多源融合的STFT-IncepNext航空发动机轴承故障诊断方法

万安平; 张华; 张今; 蒋俊杰; 王景霖; 单添敏

doi:10.13224/j.cnki.jasp.20240049

多源融合的STFT-IncepNext航空发动机轴承故障诊断方法

doi: 10.13224/j.cnki.jasp.20240049

万安平^1,,
张华^{1, 2},
张今^3,*, ,,
蒋俊杰¹,
王景霖⁴,
单添敏⁴

1.
浙大城市学院机电系，杭州 310015
2.
浙江工业大学机械工程学院，杭州 310023
3.
浙江大学机械工程学院，杭州 310027
4.
中国航空工业集团有限公司上海航空测控技术研究所故障诊断与健康管理技术航空科技重点实验室，上海 201601

基金项目: 国家自然科学基金（52372420）；航空科学基金（20183333001）

详细信息

作者简介:
万安平（1983-），男，教授，博士，主要从事复杂装备健康管理及维修决策等研究。E-mail：wanap@zucc.edu.cn

通讯作者:
张今（1995-），男，博士生，主要从事高端装备流程挖掘、数字孪生技术等研究。E-mail：cicada@zju.edu.cn

中图分类号: V263.6；TH133.33
计量
- 文章访问数: 757
- HTML浏览量: 326
- PDF量: 52
- 被引次数: 0
出版历程
- 收稿日期: 2024-01-22
- 网络出版日期: 2025-07-15

Fault diagnosis of aeroengine bearings using multi-source fusion STFT-IncepNext method

WAN Anping^1
,,
ZHANG Hua^{1, 2},
ZHANG Jin^{3,*
, ,},
JIANG Junjie¹,
WANG Jinglin⁴,
SHAN Tianmin⁴

1.
Department of Mechanical Engineering，Hangzhou City University，Hangzhou 310015，China
2.
School of Mechanical Engineering，Zhejiang University of Technology，Hangzhou 310023，China
3.
School of Mechanical Engineering，Zhejiang University，Hangzhou 310027，China
4.
Aviation Key Laboratory of Science and Technology on Fault Diagnosis and Health Management，Shanghai Aeronautical Measurement and Control Technology Research Institute，Aviation Industry Corporation of China，Limited，Shanghai 201601，China

摘要

摘要:
针对单传感器信息难以满足复杂工况下航空发动机轴承故障状态的稳定监测问题，提出一种STFT-IncepNext的航空发动机轴承故障诊断模型。首先，将同一时间窗口内的异位传感器数据进行拼接，以补充轴承在不同空间下的振动信息。其次，为了捕捉振动信号中故障成分的瞬时变化，利用短时傅里叶变换（short-time Fourier transform, STFT）将多传感器振动信号转换为时频图。最后，通过轻量化的IncepNext网络来提取时频图中蕴含的故障信息全局特征，由分类器给出识别的故障类别。实验结果表明：所提方法能够有效地增强信号的故障特征，提高轴承在不同状态下振动特征的辨识度。在特定的实验条件下，该方法实现了航空发动机轴承振动故障诊断准确率达100%，相较于STFT-EdgeNeXt、STFT-ResNeXt、STFT-ShuffleNet、STFT-ResNet18均取得了更好的性能，为航空发动机轴承故障诊断提供一种可行方法。
- 故障诊断 /
- 多传感器信息融合 /
- 短时傅里叶变换 /
- 滚动轴承 /
- 视频特征
Abstract:
To address the challenge of reliably monitoring aviation engine bearing faults under complex operating conditions with limited information from a single sensor, the STFT-IncepNext model was proposed for bearing fault diagnosis. Initially, sensor data from different positions within the same time window were concatenated to enrich the vibration information of bearings across various spatial dimensions. Subsequently, to capture the transient changes of fault components in the vibration signal, the short-time Fourier transform (STFT) was applied to convert multi-sensor vibration signals into time-frequency representations. Finally, a lightweight IncepNext network extracted global features of fault information embedded in the time-frequency representations, and a Softmax classifier identified the fault category. Experimental results demonstrated that the proposed approach effectively enhanced signal fault characteristics, and improved the discriminability of vibration features under various bearing states. Under specific experimental conditions, the method achieved an accuracy rate of 100% for diagnosing vibration faults in aeroengine bearings. Compared with STFT-EdgeNeXt, STFT-ResNeXt, STFT-ShuffleNet, and STFT-ResNet18, the method exhibited superior performance, providing a feasible approach for diagnosing faults in aeroengine bearings.
- fault diagnosis /
- multi-sensor information fusion /
- short-time Fourier transform /
- rolling bearings /
- time-frequency features

