A new method for gearbox fault diagnosis based on the self-correcting auxiliary classifier generative adversarial networks (SCACGAN) is suggested in response to the limited diversity and low quality of fault samples produced by the auxiliary classifier generative adversarial networks (ACGAN) during the small-sample gearbox fault diagnosis process, which subsequently results in low diagnostic accuracy. Firstly, an independent classifier is introduced into the auxiliary classifier generative adversarial network to mitigate the adverse impact of discriminator output errors on the quality of generated samples, and to classify the health status of different gearbox samples. Secondly, the problem of low-quality generated samples during the training phase is addressed by using the least squares function to improve the model’s generation and classification skills. Lastly, a self-correcting convolutional neural network is integrated into the generator to enhance the capability of fault feature acquisition. Experimental results demonstrate that under small-sample conditions, the proposed approach is capable of generating higher-quality fault samples, thereby improving the accuracy of gearbox fault diagnosis.

Low diagnostic accuracy results from the wide set of directed acyclic graphs that must be searched when doing structure learning on dynamic Bayesian starting networks under multidimensional input. Conventional approaches find it challenging to find the best structure. In this paper, a method is proposed to combine the R-vine Copula model with a dynamic Bayesian network (DBN) for fault diagnosis. First, the network structure space is made smaller by using the structure prediction model to filter the retrieved features and identify nodes with high correlation. Then, the first-layer tree structure of the R-vine Copula model is used combined with the transfer entropy method to construct the initial network of dynamic Bayesian network, and the DBN of the initial network is built according to the Markov process in time series for fault diagnosis, which solves the problem that it is difficult to obtain the optimal structure in the network construction under multiple features. The gearbox data of Southeast University is used for verification, and the comparison results show that the method can better learn the DBN structure, and the fit between the data and the model is high, and good diagnostic results can be obtained in fault diagnosis.

in industrial production, due to the difference in the distribution of source domain data and target domain data and the small amount of labeled fault data, the accuracy of domain adaptation-based bearing fault diagnosis algorithms proposed in the past is generally not high. In view of this, the multi-domain adaptation neural network (MDANN) fault diagnosis method was proposed in this paper, which was used for rolling bearing fault diagnosis without labeled data. Firstly, the original vibration signal was processed by using wavelet packet transformation (WPT) to reduce signal redundancy and avoid the loss of key signal features. Secondly, the multi-kernel maximum mean discrepancy (MK-MMD) algorithm was used to calculate the difference of input eigenvalues, and the network parameters of MDANN were updated by backpropagation so that the network can extract domain invariant features. Finally, in order to ensure that unlabeled target domain data can participate in network training normally, the maximum probability label was used as a pseudo-label strategy of the real label to solve the problem that unlabeled target domain data cannot be trained and enhance the acquisition of reliable diagnosis knowledge of the model. Two publicly available datasets, CWRU and PU, were used for validation. The experimental results show that the proposed method has higher diagnosis accuracy compared with common domain adaptation methods, which further shows that the method can effectively learn the transferable features and fit the discrepancy in data distribution between the two datasets.