Vibration prediction of aeroengines based on enhanced SENet model
-
摘要:
为实时监测和预警航空发动机振动状态,基于气路及振动参数,提出一种使用改进的SENet(squeeze-and-excitation network)模型,对航空发动机近未来的振动进行预测。该研究相比以往采用的实验室模拟数据和仿真数据,使用了真实的QAR(quick access recorder)数据并进行随机采样,以求更能表征发动机振动和工作参数之间的关系。同时,不仅使用其他振动信号进行验证,还在其他型号的发动机上进行测试。结果表明:针对航空发动机的振动进行预测是可行的,SENet模型可以有效并实时追踪振动的突变和波动。此外,该方法对于其他振动信号和不同类型的发动机具有一定的适用性。而且相较于以往采用的其他经典的深度模型,SENet模型在振动的预测中能得到更小的误差。实验证明,相较于以往只使用振动这个单参数进行预测,并行使用与振动相关的多参数融合进行研究更能提高预测的准确性。
Abstract:In order to monitor the vibration status of aeroengines and acquire warning signals in real-time, an enhanced SENet (squeeze-and-excitation network) model was proposed based on gas path and vibration parameters. Compared with the previous research which used datasets generated from specific lab situations and simulation data, actual QAR (quick access recorder) data were adopted for random sampling of the datasets. This technique could characterize the real operation status and the interaction of parameters better in vibration systems. The results showed that it is possible to forecast the vibration of aeroengines, and the SENet model could effectively and timely track sudden changes and the fluctuation of vibration. In addition, the applicability of this method into other vibration parameters and different types of aeroengines was tested. Furthermore, compared with other classical learning algorithms , the SENet model may obtain a smaller error in vibration forecasting. At the same time, the experiments showed that compared with previous research only focusing on the vibration, using the fusion of multi parameters could improve the accuracy of the forecast.
-
表 1 神经网络模型参数
Table 1. Parameters of neural network model
层号 参数 其他 1 7×7,64 步幅:1,
填充:33×3,Max pool 步幅:2,
填充:12 $ \left( \begin{gathered} 3 \times 3,64 \\ 3 \times 3,64 \\ \end{gathered} \right) \times 2 $ 所有卷积层:
步幅:1,
填充:13 $ \left( \begin{gathered} 3 \times 3,128 \\ 3 \times 3,128 \\ \end{gathered} \right) \times 2 $ 第1个卷积层:
步幅:(1, 2),
填充:14 $ \left( \begin{gathered} 3 \times 3,256 \\ 3 \times 3,256 \\ \end{gathered} \right) \times 2 $ 第1个卷积层:
步幅:(1, 2),
填充:15 $ \left( \begin{gathered} 3 \times 3,512 \\ 3 \times 3,512 \\ \end{gathered} \right) \times 2 $ 第1个卷积层:
步幅:(1, 2),
填充:16 Global average pooling (1, 1) 7 FC 512×10 表 2 多模型损失对比
Table 2. Comparison of multi model loss
模型 测试集损失 SENet 0.0203 单参数SENet 2.5070 ResNet 0.0210 VggNet 0.0261 MLP 0.0243 -
[1] 李书明, 李世栋, 张莹. 航空发动机压气机性能衰退影响因子定量分析[J]. 科学技术与工程, 2015, 15(32): 74-78, 86.LI Shuming, LI Shidong, ZHANG Ying. Quantitative analysis of aircraft engine compressor performance deterioration impact factor[J]. Science Technology and Engineering, 2015, 15(32): 74-78, 86. (in Chinese) [2] 民航资源网. 2018年1-6月中国民航安全信息统计报告[EB/OL]. [2022-03-06]. http: //news.carnoc.com/list/461/461183.html. [3] WANG H. A survey of maintenance policies of deteriorating systems[J]. European Journal of Operational Research,2002,139(3): 469-489. doi: 10.1016/S0377-2217(01)00197-7 [4] CHO D I,PARLAR M. A survey of maintenance models for multi-unit systems[J]. European Journal of Operational Research,1991,51(1): 1-23. doi: 10.1016/0377-2217(91)90141-H [5] WANG R,LIU M,MA Y. Fault estimation for aero-engine LPV systems based on LFT[J]. Asian Journal of Control,2021,23(1): 351-361. doi: 10.1002/asjc.2204 [6] 谢晓龙. 航空发动机性能评价与衰退预测方法研究[D]. 哈尔滨: 哈尔滨工业大学, 2016.XIE Xiaolong. Research on aero engine performance evaluation and recession prediction method[D]. Harbin: Harbin Institute of technology, 2016. (in Chinese) [7] 曹惠玲,罗立霄,曲春刚,等. 基于LS-SVM的航空发动机喘振故障诊断研究[J]. 热能动力工程,2013,28(1): 23-27,107. CAO Huiling,LUO Lixiao,QU Chungang,et al. Aero engine surge fault diagnosis based on LS-SVM[J]. Thermal Power Engineering,2013,28(1): 23-27,107. (in Chinese [8] 付旭云,陕振勇,李臻,等. 时变模糊神经网络及其在航空发动机排气温度预测中的应用[J]. 计算机集成制造系统,2014,20(4): 919-925. FU Xuyun,SHAN Zhenyong,LI Zhen,et al. Time varying fuzzy neural network and its application in aeroengine exhaust temperature prediction[J]. Computer Integrated Manufacturing System,2014,20(4): 919-925. (in Chinese doi: 10.13196/j.cims.2014.04.fuxuyun.0919.7.20140423 [9] 李书明,任沛,黄燕晓. 航空发动机基线方程的拟合[J]. 