中文摘要：

针对多元工序质量相关性诊断问题,提出一种基于分组技术的诊断算法。利用相关性分解定理,将质量分量的整体相关关系分解为一系列分量对之间的相关关系;以分量之间的相关关系为依据,利用因子分析法对质量分量进行分组,使同一组内分量之间的相关关系尽可能大,不同组间的分量相关关系尽可能小;在忽略组间相关性的前提下,对同一组内的分量对建立相应的T2控制图,构成多元工序质量相关性诊断模型。理论分析和实践表明,对于分量之间的相关关系存在较大差异的多元工序质量,分组技术可以大幅度降低相关性诊断体系的规模。

英文摘要：

Aiming at the problem of correlation diagnosis in multi-variate process quality management, a diagnosis algorithm based on grouping technology was proposed. The integral relationship of all quality components was decomposed as a series of component pair＇s correlationship by using correlation analysis theorem. According to the relationship between components, the quality components were grouped by factor analysis method to maximize the correlations in same group and minimize the ones between groups. On the basis of neglecting the correlations between groups, the corresponding T2 control charts were built in same group of component pairs to form the diagnosis model. Theoretical analysis and practice proved that the grouping technology could reduce the scale of correlation diagnosis system greatly for the multivariate process quality whose relationships of different components had significant differences.