基于自动特征工程的飞行器轴承故障诊断

doi:10.11949/0438-1157.20201539

化工学报 ›› 2021, Vol. 72 ›› Issue (S1): 430-436.DOI: 10.11949/0438-1157.20201539

基于自动特征工程的飞行器轴承故障诊断

张弛¹(),李浩¹,胡海涛¹(),朱翀²,张玉莹²,南国鹏²,舒悦³

^1.上海交通大学制冷与低温工程研究所，上海 200240
^2.中国商用飞机有限责任公司上海飞机设计研究院，上海 201210
^3.压缩机技术国家重点实验室（压缩机技术安徽省实验室），安徽合肥 230031

收稿日期:2020-11-01 修回日期:2021-01-25 出版日期:2021-06-20 发布日期:2021-06-20
通讯作者: 胡海涛
作者简介:张弛（1997—），男，硕士研究生，595723288zc@sjtu.edu.cn
基金资助:
国家自然科学基金项目(51976115);上海市科委科技创新行动计划项目(19142203000);压缩机技术国家重点实验室项目(SKL-YSJ201904)

Aircraft bearing fault diagnosis based on automatic feature engineering

ZHANG Chi¹(),LI Hao¹,HU Haitao¹(),ZHU Chong²,ZHANG Yuying²,NAN Guopeng²,SHU Yue³

^1.Institute of Refrigeration and Cryogenics, Shanghai Jiao Tong University, Shanghai 200240, China
^2.Shanghai Aircraft Design and Research Institute, Commercial Aircraft Corporation of China, Ltd. , Shanghai 201210, China
^3.State Key Laboratory of Compressor Technology (Anhui Laboratory of Compressor Technology), Hefei 230031, Anhui, China

Received:2020-11-01 Revised:2021-01-25 Online:2021-06-20 Published:2021-06-20
Contact: HU Haitao

摘要/Abstract

摘要：

针对飞行器轴承信号单一且噪声多、需要针对性特征以及需要高可解释性的问题，开发了涵盖具有自适应噪声的完全集合经验模态分解（CEEMDAN）、自动特征工程以及随机森林的故障诊断模型，模型核心为自动进行特征生成以及提取的特征工程。通过该特征工程能够根据不同对象的信号差异，自动提取出不同对象的有效特征，具备对象间的通用性，且该特征工程可根据样本量的不同调整有效特征的数量，丰富特征空间，具备灵活的可扩展性。验证表明，该涵盖自动特征工程的模型的故障分类准确率为95.32%，可较好地在大样本量下区分压缩机轴承上的不同故障。

关键词: 故障诊断, 算法, 集成, 自动特征工程, 轴承, 模态分解

Abstract:

Aiming at the problems that bearing signals in aircraft are simple and mixed with many noises, which require targeted features and high interpretability, a fault diagnosis model composed of complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and automatic feature engineering and random forest was developed. Bearing vibration signal is converted into intrinsic mode functions (IMF) through the decomposition method. The core of the model is the feature engineering that automatically performs feature generation and extraction based on 65 kinds of manually designed structural features. This feature engineering can automatically extract the effective features of different objects according to the signal difference of different objects, which has universality between objects. Besides, it can adjust the number of effective features according to different sample sizes, enrich the feature space, and have flexible scalability. Validation shows that the fault classification accuracy of the model based on automatic feature engineering and random forest classification model is 95.32%, which performs better than common model based on singular value entropy, energy entropy, envelope sample entropy feature engineering and support vector machines classification model. Result shows that automatic feature engineering fault diagnosis model can better distinguish different faults on compressor bearings under a large sample size.

Key words: fault diagnosis, algorithm, integration, automatic feature engineering, bearing, mode decomposition

中图分类号:

TH 17

张弛, 李浩, 胡海涛, 朱翀, 张玉莹, 南国鹏, 舒悦. 基于自动特征工程的飞行器轴承故障诊断[J]. 化工学报, 2021, 72(S1): 430-436.

ZHANG Chi, LI Hao, HU Haitao, ZHU Chong, ZHANG Yuying, NAN Guopeng, SHU Yue. Aircraft bearing fault diagnosis based on automatic feature engineering[J]. CIESC Journal, 2021, 72(S1): 430-436.

