一种基于TimeGAN和OCSVM的多元退化设备小子样数据增广方法

doi:10.12263/DZXB.20220079

电子学报 ›› 2022, Vol. 50 ›› Issue (11): 2678-2687.DOI: 10.12263/DZXB.20220079

一种基于TimeGAN和OCSVM的多元退化设备小子样数据增广方法

孙晨峰¹, 吕卫民¹, 戴洪德¹, 张浩晨²

^1.海军航空大学岸防兵学院, 山东烟台 264000
^2.西北工业大学机电学院, 陕西西安 710000

收稿日期:2022-01-13 修回日期:2022-03-15 出版日期:2022-11-25
- 作者简介:
- 孙晨峰男，1998年3月出生于山东省莱阳市.现为海军航空大学硕士研究生.E-mail: scf326228@163.com
  吕卫民（通讯作者）男，1970年7月出生于山东省莱州市.现为海军航空大学教授、博士生导师.主要研究方向为装备系统工程.
  戴洪德男，1981年11月出生于湖南省邵阳市.现为海军航空大学副教授.主要研究方向为惯性导航、滤波估计、故障诊断以及智能信息处理技术.E-mail: 13954559561@126.com
  张浩晨男，1998年8月出生于陕西省西安市.现为西北工业大学机电学院硕士研究生.主要研究方向为数字化制造及数据驱动下的智能制造.E-mail: zhanghaochen817@163.com
- 基金资助:
- 国家自然科学基金 (51975580)

A Small Sample Data Augmentation Method for Multivariate Degradation Equipment Based on TimeGAN and OCSVM

SUN Chen-feng¹, LÜ Wei-min¹, DAI Hong-de¹, ZHANG Hao-chen²

^1.Coastal Defense College，Naval Aviation University，Yantai，Shandong 264000，China
^2.School of Mechatronics，Northwestern Polytechnical University，Xi'an，Shaanxi 710000，China

Received:2022-01-13 Revised:2022-03-15 Online:2022-11-25 Published:2022-11-19

摘要/Abstract

摘要：

工作在复杂环境下的多元退化设备面临失效数据少、多源信息融合准确度低和监督学习数据不平衡等问题，对此本文提出一种基于时间序列生成对抗网络（Time-series Generative Adversarial Networks， TimeGAN）与单分类支持向量机（One-Class Support Vector Machine， OCSVM）组合模型的小子样数据增广方法.方法引入了TimeGAN模型拟合真实数据时间序列相关性，从而生成新的多元退化设备数据.本文提出了一种基于最大均值差异改进方法的可信度判据，避免强相关特征对生成数据质量评价的影响，通过使用T-分布随机邻近嵌入（T-distributed Stochastic Neighbor Embedding， T-SNE）和全局最大均值差异（Global Maximum Mean Discrepancy， GMMD）的组合方法，定性定量地评价生成数据的质量水平.基于训练后的OCSVM模型，对生成数据进行异常检测与剔除，进一步提高生成数据的质量.以航空发动机数据集C-MAPSS为例进行方法验证分析，通过与其他数据增强模型对比验证了所提方法的可行性和有效性.

关键词: 小子样数据, 数据增广, 多元退化设备, 时间序列生成对抗网络, 单分类支持向量机

Abstract:

The multivariate degradation equipment working in complex environments faces the problems of small amount of failure data, the low accuracy of multi-source information fusion and imbalanced supervised learning dataset. For these problems, a small sample data augmentation method based on the combination model of time-series generative adversarial networks(TimeGAN) and one class support vector machine(OCSVM) is designed. TimeGAN is introduced to fit the time series correlation and generate the new degradation data. A new credibility criterion based on improved maximum mean discrepancy is proposed to avoid the strong correlation influence for the data quality evaluation. The combination method of t-distributed stochastic neighbor embedding(T-SNE) and global maximum mean discrepancy(GMMD) is applied to evaluate the quality of generation dataset qualitatively and quantitatively. The trained OCSVM is used to detect and remove the novelty data to further improve dataset quality. The comparison of the method and other data generation models on aircraft engine dataset C-MAPSS verifies its feasibility and effectiveness.

Key words: small sample data, data augmentation, multivariate degradation equipment, time-series generative adversarial networks, one-class support vector machine

中图分类号:

TP206+.3

孙晨峰, 吕卫民, 戴洪德, 张浩晨. 一种基于TimeGAN和OCSVM的多元退化设备小子样数据增广方法[J]. 电子学报, 2022, 50(11): 2678-2687.

