GUM S1与基于贝叶斯方法的不确定度评估比较

论述了GUM S1与基于贝叶斯方法的不确定度评估,解析说明了GUM S1是一种选定参数且在特定先验信息下的贝叶斯方法。在一定的先验条件下对于线性系统,GUM S1与贝叶斯评估方法的结果是一致的;但对于非线性测量模型,两者的结果通常都不一样,其原因在于选择了不同的参数组以及先验信息的选取。通过两个测量模型的实例说明:对于线性系统,两种方法都可以使用;但对于非线性测量模型,且对被测量无先验信息时采用GUM S1所推荐方法能得到较为可靠的结果。在实际计量工作中若对被测量有了解,则可以采用贝叶斯分析方法。

Comparison of Uncertainty Evaluation between GUM S1 and Bayesian Analysis

Uncertainty evaluation method using GUM S1 and Bayesian analysis is proposed respectively. It is shown analytically that the GUM S1 solution is a special Bayesian analysis with certain parametrization and prior knowledge. When the model relation is linear and improper prior, GUM S1 and the Bayesian analysis yield the same results different results will emerge in general for non-linear model. The difference between the approaches is because of different parametrizations and different priors, which are illustrated by two examples. It is concluded that for a linear model both analyses can be applied, but for non-linear model, the GUM S1 approach may be preferred. When some prior knowledge about the measurement is available in practice work, it is proper to use the Bayesian analysis.