图/表 9

图1 轴承结构

Fig.1 Schematic diagram of bearing

表1 轴承故障类型及故障程度

Table 1 Different bearing failure types and its severity

序号	类型	样本数
1	球体轻微故障	319
2	内圈轻微故障	315
3	外圈轻微故障1	318
4	外圈轻微故障2	318
5	外圈轻微故障3	317
6	球体中等故障	317
7	内圈中等故障	317
8	外圈中等故障	317
9	球体严重故障	317
10	内圈严重故障	318
11	外圈严重故障1	317
12	外圈严重故障2	316
13	外圈严重故障3	318
14	正常	635

图2 CEEMDAN 计算流程

Fig.2 Flow chart of CEEMDAN method

表2 不同的离散特征量

Table 2 Discrete features of IMF

序号	名称	计算公式
1	平方和	$∑ i = 1 N x i 2$
2	一阶差分绝对值之和	$∑ i = 1 N - 1 x i + 1 - x i$
3	最大值位置	$a r g m a x x i$
4	最小值位置	$a r g m i n x i$
5	自回归系数	$X t = ψ 0 + ∑ i = 1 k ψ i X t - i + ε t$
6	Ricker小波分析	$2 3 a π 1 / 4 1 - x 2 a 2 e x p (- x 2 2 a 2)$
7	傅里叶变换系数	$A k = ∑ m = 0 n - 1 a m e x p - 2 π i m k n$
8	平均值	$1 n ∑ i = 1 n x i$
9	连续两点值的变化的绝对值的平均值	$1 n ∑ i = 1, . . ., n - 1 x i + 1 - x i$
…	…	…
65	自相关系数的聚合特征	$1 n - l σ 2 ∑ i = 1 n - l x i - μ x i + l - μ$

表2 不同的离散特征量

Table 2 Discrete features of IMF

序号	名称	计算公式
1	平方和	$∑ i = 1 N x i 2$
2	一阶差分绝对值之和	$∑ i = 1 N - 1 x i + 1 - x i$
3	最大值位置	$a r g m a x x i$
4	最小值位置	$a r g m i n x i$
5	自回归系数	$X t = ψ 0 + ∑ i = 1 k ψ i X t - i + ε t$
6	Ricker小波分析	$2 3 a π 1 / 4 1 - x 2 a 2 e x p (- x 2 2 a 2)$
7	傅里叶变换系数	$A k = ∑ m = 0 n - 1 a m e x p - 2 π i m k n$
8	平均值	$1 n ∑ i = 1 n x i$
9	连续两点值的变化的绝对值的平均值	$1 n ∑ i = 1, . . ., n - 1 x i + 1 - x i$
…	…	…
65	自相关系数的聚合特征	$1 n - l σ 2 ∑ i = 1 n - l x i - μ x i + l - μ$

表3 候选特征量计算

Table 3 Candidate features

序号	参数名称	P值
1	IMF₃的标准差大于0.2[max(x)-min(x)]	3.8440×10^-69
2	IMF₁中小于平均值的最长子序列长度	1.5204×10^-66
3	IMF₁中大于平均值的最长子序列长度	6.7620×10^-66
4	IMF₂中支持10的峰值数量	1.5667×10^-59
5	IMF₃中支持10的峰值数量	2.2018×10^-59
6	IMF₁中支持5的峰值数量	8.0154×10^-59
7	IMF₁的0交叉数量	1.5327×10^-58
8	IMF₄的自回归系数	1.6436×10^-58
…	…	…
18	IMF₂的复杂性指标	1.7666×10^-58
19	IMF₂的最小值	1.7772×10^-58
20	IMF₁的最大值	1.8096×10^-58

图3 随机森林中决策树的结构

Fig.3 Schematic diagram of a trained decision tree in random forest

图4 随机森林对于不同故障的分类正确性

Fig.4 Accuracy of random forest for different faults

表4 不同特征工程对准确率的影响

Table 4 Impact of different feature engineering on accuracy

模型组合	测试集准确率
CEEMDAN+自动特征工程	95.32%
CEEMDAN+奇异值熵	92.07%
CEEMDAN+能量熵	91.08%
CEEMDAN+{能量熵，奇异熵，包络样本熵}	94.52%

