Fault Diagnosis of Planetary Gear Box Based on Gramian Angular Fields and Deep Convolutional Generative Adversarial Network
Aiming at the problems of poor generalization ability and low diagnosis accuracy due to uneven sample distribution in planetary gearbox fault diagnosis,data enhancement was carried out based on the combination of Gramian angular field coding technology and deep convolutional generative adversarial network,and Alexnet convolution neural network was fused for fault diagnosis.The collected one-dimensional vibration signal is converted into Gramian angular fields.The training set and test set are divided proportionally.And the samples and random vectors of the training set are input into the deep convolutional generative adversarial network model to alternately train the generator and discriminator until the Nash balance is reached.Then,the samples similar to the original samples are generated,and the augmentation of fault samples is realized.Finally,the original samples and the generated augmented samples are used to train the convolutional neural network classification model to complete the fault identification of the planetary gearbox.The experimental analysis results show that the proposed method can effectively improve the fault diagnosis accuracy of the planetary gearbox under the condition of uneven samples,the accuracy is up to 99.15%,and meanwhile it has faster convergence speed.
万方数据
维普
