What question did this study set out to answer?

This research aims to enhance the prediction of faults in induction motors using deep learning techniques.

June 4, 2026Open Access

Motor Fault Prediction Based on Graph Convolutional Networks with Signal Association Mining

Key Points

This research aims to enhance the prediction of faults in induction motors using deep learning techniques.
Proposed a graph convolutional network with multi-head attention for fault classification.
Used time series fault data to encode and process input node data.
Classified fault types using a recurrent neural network and nonlinear layer.
Achieved competitive performance on specific datasets with the proposed method.
Demonstrated enhanced fault classification accuracy compared to traditional methods.

Abstract

Vibration signals are often used to analyze and recognize the characters of faults that verify in the operation of induction motors. At present, the processing method of fault data is mainly through expert judgment and analysis, and the processing cost is high. In recent years, induction motor fault data processing models with deep learning method have been utilized in induction motor fault diagnosis tasks. In the study, a type of deep learning algorithm of graph convolutional neural network based on multi-head attention enhancement is proposed to determine the fault type of induction motor by using time series fault data. The proposed method firstly obtains the extended time sequence feature representation, encodes the input fault node data through utilizing graph convolutional network, and finally realizes the obtained fault type classification of induction motor through recurrent neural network and nonlinear layer. The corresponding experimental results shown that the proposed approach could achieve the competitive performance in certain dataset.

Demander à l'IA

Bookmark

View Full Paper