基于CNN-MSLSTM的航空发动机滑油监测方法研究

doi:10.3969/j.issn.1000-1158.2021.02.16

摘要
图/表
参考文献(17)
相关文章 (15)

全文: PDF (2286 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要针对传统的数据特征提取方法难以提取ECT滑油监测数据有效特征的缺陷,提出了一种基于卷积神经网络(convolutional neural network,CNN)和多尺度长短期记忆(multi-scales long short-term memory,MSLSTM)神经网络的双通道网络模型CNN-MSLSTM。将多尺度学习融入LSTM中,CNN和MSLSTM作为两个通道,并行学习数据在空间维度和时间维度的特征,通过注意力机制融合,利用softmax分类器输出发动机的磨损状态。实验结果表明:3尺度的CNN-MSLSTM对ECT数据样本的分类准确率达到98%,F1评分达到98.62%,单组数据的测量时间仅为0.2036ms,总体性能优于单一的CNN和LSTM网络。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	马敏
	王涛

关键词 ：计量学;滑油监测;磨损状态;航空发动机, 电容层析成像;卷积神经网络;长短期记忆

Abstract：Aiming at the defect of traditional data feature extraction method that is difficult to extract the effective features of ECT oil monitoring data,a two-channel network model CNN-MSLSTM based on convolutional neural network (CNN) and multi-scale long-short-term memory (MSLSTM) neural network was proposed. The multi-scale learning was integrated into LSTM. CNN and MSLSTM were used as two channels to learn the characteristics of data in spatial dimension and time dimension. Through the attention mechanism fusion, the wear state of the engine was output by using softmax classifier. The experimental results showed that the classification accuracy of the 3 scale CNN-MSLSTM for ECT data samples is 98%, the F1 score is 98.62%, and the measurement time of single data is only 0.2036ms. The overall performance is better than the single CNN and LSTM networks.

Key words： metrology oil monitoring wear state aeroengine ECT CNN LSTM

收稿日期: 2019-03-12 发布日期: 2021-02-18

PACS:	TB973
	TG115.28

基金资助:国家自然科学基金(61401466);中央高校基金项目(3122013C007);国家自然科学基金委员会与中国民用航空局联合资助项目(U1733119);民航科技资助项目(20150220)

通讯作者: 王涛 E-mail: 1610164135@qq.com

作者简介: 马敏(1971-),女,安徽霍邱人,中国民航大学教授,主要研究方向为无损检测技术及新型传感器。Email: mm5739@163.com

引用本文:

马敏,王涛. 基于CNN-MSLSTM的航空发动机滑油监测方法研究[J]. 计量学报, 2021, 42(2): 232-238.
MA Min,WANG Tao. Research on Monitoring Method of AeroengineLubricating Oil Based on CNN-MSLSTM. Acta Metrologica Sinica, 2021, 42(2): 232-238.

链接本文:

http://jlxb.china-csm.org:81/Jwk_jlxb/CN/10.3969/j.issn.1000-1158.2021.02.16 或 http://jlxb.china-csm.org:81/Jwk_jlxb/CN/Y2021/V42/I2/232

