基于FMD-MAK-CNN算法的轴承故障诊断OA

Due to the low bearing fault diagnosis accuracy,a bearing fault diagnosis model based on-feature mode decomposition-maximum autocorrelated kurtosis-convolutional neural network(FMD-MAK-CNN)was proposed.Firstly,the feature mode decomposition(FMD)algorithm was employed to decompose the bearing fault signals into several modes containing abundant fault features.Secondly,the autocorrelated kurtosis(AK)of each mode was calculated,and the mode corresponding to the maximum autocorrelated kurtosis(MAK)was chosen as the signal to be analyzed.Feature vectors were constructed based on its time domain characteristics.The feature vector matrix was then formed by com-bining the feature vectors with signal labels.Finally,the feature vector matrix was divided into a train-ing set and a testing set in a ratio of 8∶2.It was fed into the convolutional neural network(CNN)for training and testing,thus achieving bearing fault diagnosis.By using two sets of experimental data from the bearing inner ring and the bearing outer ring,the effectiveness of the FMD-MAK-CNN model for fault diagnosis is verified.The average fault diagnosis accuracy rates are 97.50％and 97.83％,respec-tively.Under the same model conditions,compared with the EMD-MAK-CNN fault diagnosis model,the average fault diagnosis accuracy rates have improved by 16.17％and 16.66％,respectively.Com-pared with the FMD-EE-CNN fault diagnosis model,the average fault diagnosis accuracy rates have increased by 16.67％and 16.83％,respectively.