电子学报 ›› 2015, Vol. 43 ›› Issue (6): 1119-1126.DOI: 10.3969/j.issn.0372-2112.2015.06.013

• 学术论文 • 上一篇    下一篇

基于随机退化数据建模的设备剩余寿命自适应预测方法

孙国玺1, 张清华1, 文成林2, 段志宏1   

  1. 1. 广东石油化工学院广东省石化装备故障诊断重点实验室, 广东茂名 525000;
    2. 杭州电子科技大学自动化学院, 浙江杭州 310018
  • 收稿日期:2014-02-25 修回日期:2014-12-22 出版日期:2015-06-25 发布日期:2015-06-25
  • 作者简介:孙国玺 男,1972年10月出生,黑龙江友谊县人, 副教授,硕士生导师.2006年毕业于华南理工大学信号与系统专业,获工学博士学位,现为广东石油化工学院广东省石化装备故障诊断重点实验室副主任,主要从事石化装备的寿命预测、故障诊断和健康管理等方向的研究. E-mail:sguoxi@126.com; 张清华 男,1965年3月出生,广东梅州人,教授、博士生导师.1995年在华南理工大学获工业自动化专业硕士学位,2004年获控制理论与控制工程专业博士学位.现任广东省石化装备故障诊断重点实验室主任,主要致力于石化生产过程和生产装备的监测与故障诊断、复杂系统优化控制与仿真等方面的应用研究. E-mail:fengliangren@tom.com
  • 基金资助:

    国家自然科学基金(No.61473094,No.61174113);广东省战略性新兴产业核心技术攻关(No.2012A090100019);广东省普通高校特色创新项目(No.2014631041)

A Stochastic Degradation Modeling Based Adaptive Prognostic Approach for Equipment

SUN Guo-xi1, ZHANG Qing-hua1, WEN Cheng-lin2, DUAN Zhi-hong1   

  1. 1. Guangdong Petrochemical Equipment Fault Diagnosis Key Laboratory, Guangdong University of Petrochemical Technology, Maoming, Guangdong 525000, China;
    2. School of Automation, Hangzhou Dianzi University, Hangzhou, Zhejiang 310018
  • Received:2014-02-25 Revised:2014-12-22 Online:2015-06-25 Published:2015-06-25

摘要:

针对现有剩余寿命预测研究中需要多个同类设备历史数据离线估计模型参数的问题,本文提出了一种基于退化数据建模的服役设备剩余寿命自适应预测方法.该方法,利用指数随机退化模型来建模设备的退化过程,基于退化监测数据运用Bayesian方法更新模型的随机参数,进而得到剩余寿命的概率分布函数及点估计.区别于现有方法,本文方法基于设备到当前时刻的监测数据,利用期望最大化算法对模型中的非随机未知参数进行在线估计,由此无需多个同类设备历史数据.最后,通过数值仿真与实例分析,验证了本文方法在剩余寿命预测时的有效性.

关键词: 寿命预测, 退化, Bayesian方法, 期望最大化

Abstract:

Current prognostic studies are usually based on historical degradation data, which are collected off line from different devices in a population with the same type.However, such data are not always available in practice.Toward this end, this paper presents a degradation modeling based adaptive remaining useful life prediction method for equipments in service.In the presented method, we use an exponential-like stochastic degradation model to represent the degradation process of equipments.Then, based on the monitored data during the degradation process, Bayesian approach is applied to update the stochastic parameters in the model, so the probability distribution of the predicted remaining useful life is derived as well as its point estimation.Differing from current studies, all unknown non-stochastic parameters in the model are estimated by expectation maximization algorithm, without requiring historical degradation data of multiple devices.Finally, numerical simulations and case study results substantiate the superiority of the presented method in predicting the remaining useful life.

Key words: lifetime prognosis, degradation, Bayesian method, expectation maximization

中图分类号: