SLOW ROTATING BEARING CONDITION ASSESSMENT BASED ON BAYESIAN GAUSSIAN MIXTURE REGRESSION

S. A. Aye, P. S. Heyns, C. J. H. Thiart

Abstract

This paper presents the condition monitoring of slowly rotating bearing using experimental data from acoustic emission signal. The condition monitoring methodology is based on a nonlinear parametric Bayesian technique, Gaussian Mixture Regression which is expected to accurately diagnose bearing damage under fluctuating load and speed conditions. The proposed model has the ability to model high dimensional or multi-modal data and retains the flexibility of nonparametric approach. Therefore, the Gaussian Mixture regression method is applied to the condition monitoring of slowly rotating bearing in this study. Results show that the GMR approach is an appropriate, powerful, cost effective and an easy-to-tune regression technique for monitoring and predicting slowly rotating bearing damage under fluctuating speed and loading conditions.

Keywords

condition monitoring, Gaussian mixture regression, slow rotating bearings, damage

Download

IMEKO-TC22-2014-015.pdf

DOI

-

IMEKO TC

TC22 - Vibration Measurement

Back to the proceedings