The transient concept of bearings: a novel strategy for RUL prediction

Bearings serve as integral components in mechanical devices, providing stability during mechanical transmission and reducing friction coefficients. Hence, the precise prediction of bearing remaining useful life (RUL) is paramount for the health monitoring of mechanical systems. However, traditional techniques which utilize linear degradation processes for constructing health index models often fail to adequately portray the complex relationship between degradation and time. To rectify this, we introduce The Transient Concept of Bearings and determine the degradation rate predicated on this novel concept. We construct a degradation rate model for bearings using a K-means-transformer network and leverage transfer learning methodologies to predict the RUL of bearings. Validation of the proposed concepts and demonstration of their accuracy are achieved using the PHM2012 challenge dataset, even amidst incomplete data scenarios. When compared to existing RUL prediction models, our approach not only significantly improves prediction accuracy but also sheds valuable insights into the bearing degradation process.