表5 不同分类算法的准确率

Table 5 Accuracy of different classification algorithm

分类算法	测试集准确率
逻辑回归	90.18%
SVM	92.54%
单隐层神经网络	89.70%
决策树	93.81%
XGBoost	95.42%

参考文献 16

1	卢浩贤, 肖彪, 何林, 等. 制冷空调系统中压缩机缺氟可靠性的试验研究[J]. 环境技术, 2018, 36(2): 55-59, 81.
	Lu H X, Xiao B, He L, et al. Research on reliability experiment of compressor refrigerant leakage for air conditioning system [J]. Environmental Technology, 2018, 36(2): 55-59, 81.
2	王录雁, 王强, 张梅军, 等. 基于EMD的滚动轴承故障灰色诊断方法[J]. 振动与冲击, 2014, 33(3): 197-202.
	Wang L Y, Wang Q, Zhang M J, et al. A grey fault diagnosis method for rolling bearings based on EMD [J]. Journal of Vibration and Shock, 2014, 33(3): 197-202.
3	陈俊洵, 程龙生, 胡绍林, 等. 基于EMD的改进马田系统的滚动轴承故障诊断[J]. 振动与冲击, 2017, 36(5): 151-156.
	Chen J X, Cheng L S, Hu S L, et al. Fault diagnosis of rolling bearings using modified Mahalanobis-Taguchi system based on EMD [J]. Journal of Vibration and Shock, 2017, 36(5): 151-156.
4	潘洋洋, 何伟, 朱丹宸. 基于CEEMD与IMCKD的滚动轴承故障诊断方法[J]. 机电工程技术, 2019, 48(10): 98-102.
	Pan Y Y, He W, Zhu D C. Fault diagnosis method for rolling bearings based on CEEMD and IMCKD [J]. Mechanical & Electrical Engineering Technology, 2019, 48(10): 98-102.
5	卓仁雄, 肖金凤. 基于改进的集合经验模态分解的电动机滚动轴承故障诊断研究[J]. 机械制造与自动化, 2019, 48(1): 36-39.
	Zhuo R X, Xiao J F. Research on fault diagnosis method of motor bearing based on improved EEMD and SVM [J]. Machine Building & Automation, 2019, 48(1): 36-39.
6	Kong Y, Wang T Y, Chu F L. Meshing frequency modulation assisted empirical wavelet transform for fault diagnosis of wind turbine planetary ring gear [J]. Renewable Energy, 2019, 132: 1373-1388.
7	Torres M E, Colominas M A, Schlotthauer G, et al. A complete ensemble empirical mode decomposition with adaptive noise [C]// 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2011: 4144-4147.
8	The Case Western Reserve University Bearing Data Center Website. Bearing Data Center Seeded Fault Test Data [DB/OL]. [2001-11-12]. .
9	Paquet C, Lavieille M, Schuster W. Aircraft supplemental cooling system noise propagation and radiation: comparisons between acoustics numerical modeling and testing [C]// 17th AIAA/CEAS Aeroacoustics Conference (32nd AIAA Aeroacoustics Conference). AIAA, 2011.
10	Deng M Q, Deng A D, Zhu J, et al. Bandwidth Fourier decomposition and its application in incipient fault identification of rolling bearings [J]. Measurement Science and Technology, 2020, 31(1): 015012.
11	Wang L C, Der S Z, Nasrabadi N M. Automatic target recognition using a feature-decomposition and data-decomposition modular neural network [C]// Proc. SPIE3307, Applications of Artificial Neural Networks in Image Processing Ⅲ. 1998, 3307: 2-13.
12	Lilliefors H W. On the Kolmogorov-Smirnov test for normality with mean and variance unknown [J]. Journal of the American Statistical Association, 1967, 62(318): 399-402.
13	Thissen D, Steinberg L, Kuang D. Quick and easy implementation of the Benjamini-Hochberg procedure for controlling the false positive rate in multiple comparisons [J]. Journal of Educational and Behavioral Statistics, 2002, 27(1): 77-83.
14	刘晓婉. 基于CEEMDAN模糊熵和SVM的滚动轴承故障诊断[J]. 现代制造技术与装备, 2019, (9): 178-179.
	Liu X W. Fault diagnosis of rolling bearing based on CEMDAN fuzzy entropy and SVM [J]. Modern Manufacturing Technology and Equipment, 2019, (9): 178-179.
15	Friedl M A, Brodley C E. Decision tree classification of land cover from remotely sensed data [J]. Remote Sensing of Environment, 1997, 61(3): 399-409.
16	Liaw A, Wiener M. Classification and regression by RandomForest [J]. R News, 2002, 2(3): 18-22.

