论文部分内容阅读
针对电子整机系统结构复杂,失效机理众多,无法利用传统的加速模型外推对其寿命和可靠性特征进行分析的问题,提出一种基于顺序Dirichlet分布的分析模型,利用多应力、多水平的环境应力,对每一阶段上的失效率建立指数分布模型。通过先验信息和基于反应论的修正加速模型,给出各应力水平上的失效率先验信息,利用多变量顺序Dirichlet分布描述先验失效率概率密度函数,并根据先验信息对Dirichlet分布参数进行辨识设计和对参数物理意义进行阐述。根据恒加定数试验特点,提出似然函数的解析步骤,利用Gibbs拒绝抽样方法对Dirichlet后验分布进行推断分析,得到后验信息。最后分析一个实例,给出抽样过程和几个分位点上的失效率估计值,并比较正常状态下先验和后验的可靠度变化趋势,验证算法具有一定的效率,为电子整机寿命预测与评估提供一种新方法。
In view of the complicated structure and numerous failure mechanisms of the electronic complete machine system, it is impossible to analyze the characteristics of its life and reliability by extrapolating the traditional acceleration model. A model based on the ordered Dirichlet distribution is proposed. The multi-stress, multi-level Environmental stress, an exponential distribution model for failure rates at each stage. Based on the priori information and the modified acceleration model based on the reaction theory, the prior information of the failure rate at each stress level is given. The prior probability density function of the failure probability is described by the multivariate sequential Dirichlet distribution. The Dirichlet distribution parameters Identify the design and explain the physical meaning of the parameters. According to the characteristics of constant plus fixed number test, the analytic steps of the likelihood function are proposed, and the posterior information is obtained by using the Gibbs reject sampling method to infer the Dirichlet posterior distribution. Finally, an example is given to give the sampling process and several sub-points of the failure rate estimates, and compare the normal state of the prior and post-test reliability trends to verify the algorithm has a certain efficiency for the electronic machine life Prediction and evaluation provide a new method.