确定测试性验证试验方案的贝叶斯方法

现有的测试性验证试验方案在一定的风险约束下都需要较大的故障样本量。针对这一问题，以故障检测率(fault detection rate, FDR)为验证指标，提出利用研制阶段试验数据和专家信息制定测试性验证试验方案的贝叶斯方法。该方法首先利用研制阶段试验数据建立了产品的FDR增长模型，以此描述FDR在研制阶段的变化趋势，然后利用专家信息确定模型中的超参数，进而得到FDR的验前分布，最后依据贝叶斯最大后验风险准则制定了新的测试性验证试验方案。通过实例的对比分析表明，与经典试验方案相比，新方案样本量减少效果明显，在样本量保持不变的条件下双方风险大大降低。

Bayesian method for determination of testability demonstration test scheme

Existing testability demonstration test schemes need large fault sample size with a specified risk level. To solve this problem, taking the fault detection rate (FDR) as a target, one kind of Bayesian method for determining testability demonstration test scheme is proposed. The proposed method first establishes an FDR growth model describing the changing trend of FDR based on the test data acquired in the development stage and then employs the expert information to decide model parameters. After that, the prior distribution of FDR can be easily obtained through the above model. Finally, a new testability demonstration test scheme is defined according to the Bayesian maximum posterior risk rule. Compared with the classical counterpart, the new test scheme has an obvious effect on reducing the fault sample size, and lowering the risk of both sides greatly.