Real-Time Detection of Bearing Faults in Induction Motor Using Edge AI and Piezo Microphone

Induction motors (IM) are widely utilized in industrial settings because to their inexpensive cost, low maintenance requirements, simple design, high efficiency, and dependability. Faults in these motors can cause unscheduled production downtime, resulting in substantial operational interruptions. Effective maintenance planning, which includes periodic, preventative, and predictive measures, can considerably reduce such downtime. Bearing failures account for around $\mathbf{3 0 - 4 0 \%}$ of induction motor defects. In this study, an edge AI-based system is proposed for detecting bearing faults in induction motors. The device uses motor sound recorded by a piezo contact microphone to identify the operation of both healthy and faulty bearing. The Mel Frequency Cepstral Coefficients (MFCC) were extracted from the data using a sliding window. The classification performance of four supervised Artificial Intelligence (AI) methods (k-Nearest Neighbor, Support Vector Machine, Random Forest, and Multilayer Deep Fully Connected Neural Network) were evaluated on our dataset. All of them accurately determined both the induction motor’s bearing problems and healthy states. The proposed system diagnosis bearing fault in real time. In addition, it records the failure events in a $\log$ file for later analysis and shows each failure along with a timestamp.