Aiming at the problems of low diagnosis accuracy and unstable signal characteristics of rolling bearings, a fault diagnosis model based on Zebra algorithm optimized variational mode decomposition (ZOA-VMD) and optimized support vector machine (ZOA-SVM) was proposed. Firstly, the vibration signal is decomposed by ZOA-VMD, and the features of intrinsic mode function (IMF), energy, energy entropy, arrangement entropy and multi-scale arrangement entropy are extracted, and the feature vector is constructed. Then, the Zebra algorithm is used to optimize the core parameters of SVM, so as to avoid the uncertainty caused by artificial parameter setting, and the feature vector is input into the fault diagnosis model of ZOA-SVM, to improve the diagnosis accuracy and achieve better fault classification effect. Finally, by analyzing the simulation results of the bearing data of Case Western Reserve University in the United States, it is shown that the average test set diagnostic accuracy of the proposed ZOA-VMD and ZOA-SVM methods is up to 95%, which can effectively identify the fault types of rolling bearings, and has certain value in practical engineering applications.