采用双树复小波变换的可解释机械故障诊断方法

刘家阳; 潘元明; 李功准; 潘诗豪; 宋义哲; 余健平; 张龙

doi:10.7652/xjtuxb202608005

您当前的位置：

首页 >

文章列表页 >

采用双树复小波变换的可解释机械故障诊断方法

专题先进无损检测技术 | 更新时间：2026-07-28

- 采用双树复小波变换的可解释机械故障诊断方法
- Interpretable Mechanical Fault Diagnosis Method Based on Dual-Tree Complex Wavelet Transform
- 西安交通大学学报 2026年60卷第8期页码：56-67
- 作者机构：
  
  华东交通大学机电与车辆工程学院，330013，南昌
- 作者简介：
  
  作者简介：刘家阳（1994-），男，讲师，硕士生导师；
  张龙（通信作者），男，教授，博士生导师。
- 基金信息：
  
  江西省自然科学基金资助项目(20252BAC200081);国家自然科学基金资助项目(52565011);机械系统与振动全国重点实验室开放课题资助项目(MSV202508)
- DOI：10.7652/xjtuxb202608005
  中图分类号： TH133.3
- 收稿：2025-12-18，
  
  网络首发：2026-03-04，
  
  纸质出版：2026-08-10
- 稿件说明：
移动端阅览
刘家阳, 潘元明, 李功准, 等. 采用双树复小波变换的可解释机械故障诊断方法[J]. 西安交通大学学报, 2026,60(8):56-67.

LIU Jiayan, PAN Yuanming, LI Gongzhun, et al. Interpretable Mechanical Fault Diagnosis Method Based on Dual-Tree Complex Wavelet Transform[J]. Journal of Xi'an Jiaotong University, 2026, 60(8): 56-67.
刘家阳, 潘元明, 李功准, 等. 采用双树复小波变换的可解释机械故障诊断方法[J]. 西安交通大学学报, 2026,60(8):56-67. DOI： 10.7652/xjtuxb202608005.

LIU Jiayan, PAN Yuanming, LI Gongzhun, et al. Interpretable Mechanical Fault Diagnosis Method Based on Dual-Tree Complex Wavelet Transform[J]. Journal of Xi'an Jiaotong University, 2026, 60(8): 56-67. DOI： 10.7652/xjtuxb202608005.

摘要

针对卷积神经网络（convolutional neural network

CNN）在机械故障诊断中存在的决策过程不透明、缺乏物理可解释性的问题，提出一种基于双树复小波变换（dual-tree complex wavelet transform

DTCWT）的可解释智能诊断方法。将具有明确时频物理意义的Morlet复小波变换构建为可解释卷积层，替代传统CNN首层的随机初始化卷积核，形成DTCWT-CNN混合架构。DTCWT卷积层通过实部树与虚部树的并行结构同时提取信号的幅值与相位信息，并引入可学习的频带加权机制与软阈值去噪策略，实现对轴承故障冲击特征的自适应精准捕获。在凯斯西储大学轴承数据集以及帕德博恩大学轴承数据集上的实验表明，DTCWT-CNN在不同通道配置下的平均诊断准确率达到99%，显著优于其他对比方法。在南昌铁路局机车轴承数据集上的验证结果显示，所提出方法的平均准确率达到98%，同样表现出优异的诊断性能。通过对训练前后卷积核的累计频带分析，可视化揭示了模型自适应调整至故障特征频段的决策机制，验证了方法在增强诊断性能的同时实现了网络决策依据的物理可解释性。

Abstract

To address the opaque decision-making process and lack of physical interpretability of convolutional neural networks (CNNs) in mechanical fault diagnosis

an interpretable intelligent diagnosis method based on the dual-tree complex wavelet transform (DTCWT) was proposed. The Morlet complex wavelet transform

which carries clear time-frequency physical significance

was constructed as an interpretable convolutional layer

replacing the randomly initialized convolution kernels in the first layer of a traditional CNN

thus forming a DTCWT-CNN hybrid architecture. The DTCWT convolutional layer simultaneously extracted the amplitude and phase information of the signal through the parallel structure of real and imaginary trees

