针对故障滚动轴承在单一工况数据下训练的深度学习模型无法在复杂工况下无法实现有效的故障诊断，提出一种基于卷积神经网络的领域适配模型（Convolutional Neural Network—Domain Adaptation， CNN-DA）。卷积网络用于对故障振动信号进行高层特征提取，网络首尾加入通道注意力机制（Channel Attention Mechanism，CAM），以动态分配特征通道的权重，减小无效信息的干扰。结合领域自适应方法，将特征提取层获取到的高层故障特征进行源、目标域（Source、Target）领域适配，领域适配模块整合了全域适配（Whole Domain Adaptation）和类别域适配（Category Domain Adaptation），以使两个领域中相同故障标签的特征的数据分布逐渐趋于重合，最后将深度学习模型应用于多种不同工况迁移的场合进行训练，得到训练结果和测试结果。通过不同来源数据集的实验，在多种工况迁移下测试模型，结果表明提出的模型能够应对复杂工况变化下的滚动轴承故障检测。
Abstract
In view of the fact that the deep learning model trained under single working condition data can't realize effective fault diagnosis under complex working conditions, a convolutional neural network-domain adaptation model (CNN-DA) is proposed. Convolution network is used to extract high-level features of fault vibration signals. Channel Attention Mechanism (CAM) is added to the network at the beginning and end to dynamically allocate the weights of feature channels and reduce the interference of invalid information. Combining with the domain adaptive method, the high-level fault features obtained by the feature extraction layer are adapted in the Source and Target domains. The domain adaptation module integrates the Whole Domain Adaptation and the Category Domain Adaptation, so that the data distribution of the features of the same fault tag in the two domains tends to coincide gradually. Finally, the deep learning model is applied to a variety of different situations for training, and the training results and test results are obtained. Through experiments on data sets from different sources, the model is tested under various working conditions, and the results show that the proposed model can cope with the fault detection of rolling bearings under complex working conditions.
关键词
故障诊断 /
深度学习 /
CNN-DA /
领域自适应
{{custom_keyword}} /
Key words
Fault Diagnosis /
Deep Learning /
convolutional neural network—domain adaptation（CNN-DA） /
Domain Adaptation /
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Lu W, Liang B, Yu C, et al. Deep model based domain adaptation for fault diagnosis[J]. IEEE Transactions on Industrial Electronics, 2017, 64(3):2296-2305.
[2] Pan, S. J. Tsang, I. W. Kwok, J. T. Yang. Domain adaptation via Transfer Component Analysis[J]. IEEE Transactions on Neural Networks, 2011, 22(2):199-210.
[3] Long M, Wang J, Ding G, et al. Adaptation regularization: A general framework for transfer learning[J]. IEEE Transactions on Knowledge & Data Engineering, 2014, 26(5):1076-1089.
[4] Wang Q, Michau G, Fink O. Domian adaptive transfer learning for fault diagnosis[C]//2019 Prognostics and System Health Management Conference (PHM-Paris), IEEE. 2019.
[5] Wen Long, Gao Liang, Li Xinyu.[J]. IEEE Transactions on Systems, 2019, 49 (1) :136-144.
[6] Mingsheng Long, Yue Cao, Jianmin Wang, Michael Jordan. Learning transferable features with deep adaptation networks[C]//Proceedings of the 32nd International Conference on Machine Learning. 2015.
[7] Gretton A, Borgwardt K M, Rasch M J, et al. A kernel two-sample test[J]. Journal of Machine Learning Research, 2012, 13:723-773.
[8] Li Y, Wang N, Shi J, et al. Revisiting batch normalization for practical domain adaptation[J]. Pattern Recognition, 2016, 80: 109-117.
[9] WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[C]//15th European Conference on Computer Vision. Munich: Springer Verlag. 2018.
[10] Lou X, Loparo K A. Bearing fault diagnosis based on wavelet transform and fuzzy inference[J]. Mechanical Systems and Signal Processing (MSSP), 2004, 18(5):1077-1095.
[11] Bilbao. A benchmark data set for data-driven classification[C]//European Conference of the Prognostics and Health Management Society (PHM Society). 2016.
[12] He K, Zhang X, Ren S. Delving deep into Rectifiers: Surpassing human-level performance on ImageNet classification[C]//2015 IEEE International Conference on Computer Vision (ICCV). 2015.
[13] 朱浩, 宁芊. 基于注意力机制-Inception-CNN 模型的滚动轴承故障分类[J]. 振动与冲击，2020, 39(19):84-93.
Zhu Hao, Ning Qian. Rolling bearing fault classification based on attention mechanism-initiation-CNN model [J]. Vibration and Shock, 2020,39 (19): 84-93.
[14] Guo L, Lei Y, Xing S, et al. Deep convolutional transfer learning network: A new method for intelligent fault diagnosis of machines with unlabeled data[J]. IEEE Transactions on Industrial Electronics, 2018, 66(9):7316-7325.
[15] Y Ganin, E Ustinova, H Ajakan, P Germain, et al. Domain-adversarial training of neural networks[J]. Journal of Machine Learning Research, 2016, 17(59):1-35.
[16] Maaten L V D, Hinton G. Visualizing data using t-SNE[J]. Journal of machine learning research, 2008, 9:2579-2605.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
PDF(2079 KB)
