截尾试验下疲劳寿命分布的极值模型

针对截尾寿命试验数据缺少随机变量尾部分布和常规截尾数据分析对长寿命区高可靠度估计不准确的情况,根据极值统计学原理,给出一个结构元件疲劳寿命分布的极值模型.文中利用全样本试验数据构造截尾试验,采用传统的极大似然估计法给出寿命总体的分布形式,分析结构元件疲劳寿命的极值分布情况,建立一个新的寿命分布的极值模型,对长寿命区的寿命分布形式和高可靠度的估算有明显优势.三个算例表明,预测结果与全样本试验结果符合很好.

EXTREME-BASED FATIGUE LIFE DISTRIBUTION MODEL OF STRUCTURAL COMPONENTS IN CENSORED LIFE TESTING

In censored life tests for structural components, the cumulative density functions (CDF) of fatigue life obtained by conventional methods are not accurate and reasonable in their tail distributions, which are critical for the failure estimation of high-reliability components. Considering the stochastic nature of environmental influence and the randomicity of geometrical and mechanical properties of structural materials, fatigue life is a random variable and usually assumed to follow logarithmic normal distribution approximately. The distributions of extreme life are analyzed according to statistics of extremes that deduces the extreme values ofthe exponential type of initial distributions asymptotically follow the first (Gumbel) extreme distribution. Consequently, an extreme-based life distribution model is presented which has a more precise distribution especially in the long life region. Three examples are used to demonstrate the new life distribution model and the numerical results show that it has a better agreement with the experimental data than the conventional maximum likelihood method.