【作者】 刘强；

【导师】 周经伦；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2011， 博士

【摘要】 动量轮是卫星姿控系统的关键机电产品,要求在几年甚至十几年的任务时间内始终稳定可靠地工作。对这类卫星长寿命机电产品,基于大样本寿命试验的传统可靠性技术,难以解决其可靠性建模、试验设计与评估问题。本文提出基于失效物理的性能可靠性技术,从失效物理分析出发,综合基于失效物理的性能退化建模、基于性能退化的可靠性试验设计以及基于信息集成的可靠性评估技术,以动量轮为背景,解决卫星长寿命机电产品可靠性工作中的现实问题。论文主要研究内容如下:(1)研究基于失效物理的性能退化建模技术,解决卫星长寿命机电产品的性能可靠性建模问题。给出了一般化的由随机性模型和确定性模型构成的多项式耦合退化模型;提出了典型性能退化模型的MCMC参数估计法、两阶段参数估计法和EM迭代参数估计法;构建了基于性能退化模型的寿命分布建模方法,理顺了寿命分布和性能退化模型之间的关系,为产品寿命分布建模提供了理论支持。(2)研究基于性能退化的可靠性试验设计技术,解决试验费用昂贵、试验时间受限情形下,卫星长寿命机电产品可靠性试验中的关于检测频率、样本量、试验截止时间的方案设计问题。提出了一般化的基于性能退化的试验设计方法;针对威布尔寿命型和对数正态寿命型产品,分别提出了试验设计模型和实现算法;通过仿真分析证实了试验设计方法的合理性和有效性。(3)研究基于信息集成的可靠性评估技术,解决小子样条件下卫星长寿命机电产品的可靠性评估问题。针对实际中可能出现的各类型寿命数据、性能数据和专家判断信息等验前信息,分别提出了寿命数据的Bootstrap随机抽样挖掘方法、性能数据的EMD-AR模型随机抽样挖掘方法以及专家判断信息的参数区间随机抽样挖掘方法;给出了基于JS距离的验前信息集成模型,并构建了确定全局最优权重的模型和算法;针对威布尔寿命型和对数正态寿命型产品,分别提出了基于性能退化模型的可靠性建模方法,并结合验前信息集成方法给出了基于性能退化模型的贝叶斯可靠性评估方法。(4)基于上述理论方法,对xx型动量轮的可靠性展开实例研究。确定了动量轮的失效机理为轴承组件润滑系统失效,而导致其失效的主要因素是润滑剂的挥发和爬移;建立了动量轮的润滑剂动态微循环失效物理模型;得到了较高精度的动量轮可靠性评估结论,说明了本文的理论方法可以较好的解决动量轮的可靠性工程实际问题。更多还原

【Abstract】 As one of the most important electromechanical products embedded in the attitude control system of satellite, the Momentum Wheel has the requirement of stable working within many years. For this kind of long life products of satellite, the traditional reliability technology is not suitable for the reliability modeling, design of experiments and estimation. In this dissertation, the performance reliability technology based on physics of failure is proposed. With the view of physics of failure, the performance reliability modeling method, design of experiments method and reliability estimation method based on information fusion are integrated to address the problems existing in the reliability work of long life electromechanical products of satellite. The main research production of this dissertation includes the following aspects:(1) Study the performance reliability modeling method based on the physics of failure, solve the reliability modeling problem of long life electromechanical products of satellite. Put forward a general degradation model with the polynomial coupling form based on stochastic model and deterministic model. Propose the MCMC method, two stages method and EM iterative method for parameter estimation based on performance degradation model. Give the life distribution modeling method based on performance degradation model, and build the bridge between the life distribution model and performance degradation model, which provide the theory support for the life distribution modeling.(2) Study the method for design of experiment based on performance reliability, solve the design of experiment problem of long life electromechanical products of satellite. Propose a general optimal method for design of experiment based on performance degradation. For the Weibull life product and Lognormal life product, the optimal design model and algorithm is proposed respectively. Demonstrate the rationality and validity of these method based on simulation.(3) Study the reliability estimation method based on information fusion, solve the small sample problem of long life electromechanical products of satellite. For different prior information derived from practice, the data mining method based on Bootstrap random sampling, the data mining method based on EMD-AR model random sampling and the data mining method based on parameter interval random sampling are proposed respectively for life data, performance data and expert judgments. Give the prior information fusion model based on Jensen-Shannon divergence, and build the global optimal weight model and algorithm based on the fusion model. For the Weibull life product and Lognormal life product, the performance reliability estimation method is proposed, and the Bayesian method is proposed for the performance reliability estimation based on prior information fusion model. (4) Based on above methods, study the reliability estimation for xx-type Momentum wheel. Determine that the failure mechanism of the Momentum wheel is the failure of lubrication system of bear component, and the main factor of failure is the volatilization and creep of lubrication. Build the physics of failure model in the view of microcirculation dynamic system of lubrication. Get the more high-precision estimation for the reliability of Momentum wheel, and show that the theory of this dissertation can address the practical problem for the reliability engineering of the Momentum wheel.更多还原