For fault diagnosis problems of rolling bearing that the nuances of one-dimensional vibration data feature extraction is not obvious and real-time diagnosis speed is low

using vibration data and image data pretreatment method

the vibration signal can be converted into 2D time-frequency diagrams using the continuous wavelet transform

a kind of based on the re-parametric technology (RepVGG) structure of bearing fault diagnosis method is put forward. The multi-branch network structure of the training model is equivalent to the single-path network structure

so as to improve the accuracy and speed of the inference model. The experimental verification is done on the bearing fault data set. The results show that the RepVGG model can accurately identify bearing fault categories

and the average accuracy rate is better than other methods. Moreover

under the same experimental hardware conditions

the RepVGG model is effective and efficient. The reasoning time is reduced by 81% and 66.19% respectively

which effectively improves the speed and accuracy of fault diagnosis

and has good adaptability and superiority.