A multivariate process quality correlation diagnosis method based on grouping technique.

Area of Science:

• Industrial Engineering
• Statistical Quality Control
• Multivariate Data Analysis

Background:

• Correlation diagnosis is crucial but challenging in multivariate process quality management.
• Existing methods may struggle with highly variable correlations between quality components.

Purpose of the Study:

• To propose a novel diagnostic method for multivariate process quality correlation.
• To simplify complex correlation diagnostic models through component grouping.

Main Methods:

• Established theorems on covariance matrix properties for product quality.
• Proved a correlation decomposition theorem.
• Applied factor analysis for quality component grouping to optimize intra- and inter-group correlations.
• Developed T² control charts for grouped component pairs.

Main Results:

• The proposed grouping technique significantly reduces the size of the correlation diagnostic model.
• The method effectively handles multivariate processes with varying correlations.
• Theoretical analysis and practical applications validate the approach.

Conclusions:

• The quality component grouping method offers a generalized theoretical model for correlation diagnosis.
• This approach enhances the efficiency and applicability of multivariate quality management.
• The technique is particularly beneficial for processes with diverse component correlations.