先验不确定下零故障验证的贝叶斯稳健设计
Robust Bayesian Design of Zero-Failure Reliability Demonstration under Prior Uncertainty
DOI: 10.12677/aam.2026.1511442, PDF,   
作者: 张福原:青岛科技大学数理学院,山东 青岛;蒋昊良*:青岛科技大学信息科学技术学院,山东 青岛
关键词: 零故障试验;贝叶斯稳健设计;先验不确定性;Beta先验;Monte Carlo模拟;Zero-Failure Testing; Robust Bayesian Design; Prior Uncertainty; Beta Prior; Monte Carlo Simulation
摘要: 针对高可靠产品零故障验证中历史信息利用与先验设定不确定性并存的问题,本文提出一种贝叶斯稳健试验设计方法。基于Beta-二项模型,利用历史成功比例的Wilson区间刻画先验均值范围,以折扣有效样本量描述信息强度的不确定性,构造二维先验参数集合,并通过最坏情形后验可靠度超越概率约束确定最小试验次数。利用NIST Dataplot公开寿命数据开展数值分析,历史任务成功率为0.9717。当最低可接受可靠度为0.95、风险阈值为0.05时,经典零故障、固定先验贝叶斯和稳健贝叶斯方法所需试验次数分别为59、25和73;可靠度要求提高至0.99时,分别增至299、365和847。结果表明,历史信息并不必然节省试验量,其作用取决于先验与验证目标的相容性;稳健约束能够量化先验不确定性对应的额外试验需求。10万次Monte Carlo模拟所得通过频率与理论概率基本一致。研究为高可靠产品的试验资源配置与风险权衡提供了可解释的统计设计思路。
Abstract: To address the challenges of utilizing historical information while accounting for prior uncertainty in zero-failure reliability demonstration of highly reliable products, this study proposes a robust Bayesian test design method. Based on the Beta-binomial model, the Wilson confidence interval for the historical success proportion is used to characterize uncertainty in the prior mean, while the discounted effective sample size represents uncertainty in the strength of historical information. A two-dimensional set of prior parameters is constructed, and the minimum required sample size is determined by imposing a worst-case constraint on the posterior probability that reliability exceeds a specified threshold. Numerical analyses are conducted using publicly available lifetime data from NIST Dataplot, with a historical mission success rate of 0.9717. When the minimum acceptable reliability is 0.95 and the risk threshold is 0.05, the classical zero-failure, fixed-prior Bayesian, and robust Bayesian methods require 59, 25, and 73 tests, respectively. When the reliability requirement increases to 0.99, the corresponding sample sizes rise to 299, 365, and 847. The results indicate that historical information does not necessarily reduce testing requirements, as its effectiveness depends on its compatibility with the reliability target. The robust constraint quantifies the additional testing effort required to accommodate prior uncertainty. Monte Carlo simulations with 100,000 replications show close agreement between simulated acceptance frequencies and theoretical probabilities. This study provides an interpretable statistical design framework for balancing testing resources and risk in the reliability demonstration of highly reliable products.
文章引用:张福原, 蒋昊良. 先验不确定下零故障验证的贝叶斯稳健设计[J]. 应用数学进展, 2026, 15(11): 187-200. https://doi.org/10.12677/aam.2026.1511442

参考文献

[1] 李根成, 祁岭. 零故障定时截尾可靠性验证试验方案分析[J]. 火炮发射与控制学报, 2026, 47(4): 112-119.
[2] 何基报, 茆诗松. 对数正态分布场合无失效的BAYES验证试验方案[J]. 应用概率统计, 2000, 16(3): 239-248.
[3] 何婺晖, 唐家银. Weibull型无故障数据下综合Bayes置信可靠性评估模型[J]. 机械强度, 2020, 42(3): 610-616.
[4] 龙足腾, 郑波, 甯洋, 等. 无失效威布尔情形下基于双修正多层Bayes的可靠性评估[J]. 工程设计学报, 2024, 31(1): 59-66.
[5] 王成, 罗玉波. 二项分布参数区间估计——基于十三种方法的模拟比较[J]. 统计与管理, 2021, 36(3): 24-31.
[6] Martz Jr, H.F. and Waller, R.A. (1979) A Bayesian Zero-Failure (BAZE) Reliability Demonstration Testing Procedure. Journal of Quality Technology, 11, 128-138.
https://doi.org/10.1080/00224065.1979.11980894
[7] Coolen, F.P.A., Coolen-Schrijner, P. and Rahrouh, M. (2005) Bayesian Reliability Demonstration for Failure-Free Periods. Reliability Engineering & System Safety, 88, 81-91.
https://doi.org/10.1016/j.ress.2004.07.015
[8] Coolen, F.P.A. and Coolen-Schrijner, P. (2006) On Zero-Failure Testing for Bayesian High-Reliability Demonstration. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability, 220, 35-44.
https://doi.org/10.1243/1748006xjrr3
[9] Coolen, F.P.A. and Newby, M.J. (1994) Bayesian Reliability Analysis with Imprecise Prior Probabilities. Reliability Engineering & System Safety, 43, 75-85.
https://doi.org/10.1016/0951-8320(94)90096-5
[10] Berger, J.O., Moreno, E., Pericchi, L.R., Bayarri, M.J., Bernardo, J.M., Cano, J.A., et al. (1994) An Overview of Robust Bayesian Analysis. Test, 3, 5-124.
https://doi.org/10.1007/bf02562676
[11] Walter, G. and Coolen, F.P.A. (2018) Robust Bayesian Reliability for Complex Systems under Prior-Data Conflict. ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering, 4, Article 04018025.
https://doi.org/10.1061/ajrua6.0000974
[12] National Institute of Standards and Technology (2026) Dataplot: Datasets (Reliability/Tobias and Trindade, TOB159.DAT).
https://www.itl.nist.gov/div898/software/dataplot/datasets.htm
[13] Wilson, E.B. (1927) Probable Inference, the Law of Succession, and Statistical Inference. Journal of the American Statistical Association, 22, 209-212.
https://doi.org/10.1080/01621459.1927.10502953
[14] Hobbs, B.P., Carlin, B.P., Mandrekar, S.J. and Sargent, D.J. (2011) Hierarchical Commensurate and Power Prior Models for Adaptive Incorporation of Historical Information in Clinical Trials. Biometrics, 67, 1047-1056.
https://doi.org/10.1111/j.1541-0420.2011.01564.x