基于Transformer的多特征融合的航空发动机剩余使用寿命预测

doi:10.3969/j.issn.1000-5641.2022.05.018

摘要/Abstract

摘要：

发动机作为飞机的核心部件, 对飞机运行起着至关重要的作用. 对航空发动机做准确的剩余使用寿命预测, 能够提前进行维护诊断, 预防重大事故的发生, 节约维护成本. 针对现有的方法缺乏对不同时间步长的考虑以及不同传感器和操作条件之间关系的研究, 提出了一种基于Transformer的多编码器特征输出融合的航空发动机剩余使用寿命预测方法. 该方法选取两个不同时间长度的输入数据, 使用排列熵对传感器之间的关系进行分析, 并将操作条件数据独立提取特征. 在广泛使用的航空发动机CMAPSS(Commercial Modular Aero-Propulsion System Simulation)数据集上进行了实验验证. 实验结果表明, 该方法优于现有的先进预测方法, 可有效提高预测精度.

关键词: 深度学习, 剩余使用寿命, 航空发动机, Transformer

Abstract:

As the core components of aircraft, engines play a vital role during flight. Accurate prediction of the remaining useful life of the aeroengine can help prognostics and health management, thus preventing major accidents and saving maintenance costs. In view of the lack of consideration of different time steps and the relationship between different sensors and operating conditions in existing methods, a remaining useful life prediction method based on the Transformer was proposed, which fuses multi-feature outputs from different encoder layers. This method selects two input data with different time steps, analyzes the relationship between the sensors using permutation entropy, and extracts features independently from the operating condition data. The experimental results on the public aeroengine dataset CMAPSS (Commercial Modular Aero-Propulsion System Simulation) show that the proposed method is superior to other advanced remaining useful life prediction methods.

Key words: deep learning, remaining useful life, aeroengine, Transformer

中图分类号:

TP399

马依琳, 陶慧玲, 董启文, 王晔. 基于Transformer的多特征融合的航空发动机剩余使用寿命预测[J]. 华东师范大学学报（自然科学版）, 2022, 2022(5): 219-232.

Yilin MA, Huiling TAO, Qiwen DONG, Ye WANG. Prediction of remaining useful life of aeroengines based on the Transformer with multi-feature fusion[J]. Journal of East China Normal University(Natural Science), 2022, 2022(5): 219-232.

