南京大学学报(自然科学版) ›› 2019, Vol. 55 ›› Issue (2): 161–169.doi: 10.13232/j.cnki.jnju.2019.02.001

• •    下一篇

基于支持向量回归与Bayesian理论的性能退化产品可靠性分析

冯海林，王雨薇*   

  1. 西安电子科技大学数学与统计学院，西安，710126
  • 接受日期:2018-10-08 出版日期:2019-04-01 发布日期:2019-03-31
  • 通讯作者: 王雨薇 E-mail:870121009@qq.com
  • 基金资助:
    国家自然科学基金(71271165)

Reliability analysis of performance degradation product based on support vector regression and bayesian theory

Feng Hailin，Wang Yuwei*   

  1. School of Mathematics and Statistics，Xidian University，Xi’an，710126，China
  • Accepted:2018-10-08 Online:2019-04-01 Published:2019-03-31
  • Contact: Wang Yuwei E-mail:870121009@qq.com

PDF (PC)

830

摘要： 为提高小样本非线性下的性能退化产品可靠性评估准确性，提出一种融合性能退化数据与寿命数据的可靠性评估方法. 首先选定产品的性能退化变量并记录下与时间相关的退化数据，随后利用支持向量回归(Support Vector Regression，SVR)理论对性能退化轨迹进行拟合建模，再通过首达阈值的概念推算出产品伪失效寿命数据. 假设产品寿命服从特定的概率分布，则可求得产品的可靠性解析表达式. 为了充分利用可得到的多种信息，增加评估的可信度，结合Bayesian理论，融合可得的真实寿命数据，一旦得到新的失效寿命数据，便可更新迭代模型参数提高评估准确性. 最后通过数值仿真与陀螺仪退化数据的实例验证所提方法的可行性. 结果表明，本研究有效地解决了小样本非线性情形下性能退化产品可靠性分析不精确的难题．

关键词: 支持向量回归, 性能退化, 数据融合, Bayesian理论, 可靠性

Abstract: The purpose of this work is to improve the accuracy of product reliability analysis，when the product performance degradation in the case of small data volume and non－linear conditions. As the working time of the product increases，the physical characteristics of the product tend to undergo some physical changes. Therefore，when the life failure data is insufficient，we can predict the working life and reliability of the product by measuring these time－varying performance degradation data. Firstly，based on the characteristic of the product，the performance degradation variable most relevant to the life is selected. As the working hours increase，the amount of performance degradation of the product is recorded in an orderly manner. Then，combined with the friendly advantage of the support vector for small sample data，using Support Vector Regression(SVR) theory to fit and model performance degradation trajectory，and pseudo－failure life data of the product is derived by the concept of exceeding the failure threshold for the first time. Since the product’s life data is usually subject to a specific probability distribution，we can calculate the reliability analysis expression of the product. In actual engineering practice，there are a variety of information related to product life. In addition to the pseudo－failure data that is introduced through performance degradation data，we can also get a small amount of product true failure data. Bayesian theory is a conditional probability about random events，which can effectively combine multiple kinds of information. In order to make the most of these available information，we use Bayesian theory to fuse the available real life data to update the model parameters. In this way，once new real life data are obtained，the model parameters can be updated. Finally，the effectiveness of the proposed method is verified by numerical simulation and gyroscope degradation data. The results show that this study effectively improves the inaccuracy of reliability analysis of performance degradation products in small sample and nonlinear cases.

