基于孪生减元注意力网络的航空发动机故障诊断

王月; 赵明航; 刘雪云; 林琳; 钟诗胜

doi:10.13224/j.cnki.jasp.20210195

基于孪生减元注意力网络的航空发动机故障诊断

doi: 10.13224/j.cnki.jasp.20210195

王月¹,
赵明航^2,*, ,,
刘雪云²,
林琳¹,
钟诗胜^{1, 2}

1.
哈尔滨工业大学机电工程学院，哈尔滨 150001
2.
哈尔滨工业大学（威海）海洋工程学院，山东威海 264209

基金项目: 国家自然科学基金联合基金（U1933202）；山东省自然科学基金（ZR2020QE156）

详细信息

作者简介:
王月（1997-），女，工程师，硕士，研究方向为航空发动机健康管理与故障诊断

通讯作者:
赵明航（1991-），男，副教授，博士，研究方向为航空发动机状态监测与智能维护。 E-mail：zhaomh@hit.edu.cn

中图分类号: V263.6
计量
- 文章访问数: 816
- HTML浏览量: 273
- PDF量: 119
- 被引次数: 0
出版历程
- 收稿日期: 2021-04-26
- 网络出版日期: 2023-05-15

Aero-engine fault diagnosis based on Siamese reduced-neuron attention networks

WANG Yue¹,
ZHAO Minghang^{2,*
, ,},
LIU Xueyun²,
LIN Lin¹,
ZHONG Shisheng^{1, 2}

1.
School of Mechatronics Engineering，Harbin Institute of Technology，Harbin 150001，China
2.
School of Ocean Engineering， Harbin Institute of Technology at Weihai，Weihai Shandong 264209，China

摘要

摘要:
针对传统故障诊断方法在故障样本缺乏条件下容易遭遇过拟合，以及强噪声条件下微弱故障特征难以提取的问题，提出了一种基于孪生减元注意力网络的航空发动机故障诊断方法。根据孪生神经网络的原理，将训练样本集中的样本随机两两配对，使输入从样本变为样本对，实现样本量的扩增。在特征提取模块引入减元注意力机制。其中，注意力机制能够通过全局扫描，快速找到有用特征，并且抑制冗余特征，这与航空发动机微弱气路故障特征被噪声所淹没的情况吻合良好；减元操作可以降低模型的参数量，缓解过拟合现象。研究结果表明：该方法在某航空公司CFM56-5B/7B系列发动机的真实监测数据上取得了88.39%的平均准确率。
- 故障诊断 /
- 孪生神经网络 /
- 减元注意力 /
- 涡扇发动机 /
- 小样本
Abstract:
In view of the problems that traditional fault diagnosis methods are prone to over-fitting under the condition of insufficient fault samples, and the weak fault features are difficult to be extracted under strong noise conditions, an aero-engine fault diagnosis method based on Siamese reduced-neuron attention networks was proposed. According to the principle of Siamese neural network, pairwise coupling of the samples in the training dataset was conducted, so that the input was changed from samples to sample pairs, and the diversity of input was improved. A reduced-neuron attention mechanism was integrated into the feature extraction module. Among them, the attention mechanism can quickly find useful features through global scanning, and suppress redundant features, which was in good agreement with the situation where the weak gas path fault features of aero-engines were submerged by noise; the reduced-neuron operation can reduce the amount of parameters and alleviate overfitting. The results show that this method achieves an average accuracy of 88.39% on the real monitoring data of CMF56-5B/7B series engines of an airline.
- fault diagnosis /
- siamese neural network /
- reduced-neuron attention /
- turbofan engine /
- limited samples

HTML

图 1 孪生神经网络结构图

Figure 1. Architecture of a siamese neural network

下载: 全尺寸图片幻灯片

图 2 引入减元注意力机制的残差模块（RM-RNAM）

Figure 2. Residual module with reduced-neuron attention mechanism （RM-RNAM）

下载: 全尺寸图片幻灯片

图 3 孪生减元注意力网络结构图

Figure 3. Architecture of the siamese reduced-neuron attention network

下载: 全尺寸图片幻灯片

图 4 不同模型实验结果对比

Figure 4. Comparison of experimental results of different models

下载: 全尺寸图片幻灯片

图 5 训练过程中的损失曲线图

Figure 5. Loss curves during the training process

下载: 全尺寸图片幻灯片

图 6 测试数据上的混淆矩阵

Figure 6. Confusion matrices on the testing dataset

下载: 全尺寸图片幻灯片

表 1 航空发动机故障诊断数据集

Table 1. Aero-engine fault diagnosis set

故障编号	健康状态	训练集样本数量	测试集样本数量
1	正常状态	60	16
2	TAT故障	15	5
3	EGTI故障	15	5
4	VBV故障	15	5

下载: 导出CSV

表 2 训练集样本对和标签表

Table 2. Training set pair and label

样本1	样本2	样本对	样本对标签
正常状态	正常样本	（正常状态，正常状态）	1
正常状态	VBV故障	（正常状态，VBV故障）	0
正常状态	EGTI故障	（正常状态，EGTI故障）	0
正常状态	TAT故障	（正常状态，TAT故障）	0
VBV故障	VBV故障	（VBV故障，VBV故障）	1
VBV故障	EGTI故障	（VBV故障，EGTI故障）	0
VBV故障	TAT故障	（VBV故障，TAT故障）	0
EGTI故障	EGTI故障	（EGTI故障，EGTI故障）	1
EGTI故障	TAT故障	（EGTI故障，TAT故障）	0
TAT故障	TAT故障	（TAT故障，TAT故障）	1

