自适应辅助分类器生成式对抗网络样本生成模型及轴承故障诊断

故障样本获取困难导致的训练样本不均衡严重影响故障诊断模型的可用性及准确率，因此提出一种基于自适应辅助分类器生成式对抗网络的故障样本生成模型，通过度量判别器与生成器的相对性能自适应地调节生成器损失值，使训练收敛更快、生成数据质量更好。将所提方法、辅助分类器生成式对抗网络方法生成的数据，以及未经处理的试验原始数据作为BP分类模型的输入数据进行试验，结果表明所提方法生成数据训练的模型更优。所提方法与1D-CNN、e2e-LSTM、CFVS-SVM和FFT-CNN等方法的对比结果表明，所提方法的故障诊断准确率、信息处理时间均最优。

Bearing Fault Diagnosis Based on SA-ACGAN Data Generation Model

Unbalancing training dataset caused by the difficulty in obtaining fault samples seriously affectsed the robust and accuracy of fault diagnosis model. A data generation model was proposed based on self-adaptive auxiliary classifier GAN， which adaptively adjusted the generator loss by measuring the relative performance between discriminator and generator， accelerated the converge speed of training processes， and improved the quality of generated data. The raw data， data generated by auxiliary classifier GAN method， and data generated by proposed method were used as the input data of the BP neural network. The results show that the model trained by data of the proposed method was superior. Comparison results of the proposed method with 1D-CNN，e2e-LSTM，CFVS-SVM， and FFT-CNN fault diagnosis methods manifest that the proposed method is better in fault diagnosis accuracy and data processing time