Fault diagnosis of high-speed train rolling bearings based on dual time-frequency transformation
LIU Yuheng
SUI Jiangtao
ZHANG Xiaoning
CAO Jingchen
Abstract：[Objective]Aiming at the problems that high-speed train rolling bearing faults have strong randomness,severe noise interference and scarce fault data,which further lead to insufficient local feature extraction capability and low diagnosis accuracy of diagnostic models,a fault diagnosis method based on ResNet-18 network with dual time-frequency feature fusion was proposed.[Methods]With ResNet-18 network as the basic architecture,the original one-dimensional vibration signals were converted into two types of time-frequency diagrams by short time Fourier transform(STFT)and continuous wavelet transform(CWT)to achieve dual-channel feature extraction.Subsequently,the two extracted features were fused through a self-attention feature fusion layer.Tests were conducted using bearing data from Case Western Reserve University(CWRU,USA)and bench test data of high-speed train rolling bearings.[Results]The results show that the classification accuracy of the proposed model reaches 99.66%,demonstrating excellent diagnostic performance,which can provide reference for the engineering application of high-speed train rolling bearing fault diagnosis.
Keywords：High-speed train rolling bearingFault diagnosisTime-frequency diagramResNet-18 networkSelf-attention
Publication Date：2025-11-15
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：7( 141-147 )