Abstract
The demand for transfer learning methods for mechanical fault diagnosis has considerably progressed in recent years. However, the existing methods always depend on the maximum mean discrepancy (MMD) in measuring the domain discrepancy. But MMD can not guarantee the different domain features to be similar enough. Inspired by generative adversarial networks (GAN) and domain adversarial training of neural networks (DANN), this study presents a novel deep adaptive adversarial network (DAAN). The DAAN comprises a condition recognition module and domain adversarial learning module. The condition recognition module is constructed with a generator to extract features and classify the health condition of machinery automatically. The domain adversarial learning module is achieved with a discriminator based on Wasserstein distance to learn domain-invariant features. Then spectral normalization (SN) is employed to accelerate convergence. The effectiveness of DAAN is demonstrated through three transfer fault diagnosis experiments, and the results show that the DAAN can converge to zero after approximately 15 training epochs, and all the average testing accuracies in each case can achieve over 92%. It is expected that the proposed DAAN can effectively learn domain-invariant features to bridge the discrepancy between the data from different working conditions.
Highlights
Bearings and gears are widely used transmission parts in rotating machinery, and their failure directly affects the healthy operation of machinery and even causes serious incidents. erefore, monitoring and diagnosing the health condition of these transmission parts is crucial [1, 2]
The training accuracy is approached 100% after approximately 15 training epochs, and the testing accuracy needs approximately 47 training epochs to achieve this goal. e classifier loss curve of deep adaptive adversarial network (DAAN) is plotted in Figure 4, and the training loss in DAAN converges to zero after approximately 15 training epochs
Taking the case A⟶B as an example, the domain-invariant features learned by the DAAN are displayed in Figure 6(f ), and the mapping results obtained using the other comparison methods are shown in Figures 6(a)-6(e). e source and target domains are represented in terms of S and T, respectively. e result in Figure 6(a) demonstrates that the stacked autoencoders (SAE) model obtains good cluster results, the distribution discrepancies of the two domains are substantially large, except for the normal condition (NC) condition. us, it can not effectively classify the unlabeled target samples when the model is only trained using the source samples
Summary
Bearings and gears are widely used transmission parts in rotating machinery, and their failure directly affects the healthy operation of machinery and even causes serious incidents. erefore, monitoring and diagnosing the health condition of these transmission parts is crucial [1, 2]. Xu et al [7] used a deep convolutional neural network (CNN) to achieve a bearing fault diagnosis problem under different working conditions. An et al [8] adopted a recurrent neural network (RNN) to process variable size sequences of bearing fault samples and achieved satisfactory performance. Wang et al [10] presented a capsule neural network for bearing fault diagnosis and obtained a high classification accuracy. These methods have achieved excellent diagnosis performance, they require plenty of labeled data. Obtaining a considerable amount of labeled data is quite hard for some machines, and the probability distribution of the fault samples constantly changes due to variable speeds and loads
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