SUN Chen-feng, LÜ Wei-min, DAI Hong-de, ZHANG Hao-chen. A Small Sample Data Augmentation Method for Multivariate Degradation Equipment Based on TimeGAN and OCSVM[J]. Acta Electronica Sinica, 2022, 50(11): 2678-2687.

图/表 11

图1 TimeGAN模型结构框图

图2 小子样数据增广方法流程图

图3 发动机特征数据分布图

图4 发动机特征热力图

表1 TimeGAN模型参数表

AutoEncoder	参数	GAN模型	参数
采样时间步长	24	生成器噪声层数	32
特征维数	12	GAN隐藏层	24
隐藏层层数	1	隐层激活函数	Leaky ReLU
学习率	0.0005	判别器真实损失	1

图5 发动机时间步生成数据图

图6 发动机两类数据可视化TSNE图

表2 生成数据各特征最大均值差异表

特征	MMD	特征	MMD	特征	MMD	GMMD
Sensor2	1.1816	Sensor8	0.7846	Sensor15	0.6838	0.457
Sensor3	0.6211	Sensor12	1.2198	Sensor17	0.9967
Sensor4	0.8921	Sensor13	1.2692	Sensor20	1.0949
Sensor7	0.8894	Sensor14	0.5119	Sensor21	0.5922

表3 不同生成批次的GMMD分析表

生成批次	GMMD	生成批次	GMMD	生成批次	GMMD	$S$
Batch1	0.457	Batch4	0.546	Batch7	0.539	0.0373
Batch2	0.534	Batch5	0.561	Batch8	0.548
Batch3	0.471	Batch6	0.557	Batch9	0.531

表3 不同生成批次的GMMD分析表

生成批次	GMMD	生成批次	GMMD	生成批次	GMMD	$S$
Batch1	0.457	Batch4	0.546	Batch7	0.539	0.0373
Batch2	0.534	Batch5	0.561	Batch8	0.548
Batch3	0.471	Batch6	0.557	Batch9	0.531

图7 发动机生成数据异常检测结果图

表4 不同方法下的小子样数据生成评价分析

小子样增广方法	运算时间	最少真实数据量	模型精度	GMMD指标	K-S检验	Kruskal-Wallis H检验
Bootstrap	较快	极小子样以上	较高	0.5524	0.395	1.55
TimeGAN	慢	小子样以上	较高	0.5271(9次均值)	0.368	0.207
TimeGAN-孤立森林	慢	小子样以上	较高	0.5239	0.351	0.194
TimeGAN-OCSVM	慢	小子样以上	高	0.5220	0.347	0.196