图/表 12

图1

图2

图3

图4

表1

图5

表2

图6

表3

表4

表5

表6

参考文献 31

1	任治潞. 从“重客轻货”到“客货并重”——《“十四五”航空物流发展专项规划》解读. 大飞机, 2022, (2): 56- 60.
2	邹建军. 新发展格局下我国航空物流建设发展策略思考. 民航管理, 2020, (12): 16- 22.
3	李小龙, 徐启明. 大运行理念下飞机维修成本精益管理. 民航管理, 2021, (10): 58- 63.
4	周俊. 数据驱动的航空发动机剩余使用寿命预测方法研究 [D]. 南京: 南京航空航天大学, 2017.
5	蔡光耀, 高晶, 苗学问. 航空发动机健康管理系统发展现状及其指标体系研究. 测控技术, 2016, 35 (4): 1-5.
6	PARIS P C, ERDOGAN F. A critical analysis of crack propagation laws. Journal of Basic Engineering, 1963, 85 (4): 528- 533.
7	ZHAO F, TIAN Z, ZENG Y. Uncertainty quantification in gear remaining useful life prediction through an integrated prognostics method. IEEE Transactions on Reliability, 2013, 62 (1): 146- 159.
8	LEI Y, LI N, GONTARZ S, et al. A model-based method for remaining useful life prediction of machinery. IEEE Transactions on Reliability, 2016, 65 (3): 1314- 1326.
9	SUN J, ZUO H, WANG W, et al. Prognostics uncertainty reduction by fusing on-line monitoring data based on a state-space-based degradation model. Mechanical Systems and Signal Processing, 2014, 45 (2): 396- 407.
10	HUANG Z, XU Z, WANG W, et al. Remaining useful life prediction for a nonlinear heterogeneous wiener process model with an adaptive drift. IEEE Transactions on Reliability, 2015, 64 (2): 687- 700.
11	朱磊, 左洪福, 蔡景. 基于Wiener过程的民用航空发动机性能可靠性预测. 航空动力学报, 2013, 28, (5): 1006-1012.
12	WANG H K, HUANG H Z, LI Y F, et al. Condition-based maintenance with scheduling threshold and maintenance threshold. IEEE Transactions on Reliability, 2016, 65 (2): 513- 524.
13	PENG W, LI Y F, YANG Y J, et al. Bivariate analysis of incomplete degradation observations based on inverse gaussian processes and copulas. IEEE Transactions on Reliability, 2016, 65 (2): 624- 639.
14	GARCÍA NIETO P J, GARCÍA-GONZALO E, SÁNCHEZ LASHERAS F, et al. Hybrid PSO–SVM-based method for forecasting of the remaining useful life for aircraft engines and evaluation of its reliability. Reliability Engineering and System Safety, 2015, 138, 219- 231.
15	KHELIF R, CHEBEL-MORELLO B, MALINOWSKI S, et al. Direct remaining useful life estimation based on support vector regression. IEEE Transactions on Industrial Electronics, 2017, 64 (3): 2276- 2285.
16	ZHENG S, RISTOVSKI K, FARAHAT A, et al. Long short-term memory network for remaining useful life estimation [C]// 2017 IEEE International Conference on Prognostics and Health Management (ICPHM). IEEE, 2017: 88-95. DOI: 10.1109/ICPHM.2017.7998311.
17	REN L, CHENG X J, WANG X K, et al. Multi-scale dense gate recurrent unit networks for bearing remaining useful life prediction. Future Generation Computer Systems, 2019, 94, 601- 609.
18	WANG J J, WEN G L, YANG S P, et al. Remaining useful life estimation in prognostics using deep bidirectional LSTM neural network [C]// 2018 Prognostics and System Health Management Conference (PHM-Chongqing). IEEE, 2018: 1037-1042. DOI: 10.1109/PHM-Chongqing.2018.00184.
19	HU K, CHENG Y W, WU J, et al. Deep bidirectional recurrent neural networks ensemble for remaining useful life prediction of aircraft engine [J]. IEEE Transactions on Cybernetics. IEEE, 2021. DOI: 10.1109/TCYB.2021.3124838.
20	LI H, LI Y, WANG Z J, et al. Remaining useful life prediction of aero-engine based on PCA-LSTM [C]// 2021 7th International Conference on Condition Monitoring of Machinery in Non-Stationary Operations (CMMNO). IEEE, 2021: 63-66.
21	LI X, DING Q, SUN J Q. Remaining useful life estimation in prognostics using deep convolution neural networks. Reliability Engineering and System Safety, 2018, 172, 1- 11.
22	LI R Z, CHU Z T, JIN W K, et al. Temporal convolutional network based regression approach for estimation of remaining useful life [C]// 2021 IEEE International Conference on Prognostics and Health Management (ICPHM). IEEE, 2021. DOI: 10.1109/ICPHM51084.2021.9486528.
23	ZENG F C, LI Y M, JIANG Y H, et al. A deep attention residual neural network-based remaining useful life prediction of machinery. Measurement, 2021, 181, 109642.
24	ABDERREZEK S, BOUROUIS A. Convolutional autoencoder and bidirectional long short-term memory to estimate remaining useful life for condition based maintenance [C]// 2021 International Conference on Networking and Advanced Systems (ICNAS). IEEE, 2021. DOI: 10.1109/ICNAS53565.2021.9628958.
25	REMADNA I, TERRISSA S L, ZEMOURI R, et al. Leveraging the power of the combination of CNN and bi-directional LSTM networks for aircraft engine RUL estimation [C]// 2020 Prognostics and Health Management Conference (PHM-Besançon). IEEE, 2020: 116-121. DOI: 10.1109/PHM-Besancon49106.2020.00025.
26	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need [C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. New York: Curran Associates Inc., 2017: 6000-6010.
27	MO Y, WU Q, LI X, et al. Remaining useful life estimation via transformer encoder enhanced by a gated convolutional unit. Journal of Intelligent Manufacturing, 2021, 32 (7): 1997- 2006.
28	ZHANG Z Z, SONG W, LI Q Q. Dual-aspect self-attention based on transformer for remaining useful life prediction [J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 2505711. DOI: 10.1109/TIM.2022.3160561.
29	LIU L, WANG S, LIU D, et al. Entropy-based sensor selection for condition monitoring and prognostics of aircraft engine. Microelectronics Reliability, 2015, 55 (9/10): 2092- 2096.
30	SAXENA A, GOEBEL K, SIMON D, et al. Damage propagation modeling for aircraft engine run-to-failure simulation [C]// 2008 International Conference on Prognostics and Health Management. IEEE, 2008. DOI: 10.1109/PHM.2008.4711414.
31	ZHAO Z Q, LIANG B, WANG X Q, et al. Remaining useful life prediction of aircraft engine based on degradation pattern learning. Reliability Engineering and System Safety, 2017, 164, 74- 83.

