| 摘要: |
| 主要开展了深度卷积生成对抗网络（DCGAN）在流体机械故障诊断方面的应用研究，建立了基于深度学习和迁移学习在小样本条件下的流体机械故障诊断方法.仿真实验结果表明：基于DCGAN生成的数据能涵盖原始数据的主要特征，可被用于对流体机械小样本故障诊断数据的扩充.结合迁移学习和深度学习的故障诊断方法，采用不同流体机械的运行数据对模型进行实验验证，结果表明：该模型解决了小样本训练中的过拟合问题，诊断准确率为98%~100%. |
| 关键词: 流体机械 气蚀诊断 内啮合齿轮泵 深度学习 深度卷积生成对抗网络(DCGAN) |
| DOI：10.3969/J.ISSN.1000-5137.2023.02.018 |
| 分类号:TH137 |
| 基金项目:国家自然科学基金（51806145） |
|
| Research on fault diagnosis method of small sample fluid machinery based on deep learning method |
|
LIU Di, LIU Yingyuan
|
|
College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 201418, China
|
| Abstract: |
| We mainly carry out the application research of deep convolution generative adversarial network (DCGAN) in the fault diagnosis of fluid machinery, and establishes a fault diagnosis model of fluid machinery under the condition of small samples based on deep learning and transfer learning. The experimental results show that the data generated based on DCGAN could cover the main features of the original data and can be used for data expansion of small sample fault diagnosis of fluid machinery. Meanwhile, the fault diagnosis model combined with transfer learning and deep learning is validated for different fluid machineries. Results show that the model solves the over-fitting problem in small sample training, and the diagnostic accuracy is 98%-100%. |
| Key words: fluid machinery cavitation diagnosis internal gear pump deep learning deep convolution generative adversarial network (DCGAN) |
