针对成败型系统在研制阶段的动态增长的可靠性评估问题,依据前期可靠性增长试验中多阶段的试验信息和专家经验,提出了基于新Dirichlet先验分布的可靠性增长的Bayes评估方法.该方法根据已有的不同阶段试验信息,利用离散AMSAA(Army Materi...针对成败型系统在研制阶段的动态增长的可靠性评估问题,依据前期可靠性增长试验中多阶段的试验信息和专家经验,提出了基于新Dirichlet先验分布的可靠性增长的Bayes评估方法.该方法根据已有的不同阶段试验信息,利用离散AMSAA(Army Material System Analysis Activity)可靠性增长模型描述可靠性增长试验中不同阶段可靠性的增长趋势,并对各个阶段的可靠性进行评估;针对下一阶段建立基于新Dirichlet分布的可靠性先验分布,并根据下一阶段可靠度的估计值采用最大熵模型给出先验分布参数估计;在获得现场试验数据的条件下,给出下一阶段的可靠度后验估计,并讨论不同区间范围对后验可靠性的影响.最后通过实例分析,表明相较于直接利用离散AMSAA模型和Beta先验分布,该方法在合适的专家经验的指导下能够给出更为准确的可靠性评估.展开更多
An advanced reliability growth model, i.e. exponential model, was presented to estimate the model parameters for multi-systems, which was synchronously tested, synchronously censored, and synchronously improved. In th...An advanced reliability growth model, i.e. exponential model, was presented to estimate the model parameters for multi-systems, which was synchronously tested, synchronously censored, and synchronously improved. In the presented method, the data during the reliability growth process were taken into consideration sufficiently, including the failure numbers, safety numbers and failure time at each censored time. If the multi-systems were synchronously improved for many times, and the reliability growth of each system fitted AMSAA (Army Material Systems Analysis Activity) model, the failure time of each system could be considered rationally as an exponential distribution between two adjoining censored times. The nonparametric method was employed to obtain the reliability at each censored time of the synchronous multi-systems. The point estimations of the model parameters, a and b, were given by the least square method. The confidence interval for the parameter b was given as well. An engineering illustration was used to compare the result of the presented method with those of the available models. The result shows that the presented exponential growth model fits AMSAA-BISE (Army Material Systems Analysis Activity-Beijing Institute of Structure and Environment) model rather well, and two models are suitable to estimate the reliability growth for the synchronously developed multi-systems.展开更多
文摘针对成败型系统在研制阶段的动态增长的可靠性评估问题,依据前期可靠性增长试验中多阶段的试验信息和专家经验,提出了基于新Dirichlet先验分布的可靠性增长的Bayes评估方法.该方法根据已有的不同阶段试验信息,利用离散AMSAA(Army Material System Analysis Activity)可靠性增长模型描述可靠性增长试验中不同阶段可靠性的增长趋势,并对各个阶段的可靠性进行评估;针对下一阶段建立基于新Dirichlet分布的可靠性先验分布,并根据下一阶段可靠度的估计值采用最大熵模型给出先验分布参数估计;在获得现场试验数据的条件下,给出下一阶段的可靠度后验估计,并讨论不同区间范围对后验可靠性的影响.最后通过实例分析,表明相较于直接利用离散AMSAA模型和Beta先验分布,该方法在合适的专家经验的指导下能够给出更为准确的可靠性评估.
文摘An advanced reliability growth model, i.e. exponential model, was presented to estimate the model parameters for multi-systems, which was synchronously tested, synchronously censored, and synchronously improved. In the presented method, the data during the reliability growth process were taken into consideration sufficiently, including the failure numbers, safety numbers and failure time at each censored time. If the multi-systems were synchronously improved for many times, and the reliability growth of each system fitted AMSAA (Army Material Systems Analysis Activity) model, the failure time of each system could be considered rationally as an exponential distribution between two adjoining censored times. The nonparametric method was employed to obtain the reliability at each censored time of the synchronous multi-systems. The point estimations of the model parameters, a and b, were given by the least square method. The confidence interval for the parameter b was given as well. An engineering illustration was used to compare the result of the presented method with those of the available models. The result shows that the presented exponential growth model fits AMSAA-BISE (Army Material Systems Analysis Activity-Beijing Institute of Structure and Environment) model rather well, and two models are suitable to estimate the reliability growth for the synchronously developed multi-systems.