Page 287 - 《振动工程学报》2026年第2期
P. 287
第 2 期 向 玲,等:基于多尺度残差动态域适应网络的不同工况下转子故障诊断方法 603
行短时傅里叶变换,将得到的二维时频域图像作为 deep learning method[J]. Applied Sciences, 2023, 13( 3) :
模型的输入,利用所提出的多尺度残差网络对数据 1327.
进行特征提取,并在全连接层通过动态分布域适应 [6] 李泽东,李志农,陶俊勇,等. 基于全矢谱-深度置信网络
方法对齐源域和目标域的分布差异。运用转子试验 的转子故障诊断方法研究 [J]. 兵器装备工程学报,2022,
台采集的故障数据对所提方法性能进行了验证,试 43(1):48-54.
验结果表明: LI Zedong,LI Zhinong,TAO Junyong,et al. Rotor fault
(1)所提多尺度残差神经网络融合了多尺度卷 diagnosis based on full vector spectrum and deep belief
积和空间可分离卷积,提取到丰富的转子故障特征, network[J]. Journal of Ordnance Equipment Engineering,
增强了模型对于故障特征的感知能力; 2022,43(1):48-54.
(2)所提方法将动态域适应策略引入到多尺度 [7] KIM M,JUNG J H,KO J U,et al. Direct connection-based
残差网络中,动态评估迁移学习中边缘分布和条件
convolutional neural network (DC-CNN) for fault diagnosis
分布的相对重要性,有效地缩小源域和目标域之间
of rotor systems[J]. IEEE Access,2020,8:172043-172056.
的差异,对齐源域和目标域的分布特征,在不同的试
[8] ZHANG Q L,HE Q S,QIN J Y,et al. Application of fault
验 台 上, 进 行 跨 工 况 迁 移 试 验 的 诊 断 能 力 均 优 于
diagnosis method combining finite element method and trans-
CMMD、 DDC、 DCORAL、 DAN、 JAN 这 5 种 传 统 域
fer learning for insufficient turbine rotor fault samples[J].
适应方法,所提方法提升了跨域诊断精度。
Entropy,2023,25(3):414.
[9] 唐竞鹏,王红军,钟建琳,等. 基于 WDCNN-SVM 深度迁
参考文献:
移学习的燃气轮机转子故障诊断方法 [J]. 电子测量与仪器
学报,2021,35(11):115-123.
[1] JANSSENS O, SLAVKOVIKJ V, VERVISCH B, et al.
TANG Jingpeng,WANG Hongjun,ZHONG Jianlin,et al.
Convolutional neural network based fault detection for rotat-
Gas turbine rotor fault diagnosis method based on WDCNN-
ing machinery[J]. Journal of Sound and Vibration, 2016,
SVM deep transfer learning[J]. Journal of Electronic Measure-
377:331-345.
ment and Instrumentation,2021,35(11):115-123.
[2] 宫文峰,陈辉,张泽辉,等. 基于改进卷积神经网络的滚
[10] TZENG E,HOFFMAN J,ZHANG N,et al. Deep domain
动 轴 承 智 能 故 障 诊 断 研究 [J]. 振 动 工 程 学 报 , 2020,
confusion:maximizing for domain invariance[EB/OL]. arXiv
33(2):400-413.
preprint arXiv: 1412.3474, 2014.
GONG Wenfeng,CHEN Hui,ZHANG Zehui,et al. Intelli-
[11] LONG M,CAO Y,WANG J,et al. Learning transferable
gent fault diagnosis for rolling bearing based on improved
features with deep adaptation networks[EB/OL]. arXiv
convolutional neural network[J]. Journal of Vibration Engi-
preprint arXiv: 1502.0279,2015.
neering,2020,33(2):400-413.
[12] LONG M,WANG J M,DING G G,et al. Transfer feature
[3] 朱广贺,朱智强,袁逸萍. 基于改进 CNN 的旋转设备转子
learning with joint distribution adaptation[C]//Proceedings of
振动故障预测仿真 [J]. 计算机仿真,2023,40(6):533- IEEE International Conference on Computer Vision. Sydney,
537.
Australia:IEEE,2014:2200-2207.
ZHU Guanghe,ZHU Zhiqiang,YUAN Yiping. Simulation [13] SUN B C,FENG J S,SAENKO K. Return of of frustrat-
of rotor vibration fault prediction for rotating equipment based ingly easy domain adaptation[J]. Proceedings of the AAAI
on improved CNN[J]. Computer Simulation,2023,40(6): Conference on Artificial Intelligence,2016,30(1):2058-
533-537. 2065.
[4] 万周,何俊增,姜东,等. 基于参数优化 SDP 分析的转子 [14] XU Y J, LIU J Z, WAN Z, et al. Rotor fault diagnosis
故障诊断方法 [J]. 振动与冲击,2023,42(1):81-88. using domain-adversarial neural network with time-frequency
WAN Zhou,HE Junzeng,JIANG Dong,et al. Rotor fault analysis[J]. Machines,2022,10(8):610.
diagnosis method based on parametric optimization SDP anal- [15] LIU S C,WANG H J,TANG J P,et al. Research on fault
ysis[J]. Journal of Vibration and Shock,2023,42(1):81- diagnosis of gas turbine rotor based on adversarial discrimina-
88. tive domain adaption transfer learning[J]. Measurement,
[5] WANG C Y,ZHENG Z L,GUO D,et al. An experimental 2022,196:111174.
setup to detect the crack fault of asymmetric rotors based on a [16] 朱琰,黄敏,王小静,等. 基于深度迁移学习的滑动轴承-

