基于改进ECANet-TCN和迁移学习的轴承剩余寿命预测

王焱1,2,丁华1,2,孙晓春1,2,李莉1,2,刘泽平2,3,楚寒驰1

振动与冲击 ›› 2023, Vol. 42 ›› Issue (21) : 149-159.

PDF(3149 KB)
PDF(3149 KB)
振动与冲击 ›› 2023, Vol. 42 ›› Issue (21) : 149-159.
论文

基于改进ECANet-TCN和迁移学习的轴承剩余寿命预测

  • 王焱1,2,丁华1,2,孙晓春1,2,李莉1,2,刘泽平2,3,楚寒驰1
作者信息 +

Bearing residual life prediction based on improved ECANet-TCN and transfer learning

  • WANG Yan1,2, DING Hua1,2, SUN Xiaochun1,2, LI Li1,2, LIU Zeping2,3, CHU Hanchi1
Author information +
文章历史 +

摘要

针对轴承运行工况不同、有效数据少、数据无标签、预测准确度低等问题,提出一种基于改进时间卷积网络的迁移学习轴承寿命预测模型,将模型在源域上学习的寿命预测知识迁移到目标域,可用小样本无标签数据训练出迁移模型。首先,采用有效通道注意力模块对源域数据特征重新标定;其次,使用时间卷积网络(temporal convolutional network, TCN)学习特征信息,并训练出最优源域模型;最后,利用源域数据、源域模型和目标域数据训练出迁移模型,迁移模型可以对不同设备不同工况信号进行剩余寿命预测。在IEEE PHM Challenge 2012轴承全寿命数据集和XJTU-SY 滚动轴承加速寿命数据集上开展对比实验,结果表明,该方法可以更好地挖掘轴承内在退化趋势,有效提高剩余使用寿命预测精度,对比现有流行预测方法预测误差降低40.1%~77.8%,证明了该方法在不同设备不同工况条件下剩余寿命预测的有效性和可行性。

Abstract

In order to solve the problems of different operating conditions, less effective data, no labels, and low prediction accuracy, a life prediction model of transfer learning bearing based on an improved time convolution network is proposed. The life prediction knowledge learned by the model from the source domain is transferred to the target domain, and the transfer model can be trained with small samples of unlabeled data. Firstly, the efficient channel attention network is adopted to recalibrate the data features of the source domain; Secondly, the temporal convolutional network (TCN) is used to learn the feature information and train the optimal source domain model; Finally, the migration model is trained by source domain data, source domain model and target domain data. The transfer model can predict the remaining life of signals from different equipment and different working conditions. Comparative experiments were carried out on the IEEE PHM Challenge 2012 bearing life data set and the XJTU-SY rolling bearing accelerated life data set. The results show that this method can better mine the internal degradation trend of bearings, and effectively improve the prediction accuracy of remaining useful life. Compared with the existing popular prediction methods, the prediction error has been reduced by 40.1%~77.8%. This method is proven to be effective and feasible in predicting the residual life of different equipment and different working conditions.

关键词

轴承 / 剩余寿命预测 / ECANet / 时间卷积网络 / 迁移学习

Key words

bearing / prediction of residual life / ECANet / time convolution network / transfer learning

引用本文

导出引用
王焱1,2,丁华1,2,孙晓春1,2,李莉1,2,刘泽平2,3,楚寒驰1. 基于改进ECANet-TCN和迁移学习的轴承剩余寿命预测[J]. 振动与冲击, 2023, 42(21): 149-159
WANG Yan1,2, DING Hua1,2, SUN Xiaochun1,2, LI Li1,2, LIU Zeping2,3, CHU Hanchi1. Bearing residual life prediction based on improved ECANet-TCN and transfer learning[J]. Journal of Vibration and Shock, 2023, 42(21): 149-159

