Bearing Fault Diagnosis Model Based on ALA-MDLS

A combined MTF-ALA-DarkNet19-LSTM-SA hybrid model is suggested to increase the accuracy of diagnosing rolling bearing faults. The self-attention mechanism (SA) was added to improve the main features, and an LSTM network was incorporated into the output layer to capture the time dependence in order to overcome the limitation of DarkNet19 feature expression. Markov Transform Field (MTF) was used to convert the vibration signal into a time-series correlation image, and the artificial lemming algorithm (ALA) was used with MTF to optimize the network parameters. The suggested model successfully incorporates spatiotemporal data, according to experiments, and achieves 97.5 % classification accuracy. This offers a new method for the intelligent diagnosis of industrial bearings.