作者：杨健, 廖晨茜, 蔡晋辉, 曾九孙

作者单位：中国计量大学计量测试工程学院，浙江 杭州 310018

关键词：对抗迁移学习;故障诊断;一维卷积结构;域判别器

摘要：

为解决故障诊断中标签不足的问题，该文以滚动轴承作为对象提出一种改进的对抗迁移学习模型。该模型通过一维卷积结构提取时间信号特征，能够直接处理时域振动信号，并通过大尺寸卷积核抑制噪声的干扰；在对抗迁移学习的域判别器中采用卷积结构替换全连接神经网络，以对抗训练的方式减少训练数据和测试数据间的分布差异，以提高故障诊断精度。将改进后的模型应用于两个滚动轴承故障诊断案例中，通过添加不同信噪比的噪声信号验证提出的模型具有良好的抗干扰能力，同时以故障分类正确率作为指标，验证该模型具有更高的诊断精度和鲁棒性。

Rolling bearing fault diagnosis based on improved adversarial transfer learning
YANG Jian, LIAO Chenxi, CAI Jinhui, ZENG Jiusun
College of Metrology & Measurement Engineering, China Jiliang University, Hangzhou 310018, China
Abstract: In order to solve the problem of insufficient labels in fault diagnosis, this paper proposes an improved adversarial transfer learning model based on rolling bearings. The model extracts time signal features through a one-dimensional convolution structure, can directly process time-domain vibration signals, and suppress noise interference through a large-size convolution kernel; the convolution structure is used to replace the fully connected nerve in the domain discriminator against transfer learning, the network reduces the distribution difference between training data and test data in a way of adversarial training, so as to improve the accuracy of fault diagnosis. The improved model was applied to two rolling bearing fault diagnosis cases. The addition of noise signals with different signal-to-noise ratios verified that the proposed model has good anti-interference ability. At the same time, the correct rate of fault classification is used as an indicator to verify that the model has higher diagnostic accuracy and robustness.
Keywords: adversarial transfer learning;fault diagnosis;one-dimensional convolution;domain discri-mination
2022, 48(5):96-101  收稿日期: 2021-03-22;收到修改稿日期: 2021-04-21
基金项目: 国家重点研发计划资助项目（2018YFF0214701）；国家自然科学基金资助项目（61673358）
作者简介: 杨健（1995-），男，安徽滁州市人，硕士研究生，专业方向为故障诊断
参考文献
[1] YAN R, GAO R X, CHEN X. Wavelets for fault diagnosis of rotary machines: A review with applications[J]. Signal processing, 2014, 96: 1-15
[2] 乔福宇,马良玉,马永光. 基于功率曲线分析与神经网络的风电机组故障预警方法[J]. 中国测试, 2020, 46(8): 44-50
[3] GAN M, WANG C. Construction of hierarchical diagnosis network based on deep learning and its application in the fault pattern recognition of rolling element bearings[J]. Mechanical Systems and Signal Processing, 2016, 72: 92-104
[4] 周林春,陈春俊. 复数卷积神经网络滚动轴承故障诊断研究[J]. 中国测试, 2020, 46(11): 109-115
[5] TAN C, SUN F, KONG T, et al. A survey on deep transfer learning[C]//International conference on artificial neural networks. Springer, Cham, 2018: 270-279.
[6] LI J, HUANG R, LI W. Intelligent FaultTransfer Learning based on Stacked AutoEncoder[J]. Procedia Manufacturing, 2020, 49: 75-80
[7] GOODFELLOW I, POUGET-ABADIE J, MIRZA- M, et al. Generative adversarial nets[C]// Advances in neural information processing systems. 2014: 2672-2680.
[8] LI F, TANG T, TANG B, et al. Deep convolution domain-adversarial transfer learning for fault diagnosis of rolling bearings[J]. Measurement, 2021, 169: 108339
[9] GANIN Y, USTINOVA E, AJAKAN H, et al. Domain-adversarial training of neural networks[J]. The Journal of Machine Learning Research, 2016, 17(1): 2096-2030
[10] WANG J. Transfer learning tutorial[DB/OL]. 2018, https://github.com/jindongwang/transferlearning-tutorial.
[11] ZHANG W, PENG G, LI C, et al. A new deep learning model for fault diagnosis with good anti-noise and domain adaptation ability on raw vibration signals[J]. Sensors, 2017, 17(2): 425
[12] ZHAO Z, ZHANG Q, YU X, et al. Uns-upervised deep transfer learning for intelligent fault diagnosis: an open source and comparative study[J]. arXiv preprt arXiv, 2019: 1912
[13] 李可. 江南大学轴承数据[EB/OL]. http://mad-net.org:8765/explore.html?t=0.5398991360140968.
[14] Case Western Reserve University. Case Western Reserve University Bearing Data [EB/OL]. https://csegroups.case.edu/bearingdatacenter/pages/apparatus-procedures.
[15] RADFORD A, METZ L, CHINTALA S. Unsupervised representation learning withdeep convolutional generative adversarial networks[C]// ICLR. 2016.