Aiming at the nonlinear characteristics of rolling bearing fault signal and the vibration signal with different types of faults will have different morphological features, a rolling bearing fault diagnosis method based on improved variational mode decomposition (VMD) morphological spectrum and fuzzy C-means clustering was proposed. In this method, the original rolling bearing vibration signals were decomposed by VMD method, aiming at the selection of key parameters in the decomposition process, the selection method of relevant parameters was proposed and the energy fluctuation coefficients of each intrinsic mode function (IMF) were calculated to obtain the modal components sensitive to the signal characteristic information for reconstruction. Then, the morphological spectrum of the reconstructed signal was calculated to reflect the morphological characteristics of the signal. Finally, the working state and fault type of rolling bearing were diagnosed by FCM algorithm. Through analyzing the rolling bearing vibration signals, the proposed method was compared with the fault diagnosis methods based on original vibration signal morphological spectrum, empirical mode decomposition morphological spectrum and ensemble empirical mode decomposition morphological spectrum. The results show that this method can effectively extract the fault characteristics of rolling bearing signals and the accurate diagnosis of fault types is realized.