【作者】 韩宁；

【导师】 宋月；

【作者基本信息】 西安电子科技大学 ， 概率论与数理统计， 2014， 硕士

【摘要】 剩余使用寿命(Remaining Useful Life,RUL)预测是视情维修(Condition Based Maintenance,CBM)过程中核心问题之一,也是产品再制造过程中一个不可回避的关键问题。这一问题已经成为系统故障预测和健康管理(Prognostics and Health Management,PHM)领域的研究热点和挑战。根据剩余寿命预测结果分析,可以提高系统或设备可用性和可靠性,同时可以减少维修保障费用,降低失效事件发生的风险,降低或避免故障造成的重大损失,具有重要的研究和实用价值。首先,本文综述基于统计分析的剩余寿命预测的方法,主要分为回归模型的方法,马尔科夫(Markov)模型的方法,维纳过程的方法以及随机滤波的方法。文献中,针对维纳过程的方法对剩余寿命进行预测,提出了随机过程的线性退化模型和非线性退化模型,利用了贝叶斯估计和EM算法对模型的参数进行估计,得到了剩余寿命的概率密度函数。其次,在随机系数回归模型的基础上,为了获得更精准的剩余寿命的预测,建立了带测量误差的剩余寿命预测模型。本文运用了贝叶斯估计和EM算法相结合的方法对参数进行估计,得到了剩余寿命的概率密度函数。数值实验表明了测量误差对剩余寿命的影响,也说明了带测量误差的剩余寿命预测模型的有效性。最后,基于冲击载荷对设备的性能退化的影响,使得设备性能或者寿命加速达到阈值,建立了一个考虑冲击载荷和自然退化相结合的剩余寿命预测模型。估计模型参数运用了贝叶斯估计和EM算法相结合的方法,得到了剩余寿命的概率密度函数。实验说明了提出的模型更为符合实际情况,预测的结果也更加精确,也说明了该模型的可靠性与优越性。更多还原

【Abstract】 Remaining useful life(RUL) prediction is one of core problems of the Condition Based Maintenance(CBM) process, and it is also an inevitable key problem during product remanufacturing process. This issue has become a research focus and challenge in Prognostics and Health Management(PHM) areas. According to the analysis of RUL results can improve availability and reliability of device, reduces maintenance and support costs, reduces the risk of failure events, and reduces or avoids failure caused significant damage, so it has important research and practical value.Firstly, this paper outlines some methods about the remaining useful life prediction based on statistical analysis, the method mainly include regression model, the Markov model, Wiener process and stochastic filtering. The literatures present a linear degradation model and a nonlinear exponential degradation model based on the Wiener process. The parameters to be estimated by Bayesian estimation and EM algorithm, and the probability density function of remaining useful life is obtained.Secondly, on the basis of regression model with random coefficients, in order to obtain more accurate residual useful life. This paper uses Bayesian estimation and EM algorithm to estimate parameters. Numerical experiment shows the effectiveness of the method of RUL model with measurement error.Finally, due to shock load impacts on the performance degradation of device, the remaining useful life model considering the shock load and natural degradation rules is established. This paper uses Bayesian estimation and EM algorithm to estimate parameters, and gets the probability density function of residual useful life. Experiment shows that model is more consistent with the actual situation, the predicted results are more accurate, and it also shows the reliability and superiority.更多还原