and incorporated a learnable frequency-band weighting mechanism and a soft-thresholding denoising strategy to achieve adaptive and precise capture of bearing fault impact features. Experiments were conducted on the Case Western Reserve University bearing dataset and the Universität Paderborn bearing dataset. The results show that the DTCWT-CNN achieves an average diagnostic accuracy of 99% under different channel configurations

which is significantly superior to other comparative methods. Validation was performed on the locomotive bearing dataset from the Nanchang Railway Bureau

and the results show that the average accuracy of the proposed method reaches 98% with excellent diagnostic performance. Through cumulative frequency-band analysis of convolution kernels before and after training

the model's decision-making mechanism of adaptively adjusting to fault-characteristic frequency bands is revealed visually

which verifies that the proposed method not only enhances diagnostic performance but also realizes physical interpretability for the network's decision basis.

关键词

Keywords

references

Zhang Long, Wang Jinbo, Xiao Qian, et al. A bearing fault diagnosis method with implementation of multiscale time-frequency feature fusion and bidirectional information interaction [J]. IEEE Sensors Journal, 2025,25（13）:23740-23756.

张龙，赵丽娟，王朝兵，等.自适应频域字典的机车轮对轴承稀疏诊断方法[J].铁道科学与工程学报，2023，20（4）：1456-1468.

Zhang Long, Zhao Lijuan, Wang Chaobing, et al. Adaptive frequency domain dictionary for sparse diagnosis of locomotive wheelset bearing[J]. Journal of Railway Science and Engineering, 2023, 20 (4): 14561468.

张龙，胡燕青，赵丽娟，等.采用递归图编码技术与残差网络的滚动轴承故障诊断[J].西安交通大学学报，2023，57（2）：110-120.

Zhang Long, Hu Yanqing, Zhao Lijuan, et al. Fault diagnosis of rolling bearings using recurrence plot coding technique and residual network[J]. Journal of Xi'an Jiaotong University, 2023, 57 (2): 110-120.

张龙，刘皓阳，张号，等.改进共空间模式与多源特征融合的轴承智能诊断方法[J].西安交通大学学报，2023，57（8）：127-137.

Zhang Long, Liu Haoyang, Zhang Hao, et al. Intelligent bearing fault diagnosis using modified common spatial pattern and multi-source feature fusion[J]. Journal of Xi'an Jiaotong University, 2023, 57 (8)：127-137.

张龙，肖逸文，周神赐，等.一种双阻尼小波赋能的可解释卷积神经网络在轴承故障诊断中的应用[J].西安交通大学学报，2025,59（3）:210-221.

Zhang Long, Xiao Yiwen, Zhou Shenci, et al. Application of interpretable convolutional neural network enabled by bi-damped wavelets in bearing fault diagnosis[J]. Journal of Xi'an Jiaotong University,2025,59（3）：210-221.

陈钱，陈康康，董兴建，等.一种面向机械设备故障诊断的可解释卷积神经网络[J].机械工程学报，2024，60（12）：65-76.

Chen Qian, Chen Kangkang, Dong Xingjian, et al. Interpretable convolutional neural network for mechanical equipment fault diagnosis[J]. Journal of Mechanical Engineering, 2024, 60 (12): 65-76.

Zhang Yu, Tiňo P, Leonardis A, et al. A survey on neural network interpretability [J]. IEEE Transactions on Emerging Topics in Computational Intelligence,2021,5（5）:726-742.

Brito L C, Susto G A, Brito J N, et al. An explainable artificial intelligence approach for unsupervised fault detection and diagnosis in rotating machinery [J].Mechanical Systems and Signal Processing,2022,163：108105.

Jiang Fei, Kuang Yicong, Li Tao, et al. Towards enhanced interpretability: a mechanism-driven domain adaptation model for bearing fault diagnosis across operating conditions [J].Mechanical Systems and Signal Processing,2025,225:112244.

Peng Ying, Shao Haidong, Xiao Yiming, et al. A systematic review on interpretability research of intelligent fault diagnosis models [J].Measurement Science and Technology,2025,36（1）:012009.

Selvaraju R R, Cogswell M, Das A, et al. GradCAM:visual explanations from deep networks via gradient-based localization [C]//2017 IEEE International Conference on Computer Vision（ICCV）.Piscataway, NJ, USA:IEEE,2017:618-626.