HTML

图 1 STFT-IncepNext网络模型

Figure 1. Network model of STFT-IncepNext

下载: 全尺寸图片幻灯片

图 2 卷积编码器模块

Figure 2. Convolution encoder model

下载: 全尺寸图片幻灯片

图 3 SDTA编码器模块

Figure 3. SDTA encoder module

下载: 全尺寸图片幻灯片

图 4 多尺度分支卷积模块

Figure 4. Multi-scale branching convolution module

下载: 全尺寸图片幻灯片

图 5 直升机行星传动仿真实验平台

Figure 5. Helicopter planetary gearboxes transmission platform

下载: 全尺寸图片幻灯片

图 6 不同故障类型时频图

Figure 6. Time frequency diagram of different fault types

下载: 全尺寸图片幻灯片

图 7 STFT-IncepNext模型训练流程

Figure 7. STFT-IncepNext model training process

下载: 全尺寸图片幻灯片

图 8 不同传感器集合对模型准确率的影响

Figure 8. Influence of different sensor sets on model accuracy

下载: 全尺寸图片幻灯片

图 9 训练过程中损失函数和准确率

Figure 9. Loss function and accuracy during training process

下载: 全尺寸图片幻灯片

图 10 原始数据与分类层的特征可视化结果

Figure 10. Results of feature visualization of raw data and classification layer

下载: 全尺寸图片幻灯片

图 11 模型准确率差值

Figure 11. Model accuracy difference

下载: 全尺寸图片幻灯片

表 1 传感器布置点位

Table 1. Sensor placement points

传感器类型	采集仪对应位置	安装位置
转速	NI_1	驱动电动机输出轴
加速度1	NI_2（轴向）	平行齿轮箱中间轴轴承端盖
加速度2	NI_3（水平径向）
加速度3	NI_4（垂直径向）
加速度4	NI_5	平行齿轮箱右侧箱体（靠上轴承）
加速度5	NI_6	承重台面（靠平行齿轮箱）

下载: 导出CSV

表 2 不同工况下的样本配置

Table 2. Sample configuration under different operating conditions

工况编号	转频/Hz	负载/（N·m）	故障模式	样本量	样本点数
1	34	22	5	200000	1000
2	34	44	5	200000	1000
3	57	22	5	200000	1000
4	57	44	5	200000	1000

下载: 导出CSV

表 3 直升机传动系统实验台数据集

Table 3. Helicopter transmission system test bench dataset

标签	故障类型	工况模式	样本量
0	正常	4	800
1	滚动体	4	800
2	联合	4	800
3	内圈	4	800
4	外圈	4	800

下载: 导出CSV

表 4 不同诊断方法的准确率对比

Table 4. Comparison of accuracy between different diagnostic methods

图像生成方法	模型方法	单传感器准确率/%	多传感器准确率/%	模型参数数量/10⁶
连续小波变换	CWT-IncepNext	79.83	87.50	1.46
	CWT-EdgeNeXt	73.44	84.55	1.16
	CWT-ResNet18	61.29	78.65	1.18
	CWT-ShuffleNet	70.63	73.96	1.26
	CWT-ResNeXt	70.49	86.11	22.99
格拉姆角场	GAF-IncepNext	73.26	82.99	1.46
	GAF-EdgeNeXt	54.86	67.36	1.16
	GAF-ResNet18	63.37	78.99	1.18
	GAF-ShuffleNet	56.08	72.74	1.26
	GAF-ResNeXt	66.84	79.34	22.99
短时傅里叶变换	STFT-IncepNext	98.33	100	1.46
	STFT-EdgeNeXt	98.17	99.83	1.16
	STFT-ResNet18	98.00	99.67	1.18
	STFT-ShuffleNet	78.83	91.33	1.26
	STFT-ResNeXt	97.50	99.83	22.99

