摘要
提出一种基于样本之间最小马氏距离的样本平均方法 ,从总体正常历史采样数据样本集合中 ,构造新的数据样本集 ,建立简化多元统计模型 .然后通过判断两数据集的质心偏移和协方差的差异程度来检验新的数据样本集对总体样本集的可代表性 ,从而达到用较少的有效样本代表总体样本统计特征的目的 .仿真结果表明用本文提出的简化多元统计模型进行故障诊断的效果与传统模型相同 。
An averaging sample approach based on the shortest Mahalanobis Distances (MD) among the samples is proposed. A new data set of samples is constructed from the data set of total normal historical samples for building multivariate statistical models. The representativity of the data set of new samples related to the total samples can be examined through comparing the deviation of centroids and difference of the covariance between the two data sets so that a goal is achieved which the statistical characteristi...
出处
《信息与控制》
CSCD
北大核心
2001年第S1期676-680,共5页
Information and Control
关键词
马氏距离
主元分析
多元统计模型
多元校验
可代表性
故障诊断
mahalanobis distance
principal component analysis
multivariate statistical model
multivariate calibration
representativity
fault diagnosis
