滚动轴承作为旋转机械的关键传动部件，当其发生故障时将直接影响设备的安全可靠运行，因此对滚动轴承开展故障诊断研究具有重要意义。在轴承的全生命周期中，故障数据是非常有限的，且故障数据与正常数据往往是高度不平衡的，这限制了数据驱动的故障诊断的准确性。本文针对实际中不平衡小样本数据，设计单一和复合故障模式，构建小样本数据集和不平衡数据集，研究滚动轴承故障数据增强模型，同时针对复合故障诊断精度不高的问题，开展滚动轴承故障诊断模型改进研究。

（1）针对滚动轴承故障诊断中样本数量少导致诊断模型精度低的问题，开展基于梯度惩罚Wasserstein生成对抗网络（Wasserstein Generative Adversarial Network-Gradient Penalty，WGAN-GP）的小样本数据增强模型研究。针对滚动轴承单一故障和复合故障模式下的振动图像，构建小样本数据集。采用WGAN-GP模型进行数据增强，并采用最大均值差异（Maximum Mean Discrepancy，MMD）对生成样本进行评价，得到增强数据集。研究不同数量的原始样本和不同的扩容比例对诊断模型性能的影响，同时对比深度卷积生成对抗网络、Wasserstein GAN、WGAN-GP增强模型的样本生成质量。结果表明基于WGAN-GP的数据增强方法能够在小样本下实现滚动轴承数据的高质量扩容。

（2）针对滚动轴承故障诊断的样本类间不平衡和部分类别样本数量极少的问题，开展融合重采样的条件Wasserstein GAN-GP（Conditional WGAN-GP，CWGAN-GP）不平衡数据增强模型研究。针对滚动轴承单一故障、复合故障和混合故障模式下的振动图像，构建不平衡数据集。采用重采样方法与CWGAN-GP模型，构建原始不平衡数据集、增强数据集、重采样数据集和重采样后再增强数据集，在常规不平衡、较不平衡、类间不平衡和极端不平衡四种情况下进行模型训练和验证。结果表明只用数据增强模型可以有效改善数据不平衡问题，从而提高诊断准确率；但当数据呈极不平衡时，仅用数据增强模型生成的样本难以满足诊断需求，此时结合重采样方法可以改善样本生成质量，提高诊断准确率，最终实现数据极不平衡下的高精度诊断。

（3）针对滚动轴承复合故障诊断精度低的问题，开展基于Inception模块改进卷积神经网络（Convolutional Neural Network，CNN）的滚动轴承智能故障诊断研究。采用改进的Inception模块替代卷积神经网络的卷积层，通过使用不同的卷积核提取不同尺度的信息，再结合CWGAN-GP，完成复杂故障数据集上的滚动轴承智能故障诊断。结果表明Inception模块改进Lenet-5网络模型能够完成复杂故障下滚动轴承故障准确诊断。

本文以滚动轴承为研究对象，完成了小样本和不平衡样本条件下的振动图像数据集高质量扩容研究，并在此基础上建立了改进的故障诊断深度学习模型，通过不同故障模式下的实验验证，证明了所提出的滚动轴承智能故障诊断方法是可行且有效的。

Rolling bearings are key transmission components of rotating machinery and when they fail, they will directly affect the safe and reliable operation of the equipment, so it is important to carry out fault diagnosis research on rolling bearings. Fault data are very limited during the full life cycle of a bearing and it is often highly unbalanced with normal data, which limits the accuracy of data-driven fault diagnosis. In this paper, aiming at unbalanced small sample data in practice, single and compound fault modes are designed, small sample data sets and unbalanced data sets are constructed, fault data enhancement model of rolling bearing is studied, and fault diagnosis model improvement research of rolling bearing is carried out to solve the problem of low precision of complex fault diagnosis.

(1) To solve the problem of low accuracy of diagnostic model due to the small number of samples in rolling bearing fault diagnosis, a small sample data enhancement model based on gradient penalty Wasserstein Generative Adversarial Network-Gradient Penalty (WGAN-GP) was developed. A small sample data set is constructed for vibration images of single failure and complex failure modes of rolling bearings. WGAN-GP model is used to enhance the data, and Maximum Mean Discrepancy (MMD) is used to evaluate the generated samples to obtain the expanded data set. The effects of different numbers of original samples and different expansion ratios on the performance of diagnostic model were studied, and the sample generation quality of DCGAN, Wasserstein GAN and WGAN-GP enhanced models were compared. The results show that the data enhancement method based on WGAN-GP can enlarge the capacity of rolling bearing data in a small sample.

