In the era of big data, data‐driven methods mainly based on deep learning have been widely used in the field of intelligent fault diagnosis. Traditional neural networks tend to be more subjective when classifying fault time‐frequency graphs, such as pooling layer, and ignore the location relationship of features. The newly proposed neural network named capsules network takes into account the size and location of the image. Inspired by this, capsules network combined with the Xception module (XCN) is applied in intelligent fault diagnosis, so as to improve the classification accuracy of intelligent fault diagnosis. Firstly, the fault time‐frequency graphs are obtained by wavelet time‐frequency analysis. Then the time‐frequency graphs data which are adjusted the pixel size are input into XCN for training. In order to accelerate the learning rate, the parameters which have bigger change are punished by cost function in the process of training. After the operation of dynamic routing, the length of the capsule is used to classify the types of faults and get the classification of loss. Then the longest capsule is used to reconstruct fault time‐frequency graphs which are used to measure the reconstruction of loss. In order to determine the convergence condition, the three losses are combined through the weight coefficient. Finally, the proposed model and the traditional methods are, respectively, trained and tested under laboratory conditions and actual wind turbine gearbox conditions to verify the classification ability and reliable ability.

In industrial production, it is highly essential to extract faults in gearbox accurately. Specifically, in a strong noise environment, it is difficult to extract the fault features accurately. LMD (local mean decomposition) is widely used as an adaptive decomposition method in fault diagnosis. In order to improve the mode mixing of LMD, ELMD (ensemble Local Mean Decomposition) is proposed as local mode mixing exists in noisy environment, but white noise added in ELMD cannot be completely neutralized leading to the influence of increased white noise on PF (product function) component. This further leads to the increase in reconstruction errors. Therefore, this paper proposes a composite fault diagnosis method for gearboxes based on an improved ensemble local mean decomposition. The idea is to add white noise in pairs to optimize ELMD, defined as CELMD (Complementary Ensemble Local Mean Decomposition) then remove the decomposed high noise component by PE (Permutation Entropy) while applying the SG (Savitzky‐Golay) filter to smooth out the low noise in PFs. The method is applied to both simulated signal and experimental signal, which overcomes mode mixing phenomenon and reduces reconstruction error. At the same time, this method avoids the occurrence of pseudocomponents and reduces the amount of calculation. Compared with LMD, ELMD, CELMD, and CELMDAN, it shows that improved ensemble local mean decomposition method is an effective method for extracting composite fault features.

In recent years, a new method of fault diagnosis, named variational mode decomposition (VMD), has been widely used in industrial production, but the decomposition accuracy of VMD is determined by two parameters, which are respectively the decomposition layer number k and the penalty factor α, if the parameters are not properly selected, there will be over-decomposition or under-decomposition. In order to find an approach to determine the parameters adaptively, a method to optimize VMD by using the immune fruit fly optimization algorithm (IFOA) is proposed in this paper. In this method, permutation entropy is used as the fitness function, firstly, the immune fruit fly optimization algorithm is used to search the combined parameters of k and α in VMD, searching for the best combination parameters of k and α by iteration, and then uses the combined parameters to perform VMD, finally, the center frequency is determined through frequency spectrum analysis. The method mentioned is applied to the fault extraction of a simulated signal and a measured signal of a wind turbine gearbox, and the fault frequency is successfully extracted. Using ensemble empirical mode decomposition (EEMD) and singular spectrum decomposition (SSD) to compare with the proposed method, which validated feasibility of the proposed method.

In practical engineering, significant noise and amplitude fluctuations in bearing vibration data, hinder the accuracy of fault identification. In order to overcome these difficulties, reduce external interference, and accurately obtain effective bearing fault information in complex environments, a new technique of adaptive SVMD and RCMDSE-IDHT bearing fault diagnosis is proposed. Firstly, the article proposes to adopt the Crested Porcupine Optimizer (CPO) algorithm to optimise the Successive Variational Modal Decomposition (SVMD) parameters to form the adaptive SVMD (ASVMD) algorithm; Subsequently, the original bearing vibration signals are decomposed using the ASVMD to form a number of Intrinsic Mode Function (IMF) signals; Secondly, the Pearson correlation coefficient is used to filter the IMF signal and reconstruct so as to remove the noise from the original signal; In order to further improve the fault identification accuracy, the article proposes to optimise and improve the Differential Symbolic Entropy (DSE) to form the Refined Composite Multi-scale Differential Symbolic Entropy (RCMDSE); and the RCMDSE is used for comprehensive feature extraction of the reconstructed vibration signal; Finally, for the defects of the DHT classifier, the article proposes to adopt the Spearman's correlation coefficient and the entropy weight method to improve the traditional Mahalanobis Distance (MD) to form the improve Weighted Mahalanobis Distance (IWMD), which uses the advantages of IWMD to improve the Self-organizing Divisive Hierarchical Voronoi Tessellation (DHT) classifier to form improve DHT (IDHT) classifier. In order to test the accuracy as well as the effectiveness of the new method proposed in the article, the bearing data was used for test analysis, and the accuracy of the failure recognition accuracy was as high as 99.237%through test analysis. To verify the superiority of the method mentioned, compared with the six fault recognition models, the results showed that the accuracy of the diagnosis of the new method of faults in this paper increased by 0.72% to 11.75%. At the same time, to validate the extensiveness of the method proposed in this paper, the article utilizes the Southeast University bearing dataset for validation.