航空学报 > 2015, Vol. 36 Issue (2): 564-574   doi: 10.7527/S1000-6893.2014.0312

基于随机Wiener过程的航空发动机剩余寿命预测

刘君强, 谢吉伟, 左洪福, 张马兰   

  1. 南京航空航天大学 民航学院, 南京 211106
  • 收稿日期:2014-03-12 修回日期:2014-11-05 出版日期:2015-02-15 发布日期:2014-11-15
  • 通讯作者: 刘君强,Tel.: 025-84896260 E-mail: liujunqiang@nuaa.edu.cn E-mail:liujunqiang@nuaa.edu.cn
  • 作者简介:刘君强 男,博士,讲师。主要研究方向:交通信息工程及控制、航空发动机健康管理等。Tel:025-84896260 E-mail:liujunqiang@nuaa.edu.cn;谢吉伟 男,硕士研究生。主要研究方向:交通信息工程及控制。E-mail:xiejiwei@nuaa.edu.cn;左洪福 男,博士,教授,博士生导师。主要研究方向:可靠性工程、维修理论、故障诊断与监控等。Tel:025-84893550 E-mail:rms@nuaa.edu.cn;张马兰 女,硕士研究生。主要研究方向:交通信息工程及控制。E-mail:zhangmalan@126.com
  • 基金资助:

    国家自然科学基金(61232002,60939003);中国博士后科学基金(2012M1081,2013T60537);江苏省博士后科学基金(1031107C);中央高校基本科研业务专项资金(NS2014066);江苏高校哲学社会科学研究项目(2014SJD041)

Residual lifetime prediction for aeroengines based on Wiener process with random effects

LIU Junqiang, XIE Jiwei, ZUO Hongfu, ZHANG Malan   

  1. College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China
  • Received:2014-03-12 Revised:2014-11-05 Online:2015-02-15 Published:2014-11-15
  • Supported by:

    National Natural Science Foundation of China (61232002, 60939003); China Postdoctoral Science Foundation (2012M1081, 2013T60537); Postdoctoral Science Foundation of Jiangsu Province of China(1031107C); Fundamental Research Funds for the Central Universities (NS2014066); Philosophy and Social Science Research Projects in Colleges and Universities in Jiangsu (2014SJD041)

摘要:

针对目前剩余寿命(RL)预测方法没有综合考虑发动机个体性能退化的差异性和多阶段性的问题,提出了基于多阶段性能退化模型预测航空发动机剩余寿命的方法。首先,该方法采用多阶段Wiener过程对航空发动机进行退化建模,并假设退化模型参数服从随机分布来描述发动机个体的差异性。然后,根据历史性能退化数据与历史失效时间数据,利用期望最大化算法对模型参数的先验分布进行估计。当获得单台发动机的实时退化数据后,使用Bayesian方法对模型参数进行更新,从而实时更新航空发动机的RL分布,最终实现对单台航空发动机的RL预测。实验结果表明,该方法预测精度较高,能为航空发动机维修计划的制定提供依据。

关键词: Wiener过程, 剩余寿命, 信息融合, 期望最大化算法, Bayesian方法

Abstract:

There are few models in consideration of the unit-to-unit variability and multi-phase variability simultaneously in the residual lifetime (RL) prediction for aeroengines, so propose a Wiener process-based degradation modeling method for RUL prediction considering the above-mentioned factors. First, this method models the degradation path for aeroengines conditioned on multi-phased Wiener process. Then, historical degradation data and failure-time data are fused to derive the prior distribution of the unknown parameters for degradation model and expectation maximization algorithm is used to estimate the hyper-parameters of the prior distribution. Once the real-time degradation data are available, posterior distribution of the parameters is updated through a Bayesian method. Lastly, the RL prediction is obtained based on the updated parameters. Experiment shows that the method can improve the accuracy of the RL prediction and can provide the decision-maker with enough information to perform necessary maintenance actions prior to the failure.

Key words: Wiener process, residual lifetime, information fused, expectation maximization algorithm, Bayesian method

中图分类号: