基于多尺度知识蒸馏与增量学习的滚动轴承故障诊断方法

摘要
图/表
参考文献(24)
相关文章 (15)

全文: PDF (1584 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要为了缓解单任务轴承故障诊断方法在不同工况诊断时产生的灾难性遗忘问题，提出一种基于多尺度知识蒸馏与增量学习(multi-scale knowledge distillation and continual learning, CL-MSKD)的滚动轴承故障诊断方法。以一维卷积神经网络作为CL-MSKD主要框架，余弦归一化层作为多任务共享的分类器，通过标签与特征两个尺度的知识蒸馏实现模型知识的保存与传递。CL-MSKD能够以一个统一结构的网络模型对在不同工况下的轴承故障进行诊断，通过知识压缩方法不断地学习和保存知识，最终缓解增量阶段产生的灾难性遗忘问题，提升跨工况场景下轴承故障诊断性能。实验表明，CL-MSKD能够有效缓解灾难性遗忘并保持良好的诊断效果。在任务环境差异较大的情况下，准确率指标仍能到达97.09%，与其他增量方法相比稳定性更好，精度更高。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	夏逸飞1
	2
	皋军1
	邵星1
	王翠香1

关键词 ：增量学习, 知识蒸馏, 卷积神经网络, 轴承故障诊断, 共享分类器

Abstract：In order to alleviate the catastrophic forgetting problem caused by the single-task bearing fault diagnosis method under different working conditions, a rolling bearing fault diagnosis method based on multi-scale knowledge distillation and continual learning (CL-MSKD) is proposed. The one-dimensional convolutional neural network is used as the main framework of CL-MSKD, and the cosine normalization layer is used as a multi-task shared classifier. The model knowledge is preserved and transmitted through the knowledge distillation of label and feature scales. CL-MSKD can diagnose bearing faults under different working conditions with a unified structure network model, continuously learn and save knowledge through knowledge compression method, and finally alleviate the catastrophic forgetting problem in the incremental stage, and improve the bearing fault diagnosis performance under cross-working conditions. The experiment show that CL-MSKD can effectively alleviate catastrophic amnesia and maintain good diagnostic effect. In the case of large differences in task environments, the accuracy index can still reach 97.09%, which is better stability and higher precision than other incremental methods.

Key words： continual learning knowledge distillation convolutional neural network bearing fault diagnosis shared classifier

收稿日期: 2023-09-22 出版日期: 2024-06-28

引用本文:

夏逸飞1,2，皋军1，邵星1，王翠香1. 基于多尺度知识蒸馏与增量学习的滚动轴承故障诊断方法[J]. 振动与冲击, 2024, 43(12): 276-285.
XIA Yifei1,2, GAO Jun1, SHAO Xing1, WANG Cuixiang1. A rolling bearing fault diagnosis method based on multi-scale knowledge distillation and continual learning. JOURNAL OF VIBRATION AND SHOCK, 2024, 43(12): 276-285.

链接本文:

