基于极小子样的某列车齿轮箱箱体疲劳寿命可靠性评估

Reliability evaluation of gearbox based on extremely small sample data

针对采用Bootstrap方法对极小子样进行可靠性评估时,在重抽样的过程中样本向均值集中导致评估结果不够准确的问题,提出一种改进的Bootstrap可靠性评估方法。采用虚拟增广样本法对极小子样进行增广,增广至小样本。然后将增广后的小样本按照大小进行排序并分组,运用四分位差法和Bootstrap法对分组后的数据进行可靠性评估。最后,将该方法用于某列车齿轮箱箱体疲劳寿命的可靠性评估。评估结果表明:该方法可以较好地解决Bootstrap法无法对极小子样进行评估的问题,改善了Bootstrap法的抽样结果向均值集中的状况。

Aiming at the problem of using Bootstrap method to evaluate the reliability of very small sub-samples,the concentration of samples in the mean value during the resampling process leads to inaccurate evaluation results.The paper presents an improved Bootstrap method.Using the virtual augmented sample method augment the very small sub-samples to small samples.Then according to size the augmented small samples are sorted and grouped,and using the interquartile range method and Bootstrap method evaluate the grouped data for reliability.Finally,apply this method to reliability evaluation of gearbox based on extremely small sample data.Compared to the classical Bootstrap method,this method can solve the problem that the Bootstrap method cannot evaluate the extremely small sample data,and improves the situation that the re-sampling data of the bootstrap is concentrated to the mean.