(2) To solve the problems of unbalance between samples in rolling bearing fault diagnosis and the scarcity of samples in some categories, the study on the imbalance data enhancement model of Wasserstein GAN-GP (Conditional WGAN-GP) which fuses resampling was carried out. An unbalanced data set is constructed for vibration images of single failure, compound failure and mixed failure modes of rolling bearings. Using resampling method and CWGAN-GP model, the original imbalance data set, expanded data set, resampled data set and resampled data set are constructed, and then the model is trained and validated under four situations: conventional imbalance, relatively imbalance, inter-class imbalance and extreme imbalance. The results show that data imbalance can be effectively improved by using only data enhancement model, thus improving the accuracy of diagnosis. However, when the data is extremely unbalanced, it is difficult to satisfy the diagnostic requirements only by using the sample generated by the data enhancement model. At this time, the combination of resampling method can improve the quality of sample generation, improve the diagnostic accuracy and finally achieve high-precision diagnosis under the extremely unbalanced data.

(3) To solve the problem of low precision in complex fault diagnosis of rolling bearings, the research on intelligent fault diagnosis of rolling bearings based on Inception module improved Convolutional Neural Network (CNN) is carried out. The improved Inception module is used instead of the convolution layer of the convolution neural network. By using different convolution cores to extract information of different scales and combining with CWGAN-GP, intelligent fault diagnosis of rolling bearing on complex fault data sets is completed. The result shows that Inception module improves Lenet-5 network model to accurately diagnose rolling bearing faults under complex faults.

In this paper, the rolling bearing is taken as the research object, and the research on high-quality expansion of vibration image data set with small and unbalanced samples is completed. Based on this, an improved deep learning model for fault diagnosis is established. The experimental verification in different fault modes proves that the proposed intelligent fault diagnosis method for rolling bearing is feasible and effective.

[1]彭宇, 刘大同, 彭喜元. 故障预测与健康管理技术综述[J]. 电子测量与仪器学报, 2010, 24(01): 1-9.

[2]Lei Yaguo, Yang Bin, Jiang Xinwei, et al. Applications of machine learning to machine fault diagnosis: A review and roadmap[J]. Mechanical Systems and Signal Processing, 2020, 138: 106587.

[3]李晗, 萧德云. 基于数据驱动的故障诊断方法综述[J]. 控制与决策, 2011, 26(01): 1-9+16.

[4]Zhao Zhen, Liu Peter Xiaoping, Gao Jinfeng. Model-based fault diagnosis methods for systems with stochastic process - A survey[J]. Neurocomputing, 2022, 513: 137-152.

[5]何正嘉, 曹宏瑞, 訾艳阳, 等. 机械设备运行可靠性评估的发展与思考[J]. 机械工程学报, 2014, 50(2): 171-186.

[6]Lv Haixin, Chen Jinglong, Pan Tongyang, et al. Attention mechanism in intelligent fault diagnosis of machinery: A review of technique and application[J]. Measurement, 2022, 199: 111594.

[7]Li Weihua, Huang Ruyi, Li Jipu, et al. A perspective survey on deep transfer learning for fault diagnosis in industrial scenarios: Theories, applications and challenges[J]. Mechanical Systems and Signal Processing, 2022, 167: 108487.

[8]陈鹏. 滚动轴承故障诊断及性能退化评估方法研究[D]. 兰州: 兰州理工大学, 2021.

[9]Yang Zheng, Xu Binbin, Luo Wei, et al. Autoencoder-based representation learning and its application in intelligent fault diagnosis: A review[J]. Measurement, 2022, 189: 110460.

[10]Henriquez P , Alonso J B , Ferrer M A , et al. Review of Automatic Fault Diagnosis Systems Using Audio and Vibration Signals[J]. IEEE Transactions on Systems Man & Cybernetics Systems, 2014, 44(5): 642-652.

