Prognostics of slow speed bearings using a composite integrated Gaussian process regression model

Abstract

Prognostics of manufacturing systems enables improved maintenance scheduling and cost reduction through reduced downtime, improved allocation of maintenance resources and reduced consequential costs of breakdowns. Prognostics are necessary for predictive maintenance of bearings in manufacturing systems. The findings show that in general the composite integrated GPR models perform better than the simple mean simple covariance GPR models, irrespective of whether the training or test sets are dependent or independent. In this investigation the Affine Mean GPR (AMGPR) was found to be the most effective prognostic model for prognostics of slow speed bearings on both dependent and independent data samples.

Keywords:

• prognostics
• Gaussian process regressions
• slow speed bearings
• mean function
• covariance function