Diagnóstico de falhas em máquinas elétricas rotativas utilizando técnicas de ensemble learning

Abstract

This article analyzes the problem of classifying faults in rotating electrical machines. For this, a comparison between Random Forest (RF) and Gradient Boosting (GB) ensemble learning techniques was performed in order to analyze the performance of these algorithms. The database used is composed of eight mechanical variables related to motor operation under failure and non-failure conditions. The feature extraction technique employed was Root Mean Square (RMS). Therefore, the simulations performed of the algorithms resulted in high success rates, with Gradient Boosting obtaining the best performance, with an accuracy higher than 99%, which reinforces the great applicability of these algorithms in problems of this nature.