http://jvs.sjtu.edu.cn/CN/ 或 http://jvs.sjtu.edu.cn/CN/Y2024/V43/I12/276

[1] Li J, Liu Y, Li Q. Intelligent fault diagnosis of rolling bearings under imbalanced data conditions using attention-based deep learning method[J]. Measurement, 2022, 189: 110500. [2] Lv Y, Zhao W, Zhao Z, et al. Vibration signal-based early fault prognosis: Status quo and applications[J]. Advanced Engineering Informatics, 2022, 52: 101609. [3] 金江涛,许子非,李春等.基于卷积双向长短期记忆网络与混沌理论的滚动轴承故障诊断[J].振动与冲击,2022,41(17):160-169. Jin Jiangtao, Xu Zifei, Li Chun et al. Fault diagnosis of rolling bearing based on Convolutional bidirectional long Short Term Memory Network and Chaos Theory [J]. Vibration and shock, 2022, 9 (17) : 160-169. [4] 程建刚,毕凤荣,张立鹏等.基于多重注意力-卷积神经网络-双向门控循环单元的机械故障诊断方法研究[J].内燃机工程,2021,42(04):77-83+92. Cheng Jiangang, Bi Fengrong, Zhang Lipeng et al. Research on Mechanical fault diagnosis based on Multiple attention-Convolutional neural Networks-bidirectional gated cycle units [J]. Internal combustion engine engineering, 2021, and (4) : 77-83 + 92. [5] Yu J B. Evolutionary manifold regularized stacked denoising autoencoders for gearbox fault diagnosis[J]. Knowledge-Based Systems, 2019, 178: 111-122. [6] Gao D, Zhu Y, Ren Z, et al. A novel weak fault diagnosis method for rolling bearings based on LSTM considering quasi-periodicity[J]. Knowledge-Based Systems, 2021, 231: 107413. [7] 丁承君,冯玉伯,王曼娜.基于变分模态分解与深度卷积神经网络的滚动轴承故障诊断[J].振动与冲击,2021,40(02):287-296. Ding Chengjun, Feng Yubo, WANG Manna. Fault diagnosis of rolling bearing based on variational mode decomposition and Deep Convolutional Neural Network [J]. Vibration and shock, 2021, 40 (02) : 287-296.. [8] Wang H, Liu Z, Peng D, et al. Attention-guided joint learning CNN with noise robustness for bearing fault diagnosis and vibration signal denoising[J]. ISA Transactions, 2022, 128: 470-484. [9] Yan S, Xie J, He X. Der: Dynamically expandable representation for class Continual learning[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2021: 3014-3023. [10] Luo S, Huang X, Wang Y, et al. Transfer learning based on improved stacked autoencoder for bearing fault diagnosis[J]. Knowledge-Based Systems, 2022, 256: 109846. [11] Yu W, Zhao C. Broad convolutional neural network based industrial process fault diagnosis with Continual learning capability[J]. IEEE Transactions on Industrial Electronics, 2019, 67(6): 5081-5091. [12] Tian S, Li W, Ning X, et al. Continuous transfer of neural network representational similarity for incremental learning[J]. Neurocomputing, 2023, 545: 126300. [13] Fu Z, Wang Z, Xu X, et al. Semantic alignment with self-supervision for class incremental learning[J]. Knowledge-Based Systems, 2023: 111114. [14] Chen Z, Deng L, Gou J, et al. Building and road detection from remote sensing images based on weights adaptive multi-teacher collaborative distillation using a fused knowledge[J]. International Journal of Applied Earth Observation and Geoinformation, 2023, 124: 103522. [15] Du Y, Niu J, Wang Y, et al. Multi-Stage Knowledge Distillation for Sequential Recommendation with Interest Knowledge[J]. Available at SSRN 4432610. [16] Lee H, Park Y, Seo H, et al. Self-knowledge distillation via dropout[J]. Computer Vision and Image Understanding, 2023, 233: 103720. [17] Yang Z, Long J, Zi Y, et al. Continual novelty identification from initially one-class learning to unknown abnormality classification[J]. IEEE Transactions on Industrial Electronics, 2021, 69(7): 7394-7404. [18] Fu Y, Cao H, Chen X, et al. Task-incremental broad learning system for multi-component intelligent fault diagnosis of machinery[J]. Knowledge-Based Systems, 2022, 246: 108730.Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//International conference on machine learning. pmlr, 2015: 448-456Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//International conference on machine learning. pmlr, 2015: 448-456. [19] Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//International conference on machine learning. pmlr, 2015: 448-456. [20] Kim S, Kim D, Cho M, et al. Proxy anchor loss for deep metric learning[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 3238-3247. [21] Douillard A, Cord M, Ollion C, et al. Podnet: Pooled outputs distillation for small-tasks Continual learning[C]//Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XX 16. Springer International Publishing, 2020: 86-102.Rebuffi S A, Kolesnikov A, Sperl G, et al. icarl: Continual classifier and representation learning[C]//Proceedings of the IEEE conference on Computer Vision and Pattern Recognition. 2017: 2001-2010. [22] Rebuffi S A,Kolesnikov A,Sperl G,et al.icarl:Incremental classifier and representation learning[C]. Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, Honolulu,USA,2017:2001-2010. [23] Li K, Xiong M, Li F, et al. A novel fault diagnosis algorithm for rotating machinery based on a sparsity and neighborhood preserving deep extreme learning machine[J]. Neurocomputing, 2019, 350: 261-270. [24] Li Z, Hoiem D. Learning without forgetting[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 40(12): 2935-2947.