Remaining useful life estimation using deep metric transfer learning for kernel regression
Accurate estimation of remaining useful life (RUL) is indispensable for the safe operation of
rotating machinery, reducing maintenance costs and unnecessary downtime. Numerous
data-driven models have been reported to predict the RUL of bearings using historical data.
However, it is still very challenging to predict the RUL of bearings under different operating
conditions. It is necessary to propose a model which can extract domain invariant deep
features and accurately predict the RUL of bearings under new operating condition. In this …
rotating machinery, reducing maintenance costs and unnecessary downtime. Numerous
data-driven models have been reported to predict the RUL of bearings using historical data.
However, it is still very challenging to predict the RUL of bearings under different operating
conditions. It is necessary to propose a model which can extract domain invariant deep
features and accurately predict the RUL of bearings under new operating condition. In this …
Abstract
Accurate estimation of remaining useful life (RUL) is indispensable for the safe operation of rotating machinery, reducing maintenance costs and unnecessary downtime. Numerous data-driven models have been reported to predict the RUL of bearings using historical data. However, it is still very challenging to predict the RUL of bearings under different operating conditions. It is necessary to propose a model which can extract domain invariant deep features and accurately predict the RUL of bearings under new operating condition. In this paper, a novel method called deep transfer metric learning for kernel regression (DTMLKR) is proposed and applied to the RUL prediction of bearings under multiple operating conditions. This method combines deep metric learning with transfer learning (TL) to solve regression problems. Case studies on the IEEE PHM Challenge 2012 dataset demonstrate the effectiveness of the proposed method. Compared with other state-of-the-art methods, the superiority of the proposed method is verified.
Elsevier