Dynamic Reweighted Domain Adaption for Cross-Domain Bearing Fault Diagnosis

Abstract

Certain progress has been made in fault diagnosis under cross-domain scenarios recently. Most researchers have paid almost all their attention to promoting domain adaptation in a common space. However, several challenges that will cause negative transfer have been ignored. In this paper, a reweighting method is proposed to overcome this difficulty from two aspects. First, extracted features differ greatly from one another in promoting positive transfer, and measuring the difference is important. Measured by conditional entropy, the weight of adversarial losses for those well aligned features are reduced. Second, the balance between domain adaptation and class discrimination greatly influences the transferring task. Here, a dynamic weight strategy is adopted to compute the balance factor. Consideration is made from the perspective of maximum mean discrepancy and multiclass linear discriminant analysis. The first item is supposed to measure the degree of the domain adaptation between source and the target domain, and the second is supposed to show the classification performance of the classifier on the learned features in the current training epoch. Finally, extensive experiments on several bearing fault diagnosis datasets are conducted. The performance shows that our model has an obvious advantage compared with other common transferring algorithms.

Adversarial Deep Transfer Learning in Fault Diagnosis: Progress, Challenges, and Future Prospects

“…However, these data still conform to the same machine health condition, and there are common basic characteristics among them. Therefore, it is feasible and promising to study the knowledge transfer of fault diagnosis among different sensors [ 210 , 211 , 212 ]. The fourth is different machines; Guo et al [ 213 ] combined adversarial discriminant methods and difference-based methods, using MMD to narrow the differences between features after acquiring domain-independent features in the shared space through adversarial training, which is one of the earliest jobs of different machines.…”

Markov Transition Field Enhanced Deep Domain Adaptation Network for Milling Tool Condition Monitoring

“…Traditional machine learning algorithms perform poorly when the training and test data come from different distributions. In this case, domain adaptation becomes useful [36]. A domain consists of a data space, x, and a probability distribution, P X , on its samples X ∈ x. Domain adaptation means to adapt useful knowledge from a source domain, S, to a target domain, T. Specifically, we are provided a source dataset (X S ,…”

Ubiquitous Domain Adaptation at the Edge for Vibration-Based Machine Status Monitoring

“…Zhang et al demonstrated the use of cross-domain adversarial learning for machine fault diagnosis using training and testing data collected under different operating conditions [16]. Interestingly, Meng et al [17] computed a balance factor between domain adaptation and class discrimination during bearing fault diagnosis in machines. They chose a dynamic weighted strategy to reduce instances of negative transfer during domain adaptation.…”

“…Interestingly, our proposed approach also addresses sample selection bias in intermittent error-prone data arising from sensor quality and noise. Lin et al [12] and Meng et al's [17] approaches are strong motivators for the use of cross-domain learning in our machine status monitoring work. However, using CNNs and transfer learning and the associated need for complex computing resources makes their approach somewhat complex for use with lowcost and often constrained edge processing systems.…”