基于K近邻算法的数据融合与改进图卷积神经网络的电机轴承故障诊断

打开文本图片集

DOI：10.15938/j. emc.2025.05.002

中图分类号：TM307；U226.8 文献标志码：A 文章编号：1007-449X（2025）05-0012-07

Fault diagnosis of motor bearing based on data fusion and improved graph convolutional network

SUN Liling， TANG Liyu， XU Boqiang （School of Electrical and Electric Engineering，North China Electric Power University，Baoding O710O3，China）

Abstract：To address the issues of low diagnostic accuracy when using single-type data for motor bearing fault diagnosis and the over-smoothing phenomenon in graph convolutional networks，a motor bearing fault diagnosis method based on multi-data fusion and an improved graph convolutional network was proposed. Initially，the vibration signal of the motor bearing and the current signal of the motor were transformed into the frequency domain using the fast Fourier transform. Subsequently，each frequency was treated as a node，with the corresponding vibration and current signals serving as node features. Based on the K-nearest neighbor graph construction method， the vibration and current signals were fused into graph-structured data.This graph data was then fed into an improved graph convolutional network，enhanced by the addition of an initial residual connection module，for training to yield diagnostic results. Comparative experiments for motor bearing fault diagnosis were conducted on the Paderborn dataset using the proposed method and various other models. The experimental results demonstrate that the proposed model achieves a fault identification accuracy of 98.6% ，outperforming the comparison methods，thereby validating effectiveness of the proposed data fusion approach and the improved graph convolutional network.

Keywords： deep learning； fault diagnosis； graph convolutional network ；motor bearing； fast Fouriertransform； data fusion； current data

0引言

滚动轴承是电机的重要组成部件，其健康状况直接影响电机的运行安全与性能。（剩余11660字）