[11]工业和信息化部信息电信研究院. 工信部电信研究院大数据白皮书[EB/OL]. http://www.cctime.com/html/2014-5-12/20145121139179652.htm, [2014-05-12].

[12]李彦夫, 韩特. 基于深度学习的工业装备PHM研究综述[J]. 振动.测试与诊断, 2022, 42(05): 835-847+1029.

[13]Zhu Zhiqin, Lei Yangbo, Qi Guanqiu, et al. A review of the application of deep learning in intelligent fault diagnosis of rotating machinery[J]. Measurement, 2023, 206: 112346.

[14]吴春志, 冯辅周, 吴守军, 等. 深度学习在旋转机械设备故障诊断中的应用研究综述[J]. 噪声与振动控制, 2019, 39(05): 1-7.

[15]樊红卫, 张旭辉, 曹现刚, 等. 智慧矿山背景下我国煤矿机械故障诊断研究现状与展望[J]. 振动与冲击, 2020, 39(24): 194-204.

[16]Sun Yongjian, Wang Jian, Wang Xiaohong. Fault diagnosis of mechanical equipment in high energy consumption industries in China: A review[J]. Mechanical Systems and Signal Processing, 2023, 186: 109833.

[17]司伟伟, 岑健, 伍银波, 等. 小样本轴承故障诊断研究综述[J]. 计算机工程与应用, 2023, 59(06): 45-56.

[18]李舜酩, 郭海东, 李殿荣. 振动信号处理方法综述[J]. 仪器仪表学报, 2013, 34(08): 1907-1915.

[19]胡智勇, 胡杰鑫, 谢里阳, 等. 滚动轴承振动信号处理方法综述[J]. 中国工程机械学报, 2016, 14(06): 525-531.

[20]林水泉. 基于旋转机械滚动轴承的时域故障诊断方法[J]. 自动化技术与应用, 2020, 39(08): 1-5+35.

[21]程秀芳, 王鹏. 基于时域和频域分析的滚动轴承故障诊断[J]. 华北理工大学学报(自然科学版), 2020, 42(01): 58-64.

[22]郭庆丰, 王成栋, 刘佩森. 时域指标和峭度分析法在滚动轴承故障诊断中的应用[J].机械传动, 2016, 40(11): 172-175.

[23]刘鲲鹏, 白云川, 吕麒鹏, 等. 基于角域功率谱估计的滚动轴承故障诊断[J]. 内燃机与配件, 2018(19): 88-89.

[24]代士超, 郭瑜, 伍星. 基于同步平均与倒频谱编辑的齿轮箱滚动轴承故障特征量提取[J]. 振动与冲击, 2015, 34(21): 205-209.

[25]D. Abboud, J. Antoni, S. Sieg-Zieba, et al. Envelope analysis of rotating machine vibrations in variable speed conditions: A comprehensive treatment[J]. Mechanical Systems and Signal Processing, 2017, 84: 200-226.

[26]毛清华, 张勇强, 赵晓勇, 等. 变速工况下采煤机行星齿轮传动系统故障诊断[J]. 工矿自动化, 2021, 47(07): 8-13.

[27]唐先广, 郭瑜, 丁彦春, 等. 基于短时傅里叶变换和独立分量分析的滚动轴承包络分析[J]. 机械强度, 2012, 34(01): 1-5.

[28]Fan Hongwei, Yan Yang, Zhang Xuhui, et al. Composite Fault Diagnosis of Rolling Bearing Based on Optimized Wavelet Packet AR Spectrum Energy Entropy Combined with Adaptive No Velocity Term PSO-SOM-BPNN[J]. Journal of Sensors, 2021(2021): 4138652.

[29]杨剑锋, 乔佩蕊, 李永梅, 等. 机器学习分类问题及算法研究综述[J]. 统计与决策, 2019, 35(06): 36-40.

[30]李旭然, 丁晓红. 机器学习的五大类别及其主要算法综述[J]. 软件导刊, 2019, 18(07): 4-9.

[31]Choudakkanavar Gangavva, Mangai J.Alamelu, Bansal Mohit. MFCC based ensemble learning method for multiple fault diagnosis of roller bearing[J]. International Journal of Information Technology, 2022, 14(5) :2741-2751.

