Rotating mechanical vibration signal has strong nonlinearity,non-stationary characteristics,complementary set empirical mode decomposition (CEEMD) overcomes the shortcomings of traditional EEMD,provides the characterization of different scales of signal from coarse to fine.A fault diagnosis method based on multi-scale weighted CEEMD was proposed for the difference in fault characteristics.The vibration signal was decomposed into a series of intrinsic modal functions (IMFs) using the complementary set empirical mode,then the steepness value of each IMF component was obtained,the weight of each component steepness was calculated,and the signal was reconstructed according to the weight value of each component.The data samples were divided into training sets,validation sets and test sets.The training sets were input into the one-dimensional convolutional neural network to learn and update the network parameters.Then the validation sets were used to verify the optimal diagnostic model.Finally the diagnostic models were tested by the test set.The model verification was carried out by two sets of experiments,the motor bearing data set and the gearbox data set,and the diagnostic accuracy was 99.98% and 99.73% respectively.The results show that the proposed method can quickly and accurately diagnose different fault types,and has a high fault diagnosis accuracy and robustness.