［1］左洪福. 发动机磨损状态检测与故障诊断技术［M］. 北京: 航空工业出版社, 1996.
［2］陈果. 航空器检测与诊断技术导论［M］. 北京: 中国民航出版社, 2007.
［3］ Anderson D P.磨粒图谱［M］.金元生, 杨其明,译. 北京: 机械工业出版社, 1987.
［4］李忠, 曾昭翔. 基于BP神经网络的磨损微粒智能识别［J］. 北方交通大学学报, 1998, 22(1): 86-91.
Li Z, Zeng Z X. Intelligent Recognition of Wear Particles Based on BP Neural Network ［J］. [WTBX][STBX]Journal of Northern Jiaotong University[STBZ][WTBZ], 1998, 22(1): 86-91.
［5］涂群章, 左洪福, 李艳军. 发动机磨损颗粒图像在线检测技术［J］. 内燃机工程, 2003, 24(5): 70-73.
Tu Q Z, Zuo H F, Li Y J. Online Detection Technology of Engine Wear Particle Image ［J］. [WTBX][STBX]Chinese Internal Combustion Engine Engineering[STBZ][WTBZ], 2003, 24(5): 70-73.
［6］李华强, 费逸伟. 基于Matlab聚类分析的磨粒分类识别研究［J］. 润滑与密封, 2005, (3): 84-85.
Li H Q, Fei Y W. Study on Classification and Recognition of Abrasive Particles Based on Matlab Cluster Analysis ［J］. [WTBX][STBX]Lubrication & Sealing[STBZ][WTBZ], 2005, (3): 84-85.
［7］刘国光. 基于改进支持向量机的磨粒铁谱识别的研究［J］. 润滑与密封, 2005, (3): 94-98.
Liu G G. Study on the Identification of Abrasive Iron Spectra Based on Improved Support Vector Machine ［J］. [WTBX][STBX]Lubrication Engineering[STBZ][WTBZ], 2005, (3): 94-98.
［8］曹愈远, 张建, 李艳军, 等. 基于模糊粗糙集和SVM的航空发动机故障诊断［J］. 振动.测试与诊断, 2017, 37(1): 169-173.
Cao Y Y, Zhang J, Li Y J, et al. Aeroengine Fault Diagnosis Based on Fuzzy Rough Set and SVM ［J］. [WTBX][STBX]Journal of Vibration, Measurement & Diagnosis[STBZ][WTBZ], 2017, 37(1): 169-173.
［9］吕伯平, 陈明华. 航空滑油检测技术［M］. 北京: 航空工业出版社, 2006:1-6.
［10］王涛, 薛薇, 吕淮北. 航空发动机控制系统传感器故障诊断仿真研究［J］. 计算机仿真. 2016,33(2): 56-60.
Wang T, Xue W, Lv H B. Simulation Research on Sensor Fault Diagnosis of Aeroengine Control System ［J］. [WTBX][STBX]Computer Simulation[STBZ][WTBZ], 2016,33(2): 56-60.
［11］ Stone K, Keller J M. Convolution Neural Network Approach forBuried Target Recognition in FL-LWIR imagery ［C］//International Society for Optics and Photonics. Detection and Sensing of Mines, Explosive Objects, and Obscured Targets XIX.Baltimore, MD(US),2014:907219.1-907219.15
［12］ Li F M, Chen K, Liu X H. 3D Face Reconstruction Based on Convolutional Neural Network ［C］//IEEE. 2017 10th International Conference on Intelligent Computation Technology and Automation(ICICTA). 2017:71-74.
［13］ Kappeler A, Yoo S, Dai Q, et al. Video super-resolution with convolutional neural networks ［J］. [WTBX][STBX]IEEE Trans on Computational Imaging[STBZ][WTBZ], 2016, (2): 109-122.
［14］Aggarwal H K , Mani M P, Jacob M. Model based image reconstruction using deep learned priors(MODL) ［C］//IEEE. 2018 IEEE 15th International Symposium on Biomedical(ISBI 2018). 2018:671-674.
［15］程淑红,张仕军,赵考鹏. 基于卷积神经网络的生物式水质监测方法［J］. 计量学报, 2019, 40(4): 721-727.
Cheng S H, Zhang S J, Zhao K P. Biological Water Quality Monitoring Method Based on Convolution Neural Network［J］. [WTBX][STBX]Acta Metrologica Sinica[STBZ][WTBZ], 2019, 40(4): 721-727.
［16］王金甲,周雅倩,郝智. 基于注意力模型的多传感器人类活动识别［J］. 计量学报, 2019, 40(6): 958-969.
Wang J J, Zhou Y Q, Hao Z. Multi-sensor Human Activity Recognition Based on Attention Model［J］. [WTBX][STBX]Acta Metrologica Sinica[STBZ][WTBZ], 2019, 40(6): 958-969.
［17］陈德运, 黄家定, 于晓洋. 基于前项补偿的12电极的ECT数据采集系统的设计［J］. 化工自动化及仪表, 2007, 33(1): 61-65.
Chen D Y, Huang J D, Yu X Y. Design of 12-electrode ECT data acquisition system based on pre-compensation ［J］. [WTBX][STBX]Chemical Automation and Instrumentation[STBZ][WTBZ], 2007, 33(1): 61-65.