Cross-equipment transfer fault diagnosis for gearbox bearings based on multi-source feature fusion

Objectives To address the problem of low cross-equipment diagnosis accuracy caused by single monitoring signals and insufficient fault data for various types of gearbox bearings on ships and naval vessels, a cross-equipment transfer fault diagnosis method for gearbox bearings based on multi-source feature fusion is proposed. Methods First, the short-time Fourier transform (STFT) is used to extract the time-frequency features of the preprocessed fault signals. Then, a multi-source feature fusion network (MSFFTN) is established. The multi-source feature fusion module of MSFFTN is adopted to extract and conduct fusion representation of the time-frequency features of bearing rotational speed and vibration signals. Domain invariance of the fused features between the source domain and the target domain is guaranteed through network weight sharing, thereby improving the transferability of the fused features. Furthermore, an improved joint distribution adaptation (IJDA) mechanism is proposed. The marginal distribution alignment and conditional distribution alignment of the fused features are strictly enforced to reduce the distribution discrepancy of fault features between the source domain and the target domain. Finally, cross-equipment transfer fault diagnosis of gearbox bearings is realized via a classifier, and the effectiveness of the proposed method is verified using the gearbox bearing dataset from Beijing Jiaotong University and the data collected by the self-built marine gearbox fault simulation setup. Results Experimental results show that the proposed method attains an average transfer fault diagnosis accuracy of 99.4% in cross-equipment scenarios, significantly outperforming other comparative methods. Conclusions The results of this study can provide valuable references for the cross-equipment fault diagnosis of gearbox bearings in ships and naval vessels monitored by multi-source sensors.