LSTM-FCN Neural Network-based Fault Identification for Ship Electrified Direct-drive Propulsion Systems
Long-term operation of ship propulsion systems in harsh underwater environments makes them inevitably en-counter operational faults during navigation which consequently harm stability of the entire ship.In view of this the present study focused on a LSTM-FCN neural network-based fault identification method for ship electrified direct-drive propulsion systems.The major work entailed extracting frequency components from the stator current signals as fault characteristics,combining LSTM and FCN to construct a hybrid neural network model,and inputting the extracted characteristic compo-nents into the model to achieve fault type identification.The proposed method achieved in the experiment a high fault iden-tification accuracy of 98.43%,and thereby was verified feasible and reliable.
万方数据
维普
