基于多寿命分布模型的贝叶斯建模与分析

Bayesian modeling and optimization based on multiple life distribution models

在可靠性寿命试验的数据分析中,当存在多种失效模式时,通常采用更加灵活的模型来拟合试验数据.此外,在加速寿命试验中通常会不可避免地产生删失数据。删失数据的存在使得可靠性模型的参数估计更加不可靠。针对上述问题,基于贝叶斯建模方法提出了一种考虑随机效应的多寿命分布模型.针对多种失效模式的情况,结合GLFP(general limited failure population)模型构建了贝叶斯模型考虑了随机效应对平均失效时间的影响,构建了平均失效时间与试验因子之间的函数关系.在贝叶斯建模与优化的框架下提出一种基于贝叶斯后验概率的可靠性改进方法.对某恒温器可靠性试验数据的研究表明,在考虑多种失效模式和随机效应的影响时,所提方法能够获得更加稳健和可靠的研究结果.

In the data analysis of reliability life experiments,a more flexible model is usually used to fit the experiment data when multiple failure modes exist.In addition,censored data are inevitably generated in ac-celerated life experiments.The existence of censored data makes the parameter estimation of reliability models more unreliable.To address the above problems,this paper proposes a multi-life distribution model based on Bayesian modeling approach considering random effects.Firstly,a Bayesian model is constructed for the case of multiple failure modes combined with the GLFP(general limited failure population)model.Secondly,the influence of random effects on the average failure time is considered,and a functional relationship between the average failure time and the experiment factors is constructed.Finally,a reliability improvement method based on Bayesian posterior probability is proposed in the framework of Bayesian modeling and optimization.The study of a thermostat reliability experiment data shows that the proposed method can obtain more robust and reliable research results when considering the effects of multiple failure modes and random effects.