基于LSTM分类器的航空发动机预测性维护模型

doi:10.12305/j.issn.1001-506X.2022.03.39

摘要/Abstract

摘要：

利用传感器数据进行预测性维护是航空发动机故障预测与健康管理(prognostic and health management, PHM)的关键问题。针对发动机剩余寿命预测准确性低的问题, 提出基于长短期记忆网络(long short-term memory network, LSTM)分类器的预测性维护模型。LSTM分类器通过门控单元对长时间序列信息进行充分筛选, 并将有效信息用于时间序列预测。首先, 采用滑动时间窗口制备训练样本。其次, 将预处理后的样本输入LSTM, 预测设备在特定时间窗口内的失效概率。然后, 通过调整窗口大小, 得到最优性能的二分类模型, 以更好地适应预测维护需求。最后, 利用美国航空航天局C-MAPSS数据集验证了该模型的有效性, 相比于已有分类方法, 其在剩余使用寿命分类方面更加准确。

关键词: 故障预测与健康管理, 预测性维护, 长短期记忆网络, 时间窗, 二分类

Abstract:

Predictive maintenance using sensor data is a key issue in aeroengine prognostic and health management (PHM). Aiming at the problem of low accuracy of remaining useful life prediction of aeroengine, a predictive maintenance model based on long short-term memory network (LSTM) classifier is proposed. The LSTM classifier fully screens the long time sequence information through the gating unit, and uses the effective information for time sequence prediction. Firstly, a sliding time window is used to prepare training samples. Secondly, the pre-processed samples are input into the LSTM to predict the failure probability of the equipment in a specific time window. Then, by adjusting the window size, a two-class model with the best performance is obtained to better adapt to predictive maintenance requirements. Finally, the National Aeronautics and Space Administration C-MAPSS data set is used to verify the effectiveness of the model. Compared with the existing classification methods, the proposed model is more accurate in rumaining useful life classification.

Key words: prognostic and health management (PHM), predictive maintenance, long short-term memory network (LSTM), time window, binary classification

中图分类号:

TP399

蔺瑞管, 王华伟, 车畅畅, 倪晓梅, 熊明兰. 基于LSTM分类器的航空发动机预测性维护模型[J]. 系统工程与电子技术, 2022, 44(3): 1052-1059.

Ruiguan LIN, Huawei WANG, Changchang CHE, Xiaomei NI, Minglan XIONG. Predictive maintenance model of aeroengine based on LSTM classifier[J]. Systems Engineering and Electronics, 2022, 44(3): 1052-1059.