下载: 导出CSV

参考文献(25)

[1]	WANG Xing, ZHANG Han, DU Zhaohui. Multiscale noise reduction attention network for aeroengine bearing fault diagnosis[J]. IEEE Transactions on Instrumentation and Measurement, 2023, 72: 3513810.
[2]	张向阳, 陈果, 郝腾飞, 等. 基于机匣信号的滚动轴承故障卷积神经网络诊断方法[J]. 航空动力学报, 2019, 34(12): 2729-2737. ZHANG Xiangyang, CHEN Guo, HAO Tengfei, et al. Convolutional neural network diagnosis method of rolling bearing fault based on casing signal[J]. Journal of Aerospace Power, 2019, 34(12): 2729-2737. (in Chinese ZHANG Xiangyang, CHEN Guo, HAO Tengfei, et al. Convolutional neural network diagnosis method of rolling bearing fault based on casing signal[J]. Journal of Aerospace Power, 2019, 34(12): 2729-2737. (in Chinese)
[3]	LI Yunfeng, LI Ming, YAN Zhong, et al. Application of life cycle of aeroengine mainshaft bearing based on digital twin[J]. Processes, 2023, 11(6): 1768. doi: 10.3390/pr11061768
[4]	梁兴, 罗远兴, 邓飞, 等. 多传感器数据下基于MFDFA-BP的离心泵空化致振故障分析[J]. 振动与冲击, 2022, 41(17): 238-243, 281. LIANG Xing, LUO Yuanxing, DENG Fei, et al. Cavitation induced vibration fault analysis of centrifugal pump based on MFDFA-BP under multi-sensor data[J]. Journal of Vibration and Shock, 2022, 41(17): 238-243, 281. (in Chinese LIANG Xing, LUO Yuanxing, DENG Fei, et al. Cavitation induced vibration fault analysis of centrifugal pump based on MFDFA-BP under multi-sensor data[J]. Journal of Vibration and Shock, 2022, 41(17): 238-243, 281. (in Chinese)
[5]	秦丽, 孟召丽, 吴永亮. 基于振动传感器性能退化数据的可靠性评估[J]. 传感器与微系统, 2009, 28(9): 53-55. QIN Li, MENG Zhaoli, WU Yongliang. Reliability assessment based on performance degradation data for vibration sensors[J]. Transducer and Microsystem Technologies, 2009, 28(9): 53-55. (in Chinese QIN Li, MENG Zhaoli, WU Yongliang. Reliability assessment based on performance degradation data for vibration sensors[J]. Transducer and Microsystem Technologies, 2009, 28(9): 53-55. (in Chinese)
[6]	郭盼盼, 张文斌, 崔奔, 等. 基于增强深度卷积神经网络的滚动轴承多工况故障诊断方法[J]. 振动工程学报, 2025, 38(1): 96-108. GUO Panpan, ZHANG Wenbin, CUI Ben, et al. Multi-condition fault diagnosis method of rolling bearing based on enhanced deep convolutional neural network[J]. Journal of Vibration Engineering, 2025, 38(1): 96-108. (in Chinese GUO Panpan, ZHANG Wenbin, CUI Ben, et al. Multi-condition fault diagnosis method of rolling bearing based on enhanced deep convolutional neural network[J]. Journal of Vibration Engineering, 2025, 38(1): 96-108. (in Chinese)
[7]	邢蓉, 高丙朋, 侯培浩, 等. 基于MSCNN与STFT的滚动轴承故障诊断研究[J]. 机械传动, 2020, 44(7): 41-45, 58. XING Rong, GAO Bingpeng, HOU Peihao, et al. Research of fault diagnosis of rolling bearing based on MSCNN and STFT[J]. Journal of Mechanical Transmission, 2020, 44(7): 41-45, 58. (in Chinese XING Rong, GAO Bingpeng, HOU Peihao, et al. Research of fault diagnosis of rolling bearing based on MSCNN and STFT[J]. Journal of Mechanical Transmission, 2020, 44(7): 41-45, 58. (in Chinese)
