针对滚动轴承在复杂环境下故障类型耦合及数据获取难题，提出了一种基于语义融合零样本学习的轴承复合故障诊断模型。在训练阶段，采用语义自编码器（semantic auto encoder,SAE）来建立视觉空间与语义空间之间的联系，从而缓解域偏移问题，测试阶段通过相似度计算来识别未知故障。不同于传统方法，提出了一种语义融合的编码策略，将轴承故障的振动幅值和频率特征转化为具体的语义表示。这种方法保留了丰富的物理信息，并通过融合这些信息增强了故障类型间的语义差异，从而显著提高了复合故障分类的准确性。此外，模型结合卷积神经网络（Convolutional Neural Network,CNN）和自适应边缘中心损失（Adaptive Margin Center Loss,AMCL），进一步优化了故障特征提取，更精准地捕捉轴承的复合故障特征。实验结果表明，该模型在复合故障识别上取得了87.96%的准确率，优于对比模型。
Abstract
To address the challenges of fault type coupling and data acquisition for rolling bearings in complex environments, researchers proposed a bearing compound fault diagnosis model based on semantic fusion zero sample learning. During training, a Semantic Autoencoder (SAE) establishes a link between visual space and semantic space, mitigating the domain migration issue. In testing, the model identifies unknown faults through similarity calculations. This approach introduces a semantic fusion encoding strategy, transforming the vibration amplitude and frequency characteristics of bearing faults into distinct semantic representations. This strategy retains extensive physical information and enhances semantic differences among fault types by fusing this data, thus significantly boosting the accuracy of composite fault classification. Moreover, the integration of a Convolutional Neural Network (CNN) with Adaptive Margin Center Loss (AMCL) optimizes fault feature extraction, capturing compound fault characteristics of bearings more accurately. Experimental results indicate an accuracy of 87.96%, surpassing that of the comparison model.
关键词
滚动轴承 /
故障诊断 /
语义融合 /
零样本学习 /
语义自编码器
{{custom_keyword}} /
Key words
Rolling bearing /
Fault diagnosis /
Semantic fusion /
Zero sample learning /
Semantic autoencoder
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 刘敏,程军圣,谢小平,等.基于改进的辛周期模态分解的滚动轴承复合故障诊断方法[J].振动与冲击,2024,43(14):47-56.
LIU Min, CHENG Junsheng, XIE Xiaoping, et al. Diagnosis method for compound faults of rolling bearings based on an improved symplectic periodic modal decomposition [J]. Journal of Vibration and Shock, 2024, 43(14): 47-56.
[2] 杨岗,徐五一,邓琴,等.基于振动信号的滚动轴承复合故障诊断研究综述[J].西华大学学报(自然科学版),2024,43(01):48-69.
YANG Gang, XU Wuyi, DENG Qin, et al. Review of compound fault diagnosis of rolling bearings based on vibration signals [J]. Journal of Xihua University (Natural Science Edition), 2024, 43(1): 48-69.
[3] Jin Y, Qin C, Huang Y, et al. Actual bearing compound fault diagnosis based on active learning and decoupling attentional residual network[J]. Measurement, 2021, 173: 108500.
[4] 甄冬,孙赫明,冯国金,等.基于包络谱语义构建的零样本滚动轴承复合故障诊断方法[J].振动与冲击,2024,43(14):189-200+283.
ZHEN Dong, SUN Heming, FENG Guojin, et al. Zero-shot compound fault diagnosis method for rolling bearings based on envelope spectrum semantic construction [J]. Journal of Vibration and Shock, 2024, 43(14): 189-200+283.
[5] Lin Y H, Chang L. An online transfer learning framework for time-varying distribution data prediction[J]. IEEE Transactions on Industrial Electronics, 2021, 69(6): 6278-6287.
[6] Xu J, Liang S, Ding X, et al. A zero-shot fault semantics learning model for compound fault diagnosis[J]. Expert Systems with Applications, 2023, 221: 119642.
[7] 刘潇波.基于深度学习的风电机组传动链故障智能诊断[D].华北电力大学(北京),2022.
LIU Xiaobo. Intelligent fault diagnosis of wind turbine drivetrain based on deep learning [D]. North China Electric Power University (Beijing), 2022.
[8] 赵健程,冯良骏,岳嘉祺,等.从零样本学习理论模型到工业应用——动机、演变与挑战[J].控制与决策,2024,39(09):2833-2857.
Zhao Jiancheng, Feng Liangjun, Yue Jiaqi, et al. From Zero-Shot Learning Theoretical Models to Industrial Applications: Motivation, Evolution, and Challenges[J]. Control and Decision, 2024, 39(09): 2833-2857.
[9] Feng L, Zhao C. Fault description based attribute transfer for zero-sample industrial fault diagnosis[J]. IEEE Transactions on Industrial Informatics, 2020, 17(3): 1852-1862.
[10] 赵晓平,吕凯扬,邵凡,等.基于属性描述的零样本滚动轴承故障诊断[J].振动与冲击,2022,41(15):105-115
ZHAO Xiaoping, LU Kaiyang, SHAO Fan, et al. Zero-shot fault diagnosis of rolling bearings based on attribute description [J]. Journal of Vibration and Shock, 2022, 41(15): 105-115.
[11] Lv H, Chen J, Pan T, et al. Hybrid attribute conditional adversarial denoising autoencoder for zero-shot classification of mechanical intelligent fault diagnosis[J]. Applied Soft Computing, 2020, 95: 106577.
[12] Xu J, Zhou L, Zhao W, et al. Zero-shot learning for compound fault diagnosis of bearings[J]. Expert Systems with Applications, 2022, 190: 116197.
[13] 南玲博,陈帝伊,张润强,等.基于语义嵌入空间的离心泵未知复合故障识别[J].振动与冲击,2024,43(13):61-69+77.
NAN Lingbo, CHEN Diyi, ZHANG Runqiang, et al. Unknown compound fault identification of centrifugal pumps based on semantic embedding space [J]. Journal of Vibration and Shock, 2024, 43(13): 61-69+77.
[14] Yang B, Sun H. A zero-shot learning fault diagnosis method of rolling bearing based on extended semantic information under unknown conditions[J]. Journal of the Brazilian Society of Mechanical Sciences and Engineering, 2023, 45(1): 35.
[15] Gu J, Peng Y, Lu H, et al. A novel fault diagnosis method of rotating machinery via VMD, CWT and improved CNN[J]. Measurement, 2022, 200: 111635.
[16] Xiang C, Ren Z, Shi P, et al. Data‐Driven Fault Diagnosis for Rolling Bearing Based on DIT‐FFT and XGBoost[J]. Complexity, 2021, 2021(1): 4941966.
[17] 周龙.基于零样本学习的轴承复合故障诊断方法研究[D].合肥工业大学,2022.
ZHOU Long. Research on compound fault diagnosis of bearings based on zero-shot learning [D]. Hefei University of Technology, 2022.
[18] 魏聪聪.基于GWO-FCM和零样本学习的输油泵故障诊断模型自学习方法研究[D].北京化工大学,2023.
WEI Congcong. Research on self-learning method of fault diagnosis model for oil transfer pumps based on GWO-FCM and zero-shot learning [D]. Beijing University of Chemical Technology, 2023.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
PDF(2175 KB)
