摘要
本文在逐步Ⅰ型混合截尾下,研究复合Rayleigh分布竞争失效产品部分步加寿命试验的统计分析问题。首先,基于复合Rayleigh分布竞争失效产品和损伤失效率(TFR)模型,运用极大似然理论和渐近似然理论,给出未知参数和加速因子的极大似然估计和渐近置信区间;然后,对未知参数和加速因子选取先验,利用MH抽样算法获得参数和加速因子的Bayes估计和最大后验密度置信区间(HPD);最后,通过Monte Carlo模拟对2种估计方法进行比较。结果表明:贝叶斯估计效果整体优于极大似然估计(MLE),在相同置信水平下,基于Bayes估计的HPD置信区间长度略短于MLE的渐近置信区间长度。
Under Type I hybrid censoring,the statistical analysis of step stress partially accelerated life tests for compound Rayleigh distribution competing failure products is studied.Based on the compound Rayleigh distribution competing failure products and tampered failure rate(TFR)model,the maximum likelihood estimation and asymptotic confidence interval of unknown parameters and acceleration factors are given by using the maximum likelihood theory and asymptotic approximation theory.The prior information of the unknown parameters and acceleration factors is selected,and the Bayesian estimation and the highest posterior probability density confidence interval(HPD)of the unknown parameters and acceleration factors are obtained by using the MH sampling algorithm.Finally,the two estimation methods are compared by Monte Carlo simulation.The results indicate that Bayesian estimation overall outperforms maximum likelihood estimation(MLE).At the same confidence level,the length of HPD based on Bayes estimation is superior to that of the asymptotic confidence interval based on MLE.
作者
朱艳
蔡静
龙芳
ZHU Yan;CAI Jing;LONG Fang(School of Data Science and Information Engineering,Guizhou Minzu University,Guiyang Guizhou 550025,China)
出处
《广西师范大学学报（自然科学版）》
CAS
北大核心
2024年第3期159-169,共11页
Journal of Guangxi Normal University:Natural Science Edition
基金
国家自然科学基金(11901134)
贵州省科技厅基金项目(黔科基础-ZK[2024]一般509)。
关键词
竞争失效
部分步加寿命试验
复合Rayleigh分布
极大似然估计
贝叶斯估计
competing failure
step stress partially accelerated life tests
compound Rayleigh distribution
maximum likelihood estimation
Bayesian estimation
