Cross-equipment fault diagnosis for marine gearbox bearings based on multi-source feature fusion and improved domain adaptation

To address the issue of low cross-equipment diagnosis accuracy resulting from reliance on a single monitoring signal and from insufficient fault data for gearbox bearings of various models in ships, a cross-equipment fault diagnosis method for marine gearbox bearings based on multi-source feature fusion and improved domain adaptation is proposed.

First, the short-time Fourier transform (STFT) is employed to extract the time-frequency features from the preprocessed fault signals. Subsequently, a multi-source feature fusion transfer network (MSFFTN) is developed. The multi-source feature fusion module of MSFFTN is used to extract and fuse the time-frequency representations of bearing rotational speed and vibration signals. Domain invariance of the fused features between the source and target domains is maintained through network weight sharing, thereby enhancing the transferability of the fused features. Furthermore, an improved joint distribution adaptation (IJDA) mechanism is proposed. Both the marginal and conditional distributions of the fused features are explicitly aligned to reduce the distribution discrepancy of the fault features between the source and target domains. Finally, cross-equipment transfer fault diagnosis of gearbox bearings is achieved using a classifier. The effectiveness of the proposed method is validated using the gearbox bearing dataset from Beijing Jiaotong University and the experimental data acquired on a purpose-built marine gearbox fault simulation test rig.

Experiments under multiple noise levels demonstrate that the proposed method attains an average cross-equipment fault diagnosis accuracy of 95.6%, significantly outperforming the compared methods.

The findings of this study offer valuable insights into cross-equipment fault diagnosis of gearbox bearings in ships monitored by multi-source sensors.