Use of Multiscale Permutation Entropy Feature Selection and Supervised Classifiers for Bearing Failures Diagnosis

Abstract

Entropy measurements are an accessible tool to perform irregularity and uncertainty measurements present in time series. In signal processing, the Multiscale Permutation Entropy (MPE) is recently presented as a methodology of characterization capable of measuring randomness and non-linear dynamics existing in non-stationary time series, such as mechanical vibration signals. In this article, the MPE is combined with diverse feature selection techniques and multiple classifiers based on machine learning aiming to detect different operative states in an internal combustion engine. The best combination is selected from the evaluation of parameters like precision and computational time. Finally, the proposed methodology is ratified through effectively performing a bearing fault diagnosis with a high precision rate and a reduced computational time.