Edge AI for Predictive Maintenance and Real-Time Fault Diagnosis in Industrial Motors: A Review

Abstract

Industrial motors are essential in smart manufacturing, utilities, transportation, process automation and energy systems. A brief and unnoticeable motor fault can result in production delay, increased energy cost, damage ases to connected equipment or production line and higher maintenance cost. This paper presents a compact IEEE-style review of Edge-AI-enabled predictive maintenance and real-time fault diagnosis for industrial motors with emphasis on MATLAB implementation. The current review unifies several domains including fault mechanisms, current, voltage, vibration, acoustic and thermal sensing technologies, benchmark datasets, signal processing, machine-learning and deep-learning based models and edge-deployment constraints. The most practical diagnostic channels continue to be current-signature and vibration methods, while acoustic and thermal measurements provide the complementary information for early failure analysis of bearings, stators and loads. The review pointed out common flaws, including unclear dataset split, limited validation on unseen machines, an over-reliance on accuracy just by itself, and the poor reporting of latency, memory use and false alarms. A system is presented where acquisition, filtering, extraction of features, validation safe against leakage, inference on an edge device and generation of maintenance action are all treated as one from the end-to-end system used as part of the MATLAB Edge-AI workflow. Dependable deployment calls for interpretability of datasets, automated dispatch, information on cast drift, and SCADA system integration.