The work aims to propose a rolling bearing fault diagnosis method based on DS-PCA model to analyze the common failure faults of rolling bearings caused by the long-term continuous working characteristics of packaging equipment and construct a DS-PCA fault diagnosis model to realize the online fault diagnosis of rolling bearings in packaging equipment. Firstly, DS evidence theory was used to fuse the collected radial vibration data and axial vibration data of rolling bearings to make the information have overall completeness. At the same time, the collected signals were denoised under the decision rules to eliminate the interference noise information. Then, the principal component analysis (PCA) was used to calculate the Q and T2 statistics of the fused vibration signal data, and the empirical threshold was established by calculating the Q and T2 statistics of the vibration signal of the failed bearings. Finally, according to the comparison between the real-time statistics and empirical threshold, it was judged whether the rolling bearings had failure. Through the analysis and calculation of the experimental data of rolling bearings published by Case Western Reserve University, the accuracy of rolling bearing fault diagnosis reached 94%. The method meets the requirements of packaging equipment fault diagnosis, and its application will help improve the production quality and efficiency of packaging enterprises.