[32]Li Lin, Meng Weilun, Liu Xiaodong, et al. Research on Rolling Bearing Fault Diagnosis Based on Variational Modal Decomposition Parameter Optimization and an Improved Support Vector Machine[J]. Electronics, 2023, 12(6): 1290-1301.

[33]Liu Zengkai, Lv Kanglei, Zheng Chao, et al. A fault diagnosis method for rolling element bearings based on ICEEMDAN and Bayesian network[J]. Journal of Mechanical Science and Technology, 2022, 36(5): 2201-2212.

[34]Arjun Kumar Shah, Ashish Yadav, H.Malik. EMD and ANN based intelligent model for bearing fault diagnosis[J]. Journal of Intelligent & Fuzzy Systems, 2018, 35(5): 5391-5402.

[35]路敦利, 宁芊, 杨晓敏. KNN-朴素贝叶斯算法的滚动轴承故障诊断[J]. 计算机测量与控制, 2018, 26(06): 21-23+27.

[36]李军宁, 罗文广, 陈武阁. 面向振动信号的滚动轴承故障诊断算法综述[J]. 西安工业大学学报, 2022, 42(02): 105-122.

[37]Elasha Faris, Teixeira Joao Amaral. Fault diagnosis and health management of bearings in rotating equipment based on vibration analysis – a review[J]. Journal of Vibroengineering, 2021, 24(1): 46-74.

[38]雷亚国, 贾峰, 孔德同, 等. 大数据下机械智能故障诊断的机遇与挑战[J]. 机械工程学报, 2018, 54(05): 94-104.

[39]文成林, 吕菲亚. 基于深度学习的故障诊断方法综述[J]. 电子与信息学报, 2020, 42(01): 234-248.

[40]张西宁, 郭清林, 刘书语. 深度学习技术及其故障诊断应用分析与展望[J]. 西安交通大学学报, 2020, 54(12): 1-13.

[41]宫文峰, 陈辉, 张美玲, 等. 基于深度学习的电机轴承微小故障智能诊断方法[J]. 仪器仪表学报, 2020, 41(01): 195-205.

[42]周兴康, 余建波. 基于深度一维残差卷积自编码网络的齿轮箱故障诊断[J]. 机械工程学报, 2020, 56(07): 96-108.

[43]Hongwei Fan, Ceyi Xue, Xuhui Zhang, et al. Vibration images-driven fault diagnosis based on CNN and transfer learning of rolling bearing under strong noise[J]. Shock and Vibration, 2021, 2021(2): 1-16.

[44]Xining Zhang, Shuyu Liu, Lin Li, et al. Multiscale holospectrum convolutional neural network-based fault diagnosis of rolling bearings with variable operating conditions[J]. Measurement Science and Technology, 2021, 32(10): 105027.

[45]Zhang Ying, Xing Kangshuo, Bai Ruxue, et al. An enhanced convolutional neural network for bearing fault diagnosis based on time-frequency image[J]. Measurement, 2020,157: 107667.

[46]Janssens Olivier, Slavkovikj Viktor, Vervisch Bram, et al. Convolutional neural network based fault detection for rotating machinery[J]. Joumal of Sound and Vibration, 2016, 377: 331-345.

[47]林景栋, 吴欣怡, 柴毅, 等. 卷积神经网络结构优化综述[J]. 自动化学报, 2020, 46(01): 24-37.

[48]Jinyang Jiao, Ming Zhao, Jing Lin, et al. A comprehensive review on convolutional neural network in machine fault diagnosis[J]. Neurocomputing, 2020, 417: 36-63.

[49]Changchang Che, Huawei Wang, Xiaomei Ni, et al. Intelligent fault diagnosis method of rolling bearing based on stacked denoising autoencoder and convolutional neural network[J]. Industrial Lubrication and Tribology, 2020, 72(7): 947-953.

[50]Goodfellow I, Pouget A, Mirza M, et al. Generative adversarial networks[EB/OL]. https://arxiv.org/abs/1406.2661, [2014-06-10].

[51]陈淮源, 张广驰, 陈高, 等. 基于深度学习的图像风格迁移研究进展[J]. 计算机工程与应用, 2021, 57(11): 37-45.

