FCWT-DDIM-SwinT Identification Method of Rolling Bearing Failures Under Sample Uneven Conditions
SUN Xianghai
QIU Ming
LI Junxing
ZHANG Songlin
LIU Zhiwei
LIU Jingtao
GAO Rui
Abstract：To address the problem of low accuracy caused by sample imbalance in the fault identification of rolling bearings,a fault identification method combining Denoising Diffusion Implicit Model(DDIM)and Swin Transformer(SwinT)is proposed.Firstly,the collected raw vibration signals of rolling bearings were processed using a fast continuous wavelet transform(FCWT)to reconstruct them into two-dimensional time-frequency images.Then,using DDIM,the original imbalanced dataset was augmented to construct a balanced dataset with evenly distributed fault sample categories.Finally,the balanced dataset was applied to the training process of the SwinT model,thereby enabling accurate diagnosis of multiple fault types in rolling bearings.The engineering examples show that the use of DDIM can effectively solve the problem of unbalanced fault samples;at the same time,comparing with other identification models,the SwinT model has a superior bearing fault identification capability.The research results provide a new way for the improvement of rolling bearing fault identification technology under sample imbalance conditions.
Keywords：rolling bearingsfast continuous wavelet transformdenoising diffusion implicit modelswin Transformerfault identification
Publication Date：2026-02-25
Online Publishing Date：2026-03-19(First online date of this platform, not the publication date of the document)
Pages：10( 53-62 )