Improved temporal network-based method for motor bearing fault diagnosis
HUANG Xingyuan
YANG Tongguang
JIANG Lingli
LI Xuejun
Abstract：Fatigue wear of the bearings in high-power variable-frequency motors causes the vibration signals to be mixed.In order to improve the accuracy of fault diagnosis,a multi-channel insulated bearing time information fusion diagnosis model was designed,and the coarse-grained features with temporal patterns were extracted from the measured fault data.The self-attention mechanism was introduced into the designed time information fusion diagnosis model for optimization,and the correlation weights between the data were calculated by constantly updating the weight coefficients.The proposed diagnostic framework demonstrated superior overall fault recognition rate of 99.1%,significantly outperforming conventional recurrent neural network architectures including GRU(94.7%),LSTM(91.2%),and RNN(88.4%).
Keywords：motor insulated bearingsfault diagnosisimproved timing networkself-attention mechanism
Publication Date：2025-09-30
Online Publishing Date：2025-10-15(First online date of this platform, not the publication date of the document)
Pages：8( 1-8 )