一种基于特征聚类和评价的轴承寿命预测新方法

李海浪,邹益胜,曾大懿,刘永志,赵市教,宋小欣

振动与冲击 ›› 2022, Vol. 41 ›› Issue (5) : 141-150.

PDF(2028 KB)
PDF(2028 KB)
振动与冲击 ›› 2022, Vol. 41 ›› Issue (5) : 141-150.
论文

一种基于特征聚类和评价的轴承寿命预测新方法

  • 李海浪,邹益胜,曾大懿,刘永志,赵市教,宋小欣
作者信息 +

A new method of bearing life prediction based on feature clustering and evaluation

  • LI Hailang, ZOU Yisheng, ZENG Dayi, LIU Yongzhi, ZHAO Shijiao, SONG Xiaoxin
Author information +
文章历史 +

摘要

在预测轴承寿命时,使提取的特征和剩余寿命保持高相关性,并使不同的特征之间保持低相关性,是有利于提升轴承寿命预测精度的。为解决单一的特征评价方法对后者考虑不足的问题,提出了一种基于相关性改进Kmeans聚类算法(Correlation-based improved Kmeans cluster algorithm, Corr-Kmeans)和初始聚类中心确定方法,并与特征评价相结合,最终提出一种基于特征聚类和评价的轴承寿命预测新方法。首先利用卷积自编码对频域信息提取初始特征,用Corr-Kmeans对初始特征按相关性进行聚类,使得聚类后的特征类内相关性高,而类间相关性低;其次,使用相关性、单调性和鲁棒性3个指标来综合评价每一类中的特征,按照筛选阈值将得分较高的特征从每一类中分别选出,组成用于训练与预测的特征子集;最后采用LSTM(Long Short-Term Memory, LSTM)网络对轴承剩余寿命进行预测。在一个轴承加速寿命实验的公开数据集上使用留一法进行验证,利用对比实验证明了所提方法在预测轴承剩余寿命上的有效性。

Abstract

When predicting the bearing life, it is beneficial to improve the accuracy of bearing life prediction by keeping high correlation between extracted features and remaining life and low correlation between different features. In order to solve the problem that the latter is insufficiently considered by a single feature evaluation method, an improved Kmeans cluster algorithm (Corr-Kmeans) based on Correlation and the initial cluster center determination method are proposed, which are combined with the feature evaluation, and finally a new method for bearing life prediction based on feature clustering and evaluation is proposed. First, the convolution self-coding is used to extract the initial features of the frequency domain information, and corr-Kmeans is used to cluster the initial features according to correlation, so that the correlation within the feature class after clustering is high, while the correlation between the classes is low. Secondly, three indicators of relevance, monotonicity and robustness were used to comprehensively evaluate the features in each category, and the features with high scores were selected from each category according to the screening threshold to form the feature subset for training and prediction. Finally, LSTM(Long Short-Term Memory) network is used to predict the remaining life of the bearing. By using the retention method on the open data set of a bearing accelerated life experiment, the effectiveness of the proposed method in predicting the remaining life of the bearing is proved by the comparative experiment.

关键词

轴承 / 寿命预测 / Corr-Kmeans / 聚类 / 特征评价

Key words

bearing / life prediction / Corr-Kmeans / Clustering / feature evaluation

引用本文

导出引用
李海浪,邹益胜,曾大懿,刘永志,赵市教,宋小欣. 一种基于特征聚类和评价的轴承寿命预测新方法[J]. 振动与冲击, 2022, 41(5): 141-150
LI Hailang, ZOU Yisheng, ZENG Dayi, LIU Yongzhi, ZHAO Shijiao, SONG Xiaoxin. A new method of bearing life prediction based on feature clustering and evaluation[J]. Journal of Vibration and Shock, 2022, 41(5): 141-150

