摘要
提出成套动态测试数据自动处理方案．首先通过两次移动平均及相应的统计检验，自动初辨数据属于平稳型、一阶非平稳型或二阶非平稳型，同时用移动回归外推法自动剔除异常数据，并用ＲＰＥ－ＫＦ自适应滤波法抑制测试随机误差．然后按初辨结果自动选定加权逐步回归或／与时序模型特征根判别的混合谱分析法及其算法参数，提取数据中的非周期或／与周期成分，并用移动Ｍａｒｐｌｅ算法拟合随机成分的时变ＡＲ模型，估计其统计特征量．已据其编制成软件，主要适用于几何量或与之相当的其它非急剧变化的物理量动态测试数据处理．
Proposes an overall scheme for the automatic data processing of dynamic measurement and it's software package.The stationary,first older or second order.nonstationary type of data can be automatically decided by twice moving averaging methods and co ～ponding test of hypothesis.Outliers can be rejicted by judgements such as moving-regression and extraploation.Adaptive filter is employed to depress the random errors of measurement.The trend components and period components of data can be obtained by weightal stepwise regression and the mixded spectral analysis respectively.The moving Marple's algorithm is used to fit the-varying AN model of the random component of data and to calculate it's time-varying characteristic variables.The software is suital for data processing of dynamic measurement of geometric and other physical quantities when the mean value and variance does not change too rapidly.
出处
《北京理工大学学报》
EI
CAS
CSCD
1995年第1期67-74,共8页
Transactions of Beijing Institute of Technology
关键词
数据处理
时变模型
动态量仪
自动检测
data processing/nonstationary,time-varying model,moving algorithm
