基于EEMD和Transformer融合模型的多元退化设备RUL预测

doi:10.12305/j.issn.1001-506X.2026.11.23

系统工程与电子技术 ›› 2026, Vol. 48 ›› Issue (11): 3932-3942.doi: 10.12305/j.issn.1001-506X.2026.11.23

基于EEMD和Transformer融合模型的多元退化设备RUL预测

刘星辰(), 吕卫民(), 孙晨峰(), 纪凯夫(), 李涛()

海军航空大学，山东烟台 264001

收稿日期:2025-07-21 修回日期:2025-12-10 接受日期:2026-01-30 出版日期:2026-10-25 发布日期:2026-04-30
通讯作者: 吕卫民 E-mail:a1323061826@163.com;a1360357537@163.com;c617391126@163.com;f534664376437@163.com;a1393434929@163.com
作者简介:刘星辰（1995—），男，博士研究生，主要研究方向为飞行器检测与故障诊断
孙晨峰（1998—），男，博士研究生，主要研究方向为飞行器系统工程
纪凯夫（1992—），男，博士研究生，主要研究方向为发动机使用工程
李　涛（1990—），男，博士研究生，主要研究方向为图像测量与视觉导航

RUL prediction of multivariate degradation equipment based on EEMD and the fusion model of Transformer

Xingchen Liu(), Weimin Lyu(), Chenfeng Sun(), Kaifu Ji(), Tao Li()

Naval Aviation University，Yantai 264001，China

Received:2025-07-21 Revised:2025-12-10 Accepted:2026-01-30 Online:2026-10-25 Published:2026-04-30
Contact: Weimin Lyu E-mail:a1323061826@163.com;a1360357537@163.com;c617391126@163.com;f534664376437@163.com;a1393434929@163.com

摘要/Abstract

摘要：

针对现有方法在处理多元退化设备中的非线性、非平稳数据时，退化特征提取能力不足的问题，提出一种集合经验模态分解（ensemble empirical mode decomposition，EEMD）和Transformer融合模型的剩余使用寿命（remaining useful life，RUL）预测方法。首先，利用EEMD算法将原始数据分解成多个本征模态函数（intrinsic mode function，IMF）分量，并通过计算能量占比筛选出其中表征性能退化的关键分量。在此基础上，构建具有多层级结构的Transformer融合模型，通过充分挖掘保留分量中的局部时序特征和全局长期依赖关系，最终实现从多源特征到RUL值的端到端精准映射。基于C-MAPSS数据集的实验表明，所提方法的均方根误差和评分函数分别低至12.51和241，并展现出良好的鲁棒性和泛化能力。

关键词: 多元退化设备, 剩余使用寿命预测, 集合经验模态分解, Transformer融合模型

Abstract:

Aiming at the problem that the existing methods have insufficient degradation feature extraction capabilities when dealing with nonlinear and nonstationary data of multivariate degradation equipment, a remaining useful life （RUL） prediction method based on ensemble empirical mode decomposition （EEMD） and the Transformer fusion model is proposed. Firstly, the original data is decomposed by the EEMD algorithm to generate multiple intrinsic mode function （IMF） components, and the key components that represent performance degradation are selected by calculating the energy ratio. On this basis, a multi-level structured Transformer fusion model is constructed. By fully mining the local temporal features and global long-term dependency relationships in the retained components, an end-to-end precise mapping from multi-source features to RUL values is ultimately achieved. Experiments conducted on the C-MAPSS dataset demonstrate that the root mean square error （RMSE） and the Score of the proposed method reach as low as 12.51 and 241, respectively, while exhibiting excellent robustness and generalization ability.

Key words: multivariate degradation equipment, remaining useful life (RUL) prediction, ensemble empirical mode decomposition (EEMD), Transformer fusion model

中图分类号:

TP18

刘星辰, 吕卫民, 孙晨峰, 纪凯夫, 李涛. 基于EEMD和Transformer融合模型的多元退化设备RUL预测[J]. 系统工程与电子技术, 2026, 48(11): 3932-3942.

