Aiming at the problem of early fault diagnosis of rolling bearing, it is pointed out that the empirical mode decomposition (EMD) has poor anti-noise effect and the limitation of end effect. Variational modal decomposition (VMD), principal component analysis (PCA) and entropy bearing fault diagnosis are put forward based on cepstral envelope. Firstly, it takes VMD to decompose the bearing vibration signal of rolling; secondly, the components are handled by PCA. Then, it computes the entropy of the components and choses whose entropy of components is less than the average to reconstruct. Finally, the reconstructed signal is analyzed by cepstral envelope. Experimental results show that the method can effectively extract the rolling bearing fault frequencies, and judge damage position of the rolling bearing, it has a good anti-noise ability.