Fault diagnosis of motor rolling bearings based on dual-channel information fusion and VCC-FuseNet
HE Yuling
ZHENG Hai
DAI Derui
ZHI Jiahe
WANG Nana
WANG Xiaolong
Abstract：[Objective]To address the insufficient collaborative utilization of multi-source signals and shallow cross-modal feature fusion in motor bearing fault diagnosis,a novel model named VCC-FuseNet based on dual-channel cross-modal attention fusion was proposed.[Methods]Firstly,vibration signals and three-phase current signals were reconstructed into Gramian angular difference fields(GADF)and fusion-type recurrence plots(RP)for unified two-dimensional visual representation.Secondly,deep fault features were extracted using dual-channel ConvNeXt backbones,and a Transformer-based cross-modal attention fusion module was incorporated to enable deep feature interaction and fusion.Finally,the method was validated on a permanent magnet synchronous generator(PMSG)test platform and a variable-speed bearing dataset.[Results]The results demonstrate that VCC-FuseNet achieves recognition accuracies of 98.75%on the PMSG platform and 96.74%on the variable-speed dataset,representing improvements of up to 8.39%and 7.28%over single-modal methods,respectively.VCC-FuseNet effectively facilitates joint diagnosis using vibration and current signals with high accuracy and robustness.
Keywords：Rolling bearingFault diagnosisMulti-source information fusionTransformerDeep learning
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：12( 105-116 )