作者：莫仁鹏, 司小胜, 李天梅, 朱旭, 胡昌华

作者单位：火箭军工程大学导弹工程学院，陕西 西安 710025

关键词：剩余寿命预测;多粗粒度;注意力;轴承

摘要：

考虑到轴承的振动信号往往分布在多个时间尺度上，该文提出一种基于多粗粒度与注意力网络的轴承剩余寿命（RUL）预测方法。首先采用多尺度粗粒度操作处理轴承的原始振动信号，从而获得蕴含更丰富退化信息的多尺度信号，在网络中以多尺度池化层来实现多尺度粗粒度操作；其次，基于大步幅卷积等网络层对多尺度信号进行深层特征提取、压缩、融合；此外，在网络中引入改进的卷积注意力模块为深层特征进行重标定，自适应地为不同通道和不同空间分配最佳权重；最后，将经注意力加权后的特征输入到前馈神经网络中映射得到RUL值。通过PRONOSTIA轴承数据进行实验分析，实验结果验证所提方法的有效性与优越性。

Remaining life prediction for bearing based on multiple coarse-grained and attention network
MO Renpeng, SI Xiaosheng, LI Tianmei, ZHU Xu, HU Changhua
School of Missile Engineering, Rocket Force Engineering University, Xi’an 710025, China
Abstract: Considering that the vibration signals of bearings are often distributed on multiple time scales, this paper proposes a bearing remaining useful life (RUL) prediction method based on multiple coarse-grained and attention networks. Firstly, multi-scale coarse-grained operations are used to process the original vibration signals of the bearing, to obtain multi-scale signals with richer degradation information and multi-scale pooling layers are used in the network to achieve multi-scale coarse-grained operations. Secondly, deep feature extraction, compression, and fusion of multi-scale signals are conducted based on network layers such as large-stride convolution. In addition, an improved convolution attention module is introduced into the network to recalibrate deep features and adaptively assign the best weights to different channels and different spaces. Finally, the attention-weighted features are input into the feedforward neural network to achieve the RUL prediction. The experimental analysis is carried out through PRONOSTIA bearing data, and the experimental results verify the effectiveness and superiority of the proposed method.
Keywords: remaining life prediction;multiple coarse-grained;attention;bearing
2021, 47(10):1-6  收稿日期: 2021-07-09;收到修改稿日期: 2021-08-10
基金项目: 国家自然科学基金项目（61773386，62073336）；国家自然科学优秀青年基金项目（61922089）
作者简介: 莫仁鹏（1997-），男，湖南邵阳县人，硕士研究生，专业方向为剩余寿命预测
参考文献
[1] 任子强, 司小胜, 胡昌华, 等. 融合多源数据的非线性退化建模与剩余寿命预测[J]. 中国测试, 2020, 46(2): 1-8
[2] 高旭东, 胡昌华, 杜党波. 线性FBM过程随机退化设备剩余寿命自适应预测[J]. 中国测试, 2021, 47(1): 111-119
[3] 张鹏, 胡昌华, 白灿, 等. 考虑随机效应的两阶段退化系统剩余寿命预测方法[J]. 中国测试, 2019, 45(1): 1-7
[4] 李天梅, 司小胜, 刘翔, 等. 大数据下数模联动的随机退化设备剩余寿命预测技术[J/OL]. 自动化学报: 1-23[2021-07-08]. https://doi.org/10.16383/j.aas.c201068.
[5] AMBADEKAR P K, CHOUDHARI C M. CNN based tool monitoring system to predict life of cutting tool[J]. SN Applied Sciences, 2020, 2(5): 1-11
[6] DEUTSCH J, HE D. Using deep learning-based approach to predict remaining useful life of rotating components[J]. IEEE Transactions on Systems, Man, and Cybernetics:Systems, 2017, 48(1): 11-20
[7] HINCHI A Z, TKIOUAT M. Rolling element bearing remaining useful life estimation based on a convolutional long-short-term memory network[J]. Procedia Computer Science, 2018, 127: 123-132
[8] PENG D, WANG H, LIU Z, et al. Multibranch and multiscale CNN for fault diagnosis of wheelset bearings under strong noise and variable load condition[J]. IEEE Transactions on Industrial Informatics, 2020, 16(7): 4949-4960
[9] JIANG G, HE H, YAN J, et al. Multiscale convolutional neural networks for fault diagnosis of wind turbine gearbox[J]. IEEE Transactions on Industrial Electronics, 2018, 66(4): 3196-3207
[10] 张志强, 杨清宇. 一种用于机械故障特征提取的多尺度稀疏滤波网络[J/OL]. 控制与决策: 1-10[2021-07-08]. https://doi.org/10.13195/j.kzyjc.2020.1418.
[11] 吴静然, 丁恩杰, 崔冉, 等. 采用多尺度注意力机制的旋转机械故障诊断方法[J]. 西安交通大学学报, 2020, 54(2): 51-58
[12] WOO S, PARK J, LEE J Y, et al. Cbam: Convolutional block attention module[C]//Proceedings of the European conference on computer vision (ECCV), 2018: 3-19.
[13] NECTOUX P, GOURIVEAU R, MEDJAHER K, et al. PRONOSTIA: An experimental platform for bearings accelerated degradation tests[C]//IEEE International Conference on Prognostics and Health Management, PHM'12. IEEE Catalog Number: CPF12PHM-CDR, 2012: 1-8.
[14] REN L, CUI J, SUN Y, et al. Multi-bearing remaining useful life collaborative prediction: A deep learning approach[J]. Journal of Manufacturing Systems, 2017, 43: 248-256
[15] GUO L, LEI Y, LI N, et al. Deep convolution feature learning for health indicator construction of bearings [C]//2017 Prognostics and System Health Management Conference (PHM-Harbin). IEEE, 2017: 1-6.