Application of Gaussian Process Regression for bearing degradation assessment

Life prediction of bearing is the urgent demand in engineering practice, and the effective bearing degradation assessment technique is beneficial to predictive maintenance. This paper presents an application of an important Bayesian machine learning method named Gaussian Process Regression (GPR) for bearing degradation assessment. The Gaussian Process (GP) model holds many advantages such as easy coding, prediction with probability interpretation and self-adaptive acquisition of hyper-parameters. In this study, the GPR model with different kinds of covariance functions is applied for assessment of bearing state of health (SOH). Two common covariance functions and a composite covariance function of GPR which is obtained by additive single standard covariance functions are discussed. The dynamic model is introduced to realize a better assessment by analyzing some important features. From the experimental results, it can be concluded that using GPR model for prognosis can achieve a high performance, and the composite covariance function can improve the prediction precision. In addition, compared with wavelet neural network (WNN), GPR model shows more excellent features. So the purposed model can be utilized in bearing degradation analysis, and meanwhile can serve as a reference for similar data-mining projects.