基于非参数统计的加速退化试验优化设计

Optimal DOE for Accelerated Degradation Test Based on Non-parametric Statistical Model

加速退化试验优化设计研究如何在有限的费用约束下设计方案以减少试验评估误差。采用参数统计分析试验数据需要假设分布函数,当假设错误时会引起较大的误差,而采用非参数统计分析,通过数据本身来寻找合适的回归模型则可避免这一问题。以试验费用为约束,进行加速退化试验优化设计,给出各优化变量:试验应力水平、各应力水平下样本量、性能监测次数和监测间隔,最后通过仿真算例说明该加速退化试验优化设计方法的应用。

To obtain more useful data within the limited cost, optimal design of experiment （DOE） for accelerated degradation test （ADT） is studied. Assuming the distribution function of parameter is needed when using parametric statistical method to analyze the accelerated degradation data, and error will be caused if assuming a wrong distribution. This problem can be solved by using non - parametric statistical method to establish suitable regression model with the accelerated degradation data . The opti- mal DOE for ADT is conducted under experimental cost constraints. The optimal variables can be provid- ed includingstress levels, sample size, number of inspection and interval of performance inspection at each stress level. Finally the ADT optimization design method is explained with a simulation example.