Early Failure Point Monitoring for Remaining Useful Life Estimation of Bearings Using Covariance Transfer Enabled T² Control Chart

T2 control chart, a prevalent statistical process control technique, has extensively employed in the realm of bearing early failure monitoring within the setting of prognostics and health management. However, insufficient historical data as well as the nonidentical distributions between training and testing date, caused by stochastic noise and fluctuating operational conditions, inevitably increase the risk of false alarms and hinder the robustness of bearing early failure monitoring. To address the challenges, a covariance transfer enabled T2 control chart (CTET2CC) is proposed for bearing early failure monitoring under varying operational conditions. First, feature extraction is conducted via the multitime scale phase space warping (MPSW) method. Thereafter, a Gaussian process is employed to derive functional relationships in the source and target domains, respectively, in order to construct the T2 profile control chart. Then, the domain-invariant statistics can be efficiently learned from a large number of source domain profile samples through singular value decomposition (SVD), which excels in capturing the essential features of profile samples, and providing a robust representation of the underlying structure. Furthermore, a covariance transfer through SVD strategy is designed by re-weighting the target statistics. Finally, the corresponding T2 control chart is reconstructed to boost the stability of the controlled phase for target task. Experimental results verify that the proposed method is expected to be an effective technique for bearing early failure monitoring.