<p>Aiming at the traditional fault diagnosis methods when using one-dimensional vibration signal feature extraction, there were problems such as loss of fault information, limitation of the reflected fault state, and low fault recognition rate, a rolling bearing fault diagnosis method based on the combination of Gram Angle Difference Field (GADF) and Re-parameterized Large Kernel Network (RepLKNet) was proposed. Firstly, the original one-dimensional vibration signal data was collected to do data preprocessing, and then the one-dimensional vibration signal was encoded into RGB three-channel fault images using GADF, which preserves the correlation of the data to time; secondly, the encoded two-dimensional image feature dataset was inputted into a RepLKNet network model to be trained to realize the identification of different faults of rolling bearings; finally, the rolling bearing fault diagnosis method was proposed using the Case Western Reserve University (CWRU) bearing dataset for experimental validation, and the correct rate of fault diagnosis was 99.73%. The experimental results show that the proposed method has higher computational efficiency, better generalization performance, and higher fault recognition accuracy than other intelligent fault diagnosis methods, which can provide a reference for the fault diagnosis of rolling bearings in the actual industry.</p>