多组SIMU误差系数验前分布的融合建模方法

Fusion Modeling Method of Prior Distribution of Error Coefficients for SIMU Based-on Multiple Source Informations

基于Bootstrap方法、随机加权法和数据融合理论 ,对某型捷联惯组误差系数的多组测试数据进行概率统计分析。首先 ,对 33个误差系数的多组验前总体和当前总体分别进行方差和均值的齐性检验 ,结果发现有一些参数通过了方差齐性检验 ,但均值基本上都不能通过齐性检验。为此 ,针对这种异源总体验前分布的建模问题 ,定义了验前信息的质量权重 ;其次 ,提出一种多组小样本数据参数的Bootstrap和随机加权融合估计方法 ,得到某型惯组多组误差系数的融合验前分布密度 ;最后结合对不同批次、不同使用单位惯组的多组动态数据的初步统计分析 。

Using Bootstrap method, random weight method and data fusion theory, the probabi lity statistical characteristics of error coefficients for SIMU (Strapdown Inert ia Measurement Unite) are analysed. First, normal property and abnormal values o f them are tested by Shapiro-Wilk method and F-test respectively. Second , the equ ality tests of variance and mean values of error coefficient are done, and found that every-group data of some error coefficients passed equality test of va rian ce, but mean values didn't. Third, for this case, we defined quality weight of p rior information that show difference between prior information and current info rmation. Four, we present a Bootstrap and random weight fusion estimation method for small sample of multiple sources with few test data. The prior p.d.f.s of e rror coefficients for some type SIMU are obtained. Last, some conclusions and ad vices are given.