Reliability assessment and prediction of rolling bearings based on hybrid noise reduction and BOA-MKRVM

To figure out the problem of reliability assessment and prediction due to the noise in rolling bearing vibration signals, bearing reliability assessment and prediction model was proposed combine intrinsic time-scale decomposition-adaptive maximum correlation kurtosis deconvolution (ITD-AMCKD) and Bayesian optimization algorithm mixed kernel relevance vector machine(BOA-MKRVM). First, the ITD-AMCKD hybrid model is come up with to decrease noise and extract valid information of bearing vibration signal; secondly, set up the reliability assessment model of bearing through logistic regression model; in the end, use BOA-MKRVM model to predict bearing degraded state, then substitute consequence into established reliability assessment model to acquire the prediction consequence of the bearing’s reliability. The test data from Xi’an Jiaotong University demonstrate availability of model in the paper.