传感器编号	表示符号	具体描述	单位	传感器编号	表示符号	具体描述	单位
1	T2	风扇入口温度	°R	12	phi	燃料流量与Ps30的比率	pps/psia
2	T24	LPC出口温度	°R	13	NRF	校正后的风扇速率	rpm
3	T30	HPC出口温度	°R	14	NRc	校正后的核心速率	rpm
4	T50	LPT出口温度	°R	15	BPR	涵道比
5	P2	风扇入口压力	psia	16	farB	燃烧室燃料空气比
6	P15	涵道压力	psia	17	htBleed	引气焓值
7	P30	HPC出口压力	psia	18	Nf_dmd	要求的风扇转速	rpm
8	Nf	风扇物理转速	rpm	19	PCNfR_dmd	要求的校正后风扇转速	rpm
9	Nc	核心机物理转速	rpm	20	W31	HPT冷却引气流量	lbm/s
10	epr	发动机压力比率		21	W32	LPT冷却引气流量	lbm/s
11	Ps30	HPC出口静态压力	psia

传感器编号	$ {E_{\rm{entropy}}} $	传感器编号	$ {E_{\rm{entropy}}} $
2	0.6197	12	0.6055
3	0.6275	13	0.6521
4	0.5891	14	0.6060
7	0.5986	15	0.6556
8	0.6482	17	0.6774
9	0.6129	20	0.6566
11	0.5980	21	0.6567

超参数	设置
学习率	0.01
批大小	256
优化器	Adam
输入嵌入层	节点数 64, 激活函数 ReLU
双时间长度编码器层	层数 2, 自注意力头数 4
传感器编码器层	层数 2, 自注意力头数 4
操作条件编码器层	层数 2, 自注意力头数 4
解码器层	层数 2, 自注意力头数 4
解码器前馈全连接层	节点数 64

方法	R_RMSE
方法	数据集1	数据集3
BiLSTM^[18]	13.65	13.74
DBRNN^[19]	17.97	19.18
DCNN^[21]	12.61	12.64
TCN^[22]	11.58	12.67
DARNN^[23]	12.04	10.18
GCU-Transformer^[27]	11.27	11.42
DAST^[28]	11.43	11.32
本文模型	11.04	10.68

方法	S_score
方法	数据集1	数据集3
BiLSTM^[18]	295.00	317.00
DBRNN^[19]	459.89	658.12
DCNN^[21]	273.70	284.10
TCN^[22]	195.10	228.20
DARNN^[23]	261.95	247.85
DAST^[28]	203.15	154.92
本文模型	169.07	195.18