Keywords:
Vibration, Anomaly, OCSVM, Electric Motor, Maintenance
Abstract
The importance of detecting failures and anomalies in electric motors reflects the improvement in the application of techniques used for maintenance on these machines. Thus, in the present work, a technique for detecting anomalies in the vibration of electric motors is proposed, based on the classification algorithm One-Class Support Vector Machine (OCSVM). An accelerometer device was used to capture the vibrations of an electric motor of 1.5 cv together with an electromagnetic brake system. In addition, a method for selecting the hyper-parameters of the algorithm was developed, obtaining a decision boundary that classified the operating data considered normal with a lower error rate than 5%
