无失效数据条件下滚动轴承的寿命与可靠性评价

发布时间：2018-06-08 16:13

本文选题：滚动轴承 + 无失效数据　；参考：《中国计量学院》2013年硕士论文

【摘要】：随着滚动轴承生产制造水平的提高，其寿命也有了很大的提高，在可靠性试验过程中，会出现大量的无失效数据。特别在一些高成本、高可靠性的滚动轴承如风电轴承、高速列车轴承等的可靠性试验中，一般会选择小样本、无失效截尾试验。在此情况下，由于缺少失效信息，传统的评价方法，如GB24607-2009规定的最佳线性无偏估计等不能对滚动轴承的可靠性进行适当的评价。针对以上问题，必须研究新的评价方法。本文在前人研究的基础上，应用Bayes方法，对上述问题进行深入的理论和仿真研究，提出了两种适合于无失效数据情况下的滚动轴承可靠性评价方法及其模型。第一章，阐述了滚动轴承寿命试验及可靠性评价的重要意义，讨论了滚动轴承寿命分布--Weibull分布相关参数和可靠性指标，介绍了课题来源，在此基础上提出了论文的研究内容与研究重点。第二章，阐述了滚动轴承寿命的定义、试验原理、寿命与计算模型，论述了滚动轴承可靠性评价的基本理论与方法，讨论了Weibull分布模型的可靠度、失效概率、失效率、可靠寿命等指标间的关系。最后讨论了Weibull分布形状参数的几何意义和物理意义以及Bayes理论和Bayes统计模型。第三章，提出了基于虚拟信息构建的评价方法及模型。在Weibull分布无失效数据下，在每个截尾时间点的可靠度估计过程中引入前一个截尾时间点无失效样本的虚拟失效信息，使得对该截尾时间点的可靠度估计具有更高的可信度与更好的稳健性。在得出每个截尾时间点可靠度的估计值后通过加权最小二乘法得出Weibull分布的两个参数。实例计算表明，当可靠度先验分布中的超参数在一定的区间变化时，本文提出的方法比其它方法具有更好的稳健性。第四章，提出了基于拟合Weibull分布形状参数历史数据作为先验信息的评价方法与模型。根据形状参数的历史试验数据，拟合出形状参数的概率分布作为先验信息。将Weibull分布转化为指数分布，根据共轭先验分布构造原则构造出指数分布中失效率的先验信息。然后，以失效率和形状参数为切入点，结合无失效试验数据，得出失效率和形状参数的Bayes估计，进而计算出Weibull分布的特征寿命的估计。最后通过一组实例来验证估计结果的准确性，并讨论估计的稳健性。第五章，基于Matlab GUI，编制了滚动轴承可靠性评价软件。本软件利用本文的方法对滚动轴承的可靠性作出评价，，该评价系统主要由3个功能模块组成：Bayes估计模块，其中包括本文第三章提出的基于虚拟信息的Bayes估计方法、第四章提出的形状参数先验分布分别为均匀分布和Weibull分布的两种估计方法、茆诗松估计方法与吴来林估计方法共5种方法；Bayes估计稳定性评价模块，主要用于计算截尾时间变化时，以上5种估计方法的稳定性；形状参数的先验分布拟合模块，此模块会根据输入的数据判断出形状参数最符合Weibull分布、正态分布、和指数分布中的哪种分布，然后拟合出最佳分布。第六章，对全文的研究方法与研究结果进行了总结，提出了未来的研究方向与本研究的不足之处。
[Abstract]:With the improvement of the rolling bearing production and manufacturing level, its life has also been greatly improved. In the process of reliability test, there will be a lot of no failure data. Especially in the reliability test of some high cost and high reliability rolling bearings, such as wind power bearing and high speed train bearing, the small sample will be selected and no failure truncation test is made. In this case, due to the lack of failure information, the traditional evaluation method, such as the optimal linear unbiased estimation of GB24607-2009, can not properly evaluate the reliability of rolling bearings. In view of the above problems, a new evaluation method must be studied. Based on the previous research, the Bayes method is applied to the above problems. Based on the theory and simulation research, two reliability evaluation methods and models for rolling bearing are presented, which are suitable for the case of zero failure data.
In the first chapter, the important significance of the rolling bearing life test and reliability evaluation is expounded. The related parameters and reliability indexes of the --Weibull distribution of the rolling bearing life distribution are discussed, and the source of the subject is introduced. On this basis, the research content and the research emphasis of the paper are put forward.
In the second chapter, the definition of the life of rolling bearing, the principle of the test, the life and the calculation model, the basic theory and method of the reliability evaluation of the rolling bearing are discussed, and the relationship between the reliability, the failure probability, the inefficiency and the reliable life of the Weibull distribution model is discussed. Finally, the geometric meaning of the shape parameter of the Weibull distribution is discussed and the geometric meaning of the shape parameter is discussed. The physical meaning and the Bayes theory and the Bayes statistical model.
In the third chapter, the evaluation method and model based on virtual information construction is proposed. Under the Weibull distribution without failure data, the virtual failure information of the previous truncated time point without failure samples is introduced in the reliability estimation process of each truncated time point, which makes the reliability estimation of the truncated time point have higher reliability and better reliability. Robustness. After estimating the reliability of each truncated time point, two parameters of the Weibull distribution are obtained by the weighted least square method. The example calculation shows that the proposed method has better robustness than the other methods when the super parameters in the reliability prior distribution vary in a certain interval.
In the fourth chapter, the evaluation method and model based on the historical data of fitting Weibull distribution shape parameters as a priori information are proposed. According to the historical data of the shape parameters, the probability distribution of the shape parameters is fitted as a priori information. The Weibull distribution is transformed into an exponential distribution, and the index is constructed according to the principle of the conjugate prior distribution. A priori information of the failure rate in the cloth is given. Then, with the failure rate and shape parameters as the entry point, the Bayes estimation of the loss of efficiency and shape parameters is obtained by combining with the non failure test data. Then the estimation of the characteristic life of the Weibull distribution is calculated. Finally, a set of examples is used to verify the accuracy of the estimation of the results, and the robustness of the estimation is also discussed.
In the fifth chapter, based on Matlab GUI, the reliability evaluation software of rolling bearing is developed. This software uses this method to evaluate the reliability of rolling bearings. The evaluation system mainly consists of 3 functional modules: Bayes estimation module, including the Bayes estimation method based on virtual information proposed in the third chapter of this paper, and the fourth chapter The prior distribution of shape parameters is two estimation methods of uniform distribution and Weibull distribution respectively. There are 5 methods in the estimation method and the Wu Lailin estimation method, and the Bayes estimation stability evaluation module is mainly used to calculate the stability of the above 5 estimation methods when the truncated time changes are calculated, and the prior distribution of shape parameters is fitted to the module, According to the input data, this module will determine the shape parameters which are most consistent with the Weibull distribution, normal distribution, and which distribution in the exponential distribution, and then fit the best distribution.
The sixth chapter summarizes the research methods and results of the full text, and points out the future research directions and shortcomings of this research.
【学位授予单位】：中国计量学院
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TH133.33

【参考文献】