图/表 14

图1

图2

表1

图3

表2

表3

表4

图4

图5

图6

图7

图8

表5

图9

参考文献 29

1	GANIN Y , USTINOVA E , AJAKAN H , et al. A remaining useful life prediction method with long-short term feature processing for aircraft engines[J]. Applied Soft Computing, 2020, 93, 106344. doi: 10.1016/j.asoc.2020.106344
2	CARVALHO T P , SOARES F , VITA R , et al. A systematic literature review of machine learning methods applied to predictive maintenance[J]. Computers & Industrial Engineering, 2019, 137, 106024.
3	马奇友, 刘可薇, 杜坚, 等. 基于深度长短期记忆网络的发动机叶片剩余寿命预测[J]. 推进技术, 2021, 42 (8): 1888- 1897.
	MA Q Y , LIU K W , DU J , et al. Prediction of residual life of engine blades based on deep short term memory network[J]. Journal of Propulsion Technology, 2021, 42 (8): 1888- 1897.
4	SNCHEZ S M , FRANGOPOL D M , PADGETT J , et al. Maintenance and operation of infrastructure systems: review[J]. Journal of Structural Engineering, 2016, 142 (9): F4016004.
5	WANG K S, WANG Y. How AI affects the future predictive maintenance: a primer of deep learning[C]//Proc. of the International Workshop of Advanced Manufacturing and Automation, 2018.
6	GOURIVEAU R , MEDJAHER K , ZERHOUNI N . From prognostics and health systems management to predictive maintenance[M]. London: Wiley, 2016.
7	NGUYEN K T P , MEDJAHER K . A new dynamic predictive maintenance framework using deep learning for failure prognostics[J]. Reliability Engineering & System Safety, 2019, 188, 251- 262.
8	YU W , KIM I Y , MECHEFSKE C . An improved similarity based prognostic algorithm for RUL estimation using an RNN auto encoder scheme[J]. Reliability Engineering & System Safety, 2020, 199, 106926.
9	XIA J , FENG Y W , LU C , et al. LSTM-based multi-layer self-attention method for remaining useful life estimation of mechanical systems[J]. Engineering Failure Analysis, 2021, 125, 105385. doi: 10.1016/j.engfailanal.2021.105385
10	LEI Y G , LI N P , GONTARZ S , et al. A model-based method for remaining useful life prediction of machinery[J]. IEEE Trans.on Reliability, 2016, 65 (3): 1314- 1326. doi: 10.1109/TR.2016.2570568
11	SI X S , WANG W B , HU C H , et al. Remaining useful life estimation-a review on the statistical data driven approaches[J]. European Journal Operational Research, 2011, 213 (1): 1- 14. doi: 10.1016/j.ejor.2010.11.018
12	LIU Y C , HU X F , ZHANG W J . Remaining useful life prediction based on health index similarity[J]. Reliability Engineering & System Safety, 2019, 185, 502- 510.
13	ZHANG J J , WANG P , YAN R Q , et al. Long short-term memory for machine remaining life prediction[J]. Journal of Manufacturing Systems, 2018, 48, 78- 86. doi: 10.1016/j.jmsy.2018.05.011
14	HOCHREITER S , SCHMIDHUBER J . Long short-term memory[J]. Neural Computation, 1997, 9 (8): 1735- 1780. doi: 10.1162/neco.1997.9.8.1735
15	CHE C C , WANG H W , FU Q , et al. Combining multiple deep learning algorithms for prognostic and health management of aircraft[J]. Aerospace Science and Technology, 2019, 94, 105423. doi: 10.1016/j.ast.2019.105423
16	TAMILSELVAN P , WANG P . Failure diagnosis using deep belief learning based health state classification[J]. Reliability Engineering & System Safety, 2013, 115, 124- 135.
17	GUO L , LI N P , JIA F , et al. A recurrent neural network based health indicator for remaining useful life prediction of bearings[J]. Neurocomputing, 2017, 240, 98- 109. doi: 10.1016/j.neucom.2017.02.045
18	HINCHI A Z , TKIOUAT M . Rolling element bearing remaining useful life estimation based on a convolutional long-short-term memory network[J]. Procedia Computer Science, 2018, 127, 123- 132. doi: 10.1016/j.procs.2018.01.106
19	ALDULAIMI A , ZABIHI S , ASIF A , et al. A multimodal and hybrid deep neural network model for remaining useful life estimation[J]. Computers in Industry, 2019, 108, 186- 196. doi: 10.1016/j.compind.2019.02.004
20	YUAN M, WU Y T, LIN L. Fault diagnosis and remaining useful life estimation of aero engine using LSTM neural network[C]//Proc. of the IEEE International Conference on Aircraft Utility Systems, 2016: 135-140.
21	张妍, 王村松, 陆宁云, 等. 基于退化特征相似性的航空发动机寿命预测[J]. 系统工程与电子技术, 2019, 41 (6): 1414- 1421.
	ZHANG Y , WANG C S , LU N Y , et al. Aeroengine life prediction based on similarity of degradation characteristics[J]. Systems Engineering and Electronics, 2019, 41 (6): 1414- 1421.
22	车畅畅, 王华伟, 倪晓梅, 等. 基于多尺度排列熵和长短时记忆神经网络的航空发动机剩余寿命预测[J]. 交通运输工程学报, 2019, 19 (5): 106- 114.
	CHE C C , WANG H , NI X M , et al. Aero-engine remaining life prediction based on multi-scale permutation entropy and long-short-term memory neural network[J]. Journal of Traffic and Transportation Engineering, 2019, 19 (5): 106- 114.
23	张永峰, 陆志强. 基于集成神经网络的剩余寿命预测[J]. 工程科学学报, 2020, 42 (10): 1372- 1380.
	ZHANG Y F , LU Z Q . Remaining useful life prediction based on an integrated neural network[J]. Chinese Journal of Engineering, 2020, 42 (10): 1372- 1380.
24	ZHANG Y Z , XIONG R , HE H W , et al. Long short-term memory recurrent neural network for remaining useful life prediction of lithiumion batteries[J]. IEEE Trans.on Vehicular Technology, 2018, 67 (7): 5695- 5705. doi: 10.1109/TVT.2018.2805189
25	SAXENA A, KAI G, SIMON D, et al. Damage propagation modeling for aircraft engine run-to-failure simulation[C]//Proc. of the International Conference on Prognostics and Health Management, 2008.
26	PEEL L. Data driven prognostics using a Kalman filter ensemble of neural network models[C]//Proc. of the International Conference on Prognostics and Health Management, 2008.
27	XIA T B , SONG Y , ZHENG Y , et al. An ensemble framework based on convolutional bi-directional LSTM with multiple time windows for remaining useful life estimation[J]. Compu-ters in Industry, 2020, 115, 103182. doi: 10.1016/j.compind.2019.103182
28	ZHANG B , ZHANG S H , LI W H . Bearing performance degradation assessment using long short-term memory recurrent network[J]. Computers in Industry, 2019, 106, 14- 29. doi: 10.1016/j.compind.2018.12.016
29	VERSTRAETE D , DROGUETT E , MODARRES M . A deep adversarial approach based on multi-sensor fusion for semi- supervised remaining useful life prognostics[J]. Sensors, 2020, 20 (1): 176- 186.

列号	内容
1	发动机单元编号
2	时间循环
3	时间循环1
4	时间循环2
5	时间循环3
6	传感器测量值1
⋮	⋮
26	传感器测量值21

实际	预测
实际	1	0
1	真正类(true positive, TP)	假正类(false positive, FP)
0	假负类(false positive, FN)	真负类(true negative, TN)

Epochs	Batch_size	sequence_length	Validation_split	Dropout	LSTM_1 units	LSTM_2 units
100	200	30	0.05	0.2	100	50

组号	w₀	w₁	时期	准确率	精确率	召回率	F值	AUC	时间/s
1	15	20	30	0.95	0.87	0.81	0.84	0.943	134
2	15	30	25	0.98	0.96	0.96	0.96	0.976	110
3	15	35	30	0.99	0.96	1	0.98	0.968	281
4	15	40	24	0.99	1	0.96	0.98	0.973	103
5	15	45	12	0.97	0.9	1	0.95	0.962	57
6	15	50	27	0.96	0.91	0.97	0.94	0.968	118
7	15	70	19	0.93	0.93	0.91	0.92	0.933	90
8	15	90	28	0.89	0.94	0.85	0.89	0.925	126
9	15	100	25	0.83	0.92	0.82	0.87	0.902	117

编号	方法	准确率	精确率	召回率	F值
1	逻辑回归	0.92	1	0.68	0.81
2	决策树	0.92	0.95	0.72	0.82
3	RF	0.91	0.94	0.68	0.79
4	SVC	0.92	0.95	0.72	0.82
5	KNN	0.92	0.95	0.72	0.82
6	Gaussian NB	0.94	0.83	0.96	0.89
7	本文方法	0.99	1	0.96	0.98