In order to improve the accuracy of weak fault features extraction of rolling bearings based on variational modal decomposition (VMD), a fault diagnosis method of rolling bearings were proposed based on parameter optimization VMD combined with singular value component and its entropy.In this method,the modal number K and the second penalty factor α of VMD were determined by an optimization algorithm,the typical intrinsic mode functions (IMF) of VMD were extracted according to cosinestandard deviation index.The IMF singular value and its entropy were calculated,and the results were used to judge the different fault states of rolling bearing. Combined with the experimental results of bearing vibration signal data from Case Western Reserve University in the U.S., it is shown that the singular value fault diagnosis method based on parameter optimization of VMD can identify different fault types of rolling bearings more clearly compared with the empirical modal decomposition singular value fault diagnosis method. The study provides a feasible diagnostic approach to distinguish the weak faults of rolling bearings.