基于多通道融合的齿轮箱故障诊断方法

doi:10.13229/j.cnki.jdxbgxb.20241377

摘要/Abstract

摘要：

针对单通道的振动信号在全面表征齿轮箱故障特征方面存在局限性的问题，提出了一种将多通道的振动信号变换为时频图后经过加权融合成RGB图像，再输入到诊断模型中的数据级融合方法MFTDCN。该方法能有效整合多个通道的信息，获得比单个振动信号更丰富的特征。首先，对多通道振动信号进行主成分分析（PCA）处理，降维成三个分量，用短时傅里叶变换（STFT）将3个分量变换为3个时频图像；然后，用Hoyer稀疏度对余弦相似度进行改进，将其应用于基于稀疏性和相似度的权重计算，利用得到的权重对三个时频图像进行加权处理，从而融合成RGB图像，作为诊断模型的输入。诊断模型部分为可变形卷积（DCN）与Transformer相结合，提取特征后进行分类。在东南大学的齿轮箱数据集和北京工业大学与北京交通大学联合发布的数据集上进行了相关实验验证，结果证实了该方法的有效性和可靠性。

关键词: 故障诊断, 数据级融合, 短时傅里叶变换, 余弦相似度, 可变形卷积, Transformer

Abstract:

To address the limitations of single-channel vibration signals in comprehensively characterizing gearbox fault features， this paper proposes a data-level fusion method called MFTDCN is proposed.This method effectively integrates information from multiple channels to obtain richer features compared to single vibration signals.The approach begins by applying Principal Component Analysis （PCA） to multi-channel vibration signals to reduce them into three components. These components are then transformed into three time-frequency images using Short-Time Fourier Transform （STFT）. An improved cosine similarity measure， based on Hoyer sparsity， is employed to calculate weights by combining sparsity and similarity metrics. These weights are used to perform weighted fusion of the three time-frequency images， resulting in a single RGB image， which is then used as input for the diagnostic model. The diagnostic model combines Deformable Convolution Networks （DCN） with Transformer architectures to extract features and classify faults. Experimental results on the gearbox data set of Southeast University and the data set jointly published by Beijing University of Technology and Beijing Jiaotong University confirm the effectiveness and reliability of the proposed method.

Key words: fault diagnosis, data-level fusion, short-time Fourier transform, cosine similarity, deformable convolution network, Transformer

中图分类号:

TH17

冯志刚,王颖,王宇. 基于多通道融合的齿轮箱故障诊断方法[J]. 吉林大学学报(工学版), 2026, 56(7): 1811-1824.

Zhi-gang FENG,Ying WANG,Yu WANG. Gear box fault diagnosis method based on multi-channel fusion[J]. Journal of Jilin University(Engineering and Technology Edition), 2026, 56(7): 1811-1824.

