Application of MANOVA and Mahalanobis Distance in quality control

Many quality control problems are multivariate in nature since multiple process or product variables are measured and monitored in modern industry. Occasionally a multiple population, multi-variable scenario is encountered in parallel line manufacturing system. This work deals with the analysis of such system using multivariate techniques such as Canonical Analysis, MANOVA and Mahalanobis Distance. Using such techniques, the significant differences between the populations and the magnitude and directions of the variations within and between the populations were determined. Also, elimination/reduction of such variations using multivariate techniques led to a drastic improvement of the production system. A methodology for application of such multivariate techniques for the purpose of quality assurance and improvement was developed here. The application of this methodology was performed on a pager production line and the results obtained show the benefits and the feasibility of the use of multivariate techniques for on-line monitoring and quality control.