Abstract: As the core component of a horizontal dynamic tensile testing machine, the health status of rolling bearings directly impacts equipment performance. However, existing bearing fault diagnosis methods show significant limitations in feature extraction accuracy and temporal information representation, making it challenging to capture complex fault pattern variations accurately. Therefore, this study proposes a new model for fault diagnosis of bearings in horizontal dynamic tensile testing machines. First, with minimum envelope entropy as the optimization objective, the SSA is used to automatically optimize VMD parameters. Then, the optimized VMD parameters are applied to decompose the bearing signal and obtain intrinsic mode functions (IMFs). Finally, a CNN is employed to extract fault features, which are then input into a BiLSTM to further express temporal features, ultimately achieving fault classification. Experimental results show that the model achieved an accuracy rate of 99.01% in fault identification.