参考文献

[1] LEl Y. ltelligentFault Diagnosis and Remaining Useful Llife Prediction of Rotating Machinery[M]. Elsevier Butteworth-Heinemann,Oxford,2016:102-108.
[2] 韩林洁,石春鹏,张建超.基于一维卷积神经网络的轴承剩余寿命预测[J].制造业自动化,2020,42(03):10-13.
Han Linjie,Shi Chunpeng,Zhang Jianchao.Remaining life prediction of bearing based on one-dimensional convolutional neural network[J].Manufacturing Automation,2020,42(03):10-13.
[3] Lei Y , Li N , Guo L , et al. Machinery health prognostics: A systematic review from data acquisition to RUL prediction[J]. Mechanical Systems and Signal Processing, 2018, 104(MAY1):799-834.
[4] 杨宇,潘海洋,程军圣.基于特征选择和RRVPMCD的滚动轴承故障诊断方法[J].振动工程学报,2014,27(04):629-636.
Yang Yu,Pan Haiyang,Cheng Junsheng.Rolling bearing fault diagnosis method based on feature selection and RRVPMCD[J].Journal of Vibration Engineering,2014,27(04):629-636
[5] 姚旭,王晓丹,张玉玺,权文.特征选择方法综述[J].控制与决策,2012,27(02):161-166+192.
Yao Xu,Wang Xiaodan,Zhang Yuxi,Quan Wen.Summary of Feature Selection Methods[J].Control and Decision,2012,27(02):161-166+192.
[6] 刘望舒,陈翔,顾庆,刘树龙,陈道蓄.软件缺陷预测中基于聚类分析的特征选择方法[J].中国科学:信息科学,2016,46(09):1298-1320.
Liu Wangshu,Chen Xiang,Gu Qing,Liu Shulong,Chen Daoxu.Feature selection method based on cluster analysis in software defect prediction[J].Science in China: Information Science,2016,46(09):1298-1320.
[7] 谷广宇,刘建敏,乔新勇,姜红元,杨浩.基于特征评价的发动机寿命预测方法研究[J].汽车工程,2020,42(01):108-113+120.
Gu Guangyu,Liu Jianmin,Qiao Xinyong,Jiang Hongyuan,Yang Hao.Research on engine life prediction method based on feature evaluation[J].Automotive Engineering,2020,42(01):108-113+120.
[8] 刘胜兰,高凌寒,杜剑维,刘晨.基于自适应顺序的滚动轴承最优特征选取与寿命预测[J].舰船科学技术,2019,41(21):71-76.
Liu Shenglan,Gao Linghan,Du Jianwei,Liu Chen.Optimal feature selection and life prediction of rolling bearing based on adaptive sequence[J].Ship Science and Technology,2019,41(21):71-76.
[9] Hall, Mark. (2000). Correlation-based Feature Selection for Discrete and Numeric Class Machine Learning. Proceedings of the 17th international conference on machine learning (ICML-2000). 359-366.
[10] 毛文涛,徐文涛,薛天宇,何玲.一种基于特征子集区分度优化的分组特征选择算法[J].小型微型计算机系统,2015,36(08):1827-1831.
Mao Wentao,Xu Wentao,Xue Tianyu,He Ling.A grouping feature selection algorithm based on the optimization of feature subset discrimination[J].Small Microcomputer System,2015,36(08):1827-1831.
[11] 蒋盛益,王连喜.基于特征相关性的特征选择[J].计算机工程与应用,2010,46(20):153-156.
Jiang Shengyi, Wang Lianxi. Feature selection based on feature correlation [J]. Computer engineering and applications,2010,46(20):153-156.
[12] 吴登磊,汪宇玲,吴小龙,金安安.基于欧氏距离的K均方聚类算法研究与应用[J].数字技术与应用,2017(04):148-150.
Wu Denglei,Wang Yuling,Wu Xiaolong,Jin Anan.Research and application of K-means clustering algorithm based on Euclidean distance[J].Digital Technology and Application,2017(04):148-150.
[13] 程艳云,周鹏.动态分配聚类中心的改进K均值聚类算法[J].计算机技术与发展,2017,27(02):33-36+41.
Cheng Yanyun,Zhou Peng.Improved K-means clustering algorithm for dynamically assigning clustering centers[J].Computer Technology and Development,2017,27(02):33-36+41.
[14] Louhichi S , Gzara M , Hanène Ben Abdallah. A density based algorithm for discovering clusters with varied density[C]// Computer Applications & Information Systems. IEEE, 2014.
[15] Zhang B , Zhang L , Xu J . Degradation Feature Selection for Remaining Useful Life Prediction of Rolling Element Bearings[J]. Quality & Reliability Engineering International, 2016, 32(2):547-554.
[16] Nectoux P , Gouriveau R , Medjaher K , et al. PRONOSTIA: An experimental platform for bearings accelerated degradation tests[C]// IEEE International Conference on Prognostics and Health Management. IEEE, 2012.
[17] 张继冬,邹益胜,邓佳林,张笑璐.基于全卷积层神经网络的轴承剩余寿命预测[J].中国机械工程,2019,30(18):2231-2235.
Zhang Jidong,Zou Yisheng,Deng Jialin,Zhang Xiaolu.Remaining life prediction of bearing based on fully convolutional neural network[J].China Mechanical Engineering,2019,30(18):2231-2235.
[18] Zhang B , Zhang L , Xu J . Degradation Feature Selection for Remaining Useful Life Prediction of Rolling Element Bearings[J]. Quality & Reliability Engineering International, 2016, 32(2):547-554.
[19] Luxburg U V . Clustering Stability: An Overview[J]. Foundations and Trends in Machine Learning, 2010, 2(3):2010.
[20] 何志文,李夕海,刘代志,张斌.基于相关性分析的特征选择方法研究[J].核电子学与探测技术,2005(06):163-166+183.
He Zhiwen,Li Xihai,Liu Daizhi,Zhang Bin.Research on Feature Selection Method Based on Correlation Analysis[J].Nuclear Electronics and Detection Technology,2005(06):163-166+183.
[21] 武斌,李璐,宋建成,曲兵妮,李宏伟,杨健康.基于相似性的机械设备剩余使用寿命预测方法[J].工矿自动化,2016,42(06):52-56.
Wu Bin, Li Lu, Song Jiancheng, QU Bingni, Li Hongwei, Yang Xingsheng. Prediction method of residual service life of mechanical equipment based on similarity [J]. Industrial and mining automation,2016,42(06):52-56.

PDF(2028 KB)

449

Accesses

0

Citation

Detail

段落导航
相关文章

/