Abstract:
Objectives Aiming at the problems that the symptom parameters from monitoring signals in a single analysis domain fail to fully characterize the operating state of the monitored object, and the model parameters of the Extreme Learning Machine (ELM) network are difficult to achieve the optimization, a fault diagnosis method for ship motor bearings is proposed, based on multi-domain information fusion and an improved ELM.
Methods First, a multi-domain feature parameter set was constructed from the vibration signals of ship motor bearings in the time domain, frequency domain and time-frequency domain. This set served as the input to the fault diagnosis model. The sparrow search algorithm was then used to optimize the model parameters of the ELM network by determining the optimal weights and thresholds, thus enhancing the fault diagnosis accuracy of ELM model. Finally, the fault states of motor bearings were identified using experimental data from a self-built test bench and open-source experimental datasets.
Results Experimental data verification based on the marine motor test bench demonstrated that the fault diagnosis model using multi-domain feature parameter sets, achieved a recognition accuracy of 100% on both the training and test sets. Verification with open-source experimental data showed that the recognition accuracy on the test set for the improved ELM model was 90.5%, which is 12.7% higher than that of the original ELM model. Additionally, the recognition accuracies on both training and test sets were higher than those of other diagnostic models.
Conclusions This study has improved the input symptom parameter set and the diagnosis model. The proposed method can effectively identify the fault states of motor bearings and demonstrates good model stability, providing a valuable reference for fault diagnosis of ship motor bearings.