基于D优化方法的CSADT设计

Planning of CSADT Based on D Optimization

考虑到试验设计人员更关注模型参数的估计精度,而传统试验优化方法以产品可靠性与寿命相关参数的预测精度为目标,提出将D优化方法引入恒定应力加速退化试验(CSADT)设计中。首先,用随机过程描述CSADT中产品性能退化的过程,通过对数似然函数,推导Fisher信息矩阵,基于D优化方法建立优化目标,以试验费用为约束条件,明确优化问题,给出最优试验变量:各应力水平、各应力下样本分配和试验时间分配。然后,应用该方法给出仿真实例。最后,通过模型参数偏差的敏感性分析,说明在一定偏差范围内,优化结果具有良好的稳健性。

Considering designers＇ interests in accuracy estimation of model parameters, D optimization was proposed to design CSADT （Constant Stress Accelerated Degradation Testing）, while traditional optimization method aims at the prediction accuracy of parameters related to reliability and lifetime of products. Stochastic process was used to describe a typical CSADT problem. The optimization problem was established by defining Fisher information matrix based on log-likelihood function. Under the constraint that the total experimental cost does not exceed a predetermined budget, optimization test variables, including stress levels, sample size, and testing time, at each stress level are given. Simulation examples were presented to demonstrate the proposed method. Sensitivity analyses showed that the optimization plan is robust within acceptable difference from the assumed value of parameters.