TITLE:
Adaptive Dynamic K-NN Spatio-Temporal Graph Convolutional Network for Bearing Fault Diagnosis
AUTHORS:
Xiuwei Hu, Ting Fang, Wenhui Chang, Cunrui Pang, Tonghui Wang
KEYWORDS:
Rolling Bearing, Fault Diagnosis, Graph Convolutional Network, Adaptive Dynamic KNN, Spatio-Temporal Feature Fusion
JOURNAL NAME:
Journal of Computer and Communications,
Vol.14 No.6,
June
25,
2026
ABSTRACT: Rolling bearing fault diagnosis is an important means to ensure the reliable operation of mechanical equipment. Aiming at the problems of poor working condition adaptability, over-smoothing in deep networks, and insufficient mining of fault temporal evolution features existing when the traditional K-nearest neighbor graph convolutional network is applied to bearing fault diagnosis, this paper proposes an adaptive dynamic K-NN spatio-temporal graph convolutional network (AdaDKNN-STGCN). Experimental validations are implemented on the CWRU bearing dataset. The proposed model attains a test accuracy of 99.78% in ten-class classification. It exhibits excellent generalization capability and anti-interference performance under noisy and variable load environments, which provides a reliable solution for intelligent fault diagnosis of rotating machinery bearings.