机械工程与自动化,2016,1(1): 153-154,157. LI Shuming,REN Pei,HUANG Yanxiao. Fitting of aero engine baseline equation[J]. Mechanical Engineering and Automation,2016,1(1): 153-154,157. (in Chinese doi: 10.3969/j.issn.1672-6413.2016.01.067 [10] 雷亚国, 贾峰, 孔德同, 等. 大数据下机械智能故障诊断的机遇与挑战[J]. 机械工程学报, 2018, 54(5): 94-104.LEI Yaguo, JIA Feng, KONG Detong, et al. Opportunities and challenges of mechanical intelligent fault diagnosis under big data[J]. Journal of Mechanical Engineering, 2018, 54(5): 94-104. (in Chinese) [11] KOHONEN T. An introduction to neural computing[J]. Neural Networks,1988,1(1): 3-16. doi: 10.1016/0893-6080(88)90020-2 [12] GREFF K,SRIVASTAVA R K,KOUTNÍK J,et al. LSTM: a search space odyssey[J]. IEEE Transactions on Neural Networks and Learning Systems,2016,28(10): 2222-2232. [13] FU R, ZHANG Z, LI L. Using LSTM and GRU neural network methods for traffic flow prediction[C]//2016 31st Youth Academic Annual Conference of Chinese Association of Automation (YAC). Wuhan: IEEE, 2016: 324-328. [14] ALBAWI S, MOHAMMED T A, AL-ZAWI S. Understanding of a convolutional neural network[C]//2017 International Conference on Engineering and Technology (ICET). Bali Island, Indonesia: IEEE, 2017: 1-6. [15] 杨洁,万安平,王景霖,等. 基于多传感器融合卷积神经网络的航空发动机轴承故障诊断[J]. 中国电机工程学报,2022,42(13): 4933-4942. YANG Jie,WANG Anping,WANG Jinglin,et al. Aeroengine bearing fault diagnosis based on convolutional neural network for multi-sensor information fusion[J]. Proceedings of the CSEE,2022,42(13): 4933-4942. (in Chinese doi: 10.13334/j.0258-8013.pcsee.211097 [16] 董永峰, 孙跃华, 高立超, 等. 基于改进一维卷积和双向长短期记忆神经网络的故障诊断方法[J]. 计算机应用, 2022, 42(4): 1207-1215.DONG Yongfeng, SUN Yuehua, GAO Lichao, et al. Fault diagnosis method based on improved one-dimensional convolutional and bidirectional long short-term memory neural networks[J]. Journal of Computer Applications, 2022, 42(4): 1207-1215. (in Chinese) [17] 王奕惟, 莫李平, 王奕首, 等. 基于全航段QAR数据和卷积神经网络的航空发动机状态辨识[J]. 航空动力学报, 2021, 36(7): 1556-1563.WANG Yiwei, MO Liping, WANG Yishou, et al. Aero-engine status identification based on full-segment QAR data and convolutional neural network[J]. Journal of Aerospace Power, 2021, 36(7): 1556-1563. (in Chinese) [18] 莫仁鹏, 李天梅, 司小胜, 等. 采用残差网络与卷积注意力机制的设备剩余寿命预测方法[J]. 西安交通大学学报, 2022, 56(4): 1-9.MO Renpeng, LI Tianmei, SI Xiaosheng, et al. Remaining useful life prediction for equipment using residual network and convolutional attention mechanism[J]. Journal of Xi’an Jiaotong University, 2022, 56(4): 1-9. (in Chinese) [19] 杨永灿, 刘韬, 柳小勤, 等. 基于注意力机制的一维卷积神经网络行星齿轮箱故障诊断[J]. 机械与电子, 2021, 39(10): 3-8.YANG Yongcan, LIU Tao, LIU Xiaoqin, et al. Fault diagnosis of gearbox based on one-dimensional convolutional neural network with attention mechanism[J]. Machinery Electronics, 2021, 39(10): 3-8. (in Chinese) [20] XU J,XU L. Health management based on fusion prognostics for avionics systems[J]. Journal of Systems Engineering and Electronics,2011,22(3): 428-436. doi: 10.3969/j.issn.1004-4132.2011.03.010 [21] JALIL N A, HWANG H J, DAWI N M. Machines learning trends, perspectives and prospects in education sector[R]. Nagoya, Japan: the 2019 3rd International Conference on Education and Multimedia Technology, 2019. [22] MITCHELL T. Key ideas in machine learning[M]. Pittsburgh, US: Carnegie Mellon University, 2017. [23] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 770-778. [24] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 7132-7141. [25] HE K, SUN J. Convolutional neural networks at constrained time cost[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2015: 5353-5360. [26] SRIVASTAVA R K, GREFF K, SCHMIDHUBER J. Highway networks[EB/OL]. (2015-11-03)[2022-03-01]. arxiv.org/abs/1505.00387 [27] SZEGEDY C, LIU W, JIA Y, et al. Going deeper with convolutions[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2015: 1-9. [28] BELL S, ZITNICK C L, BALA K, et al. Inside-outside net: detecting objects in context with skip pooling and recurrent neural networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 2874-2883. [29] LOSHCHILOV I, HUTTER F. Decoupled weight decay regularization[EB/OL]. (2019-01-04)[2022-03-01] arxiv.org/abs/1711.05101