参考文献

[1]  Ding H, Yang L, Cheng Z, et al. A remaining useful life prediction method for bearing based on deep neural networks[J]. Measurement, 2021, 172: 108878.
[2] 赵志宏,李晴,杨绍普等. 基于BiLSTM与注意力机制的剩余寿命预测研究[J]. 振动与冲击, 2022, 41(6): 44-50.
Zhao Zhi-hong,Li Qing,Yang Shao-pu, et al. Remaining useful life prediction based on BiLSTM and attention mechanism[J]. Journal of vibration and shock, 2022, 41(6): 44-50.
[3] 李海浪,邹益胜,曾大懿,等. 一种基于特征聚类和评价的轴承寿命预测新方法[J]. 振动与冲击, 2022, 41(5): 141-150.
Li Hai-lang, Zou Yi-sheng, Zeng Da-yi, et al. A new method of bearing life prediction based on feature clustering and evaluation[J].  Journal of vibration and shock, 2022, 41(5): 141-150.
[4] 张继冬, 邹益胜, 邓佳林,等. 基于全卷积层神经网络的轴承剩余寿命预测[J]. 中国机械工程, 2019, 30(18): 2231-2235.
Zhang Ji-dong,Zou Yi-sheng,Deng Jia-lin, et al. Bearing Remaining Life Prediction Based on Full Convolutional Layer Neural Networks[J]. China mechanical engineering, 2019, 30(18): 2231-2235.
[5] 刘文广, 杨兰柱, 秦波. LSTM网络在采煤机摇臂齿轮箱寿命预测中的应用[J]. 机电工程技术, 2020, 49(11): 168-170.
Liu Wen-guang,Yang Lan-zhu,Qin Bo . Application of LSTM Network in Life Prediction of Shearer Rocker Gearbox[J]. Electrical engineering technology, 2020, 49(11): 168-170.
[6] Ding H, Yang L, Yang Z. A predictive maintenance method for shearer key parts based on qualitative and quantitative analysis of monitoring data[J]. IEEE Access, 2019, 7: 108684-108702.
[7] 吴仁彪, 赵婷, 屈景怡. 基于深度SE-DenseNet的航班延误预测模型[J]. 电子与信息学报, 2019, 41(6): 1510-1517.
Wu Ren-biao, Zhao Ting, Qu Jing-yi. Flight Delay Prediction Model Based on Deep SE-DenseNet[J]. Journal of Electronics & Information Technology, 2019, 41(6): 1510-1517.
[8] 曾大懿, 杨基宏, 邹益胜,等. 基于并行多通道卷积长短时记忆网络的轴承寿命预测方法[J]. 中国机械工程, 2020, 31(20): 2454-2462+2471.
Zeng Da-yi, Yang Ji-hong, Zou Yi-sheng, et al. Bearing Life Prediction Method Based on PMCCNN-LSTM[J].China mechanical engineering, 2020, 31(20): 2454-2462+2471.
[9] 陈佳鲜,毛文涛,刘京,等.基于深度时序特征迁移的轴承剩余寿命预测方法[J].控制与决策,2021,36(07):1699-1706.
Chen Jia-xian, Mao Wen-tao, Liu Jing, et al. Remaining useful life prediction of bearing based on deep temporal feature transfer[J]. Control and Decision,2021,36(07):1699-1706.
[10] 李亚平, 李素杰, 马波,等. 基于TCN的滚动轴承振动趋势与剩余寿命预测研究[J]. 现代制造工程, 2021(9): 1671-3133.
Li Ya-ping, Li Su-jie, Ma Bo, et al. Research on vibration trend and remaining life prediction of rolling bearings based on TCN[J]. Modern Manufacturing Engineering, 2021(9): 1671-3133.
[11] 庄福振,罗平,何清,等.迁移学习研究进展[J].软件 学报,2015,26(1):26-39.
Zhuang Fu-zhen,Luo Ping,He Qing,et al.Survey on transfer learning research[J].Journal of Software,2015, 26(1):26-39.
[12] 赵靖,杨绍普,李强,刘永强.一种残差注意力迁移学习方法及其在滚动轴承故障诊断中的应用[J]. 中国机械工程, 2022: 1-14.