[8]	梁睿君, 冉文丰, 余传粮, 等. 基于CWT-CNN的齿轮箱运行故障状态识别[J]. 航空动力学报, 2021, 36(12): 2465-2473. LIANG Ruijun, RAN Wenfeng, YU Chuanliang, et al. Recognition of gearbox operation fault state based on CWT-CNN[J]. Journal of Aerospace Power, 2021, 36(12): 2465-2473. (in Chinese LIANG Ruijun, RAN Wenfeng, YU Chuanliang, et al. Recognition of gearbox operation fault state based on CWT-CNN[J]. Journal of Aerospace Power, 2021, 36(12): 2465-2473. (in Chinese)
[9]	赵小强, 郭海科. 多特征融合的滚动轴承故障诊断[J]. 农业工程学报, 2023, 39(13): 80-88. ZHAO Xiaoqiang, GUO Haike. Fault diagnosis of rolling bearings using multi-feature fusion[J]. Transactions of the Chinese Society of Agricultural Engineering, 2023, 39(13): 80-88. (in Chinese doi: 10.11975/j.issn.1002-6819.202302116 ZHAO Xiaoqiang, GUO Haike. Fault diagnosis of rolling bearings using multi-feature fusion[J]. Transactions of the Chinese Society of Agricultural Engineering, 2023, 39(13): 80-88. (in Chinese) doi: 10.11975/j.issn.1002-6819.202302116
[10]	段晓燕, 焦孟萱, 雷春丽, 等. 基于MTF-MSMCNN的小样本滚动轴承故障诊断方法[J]. 航空动力学报, 2024, 39(1): 20230517. DUAN Xiaoyan, JIAO Mengxuan, LEI Chunli, et al. A rolling bearing fault diagnosis method based on MTF-MSMCNN with small sample[J]. Journal of Aerospace Power, 2024, 39(1): 20230517. (in Chinese DUAN Xiaoyan, JIAO Mengxuan, LEI Chunli, et al. A rolling bearing fault diagnosis method based on MTF-MSMCNN with small sample[J]. Journal of Aerospace Power, 2024, 39(1): 20230517. (in Chinese)
[11]	郭俊锋, 谭宝宏, 王智明. 基于MDAM-GhostCNN的滚动轴承故障诊断方法[J]. 北京航空航天大学学报, 2025, 51(4): 1172-1184. GUO Junfeng, TAN Baohong, WANG Zhiming. Fault diagnosis method of rolling bearing based on MDAM-GhostCNN[J]. Journal of Beijing University of Aeronautics and Astronautics, 2025, 51(4): 1172-1184. (in Chinese GUO Junfeng, TAN Baohong, WANG Zhiming. Fault diagnosis method of rolling bearing based on MDAM-GhostCNN[J]. Journal of Beijing University of Aeronautics and Astronautics, 2025, 51(4): 1172-1184. (in Chinese)
[12]	刘仓, 童靳于, 包家汉, 等. 基于多传感器两级特征融合的滚动轴承故障诊断方法[J]. 振动与冲击, 2022, 41(8): 199-207, 259. LIU Cang, TONG Jinyu, BAO Jiahan, et al. A rolling bearing fault diagnosis method based on multi-sensor two-stage feature fusion[J]. Journal of Vibration and Shock, 2022, 41(8): 199-207, 259. (in Chinese LIU Cang, TONG Jinyu, BAO Jiahan, et al. A rolling bearing fault diagnosis method based on multi-sensor two-stage feature fusion[J]. Journal of Vibration and Shock, 2022, 41(8): 199-207, 259. (in Chinese)
[13]	GU Kai, ZHANG Yu, LIU Xiaobo, et al. DWT-LSTM-based fault diagnosis of rolling bearings with multi-sensors[J]. Electronics, 2021, 10(17): 2076. doi: 10.3390/electronics10172076
[14]	GUAN Yang, MENG Zong, SUN Dengyun, et al. 2MNet: Multi-sensor and multi-scale model toward accurate fault diagnosis of rolling bearing[J]. Reliability Engineering & System Safety, 2021, 216: 108017.
[15]	何艳. 基于语音特征分析的汉语方言辨识研究[D]. 江苏无锡: 江南大学, 2012. HE Yan. The research of Chinese dialect identification based on speech feature analysis[D]. Wuxi Jiangsu: Jiangnan University, 2012. (in Chinese HE Yan. The research of Chinese dialect identification based on speech feature analysis[D]. Wuxi Jiangsu: Jiangnan University, 2012. (in Chinese)
