贝叶斯原理的不确定度评定方法比较  

Comparison of Uncertainty Evaluation Method Based on Bayesian Principle

作　　者：姜瑞[1] 陈晓怀[1] 

机构地区：[1]合肥工业大学仪器科学与光电工程学院,安徽合肥230009

出　　处：《河南科技大学学报（自然科学版）》2016年第6期21-27,5,共7页Journal of Henan University of Science And Technology:Natural Science

基　　金：国家自然科学基金项目(51275148);合肥工业大学青年教师创新基金项目(JZ2014HGQC0126)

摘　　要：针对仅依据测量样本信息进行不确定度评定的局限性,利用贝叶斯信息融合原理,分别研究了基于无信息先验、共轭先验和最大熵先验分布的测量不确定度评定与更新方法,使评定过程充分融合历史先验信息和当前样本信息,提高了测量不确定度评定的可靠性。仿真实例表明:无信息先验方法没有将各组测量数据融合,其仿真结果波动最大;共轭先验方法仿真结果波动较大,经过多次数据融合逐渐趋于理论值;最大熵先验方法仿真结果波动较小,经过数据融合逐渐趋近于理论值。Aming at the limitation of the uncertainty evaluation only according to measuring sample information,the measurement uncertainty evaluation and updating method based on non-informative prior,conjugate prior and maximum entropy prior distribution were studied respectively by using the Bayesian information fusion theory. The evaluation process fully integrated the historical prior information with the current sample information,so that the reliability of the uncertainty evaluation was improved. The simulation examples show that the non-informative prior method does not integrate measurement data of each group and its simulation results fluctuates mostly. The conjugate prior method fluctuates greatly,and its simulation results gradually tend to theory value after multiple data fusion. The maximum entropy prior method fluctuates slightly and its simulation results gradually tend to theory value after data fusion.

关 键 词：不确定度评定 贝叶斯原理 无信息先验 共轭先验 最大熵先验 

分 类 号：TB92[一般工业技术—计量学]