Condition monitoring of ball bearings using Bayesian neural networks

Resumo

Rotating machines play a fundamental role in several engineering applications. In most of these ap-
plications, they are subjected to unexpected overloads that can cause the premature failure of their components.

This work aims to present a condition monitoring framework that employs the Bayesian neural networks approach,

which includes uncertainties quantification associated with the damage detection in ball bearings. Images of vi-
bration signals of ball bearings were used to feed a Bayesian neural networks and the algorithm predictions were

given in terms of the probability density function of the possibility of the vibration image belongs to a specific type
and severity of damages. The results demonstrated that Bayesian neural networks (BNN) as a powerful technique
for damage diagnosis, and it can quantify uncertainties in condition monitoring of ball bearings.