基于调和均值的疲劳寿命经验分布函数

Empirical Distribution Function of Fatigue Life Based on Harmonic Mean

该文基于调和均值的定义,应用积分概率变换,提出一个新的反映顺序疲劳寿命失效概率分布集中趋势的经验分布函数,并在此基础上给出极值型检验统计量及其不同显著度水平下的临界值;考虑疲劳寿命拟合实际情况,利用随机贝塔分布函数构建备择分布,提出不依赖分布形式的检验功效数值计算方法。理论分析和数值模拟得出基于调和均值经验分布函数的检验方法较常规的K-S检验具有更高的检验功效。

Employing the harmonic mean and the probability integral translation （PIT）, this paper analyzes the skewness and kurtosis characteristics of the failure probability distributions of order fatigue life variables, proposes a new empirical distribution function （EDF） which represents the central tendency of the failure probability of order statistics, and then constructs a new modified-EDF test statistic based on the extreme discrepancy criterion. Monte Carlo method is employed to obtain the critical values corresponding to different significance levels and to conduct test power comparisons with a distribution-free method of test power computation. The results of numerical examples demonstrate the advantages related to the use of the suggested modified-EDF goodness-of-fit test especially for data of small and moderate sample sizes over traditional K-S test.