation of industrial machinery and minimized unexpected downtime. The proposed framework employed a microwave transceiver front-end combined with a system-on-chip architecture that has integrated AI inference capability for real-time condition assessment of industrial assets. The method utilized a Convolutional Neural Network (CNN) based anomaly classification model that has processed sensor-derived microwave signal reflections to identify early-stage mechanical degradation. The system has captured high-frequency response patterns from industrial components and has translated them into diagnostic features through signal conditioning and feature extraction modules. The AI model has performed classification of normal and abnormal operational states with adaptive threshold mapping that has improved decision consistency under varying load conditions. The method processed microwave reflection signals captured from industrial components and extracted discriminative features for classification. The system achieved 94.5% accuracy, 94.0% precision, 94.2% recall, and 94.1% F1-score, while reducing inference latency to 62 ms.

Authors

S. Karthika, B. Sathananth
College of Engineering, Anna University, Chennai, India, V.S.B. College of Engineering Technical Campus, India

Keywords

Embedded System, Microwave Sensing, Predictive Maintenance, Industrial Automation, Convolutional Neural Network