In order to improve the diagnosis accuracy and generalization of bearing faults, an integrated vision transformer (ViT) model based on wavelet transform and the soft voting method is proposed in this paper. Firstly, the discrete wavelet transform (DWT) was utilized to decompose the vibration signal into the subsignals in the different frequency bands, and then these different subsignals were transformed into a time–frequency representation (TFR) map by the continuous wavelet transform (CWT) method. Secondly, the TFR maps were input with respective to the multiple individual ViT models for preliminary diagnosis analysis. Finally, the final diagnosis decision was obtained by using the soft voting method to fuse all the preliminary diagnosis results. Through multifaceted diagnosis tests of rolling bearings on different datasets, the diagnosis results demonstrate that the proposed integrated ViT model based on the soft voting method can diagnose the different fault categories and fault severities of bearings accurately, and has a higher diagnostic accuracy and generalization ability by comparison analysis with integrated CNN and individual ViT.

To solve the fault diagnosis difficulty of bearings with small data samples, a novel multi-information fusion vision transformer (ViT) model based on time–frequency representation (TFR) maps is proposed in this paper. The original vibration signal is decomposed into different scale sub-signals by the discrete wavelet transforms (DWTs), and the continuous wavelet transforms (CWTs) are used to transform these different scale sub-signals into time–frequency representation (TFR) maps, which are concatenated to input to the ViT model to diagnose the bearing fault. Through the multifaceted experiment analysis on the fault diagnosis of bearings with small data samples, the diagnosis results demonstrate that the proposed multi-information fusion ViT model can diagnose the fault of bearings with small data samples, with strong generalization and robustness; its average diagnosis accuracy achieved 99.85%, and it was superior to the other fault diagnosis methods, such as the multi-information fusion CNN, ViT model based on one-dimensional vibration signal, and ViT model based on the TFR of the original vibration signal.