Han Guangjie, Chen Jianhang, Liu Li, et al. An interpretable CNN with wavelet group policy embedded for intelligent fault diagnosis [J].IEEE Transactions on Instrumentation and Measurement,2024,73:1-15.

Zhou Qianyu, Tang Jiong. An interpretable parallel spatial CNN-LSTM architecture for fault diagnosis in rotating machinery [J].IEEE Internet of Things Journal,2024,11（19）:31730-31744.

He Chao, Shi Hongmei, Si Jin, et al. Physics-informed interpretable wavelet weight initialization and balanced dynamic adaptive threshold for intelligent fault diagnosis of rolling bearings [J].Journal of Manufacturing Systems,2023,70:579-592.

Li Sinan, Li Tianfu, Sun Chuang, et al. WPConvNet: an interpretable wavelet packet kernel-constrained convolutional network for noise-robust fault diagnosis [J]. IEEE Transactions on Neural Networks and Learning Systems,2024,35（10）:14974-14988.

Selesnick I W, Baraniuk R G, Kingsbury N C.The dual-tree complex wavelet transform[J].IEEE Signal Processing Magazine,2005,22（6）:123-151.

Wang Yanxue, He Zhengjia, Zi Yangang. Enhancement of signal denoising and multiple fault signatures detecting in rotating machinery using dual-tree complex wavelet transform [J].Mechanical Systems and Signal Processing,2010,24（1）:119-137.

Lecun Y, Boser B, Denker J S, et al. Backpropagation applied to handwritten zip code recognition[J].Neural Computation,1989,1（4）:541-551.

Gu Xiaohui, Yang Shaopu, Liu Yongqiang, et al. Compound faults detection of the rolling element bearing based on the optimal complex Morlet wavelet filter[J].Proceedings of the Institution of Mechanical Engineers: Part C Journal of Mechanical Engineering Science,2018,232（10）:1786-1801.

Smith W A, Randall R B.Rolling element bearing diagnostics using the Case Western Reserve University data:a benchmark study [J].Mechanical Systems and Signal Processing,2015,64/65:100-131.

Zhao Dengfeng, Tian Chaoyang, Fu Zhijun, et al. Multi scale convolutional neural network combining BiLSTM and attention mechanism for bearing fault diagnosis under multiple working conditions [J].Scientific Reports,2025,15（1）:13035.

Zhong Jingshu, Zheng Yu, Ruan Chengtao, et al. M-IPISincNet:an explainable multi-source physics-informed neural network based on improved SincNet for rolling bearings fault diagnosis [J].Information Fusion,2025,115:102761.

Qi Jinshui, Mo Jiaqing, Niu Yasen, et al. Recognition of fiber optic vibration signals based on Laplace wavelet transform and deep learning [J].IEEJ Transactions on Electrical and Electronic Engineering,2024,19（6）：1026-1034.

Guo Lijin, Han Bintao, Huang Qilan. Bearing fault diagnosis based on improved Morlet wavelet transform and shallow residual neural network [J].Applied Sciences,2024,14（11）:4542.

Lessmeier C, Kimotho J K, Zimmer D, et al. Condition monitoring of bearing damage in electromechanical drive systems by using motor current signals of electric motors:a benchmark data set for data-driven classification [C]//PHM Society European Conference. Rochester, New York, USA: The Prognostics and Health Management Society,2016:1-17.

Antoni J.Fast computation of the kurtogram for the detection of transient faults[J].Mechanical Systems and Signal Processing,2007,21（1）:108-124.

Shang Zuogang, Zhao Zhibin, Yan Ruqiang. Denoising fault-aware wavelet network:a signal processing informed neural network for fault diagnosis [J].Chinese Journal of Mechanical Engineering,2023,36（1）:9.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

采用双树复小波变换的可解释机械故障诊断方法

电力变压器故障诊断多智能体大模型构建与应用

采用双小波字典的齿轮箱复合故障特征提取方法

一维卷积神经网络在机械故障特征提取中的可解释性研究

蜉蝣优化双通道网络在齿轮箱故障诊断中的应用