图/表 21

图1

图2

图3

图4

图5

表1

图6

表2

表3

表4

表5

图7

图8

图9

表6

图10

表 7

图11

图12

表8

图13

参考文献 25

[1]	Saufi S R, Ahmad Z A B, Leong M S, et al. Gearbox fault diagnosis using a deep learning model with limited data sample[J]. IEEE Transactions on Industrial Informatics, 2020, 16(10): 6263-6271.
[2]	侯召国, 王华伟, 熊明兰, 等.基于迁移学习与加权多通道融合的齿轮箱故障诊断[J]. 振动与冲击,2023, 42(9): 236-246.
	Hou Zhao-guo, Wang Hua-wei, Xiong Ming-lan,et al. Gearbox fault diagnosis based on transfer learning and weighted multi-channel fusion[J]. Journal of Vibration and Shock,2023,42(9):236-246.
[3]	She D M, Yang Z C, Duan Y D, et al. A meta transfer learning-driven few-shot fault diagnosis method for combine harvester gearboxes[J]. Computers and Electronics in Agriculture, 2024, 227: 109605.
[4]	王进花, 刘秦玮, 曹洁, 等.基于SCACGAN的小样本齿轮箱故障诊断[J]. 北京航空航天大学学报, 2026, 52(3): 713-723.
	Wang Jin-hua, Liu Qin-wei, CAO Jie, et al. Fault diagnosis of gearbox with small-sample based on SCACGAN[J].Journal of Beijing University of Aeronautics and Astronautics, 2026, 52(3): 713-723.
[5]	刘辉, 李阳, 侯一民. 用于轴承故障诊断任务的轻量化卷积网络[J]. 哈尔滨理工大学学报, 2024(4): 80-88.
	Liu Hui, Li Yang, Hou Yi-ming. Lightweight convolutional network for bearing fault diagnosis[J]. Journal of Harbin University of Science and Technology,2024(4): 80-88.
[6]	Lei Y G, Jia F, Lin J, et al. An intelligent fault diagnosis method using unsupervised feature learning towards mechanical big data[J].IEEE Transactions on Industrial Electronics, 2016, 63(5): 3137-3147.
[7]	Hu Q, Si X S, Zhang Q H, et al. A rotating machinery fault diagnosis method based on multi-scale dimensionless indicators and random forests[J]. Mechanical Systems and Signal Processing, 2020, 139: 106609.
[8]	石永芳, 徐庆宏, 姜宏, 等. 基于特征差异性学习卷积神经网络的齿轮箱故障诊断方法[J].机床与液压,2023,51(24):176-183.
	Shi Yong-fang, XU Qing-hong, JIANG Hong, et al.Gearbox fault diagnosis method based on feature difference learning convolutional neural network[J]. Machine Tool & Hydraulics,2023,51(24):176-183.
[9]	Zhao D Z, Cui L L, Liu D D. Bearing weak fault feature extraction under time-varying speed conditions based on frequency matching demodulation transform[J]. IEEE/ASME Transactions on Mechatronics, 2022, 28(3): 1627-1637.
[10]	Goyal D, Dhami S S, Pabla B S. Non-contact fault diagnosis of bearings in machine learning environment[J]. IEEE Sensors Journal,2020, 20(9): 4816-4823.
[11]	Zhang N N, Wu L F, Yang J, et al. Naive bayes bearing fault diagnosis based on enhanced independence of data[J]. Sensors, 2018, 18(2): 18020463.
[12]	Zhao X L, Jia M P, Liu Z. Semisupervised deep sparse auto-encoder with local and nonlocal information for intelligent fault diagnosis of rotating machinery[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 70: 1-13.
[13]	Wang X H, Meng R X, Wang G T, et al. The research on fault diagnosis of rolling bearing based on current signal CNN-SVM[J]. Measurement Science and Technology, 2023, 34(12): 125021.
[14]	Rezaeianjouybari B, Shang Y. Deep learning for prognostics and health management: state of the art, challenges, and opportunities[J]. Measurement, 2020, 163: 107929.
[15]	Zhu X X, Hou D N, Zhou P, et al. Rotor fault diagnosis using a convolutional neural network with symmetrized dot pattern images[J]. Measurement, 2019, 138: 526-535.
[16]	Wang S Y, Tian J Y, Liang P F, et al. Single and simultaneous fault diagnosis of gearbox via wavelet transform and improved deep residual network under imbalanced data[J]. Engineering Applications of Artificial Intelligence,2024, 133: 108146.
[17]	Li C Y, Hu Y H, Jiang J W, et al. Fault diagnosis of a marine power-generation diesel engine based on the Gramian angular field and a convolutional neural network[J].Journal of Zhejiang University-SCIENCE A, 2024, 25(6): 470-482.
[18]	Ding X X, He Q B. Energy-fluctuated multiscale feature learning with deep convnet for intelligent spindle bearing fault diagnosis[J]. IEEE Transactions on Instrumentation and Measurement, 2017, 66(8): 1926-1935.
[19]	Liu D D, Cui L L, Cheng W D. A review on deep learning in planetary gearbox health state recognition: methods, applications, and dataset publication[J]. Measurement Science and Technology,2023, 35(1): 012002.
[20]	Wen L L, Li X Y, Gao L. A transfer convolutional neural network for fault diagnosis based on ResNet-50[J]. Neural Computing and Applications, 2020, 32(10): 6111-6124.
[21]	Wen L, Li X Y, Gao L, et al. A new convolutional neural network-based data-driven fault diagnosis method[J]. IEEE Transactions on Industrial Electronics, 2017, 65(7): 5990-5998.
[22]	Zhang W, Li C H, Peng G L, et al. A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load[J]. Mechanical Systems and Signal Processing, 2018, 100: 439-453.
[23]	Xia M, Li T, Xu L, et al. Fault diagnosis for rotating machinery using multiple sensors and convolutional neural networks[J]. IEEE/ASME Transactions on Mechatronics, 2017, 23(1): 101-110.
[24]	Li H M, Huang J Y, Gao M J, et al. Multi-view information fusion fault diagnosis method based on attention mechanism and convolutional neural network[J]. Applied Sciences, 2022, 12(22): 11410.
[25]	Xie T L, Huang X F, Choi S K. Intelligent mechanical fault diagnosis using multisensor fusion and convolution neural network[J]. IEEE Transactions on Industrial Informatics, 2021, 18(5): 3213-3223.