下载: 导出CSV

表 3 不同模型的实验结果对比

Table 3. Comparison of experimental results of different models %

参数	模型
参数	KNN	MLP	SMLP	SAN	SRAN
平均准确率	64.52	69.68	82.58	72.26	88.39
标准差	7.63	3.87	4.83	5.98	3.87

下载: 导出CSV

参考文献(25)

[1]	崔建国,李国庆,蒋丽英,等. 基于深度自编码网络的航空发动机故障诊断[J]. 振动、测试与诊断,2021,41(1): 85-89. CUI Jianguo,LI Guoqing,JIANG Liying,et al. Aero-engine fault diagnosis based on deep self-coding network[J]. Journal of Vibration, Measurement & Diagnosis,2021,41(1): 85-89. (in Chinese)
[2]	LI Huihui,GOU Linfeng,ZHENG Hua,et al. Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy[J]. International Journal of Aerospace Engineering,2021,2021: 8868875.1-8868875.20.
[3]	李红贤,韩延,吴敬涛,等. 基于ICA包络增强MEMD的滚动轴承故障诊断[J]. 航空动力学报,2021,36(2): 405-412. LI Hongxian,HAN Yan,WU Jingtao,et al. Rolling bearing fault diagnosis based on MEMD with ICA envelop enhancement[J]. Journal of Aerospace Power,2021,36(2): 405-412. (in Chinese)
[4]	HUANG Xianghua. Sensor fault diagnosis and reconstruction of engine control system based on autoassociative neural network[J]. Chinese Journal of Aeronautics,2004,17(1): 23-27. doi: 10.1016/S1000-9361(11)60198-2
[5]	罗辉. 基于深度特征的民航发动机气路异常检测方法研究[D]. 哈尔滨: 哈尔滨工业大学, 2020. LUO Hui. Research on the gas path anomaly detection of civil aero-engine based on deep features[J]. Harbin: Harbin Institute of Technology, 2020.
[6]	张建军,张利,穆海芳,等. 基于改进粒子群优化BP网络的发动机故障诊断方法[J]. 农业机械学报,2011,42(1): 198-203. ZHANG Jianjun,ZHANG Li,MU Haifang,et al. Improved BP-neural network of the particle swarm optimization in the research no engine fault diagnosis[J]. Transactions of the Chinese Society for Agricultural Machinery,2011,42(1): 198-203. (in Chinese)
[7]	滕伟,韩琛,赵立,等. 基于改进粒子滤波的重型燃气轮机跳机故障预测[J]. 中国机械工程,2021,32(2): 188-194. doi: 10.3969/j.issn.1004-132X.2021.02.009 TENG Wei,HAN Chen,ZHAO Li,et al. Tripping fault prediction of heavy-duty gas turbines based on improved particle filter[J]. China Mechanical Engineering,2021,32(2): 188-194. (in Chinese) doi: 10.3969/j.issn.1004-132X.2021.02.009
[8]	ZHAO Minghang,ZHONG Shisheng,FU Xuyun,et al. Deep residual shrinkage networks for fault diagnosis[J]. IEEE Transactions on Industrial Informatics,2020,16(7): 4681-4690. doi: 10.1109/TII.2019.2943898
[9]	许梦阳,黄金泉,鲁峰. 航空发动机气路故障诊断的SANNWA-PF算法[J]. 航空动力学报,2017,32(10): 2516-2525. XU Mengyang,HUANG Jinquan,LU Feng. SANNWA-PF algorithm of aero-engine gas path fault diagnosis[J]. Journal of Aerospace Power,2017,32(10): 2516-2525. (in Chinese)
[10]	鲜倪军. 基于ABC-BP神经网络的航空发动机故障诊断方法[J]. 装备制造技术,2018,281(5): 173-175, 178. doi: 10.3969/j.issn.1672-545X.2018.05.048 XIAN Nijun. Aero-engine fault diagnosis method based on ABC-BP neural network[J]. Equipment Manufacturing Technology,2018,281(5): 173-175, 178. (in Chinese) doi: 10.3969/j.issn.1672-545X.2018.05.048
[11]	GE Chiru,DU Qian,SUN Weiwei,et al. Deep residual network-based fusion framework for hyperspectral and lidar data[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2021,14(1): 2458-2472.
[12]	CHEN Renxiang,HUANG Xin,YANG Lixia. Intelligent fault diagnosis method of planetary gearboxes based on convolution neural network and discrete wavelet transform[J]. Computers in Industry,2019,106(1): 48-59.