[1]	陈哲文, 魏俊杰, 张玉明. 超临界水煤气化耦合SOFC发电系统集成及其能量转化机制[J]. 化工学报, 2023, 74(9): 3888-3902.
[2]	曹跃, 余冲, 李智, 杨明磊. 工业数据驱动的加氢裂化装置多工况切换过渡状态检测[J]. 化工学报, 2023, 74(9): 3841-3854.
[3]	尹刚, 李伊惠, 何飞, 曹文琦, 王民, 颜非亚, 向禹, 卢剑, 罗斌, 卢润廷. 基于KPCA和SVM的铝电解槽漏槽事故预警方法[J]. 化工学报, 2023, 74(8): 3419-3428.
[4]	诸程瑛, 王振雷. 基于改进深度强化学习的乙烯裂解炉操作优化[J]. 化工学报, 2023, 74(8): 3429-3437.
[5]	黄可欣, 李彤, 李桉琦, 林梅. 加装旋转叶轮T型通道流场的模态分解[J]. 化工学报, 2023, 74(7): 2848-2857.
[6]	余娅洁, 李静茹, 周树锋, 李清彪, 詹国武. 基于天然生物模板构建纳米材料及集成催化剂研究进展[J]. 化工学报, 2023, 74(7): 2735-2752.
[7]	李贵贤, 曹阿波, 孟文亮, 王东亮, 杨勇, 周怀荣. 耦合固体氧化物电解槽的CO₂制甲醇过程设计与评价研究[J]. 化工学报, 2023, 74(7): 2999-3009.
[8]	王光, 单发顺, 钱禹丞, 焦建芳. 基于集成学习传递熵的化工过程微小故障检测方法[J]. 化工学报, 2023, 74(7): 2967-2978.
[9]	周继鹏, 何文军, 李涛. 异形催化剂上乙烯催化氧化失活动力学反应工程计算[J]. 化工学报, 2023, 74(6): 2416-2426.
[10]	高学金, 姚玉卓, 韩华云, 齐咏生. 基于注意力动态卷积自编码器的发酵过程故障监测[J]. 化工学报, 2023, 74(6): 2503-2521.
[11]	明迁, 高逸, 胡剑, 李盛杰, 王金江. 热交换器泄漏故障虚拟感知方法研究[J]. 化工学报, 2023, 74(4): 1836-1846.
[12]	郑书闽, 郭鹏程, 颜建国, 王帅, 李文博, 周淇. 微小通道内过冷流动沸腾阻力特性实验及预测研究[J]. 化工学报, 2023, 74(4): 1549-1560.
[13]	顾学荣, 刘硕士, 杨思宇. 基于并行EGO和代理模型辅助的多参数优化方法研究[J]. 化工学报, 2023, 74(3): 1205-1215.
[14]	张生安, 刘桂莲. 高效太阳能电解水制氢系统及其性能的多目标优化[J]. 化工学报, 2023, 74(3): 1260-1274.
[15]	袁海鸥, 叶方俊, 张硕, 罗祎青, 袁希钢. 考虑中间换热器的能量集成精馏序列合成[J]. 化工学报, 2023, 74(2): 796-806.

基于自动特征工程的飞行器轴承故障诊断

Aircraft bearing fault diagnosis based on automatic feature engineering

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 16

相关文章 15

编辑推荐

Metrics

本文评价