In complex industrial environments, the vibration signal of the rolling bearing is covered by noise, which makes fault diagnosis inaccurate. In order to overcome the effect of noise on the signal, a rolling bearing fault diagnosis method based on the WOA-VMD (Whale Optimization Algorithm-Variational Mode Decomposition) and the GAT (Graph Attention network) is proposed to deal with end effect and mode mixing issues in signal decomposition. Firstly, the WOA is used to adaptively determine the penalty factor and decomposition layers in the VMD algorithm. Meanwhile, the optimal combination is determined and input into the VMD, which is used to decompose the original signal. Then, the Pearson correlation coefficient method is used to select IMF (Intrinsic Mode Function) components that have a high correlation with the original signal, and selected IMF components are reconstructed to remove the noise in the original signal. Finally, the KNN (K-Nearest Neighbor) method is used to construct the graph structure data. The multi-headed attention mechanism is used to construct the fault diagnosis model of the GAT rolling bearing in order to classify the signal. The results show an obvious noise reduction effect in the high-frequency part of the signal after the application of the proposed method, where a large amount of noise was removed. In the diagnosis of rolling bearing faults, the accuracy of the test set diagnosis in this study was 100%, which is higher than that of the four other compared methods, and the diagnosis accuracy rate of various faults reached 100%.

Flow Direction Algorithm (FDA) has better searching performance than some traditional optimization algorithms. To give the basic Flow Direction Algorithm more effective searching ability and avoid multiple local minima under the searching space, and enable it to obtain better search results, an improved FDA based on the Lévy flight strategy and the self-renewable method (LSRFDA) was proposed in this paper. The Lévy flight strategy and the self-renewable approach were added to the basic Flow Direction Algorithm. Random parameters generated by the Lévy flight strategy can increase the algorithm’s diversity of feasible solutions in a short calculation time and greatly enhance the operational efficiency of the algorithm. The self-renewable method lets the algorithm quickly obtain a better possible solution and jump to the local solution space. Then, this paper tested different mathematical testing functions, including low-dimensional and high-dimensional functions, and the test results were compared with those of different algorithms. This paper includes iterative figures, box plots, and search paths to show the different performances of the LSRFDA. Finally, this paper calculated different engineering optimization problems. The test results show that the proposed algorithm in this paper has better searching ability and quicker searching speed than the basic Flow Direction Algorithm.

Aiming to resolve the problem of redundant information concerning rolling bearing degradation characteristics and to tackle the difficulty faced by convolutional deep learning models in learning feature information in complex time series, a prediction model for remaining useful life based on multiscale fusion permutation entropy (MFPE) and a multiscale convolutional attention neural network (MACNN) is proposed. The original signal of the rolling bearing was extracted and decomposed by resonance sparse decomposition to obtain the high-resonance and low-resonance components. The multiscale permutation entropy of the low-resonance component was calculated. Moreover, the locally linear-embedding algorithm was used for dimensionality reduction to remove redundant information. The multiscale convolution module was constructed to learn the feature information at different time scales. The attention module was used to fuse the feature information and input it into the remaining useful life prediction module for evaluation. The appropriate network structure and parameter configuration were determined, and a multiscale convolutional attention neural network was designed to determine the remaining useful life prediction model. The results show that the method demonstrates effectiveness and superiority in degrading the feature information representation and improving the remaining useful life prediction accuracy compared with other models.