Remaining useful life prediction of rolling-element bearings based on cumulative transformation and EResNet-KSLSTM network
PAN Zuozhou
WU Yiding
HU Yicheng
PAN Xiyu
ZHAO Peng
JIANG Fei
Abstract：[Objective]Aiming at the problems of harsh working environment and weak early degradation trend of rolling bearings,which lead to great difficulty in fault feature extraction and low prediction accuracy,a reinforced diagnosis model based on cumulative transformation was proposed for the remaining useful life(RUL)prediction of rolling bearings.[Methods]Firstly,a feature enhancement method based on cumulative transformation was proposed,where the extracted features were converted into the corresponding cumulative transformation form to improve the sensitivity of the original features.Secondly,a new health index based on cumulative features was constructed,and the continuous trigger mechanism algorithm was used to divide the states of the health index to determine the initial fault occurrence point.Finally,the skip connection module of the ResNet was enhanced and additional calibration channels were added to improve the network's ability to focus on key degradation features,and then the stacked long short-term memory with Kolmogorov-Arnold network module(KSLSTM)network was used to obtain full spatiotemporal features for the accurate prediction of bearing RUL.[Results]The results show that the prediction accuracy of bearing RUL in the small-sample training environment is significantly improved by constructing cumulative transformation features and optimizing the network structure,and the effectiveness of the method is verified by simulations and tests.
Keywords：Rolling-element bearingCumulative featureEResNetKSLSTMRUL prediction
Publication Date：2026-05-31
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：13( 9-21 )