系统工程与电子技术 ›› 2019, Vol. 41 ›› Issue (5): 1162-1168.doi: 10.3969/j.issn.1001-506X.2019.05.32

• 可靠性 • 上一篇    下一篇

考虑随机失效阈值的设备剩余寿命在线预测

王泽洲, 陈云翔, 蔡忠义, 项华春, 罗承昆   

  1. 空军工程大学装备管理与无人机工程学院, 陕西 西安 710051
  • 出版日期:2019-04-30 发布日期:2019-04-30

Real-time prediction of remaining useful lifetime for equipment with random failure threshold

WANG Zezhou, CHEN Yunxiang, CAI Zhongyi, XIANG Huachun, LUO Chengkun   

  1. Equipment Management & UAV Engineering College, Air Force Engineering University, Xi’an 710051, China
  • Online:2019-04-30 Published:2019-04-30

摘要:

退化失效阈值是影响设备剩余寿命预测的重要因素。针对现有剩余寿命预测方法忽略失效阈值随机性影响的问题,提出考虑随机失效阈值的设备剩余寿命在线预测方法。首先,基于带测量误差与随机效应的非线性Wiener过程构建设备退化模型;其次,采用极大似然估计(maximum likelihood estimation, MLE)算法估计退化模型参数与随机失效阈值分布系数;然后,在考虑随机失效阈值的基础上推导出设备剩余寿命的概率密度函数(probability density function, PDF),并基于Bayesian原理实时更新退化模型参数,实现对剩余寿命的在线预测。最后,将该方法应用于陀螺仪剩余寿命在线预测分析,结果表明该方法能够有效提高剩余寿命预测的精度与准确性。

关键词: 剩余寿命预测, 随机失效阈值, 非线性Wiener过程, 随机效应, 测量误差

Abstract:

The degradation failure threshold is the important factor for the remaining useful lifetime prediction of equipment. Considering the current methods in which the influence of random failure threshold (RFT) can’t be considered, a real-time prediction method of remaining useful lifetime for equipment with RFT is proposed. Firstly, the nonlinear Wiener process with measurement error and random effect are used to model the degradation process. The maximum likelihood estimation (MLE) algorithm is used to estimate the parameters of degradation model and the distribution coefficients of RFT. Secondly, the probability density function (PDF) of remaining useful lifetime is derived by considering the RFT. Then, the parameter update method based on Bayesian theorem is presented to achieve the real-time prediction of remaining useful lifetime. Finally, the proposed method is used in the realtime prediction of gyroscope’s remaining useful lifetime and the results indicate that this method can effectively improve the precision and accuracy of remaining useful lifetime prediction.

Key words: remaining useful lifetime prediction, random failure threshold, nonlinear Wiener process, random effect, measurement error