Xingchen Liu, Weimin Lyu, Chenfeng Sun, Kaifu Ji, Tao Li. RUL prediction of multivariate degradation equipment based on EEMD and the fusion model of Transformer[J]. Systems Engineering and Electronics, 2026, 48(11): 3932-3942.

图/表 18

图1

图2

图3

图4

图5

图6

表1

表2

表3

图7

图8

图9

图10

图11

图12

表4

表5

图13

参考文献 27

1	Du Y L, Fang H C, Zuo H F, et al. STF2: a novel spatial-temporal feature fusion method for aero-engine remaining useful life prediction[J]. Expert Systems with Applications, 2025, 290, 128370. doi: 10.1016/j.eswa.2025.128370
2	Zhao J Y, Qu X D, Li Y Q, et al. Real-time prediction of battery remaining useful life using hybrid-fusion deep neural networks[J]. Energy, 2025, 328, 136618. doi: 10.1016/j.energy.2025.136618
3	Chen X, Cheng K. Cutting tool remaining useful life prediction using multi-sensor data fusion through graph neural networks and transformers[J]. Machines, 2025, 13 (11): 1027. doi: 10.3390/machines13111027
4	吴嘉俊, 苏春, 张玉茹. 基于双重自注意力机制和长短时记忆网络的剩余寿命预测[J]. 系统工程与电子技术, 2024, 46 (6): 1986. doi: 10.12305/j.issn.1001-506X.2024.06.16
5	赵昱宇, 索超, 王雨潇. 基于BSVAR的航空发动机剩余使用寿命预测方法[J]. 北京航空航天大学学报, 2025, 51 (11): 3790.
6	Zheng R Y, Yang B, Qian Y C, et al. Joint SOH and RUL estimation for lithium-ion batteries via optimal deep belief network with Bayesian algorithm[J]. Journal of Energy Storage, 2025, 114, 115891. doi: 10.1016/j.est.2025.115891
7	Ge Y, Jiang X X, Xu B L, et al. Log-cumulative feature alignment for enhanced prognosis of aero-engine remaining useful life[J]. Expert Systems with Applications, 2025, 277, 127277. doi: 10.1016/j.eswa.2025.127277
8	胡艳艳, 白雅婷. 基于贝叶斯图注意力Transformer的航空发动机剩余使用寿命概率预[J]. 工程科学学报, 2025, 47 (2): 374.
9	闫啸家, 梁伟阁, 张钢, 等. 基于RCNN-ABiLSTM的机械设备剩余寿命预测方法[J]. 系统工程与电子技术, 2023, 45 (3): 931. doi: 10.12305/j.issn.1001-506X.2023.03.35
10	Wan A P, Zhang H, Chen T, et al. A hybrid deep learning model for robust aeroengine remaining useful life prediction[J]. Signal, Image and Video Processing, 2025, 19 (7): 550. doi: 10.1007/s11760-025-04150-3
11	Guo Y Y, Si J C, Wang Y L, et al. Ensemble-empirical-mode-decomposition （EEMD） on SWH prediction: the effect of decomposed IMFs, continuous prediction duration, and data-driven models[J]. Ocean Engineering, 2025, 324, 120755. doi: 10.1016/j.oceaneng.2025.120755
12	Yao W J, Bai J J, Liao W, et al. From cnn to transformer: a review of medical image segmentation models[J]. Journal of Imaging Informatics in Medicine, 2024, 37 (4): 1529. doi: 10.1007/s10278-024-00981-7
13	Liu F G, Liu S Y, Chai Y, et al. Enhanced Mamba model with multi-head attention mechanism and learnable scaling parameters for remaining useful life prediction[J]. Scientific Reports, 2025, 15 (1): 7178. doi: 10.1038/s41598-025-91815-1
14	Zhu C Y, Tu Y B, Wei Q Y, et al. Short-term power prediction of photovoltaic power station based on LSTM-XGBoost model[J]. Solar Energy, 2025, 300, 113819. doi: 10.1016/j.solener.2025.113819
15	Gong J Q, Qu Z G, Zhu Z L, et al. Parallel TimesNet-BiLSTM model for ultra-short-term photovoltaic power forecasting using STL decomposition and auto-tuning[J]. Energy, 2025, 320, 135286. doi: 10.1016/j.energy.2025.135286
16	Han J H, Kwon Y, Yoon H. Mamba-attention: a self-supervised framework for efficient remaining useful life prediction[J]. Reliability Engineering & System Safety, 2025, 265, 111492.
17	Bai L B, An Y Q, Sun Y C. Measurement of project portfolio benefits with a GA-BP neural network group[J]. IEEE Trans. on Engineering Management, 2023, 71, 4737.
18	He C W, Li Z X, Zheng C Y, et al. A hybrid model based on a dual-attention mechanism for the prediction of remaining useful life of aircraft engines[J]. Sensors, 2025, 25 (18): 5682. doi: 10.3390/s25185682
19	Lamane H, Mouhir L, Moussadek R, et al. Interpreting machine learning models based on SHAP values in predicting suspended sediment concentration[J]. International Journal of Sediment Research, 2025, 40 (1): 91. doi: 10.1016/j.ijsrc.2024.10.002
20	Zong Y C, Nian Y, Zhang C J, et al. Hybrid Grid Search and Bayesian optimization-based random forest regression for predicting material compression pressure in manufacturing processes[J]. Engineering Applications of Artificial Intelligence, 2025, 141, 109580. doi: 10.1016/j.engappai.2024.109580
21	Zhou Z H, Long Z H, Wang R D, et al. An aircraft engine remaining useful life prediction method based on predictive vector angle minimization and feature fusion gate improved transformer model[J]. Journal of Manufacturing Systems, 2024, 76, 567. doi: 10.1016/j.jmsy.2024.08.025
22	Zhu J Y, Ma J, Wu J D. A regularized constrained two-stream convolution augmented transformer for aircraft engine remaining useful life prediction[J]. Engineering Applications of Artificial Intelligence, 2024, 133, 108161. doi: 10.1016/j.engappai.2024.108161
23	Cui W Y, Wang R, Sun T, et al. Managing remaining useful life of cyber-aeroengine systems using a graph spatio-temporal attention recurrent network with phase-lag index[J]. Energy, 2024, 308, 132924. doi: 10.1016/j.energy.2024.132924
24	Peng P, Li Y H, Guo Z Y. High-performance remaining useful life prediction for aeroengine based on combining health states and trajectory similarity[J]. Engineering Applications of Artificial Intelligence, 2025, 141, 109799. doi: 10.1016/j.engappai.2024.109799
25	Li Z F, Luo S L, Liu H Q, et al. TTSNet: Transformer–temporal convolutional network–self-attention with feature fusion for prediction of remaining useful life of aircraft engines[J]. Sensors, 2025, 25 (2): 432. doi: 10.3390/s25020432
26	Wu C C, He J L, Wang L, et al. A dual-path architecture based on time series decomposition and degradation correction for remaining useful life prediction of aero-engine[J]. Computers & Industrial Engineering, 2025, 203, 110964. doi: 10.1016/j.cie.2025.110964
27	Yuan Y L, Yang Q K, Wang G H, et al. Combined improved tuna swarm optimization with graph convolutional neural network for remaining useful life of engine[J]. Quality and Reliability Engineering International, 2025, 41 (1): 174. doi: 10.1002/qre.3651

数据集	FD001	FD002	FD003	FD004
训练集
发动机	100	260	100	249
数量
测试集
发动机	100	259	100	248
数量
工况种类	1	6	1	6
故障模式	1	1	2	2

序号	参数	符号
1	飞行高度	H
2	油门杆角度	TRA
3	马赫数	Ma
4	风扇进口温度	T2
5	低压压气机出口温度	T24
6	高压压气机出口温度	T30
7	低压涡轮出口温度	T50
8	风扇进口压力	P2
9	外涵道总压力	P15
10	高压压气机出口压力	P30
11	实际风扇转速	Nf
12	实际核心机转速	Nc
13	发动机压强比	epr
14	高压压气机出口静压	Ps30
15	燃油流量和高压压气机出口压力比值	Phi
16	修正风扇转速	NRf
17	修正核心机转速	NRc
18	涵道比	BPR
19	燃烧室燃气比	farB
20	引气焓值	htBleed
21	风扇目标转速	NF_dmd
22	修正风扇目标转速	PCNFR_dmd
23	高压涡轮冷气流量	W31
24	低压涡轮冷气流量	W32

参数名称	参数值
卷积核尺寸	3,5,7
Transformer头数	4
Transformer层数	2
BiLSTM层数	2
训练轮数	100
训练批大小	32
学习率	0.001
Dropout比例	0.1

预测方法	FD001 RMSE Score	FD002 RMSE Score	FD003 RMSE Score	FD004 RMSE Score	平均计算时间/s
PVA-FFG^[21]	15.27 559	15.30 1871	17.24 1702	17.62 2154	73.21
DSAN^[22]	14.09 427	14.65 1079	16.52 1193	15.78 1397	68.90
PLI-GSTARN^[23]	14.23 274	19.43 1989	13.81 385	19.92 2856	65.73
HS-TS^[24]	15.53 1265	26.64 6706	23.26 4725	31.01 9902	70.85
TTSNet^[25]	14.56 272	15.89 1320	14.92 573	17.06 1529	63.07
DP-TSDDC^[26]	14.17 325	17.25 1672	14.36 516	17.92 1763	87.52
IMTSO-GCN^[27]	13.92 307	14.63 1225	13.98 423	16.75 1387	75.16
所提方法	12.51 241	13.57 1013	12.73 367	14.95 1272	67.25

预测方法	FD001 RMSE Score	FD002 RMSE Score	FD003 RMSE Score	FD004 RMSE Score	平均计算时间/s
Transformer	21.32 892	23.76 2273	22.05 3359	25.63 2915	42.27
Transformer融合模型	17.29 506	19.43 1542	17.62 1503	20.42 1710	50.86
EEMD-Transformer-BiLSTM-注意力	13.46 327	14.52 1364	13.68 525	16.05 1402	61.12
EEMD-Transformer-一维CNN-注意力	13.72 315	14.93 1293	13.97 604	16.42 1559	65.59
EEMD-Transformer-一维CNN-BiLSTM	13.92 295	15.11 1370	14.21 653	16.72 1465	62.93
所提方法	12.51 241	13.57 1013	12.73 367	14.95 1272	67.25

基于EEMD和Transformer融合模型的多元退化设备RUL预测

RUL prediction of multivariate degradation equipment based on EEMD and the fusion model of Transformer

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 18

参考文献 27

相关文章 5

编辑推荐

Metrics

本文评价

[1]	王全全, 谢松霖, 顾志豪, 吴城坤, 张更新. 低信噪比下基于ICEEMDAN和HHO的协作频谱感知方法[J]. 系统工程与电子技术, 2025, 47(9): 3109-3116.
[2]	冯帅, 曹茹茹, 马愈昭, 丁超. 改进CEEMDAN结合新型小波变换的激光雷达去噪算法[J]. 系统工程与电子技术, 2024, 46(12): 4054-4061.
[3]	吴宗收, 汪立新, 沈强, 李灿, 李文华. 基于CEEMD半球谐振陀螺输出预测方法[J]. 系统工程与电子技术, 2023, 45(10): 3259-3264.
[4]	马愈昭, 刘逵, 张岩峰, 冯帅, 熊兴隆. CEEMD结合改进小波阈值的激光雷达信号去噪算法[J]. 系统工程与电子技术, 2023, 45(1): 93-100.
[5]	陈寅生, 姜守达, 刘晓东, 杨京礼, 王祁. 基于EEMD样本熵和SRC的自确认气体传感器故障诊断方法[J]. 系统工程与电子技术, 2016, 38(5): 1215-.