[13]	李洪伟,马中原,谢镇波. 基于集成神经网络的航空发动机故障诊断研究[J]. 新技术新工艺,2018,368(8): 63-66. LI Hongwei,MA Zhongyuan,XIE Zhenbo. Research on fault diagnosis of aero-engine based on integrated neural network[J]. New Technology & New Process,2018,368(8): 63-66. (in Chinese)
[14]	付宇,殷逸冰,冒慧杰,等. 基于静电监测和神经网络的航发典型故障诊断[J]. 计算机集成制造系统,2019,25(11): 2852-2862. FU Yu,YIN Yibing,MAO Huijie,et al. Fault diagnosis of aero-engine gas path based on electrostatic monitoring and neural network[J]. Computer Integrated Manufacturing Systems,2019,25(11): 2852-2862. (in Chinese)
[15]	ZHANG Tianci,CHEN Jinglong,Xie Jingsong,et al. SASLN: Signals augmented self-taught learning networks for mechanical fault diagnosis under small sample condition[J]. IEEE Transactions on Instrumentation and Measurement,2020,70(1): 1-11.
[16]	YU Ding,MA Liang,MA Jian,et al. A generative adversarial network-based intelligent fault diagnosis method for rotating machinery under small sample size conditions[J]. IEEE Access,2019,7(1): 149736-149749.
[17]	付松,钟诗胜,林琳,等. 基于迁移学习的民航发动机小样本故障诊断[J]. 计算机集成制造系统,2021,27(12): 1-19. FU Song,ZHONG Shisheng,LIN Lin,et al. Gas turbine fault diagnosis method under small sample based on transfer learning[J]. Computer Integrated Manufacturing Systems,2021,27(12): 1-19. (in Chinese)
[18]	庞梦洋,索中英,郑万泽,等. 基于RS-CART决策树的航空发动机小样本故障诊断[J]. 航空动力学报,2020,35(7): 1559-1568. PANG Mengyang,SUO Zhongying,ZHENG Wanze,et al. Small sample fault diagnosis of aeroengine based on RS-CART decision tree[J]. Journal of Aerospace Power,2020,35(7): 1559-1568. (in Chinese)
[19]	陈子兆,矫文成,孙慧贤,等. 基于改进深度置信网络的故障诊断方法[J]. 探测与控制学报,2020,42(4): 98-105. CHEN Zizhao,JIAO Wencheng,SUN Huixian,et al. Fault diagnosis method based on improved deep belief network[J]. Journal of Detection & Control,2020,42(4): 98-105. (in Chinese)
[20]	DONG Xun,GAO Hongli,GUO Liang,et al. Deep cost adaptive convolutional network: a classification method for imbalanced mechanical data[J]. IEEE Access,2020,8(1): 71486-71496.
[21]	BROMLEY J,BENTZ J W,BOTTOU L,et al. Signature verification using a “siamese” time delay neural network[J]. International Journal of Pattern Recognition and Artificial Intelligence,1993,7(4): 669-688. doi: 10.1142/S0218001493000339
[22]	王彬文,吕帅帅,杨宇. 基于能量图谱和孪生网络的导波损伤诊断方法[J]. 振动、测试与诊断,2021,41(1): 182-189, 208. WANG Binwen,LÜ Shuaishuai,YANG Yu. Guided-wave based damage diagnosing method with energy spectrum and siamese network[J]. Journal of Vibration, Measurement & Diagnosis,2021,41(1): 182-189, 208. (in Chinese)
[23]	王韵涛,盛庆华,郭晨洁,等. 基于多重孪生网络的表面缺陷识别方法[J]. 计算机应用,2020,40(增刊2): 225-229. WANG Yuntao,SHENG Qinghua,GUO Chenjie,et al. Surface defect recognition method based on multiple siamese neural network[J]. Journal of Computer Applications,2020,40(Suppl.2): 225-229. (in Chinese)
[24]	BUI T,RIBEIRO L,PONTI M,et al. Compact descriptors for sketch-based image retrieval using a triplet loss convolutional neural network[J]. Computer Vision and Image Understanding,2017,164(1): 27-37.
[25]	孔子迁,邓蕾,汤宝平,等. 基于时频融合和注意力机制的深度学习行星齿轮箱故障诊断方法[J]. 仪器仪表学报,2019,40(6): 221-227. KONG Ziqian,DENG Lei,TANG Baoping,et al. Fault diagnosis of planetary gearbox based on deep learning with time-frequency fusion and attention mechanism[J]. Chinese Journal of Scientific Instrument,2019,40(6): 221-227. (in Chinese)