相关文章 15

[1]	刘昕晖,陈晫群,吕岩. 基于深度学习的工业故障诊断综述[J]. 吉林大学学报(工学版), 2026, 56(7): 1759-1779.
[2]	曹洁,陈志峰,王进花,陈莉. 基于扩散模型和DenseNet的少样本齿轮箱故障诊断方法[J]. 吉林大学学报(工学版), 2026, 56(7): 1787-1797.
[3]	冯志刚,任梦媛,董冰,于明月. 基于多频带特征图和改进SqueezeNet的滚动轴承故障诊断[J]. 吉林大学学报(工学版), 2026, 56(1): 96-108.
[4]	于江波,翁剑成,林鹏飞,孙宇星,柴娇龙. 基于混合Transformer的对外客运枢纽抵站客流预测模型[J]. 吉林大学学报(工学版), 2025, 55(7): 2251-2259.
[5]	冯志刚,王首起,于明月. 基于变分模态提取及轻量级网络的滚动轴承故障诊断[J]. 吉林大学学报(工学版), 2025, 55(6): 1883-1891.
[6]	刘广文,谢欣月,付强,才华,王伟刚,马智勇. 基于时空模板焦点注意的Transformer目标跟踪算法[J]. 吉林大学学报(工学版), 2025, 55(3): 1037-1049.
[7]	王娜,崔月磊,李杨,王子从. 基于小波包对数能量图的滚动轴承故障诊断方法[J]. 吉林大学学报(工学版), 2025, 55(2): 494-502.
[8]	冯志刚,张志远,董冰,于明月. 基于改进VME算法和mRVM的滚动轴承小样本复合故障诊断[J]. 吉林大学学报(工学版), 2025, 55(12): 3831-3839.
[9]	王鹏,宋亚飞,王晓丹,路艳丽,向前. FATIDS：面向类不平衡样本的物联网入侵检测方法[J]. 吉林大学学报(工学版), 2025, 55(12): 3986-3999.
[10]	陈永,安卓奥博,张娇娇. 基于旋转自注意力改进Mask RCNN的桥梁裂缝检测方法[J]. 吉林大学学报(工学版), 2025, 55(11): 3660-3672.
[11]	胡宏宇,张争光,曲优,蔡沐雨,高菲,高镇海. 基于双分支和可变形卷积网络的驾驶员行为识别方法[J]. 吉林大学学报(工学版), 2025, 55(1): 93-104.
[12]	余萍,赵康,曹洁. 基于优化A-BiLSTM的滚动轴承故障诊断[J]. 吉林大学学报(工学版), 2024, 54(8): 2156-2166.
[13]	王长建,刘久明,张锦洲,李斌. 基于高速摄影技术的行星减速箱故障激光序列脉冲诊断方法[J]. 吉林大学学报(工学版), 2024, 54(7): 1869-1875.
[14]	蒋磊,王子其,崔振宇,常志勇,时小虎. 基于循环结构的视觉Transformer[J]. 吉林大学学报(工学版), 2024, 54(7): 2049-2056.
[15]	梁礼明,周珑颂,尹江,盛校棋. 融合多尺度Transformer的皮肤病变分割算法[J]. 吉林大学学报(工学版), 2024, 54(4): 1086-1098.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

轴承/齿轮箱	类型	训练集	测试集	标签
轴承数据	滚动体故障	210	90	0
	内圈故障	210	90	1
	外圈故障	210	90	2
	复合故障	210	90	3
	健康状态	210	90	4
齿轮数据	齿轮断齿	210	90	0
	齿轮缺口	210	90	1
	齿面磨损	210	90	2
	齿根裂纹	210	90	3
	健康状态	210	90	4

工况	Gray	GAF	MTF	STFT
均值	86.87	93.11	80.06	99.46
bearingset_20_0	82.16	96.16	74.24	100.00
bearingset_30_2	84.22	93.76	84.22	99.92
gearset_20_0	86.18	88.71	82.11	98.27
gearset_30_2	94.91	93.80	79.67	99.65

类型	工况 bearingset_30_2					工况gearset_30_2
类型	滚动体	内圈	外圈	复合	正常	缺口	断牙	裂纹	磨损	正常
分量1
分量2
分量3
RGB

工况	信息熵	方差权重	未改进的余弦相似度	能量权重	Hoyer稀疏度改进的余弦相似度
均值	97.88	98.09	98.90	98.22	99.46
bearingset_20_0	99.42	99.83	99.87	99.38	100.00
bearingset_30_2	99.17	99.63	100.00	99.55	99.92
gearset_20_0	94.02	93.94	96.24	95.08	98.28
gearset_30_2	98.89	98.94	99.48	98.86	99.65

工况	RGB-ResNet	Gray-LeNet-5	Raw-TICNN	Matrix-CNN	CAM_MCFCNN	MFTDCN
均值	97.00	93.81	99.12	97.62	96.89	99.46
bearingset_20_0	99.84	95.12	99.84	99.55	95.11	100.00
bearingset_30_2	99.55	99.28	97.60	99.94	94.89	99.92
gearset_20_0	92.27	89.33	99.05	95.76	98.67	98.27
gearset_30_2	96.32	91.52	99.97	95.23	98.88	99.65