[16]	LI Jiashi, XIA Xin, LI Wei, et al. Next-ViT: next generation vision transformer for efficient deployment in realistic industrial scenarios[J]. ArXiv e-Prints, 2022: arXiv: 2207.05501.
[17]	戴洋毅, 何康, 瑚琦, 等. CNN-Transformer混合模型在计算机视觉领域的研究综述[J]. 建模与仿真, 2023(4): 3657-3672. DAI Yangyi, HE Kang, HU Qi, et al. Review of CNN-transformer hybrid model in computer vision[J]. Modeling and Simulation, 2023(4): 3657-3672. (in Chinese DAI Yangyi, HE Kang, HU Qi, et al. Review of CNN-transformer hybrid model in computer vision[J]. Modeling and Simulation, 2023(4): 3657-3672. (in Chinese)
[18]	MAAZ M, SHAKER A, CHOLAKKAL H, et al. EdgeNeXt: efficiently amalgamated CNN-transformer architecture for mobile vision applications[M]//Computer Vision: ECCV 2022 Workshops. Cham, US: Springer Nature Switzerland, 2023: 3-20.
[19]	樊翔翔, 项载毓, 孙瑞雪, 等. 基于小波时频分析和Inception-BiGRU模型的盾构滚刀偏磨故障诊断[J]. 振动与冲击, 2023, 42(15): 232-240. FAN Xiangxiang, XIANG Zaiyu, SUN Ruixue, et al. Fault diagnosis of TBM hob eccentric wear based on wavelet time-frequency analysis and Inception-BiGRU model[J]. Journal of Vibration and Shock, 2023, 42(15): 232-240. (in Chinese FAN Xiangxiang, XIANG Zaiyu, SUN Ruixue, et al. Fault diagnosis of TBM hob eccentric wear based on wavelet time-frequency analysis and Inception-BiGRU model[J]. Journal of Vibration and Shock, 2023, 42(15): 232-240. (in Chinese)
[20]	LI Xueyi, KONG Xiangwei, ZHANG Jiqiang, et al. A study on fault diagnosis of bearing pitting under different speed condition based on an improved inception capsule network[J]. Measurement, 2021, 181: 109656. doi: 10.1016/j.measurement.2021.109656
[21]	李业政. 航空发动机附件齿轮箱滚动轴承故障诊断方法研究[D]. 北京: 北京化工大学, 2022. LI Yezheng. Research on fault diagnosis methonds for rolling bearings in aero-engine[D]. Beijing: Beijing University of Chemical Technology, 2022. (in Chinese LI Yezheng. Research on fault diagnosis methonds for rolling bearings in aero-engine[D]. Beijing: Beijing University of Chemical Technology, 2022. (in Chinese)
[22]	张欢. 基于多源信息融合的轴承复合故障诊断技术研究[D]. 贵阳: 贵州大学, 2020. ZHANG Huan. Research on bearing composite fault diagnosis technology based on multi-source information fusion[D]. Guiyang: Guizhou University, 2020. (in Chinese ZHANG Huan. Research on bearing composite fault diagnosis technology based on multi-source information fusion[D]. Guiyang: Guizhou University, 2020. (in Chinese)
[23]	LIU Hengchang, YAO Dechen, YANG Jianwei, et al. Lightweight convolutional neural network and its application in rolling bearing fault diagnosis under variable working conditions[J]. Sensors, 2019, 19(22): 4827. doi: 10.3390/s19224827
[24]	LIU Yanlin, DOU Shuihai, DU Yanping, et al. Gearbox fault diagnosis based on gramian angular field and CSKD-ResNeXt[J]. Electronics, 2023, 12(11): 2475. doi: 10.3390/electronics12112475
[25]	WU Guoguo, JI Xuerong, YANG Guolai, et al. Signal-to-image: rolling bearing fault diagnosis using ResNet family deep-learning models[J]. Processes, 2023, 11(5): 1527. doi: 10.3390/pr11051527