Zhao Jing, Yang Shaopu, Li Qiang. Fault Diagnosis Method of a Rolling Bearing Under Variable Working Conditions Based on Feature Transfer Learning[J]. P China mechanical engineering, 2022: 1-14.
[13] 康守强, 胡明武, 王玉静,等. 基于特征迁移学习的变工况下滚动轴承故障诊断方法[J]. 中国电机工程学报, 2019(3): 764-772+955.
Kang Shou-qiang, Hu Ming-wu, Wang Yu-jing, et al. Fault Diagnosis Method of a Rolling Bearing Under Variable Working Conditions Based on Feature Transfer Learning[J]. Proceedings of the CSEE, 2019(3): 764-772+955.
[14] 杨胜康,孔宪光,王奇斌,等.基于多源域深度迁移学习的机械故障诊断[J].振动与冲击,2022,41(09):32-40.
Yang Sheng-kang, Kong Xian-guan, Wang Qi-bin, et al. Mechanical fault diagnosis based on multi-source domain deep transfer learning [J]. Journal of vibration and shock, 2022,41(09):32-40. 
[15] 王新刚, 韩凯忠, 王超,等. 基于迁移学习的轴承剩余使用寿命预测方法[J]. 东北大学学报:自然科学版, 2021, 42(5): 665-672.
Wang Xin-gang,Han Kai-zhong,Wang Chao, et al. A New Transfer Learning Method with Residual Attention and Its Applications on Rolling Bearing Fault Diagnosis [J]. Journal of Northeastern University, 2021, 42(5): 665-672.
[16] 刘飞,陈仁文,邢凯玲,等.基于迁移学习与深度残差网络的滚动轴承快速故障诊断算法[J].振动与冲击,2022,41(03):154-164.
Liu Fei,Chen Ren-wen,Xing Kai-ling, et al. Fast fault diagnosis algorithm for rolling bearing based on transfer learning and deep residual network [J]. Journal of vibration and shock, 2022,41(03):154-164.
[17] Q. Wang, B. Wu, P. Zhu, et al. ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks [C] ∥2020 IEEE Conference on Computer Vision and Pattern Recognition. Seattle: IEEE 2020.
[18] Bai S, Kolter J Z, Klotun V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[J]. ArXiv preprint ArXiv. 1803.01271 ,2018.
[19] 庄福振,罗平,何清,史忠植.迁移学习研究进展[J].软件学报,2015,26(01):26-39.
Zhuang FZ, Luo P, He Q, Shi ZZ. Survey on transfer learning research. Journal of Software, 2015,26(1):26-39.
[20] Hw K ,  Zhang X ,  Ren S , et al. Identity Mappings in Deep Residual Networks[J]. Springer, Cham, 2016: 630-645.
[21] IEEE P.H.M. Data challenge [Online]. Available:. https://www.femto-st.fr/f/d/IEEEP.H.M.2012-Challenge-Details.pdf. 2012.
[22] Wang Biao, Lei Ya-guo, Li Nai-peng, et al. A hybrid prognostics approach for estimating remaining useful life of rolling element bearings[J]. IEEE Transactions on Reliability, 2018, 69(1): 401-412.
[23] Wang Y, Deng L, Zheng L, et al. Temporal convolutional network with soft thresholding and attention mechanism for machinery prognostics[J]. Journal of Manufacturing Systems, 2021, 60: 512-526.
[24] Li X, Zhuang W, Ma H, et al. Data alignments in machinery remaining useful life prediction using deep adversarial neural networks[J]. Knowledge-Based Systems, 2020, 197: 105843.

PDF(3149 KB)

642

Accesses

0

Citation

Detail

段落导航
相关文章

/