Key words: Support Vector Regression(SVR), performance degradation, data fusion, Bayesian theory, reliability

中图分类号: 

  • TB114.3
[1]　Dasgupta A，Pecht M. Material failure mechanisms and damage models. IEEE Transactions on Reliability，1991，40(5)：531－536.
[2] 庄东辰，茆诗松. 退化数据统计分析. 北京：中国统计出版社，2013.
[3] 王　亭，徐厚宝，李泠泽等. 性能退化数据下的VDMOS可靠性分析. 数学的实践与认识，2017，47(7)：161－167.(Wang T，Xu H B，Li L Z，et al. Reliability analysis for VDMOS with degradation data. Mathematics in Practice and Theory，2017，47(7)：161－167.)
[4] Tang S J，Yu C Q，Feng Y B，et al. Remaining useful life estimation based on Wiener degradation processes with random failure threshold. Journal of Central South University，2016，23(9)：2230－2241.
[5] Son K L，Fouladirad M，Barros A. Remaining useful lifetime estimation and noisy gamma deterioration process. Reliability Engineering & System Safety，2016，149：76－87.
[6] Santhosh T V，Gopika V，Ghosh A K，et al. An approach for reliability prediction of instrumentation & control cables by artificial neural networks and weibull theory for probabilistic safety assessment of NPPs. Reliability Engineering & System Safety，2018，170：31－44.
[7] Sun Y，Yan Q Q，Jiang Y L，et al. Reliability prediction model of further bus service based on random forest. Journal of Algorithms & Computational Technology，2017，11(4)：327－335. DOI：10. 1177/1748301817725306
[8] Huang R Q，Xi L F，Li X L，et al. Residual life predictions for ball bearings based on self－organizing map and back propagation neural network methods. Mechanical Systems and Signal Processing，2007，21(1)：193－207.
[9] Gebraeel N，Lawley M，Liu R，et al. Residual life predictions from vibration－based degradation signals：a neural network approach. IEEE Transactions on Industrial Electronics，2004，51(3)：694－700.
[10] Cortes C，Vapnik V. Support－vector networks. Machine Learning，1995，20(3)：273－297.
[11] Vapnik V N. Statistical learning theory. New York：Springs－Verlag，1998.
[12] Cristianini T J，Taylor J S. 支持向量机导论. 李国正，王猛，曾华军译. 北京：电子工业出版社，2004.
[13] 陈　龙，谭继文，姜晓瑜. 基于TESPAR和LS－SVM算法的滚动轴承退化趋势预测. 煤矿机械，2017，38(8)：18－20.(Chen L，Tan J W，Jiang X Y. Degradation trend prediction of rolling bearings based on TESPAR and LS－SVM algorithm. Coal Mine Machinery，2017，38(8)：18－20)
[14] 冯　帅，冯金富，王　聪等. 基于最小二乘支持向量机退化数据的可靠性评估. 机械强度，2016，38(3)：509－514.(Feng S，Feng J F，Wang C，et al. Method of degradation data reliability evaluation based on the least square support vector machine. Journal of Mechanical Strength，2016，38(3)：509－514.)
[15] 胡友涛，范金锁，胡昌华. v支持向量回归用于退化轨迹建模. 系统工程与电子技术，2017，39(1)：231－236.(Hu Y T，Fan J S，Hu C H. Modeling degradation path using v－support vector regression. Systems Engineering and Electronics，2017，39(1)：231－236.)
[16] Choi J，An D，Won J. Bayesian approach for structural reliability analysis and optimization using the kriging dimension reduction method. Journal of Mechanical Design，2010，132(5)：051003.