[52]罗佳, 黄晋英. 生成式对抗网络研究综述[J]. 仪器仪表学报, 2019, 40(03): 74-84.

[53]Mirza M, Simon O. Conditional generative adversarial nets[EB/OL]. https://arxiv.org/abs/1411.1784, [2014-11-06].

[54]Radford A, Metz L, Chintala S. Unsupervised representation learning with deep convolutional generative adversarial networks[EB/OL]. https://arxiv.org/abs/1511.06434, [2015-11-19].

[55]马波, 蔡伟东, 赵大力. 基于GAN样本生成技术的智能诊断方法[J]. 振动与冲击, 2020, 39(18): 153-160.

[56]Arjovsky Martin, Chintala Soumith, Bottou Léon. Wasserstein GAN[EB/OL]. https://arxiv.org/abs/1701.07875, [2017-01-26].

[57]Xin Gao, Fang Deng, Xianghu Yue. Data augmentation in fault diagnosis based on the Wasserstein generative adversarial network with gradient penalty[J]. Neurocomputing, 2020(396): 487-494.

[58]Liu Shaowei, Jiang Hongkai, Wu Zhenghong, et al. Rolling bearing fault diagnosis using variational autoencoding generative adversarial networks with deep regret analysis[J]. Measurement,2021(168): 1-14.

[59]Jingyao Wu, Zhibin Zhao, Chuang Sun, et al. Ss-InfoGAN for Class-Imbalance Classification of Bearing Faults[J]. Procedia Manufacturing, 2020, 49: 99-104.

[60]Funa Zhou, Shuai Yang, Hamido Fujita, et al. Deep learning fault diagnosis method based on global optimization GAN for unbalanced data[J]. Knowledge-Based Systems, 2020(187): 1-19.

[61]曲建岭, 余路, 袁涛, 等. 基于一维卷积神经网络的滚动轴承自适应故障诊断算法[J]. 仪器仪表学报, 2018, 39(07): 134-143.

[62]苏文胜, 王奉涛, 张志新, 等. EMD降噪和谱峭度法在滚动轴承早期故障诊断中的应用[J]. 振动与冲击, 2010, 29(03): 18-21+201.

[63]Tong Jinyu, Tang Shiyu, Wu Yi, et al. A fault diagnosis method of rolling bearing based on improved deep residual shrinkage networks[J]. Measurement, 2023, 206: 112282.

[64]Ishaan Gulrajani, Faruk Ahmed, Martin Arjovsky, et al. Improved Training of Wasserstein GANs[EB/OL]. https://arxiv.org/abs/1704.00028, [2017-03-31].

[65]Mirza Mehdi, Osindero Simon. Conditional Generative Adversarial Nets[EB/OL]. https://arxiv.org/abs/1411.1784, [2014-11-06].

[66]Lecun Y, Bottou L, Bengio Y, et al. Gradient-based learning Applied to Document Recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324.

[67]Fan H W, Gao S Q, Zhang X H, et al. Intelligent Recognition of Ferrographic Images Combining Optimal CNN with Transfer Learning Introducing Virtual Images[J]. IEEE Access, 2020, 8: 137074-137093.

[68]薛策译. 基于振动图像深度学习的滚动轴承智能故障诊断研究[D]. 西安: 西安科技大学, 2022.

[69]高随祥, 文新, 马艳军, 等. 深度学习导论与应用实践[M]. 北京: 清华大学出版社, 2019.

[70]高烁琪. 齿轮箱磨粒铁谱图像智能分类与异常检测研究[D]. 西安: 西安科技大学, 2021.

[71]邵阳. 基于数据驱动的滚动轴承故障诊断算法研究[D]. 济南: 山东大学, 2021.

[72]Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. https://arxiv.org/abs/1409.1556, [2014-09-04].

[73]白洁音, 赵瑞, 谷丰强, 等. 多目标检测和故障识别图像处理方法[J]. 高电压技术, 2019, 45(11): 3504-3511.

[74]宋克臣, 颜云辉, 陈文辉, 等. 局部二值模式方法研究与展望[J]. 自动化学报, 2013, 39(06): 730-744.

[75]黄非非. 基于LBP的人脸识别研究[D]. 重庆: 重庆大学, 2009.