参考文献 15

1	裴洪, 胡昌华, 司小胜, 等. 基于机器学习的设备剩余寿命预测方法综述[J]. 机械工程学报, 2019, 55(8): 1-13.
	PEI Hong, HU Chang-hua, SI Xiao-sheng, et al. Review of machine learning based remaining useful life prediction methods for equipment[J]. Journal of Mechanical Engineering, 2019, 47(12): 2449-2456. (in Chinese)
2	车畅畅, 王华伟, 倪晓梅, 等. 基于1D-CNN和Bi-LSTM的航空发动机剩余寿命预测[J]. 机械工程学报, 2021, 57(14): 304-312.
	CHE Chang-chang, WANG Hua-wei, NI Xiao-mei, et al. Residual life prediction of aeroengine based on 1D-CNN and Bi-LSTM[J]. Journal of Mechanical Engineering, 2021, 57(14): 304-312. (in Chinese)
3	牟含笑, 郑建飞, 胡昌华, 等. 基于CDBN与BiLSTM的多元退化设备剩余寿命预测[J]. 航空学报, 2022, 43(7): 325403.
	MOU Han-xiao, ZHENG Jian-fei, HU Chang-hua, et al. Remaining useful life prediction of multivariate degradation equipment based on CDBN and BiLSTM[J]. Acta Aeronautica et Astronautica Sinica, 2022, 43(7): 325403. (in Chinese)
4	于倩影, 李娟, 戴洪德, 等. 基于Lasso变量选择的航空发动机相似性剩余寿命预测[J/OL]. 航空动力学报. DOI:10.13224/j.cnki.jasp.20210516 .
	YU Qian-ying, LI Juan, DAI Hong-de, et al. Lasso based variable selection for similarity remaining useful life prediction of aero-engine[J/OL]. Journal of Aerospace Power. DOI:10.13224/j.cnki.jasp.20210516. (in Chinese)
5	韩淞宇, 邵海东, 姜洪开, 等. 基于提升卷积神经网络的航空发动机高速轴承智能故障诊断[J]. 航空学报, 2022, 43(9): 625479.
	HAN Song-yu, SHAO Hai-dong, JIANG Hong-kai, et al. Intelligent fault diagnosis of aero-engine high-speed bearing using enhanced convolutional neural network[J]. Acta Aeronautica et Astronautica Sinica, 2022, 43(9): 625479. (in Chinese)
6	张云, 于广, 王立平, 等. 基于性能退化数据的数控转台单子样可靠性分析[J]. 清华大学学报, 2020, 60(04): 299-305.
	ZHANG Yun, YU Guang, WANG Li-ping, et al. Performance degradation data based NC rotary table reliability predictions using a single sample[J]. Journal of Tsinghua University, 2020, 60(04): 299-305. (in Chinese)
7	叶伟, 李亚平, 陈顶, 等. 基于小样本的复杂装备"虚拟总体"生成与检验模型[J]. 中国管理科学, 2015, 23(S1): 240-244.
	YE Wei, LI Ya-ping, CHEN Ding, et al. Model for virtual overall producing and quality inspection of comples equipment based on small sample[J]. Chinese Journal of Manage Science, 2015, 23(S1): 240-244. (in Chinese)
8	王明磊, 原大宁, 刘宏昭. 二元Wiener过程下的小样本电主轴可靠性分析[J]. 机械科学与技术, 2017, 36(02): 279-285.
	WANG Ming-lei, YUAN Da-ning, LIU Hong-zhao. Reliability analysis of motorized spindle with small sample based on two-dimensional Wiener process[J]. Mechanical Science and Technology for Aerospace Engineering, 2017, 36(02): 279-285. (in Chinese)
9	张晟斐, 李天梅, 胡昌华, 等. 基于深度卷积生成对抗网络的缺失数据生成方法及其在剩余寿命预测中的应用[J]. 航空学报, 2022, 43(8): 225708.
	ZHANG Sheng-fei, LI Tian-mei, HU Chang-hua, et al. Deep convolutional generative adversarial networks based missing data generation method and its application in remaining useful life prediction[J]. Acta Aeronautica et Astronautica Sinica, 2022, 43(8): 225708. (in Chinese)
10	JOSEY M, PANG C K, LUO M, et al. Classification of imbalanced data by oversampling in kernel space of support vector machines[J]. IEEE Transactions on Neural Networks and Learning System, 2018, 29(9): 4065-4076.
11	JINSUNG Y, DANIEL J, MIHAELA S. Time-series generative adversarial networks[C]//Proceedings of the 2019 Conference and Workshop on Neural Information Processing Systems. Vancouver: NeurIPS, 2019: 5508-5518.
12	吴定海, 张培林, 任国全, 等. 基于支持向量的单类分类方法综述[J]. 计算机工程, 2011, 37(05): 187-189.
	WU Ding-hai, ZHANG Pei-lin, REN Guo-quan, et al. Review of one-class classification method based on support vector[J]. Computer Engineering, 2011, 37(5): 187-189. (in Chinese)
13	张庆朔, 何强, 张长伦, 等.模糊多核一类支持向量机[J]. 北京建筑大学学报, 2020, 36(01): 82-90.
	ZHANG Qing-shuo, HE Qiang, ZHANG Chang-lun, et al. Fuzzy multiple kernel one-class support vector machine[J]. Journal of Beijing University of Civil Engineering and Architecture, 2020, 36(01): 82-90. (in Chinese)
14	SAXENA A, GOEBEL K, SIMON D, et al. Damage propagation modeling for aircraft engine run-to-failure simulation[C]//Proceedings of the 2008 International Conference on Prognostics and Health Management. Denver: IEEE, 2008: 1-9.
15	张浩, 胡昌华, 杜党波, 等. 多状态影响下基于Bi-LSTM网络的锂电池剩余寿命预测方法[J]. 电子学报, 2022, 50(3): 619-624.
	ZHANG Hao, HU Chang-hua, DU Dang-bo, et al. Remaining useful life prediction method of lithium-ion battery based on Bi-LSTM network under multi-state influence[J]. Acta Electronica Sinica, 2022, 50(3): 619-624. (in Chinese)

一种基于TimeGAN和OCSVM的多元退化设备小子样数据增广方法

A Small Sample Data Augmentation Method for Multivariate Degradation Equipment Based on TimeGAN and OCSVM

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 15

相关文章 1

编辑推荐

Metrics