A visualization decision support tool for multivariate SPC diagnosis using marginal CUSUM glyphs

Abstract

A marginal cumulative sum (CUSUM) glyph is proposed to visualize and decompose out-of-control signals over time. The proposed visualization tool, consisting of a two-sided CUSUM and star glyphs, is capable of indicating when and which variables contribute to the cause. Complementing traditional multivariate control charts after indicating that the process under monitoring is out of control, the proposed method provides a visualization tool with additional diagnostic information. In addition, the proposed tool is capable of handling responses with high dimensions. Extensive simulation results conducted for up to 20 dimensions provide a user guideline of how to implement the proposed methodology.