Machine Learning Applied to Anomaly Detection in Electric Motors Aimed at Industry 4.0

SKRAUBA, AXEL ALFREDO; DOMINGUEZ, JUANSE SANTIAGO; KOLOWSKI, FACUNDO NICOLÁS; KOLODZIEJ, JAVIER ERNESTO; BAVASTRI, CARLOS ALBERTO; RICARDO ANDRÉS KORPYS

Oberá

Workshop; XX Reunión de Trabajo en Procesamiento de la Información y Control; 2023

Facultad de Ingeniería UNAM

In this work, we propose an anomaly detector forelectric motors based on an Autoencoder implemented on a lowcost, computationally efficient platform. The model was trainedon a dataset of 1,225 normal vibration signals, augmented withGaussian noise. Through topological optimization using geneticalgorithms, a compact Autoencoder with 10,336 parameterswas obtained, achieving average sensitivity and specificity above80 % for anomaly detection. The model can operate in realtime (reaching 1,990 µs per inference), classifying normal andanomalous vibration signals captured by an accelerometer based on microelectromechanical systems technology. The resultsdemonstrate the feasibility of employing machine learning forcost-effective, in-situ predictive maintenance in electric motors.As future work, the implementation and validation of a completeonline system are proposed, along with the development of adaptive methods to establish dynamic thresholds between normal andanomalous states.