逐步I型混合截尾下多级恒定部分加速寿命试验的统计分析-《集美大学学报（自然科学版）》

文章信息/Info

Title:: Statistical Analysis of Multi-Level Constant Stress Partially Accelerated Life Test Under Progressive Type-I Hybrid Censoring

Author(s):: FEI Tao1; CHEN Jianwei2; NI Wenqing1; (1.School of Science，Jimei University，Xiamen 361021，China;2.Department of Mathematics and Statistics，San Diego State University,San Diego 92182，USA)

Keywords:: partially accelerated life test; Weibull distribution; type-I progressively hybrid censoring; maximum likelihood estimation; Bootstrap method

摘要:: 在逐步I型混合截尾样本下，研究威布尔分布多级恒定部分加速寿命试验的参数估计问题。证明当分布参数和寿命试验的应力满足特定条件时，产品寿命形成几何过程，据此将恒定部分加速寿命试验的应力条件推广到多级。利用极大似然法得到未知参数和加速因子的极大似然估计，并分别采用渐近似然理论和Bootstrap方法构建参数的近似置信区间。最后通过蒙特卡罗模拟对模型推广前后的估计量进行均方根误差比较，并讨论不同移走方案下的参数估计效果。结果表明，多级恒定部分加速寿命试验下的估计效果较好。

Abstract:: The parametric estimation issues for a multi-level constant-partially accelerated life test (PALT) with Weibull distribution were studied under progressive type-I hybrid censored samples.Firstly，the condition under which product life follows a geometric process was derived from the distribution parameters and the stress in the life test.This result then serves as the basis for extending constant-PALT to multiple stress levels.Secondly，the maximum likelihood method was used to obtain maximum likelihood estimations of the unknown parameters and the acceleration factor.The approximate confidence intervals for the parameters were constructed using asymptotic likelihood theory and the Bootstrap method，respectively.Finally，the root mean square error of the estimators before and after the model extension was compared through Monte Carlo simulation，and the effects of parameter estimation under different removal schemes were discussed.The results indicated that the estimation outcomes under the multi-level constant-PALT are superior.