　 [17] 裴　洪，胡昌华，司小胜等. 融合寿命数据与退化数据的剩余寿命估计方法. 电光与控制，2016，23(9)：90－95.(Pei H，Hu C H，Si X S，et al. Remaining useful lifetime estimation by combining lifetime data with degradation data. Electronics Optics & Control，2016，23(9)：90－95.)
[18] 陈秀荣，李　娟，于加举. 融合寿命数据和退化数据的防喷阀剩余寿命预测. 山东科技大学学报(自然科学版)，2017，36(5)：23－28.(Chen X R，Li J，Yu J J. Remaining lifetime prediction of blowout preventer valve based on fusion of lifetime data and degradation data. Journal of Shandong University of Science and Technology(Natural Science)，2017，36(5)：23－28.)
[19] 胡友涛，胡昌华. 一种基于遗传算法优化小波支持向量回归机的实时寿命预测方法. 上海交通大学学报，2011，45(8)：1216－1220.(HU Y T，HU C H. A real time lifetime prediction method based on－wavelet support vector regression optimized by GA. Journal of Shanghai Jiao Tong University，2011，45(8)：1216－1220.)
[20] Vapnik V N. The nature of statistical learning theory. New York，NY：Springer，1995.
[21] Wang P F，Youn B D，Xi Z M，et al. Bayesian reliability analysis with evolving，insufficient，and subjective data sets. Journal of Mechanical Design，2009，131(11)：111008.
[22] 彭宝华，周经伦，孙　权等. 基于退化与寿命数据融合的产品剩余寿命预测. 系统工程与电子技术，2011，33(5)：1073－1078.(Peng B H，Zhou J L，Sun Q，et al. Residual lifetime prediction of products based on fusion of degradation data and lifetime data. Systems Engineering and Electronics，2011，33(5)：1073－1078.)
[23] Si X S，Wang W，Hu C H，et al. Remaining useful life estimation－A review on the statistical data driven approaches. European Journal of Operational Research，2011，213(1)：1－14.
[1] 黄华娟,韦修喜. 基于自适应调节极大熵的孪生支持向量回归机[J]. 南京大学学报(自然科学版), 2019, 55(6): 1030-1039.
[2]  黄　辉1，2，3*，马思佳1，2，王　庆3，任洪强1，2，3. 多参数影响下污水总氮浓度预测最优方法研究[J]. 南京大学学报(自然科学版), 2017, 53(6): 1194-.
[3]  李新玉1，徐桂云1，任世锦2*，杨茂云1，2.  基于可靠性的正则化加权软k－均值的子空间聚类[J]. 南京大学学报(自然科学版), 2017, 53(3): 525-.
[4] 陈湘军1,2* , 施静静1 , 张 鹏1 , 陈启美1
. 车载智能终端上下文感知及系统实现[J]. 南京大学学报(自然科学版), 2015, 51(2): 243-250.
[5] 张仕光1,2，米据生1,3**, 胡清华4. 粗糙-支持向量回归模型[J]. 南京大学学报(自然科学版), 2013, 49(5): 650-654.
[6] 张仕光;胡清华;谢宗霞;米据生;. 基于Beta噪声模型支持向量回归及其应用[J]. 南京大学学报(自然科学版), 2013, 49(4): 418-424.
[7]  赵研，卢奕南**，权勇
.  基于模糊积分多源数据融合的蛋白质功能预测*
[J]. 南京大学学报(自然科学版), 2012, 48(1): 63-69.
[8]  孟海宁 1 ** , 齐 勇 2 , 侯 迪 2 , 刘 亮 3.  基于系统性能周期检测的应用服务器软件再生建模与分析* 

[J]. 南京大学学报(自然科学版), 2010, 46(4): 440-447.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
[1] 魏　桐，童向荣. 基于加权启发式搜索的鲁棒性信任路径生成[J]. 南京大学学报(自然科学版), 2018, 54(6): 1161 -1170 .
[2] 汪贵庆, 袁杰, 沈庆宏. 基于精英蚁群算法的交通最优路径研究[J]. 南京大学学报(自然科学版), 2019, 55(5): 709 -717 .
[3] 刘瑞红,王晖,陶农建. 电位调制下单纳米颗粒的等离激元成像研究[J]. 南京大学学报(自然科学版), 2019, 55(5): 813 -818 .
[4] 李佳云,吴人杰. 基因转录爆发的建模研究[J]. 南京大学学报(自然科学版), 2020, 56(3): 418 -429 .

版权所有 © 2021 《南京大学学报(自然科学)》

地址：江苏省南京市鼓楼区汉口路22号，南京大学学报编辑部，210093

电话：025-83592704 , 传真：025-83592704        技术支持：北京玛格泰克科技发展有限公司