In the long-term operation process, rolling bearings often have multiple single faults or cascade faults, which are coupled with each other to form the compound fault. The complex coupling components of compound fault lead to difficultly building a correlation between compound fault feature and fault class. Moreover, for the transfer learning based on domain adaptation, compound fault feature of source domain is pretty different from that of target domain, thus knowledge of source domain is hard to be transferred into the cross-domain unsupervised learning process of target domain. To solve above mentioned problems well, this paper proposes a domain reinforcement feature adaptation methodology with correlation alignment (CA-DRFA) to complete the cross-domain compound fault diagnosis of bearings. Specifically, a deep reinforcement learning model is improved by being combined with the domain adversarial training way, and meanwhile the correlation alignment metrics is adopted to enhance the generalization performance in feature alignment. In addition, two experiments and an engineering application are performed to demonstrate that CA-DRFA outperforms other popular cross-domain fault diagnosis methods in cross cutting condition and speed transfer tasks. Overall, CA-DRFA is able to implement cross-domain compound fault diagnosis with high accuracy, and effectively reduce the cost of labeling target samples.

With the rapid development of sensor systems brought about by the industrial Internet of Things process, the need to unsupervisedly detect fault signals under multiple working conditions simultaneously has led to the emergence of multitarget domain adaptation (MTDA). This advancement is confronted by two primary challenges on domain adaptation: 1) fault feature extraction prefers target domains akin to the source domain, often sidelining others and 2) the learned features’ lack of interpretability. To address these, this article proposes the generalized-Gaussian cyclic variational inference framework (GCVIF). This framework engages the generalized-Gaussian cyclostationary distribution to initially capture the non-Gaussian and nonstationary attributes of fault signals, with an extended likelihood ratio test proposed to estimate distribution parameters. Leveraging these estimated distributions as priors, the generalized-Gaussian cyclic variational autoencoder is then developed to infer domain-shared representations. The process is steered by a specialized domain-shared representation learning principle, focusing on compact representation in the encoder and cyclostationary structure reconstruction in the decoder. Remarkably, extensive fault detection trials affirm that leveraging distributions estimated unsupervised as priors enables unbiased feature extraction, and the inference of domain-shared representations is inherently aligned with fault cyclostationary pulses simulation on the signal time domain, ensuring their direct mechanistic explainability.

The generalization performance of intelligent fault recognition models is tied to the assumption of identical distribution. Domain adaptation allows the source model to be extended to single or multitarget domains with distribution shifts. However, the reliable transfer of multitarget domain adaptation (MTDA) is inseparable from domain annotation. In this article, we consider a more pragmatic but challenging MTDA setting where domain labels are absent. This setting threatens most existing methods due to the elusive gaps and agnostic affiliation. We propose a systematic approach to the new setting. First, the category semantic destruction and self-supervised clustering are used to estimate domain labels. Second, the attack features are constructed to consolidate adaptation by gradient alignment and classifier robustness. Extensive experiments demonstrate that the new setting is quite challenging for existing methods, while the proposed method outperforms the existing methods and effectively suppresses transfer preference.

Feature selection is a key issue to pattern recognition and condition monitoring. This paper presents an investigation that uses self-organizing maps network to realize feature selection for gearbox condition monitoring. In order to visualize the trained SOM results more clearly, a novel visualization technique is introduced, which can project the high-dimensional input vectors into a 2-dimensional space and prepare a good basis for further analysis. Then with the use of the responses of every dimensional feature in SOM network neurons weights to the input data evaluated according to the Euclidean distances between them, the feature sets being sensitive to pattern recognition are selected. Gearbox vibration signals measured under different operating conditions are analyzed with the method. The results demonstrate that the method selects sensitive feature sets effectively and has a good potential for gearbox condition monitoring in practice.