Analysis of Multivariate Industrial Process Data for Quality Pattern Detection in Masterbatch Coloring

L. Hisgen, T. Kubik (NanoP, THM University of Applied Sciences, Germany), S. Garbe (G.E. Habich Farben GmbH, Germany), J. Fischer (superus Datenmanagement GmbH, Germany), A. Kloes (NanoP, THM University of Applied Sciences, Germany), B. Iniguez (DEEEA, Universitat Rovira i Virgili, Spain), M. Schwarz (NanoP, THM University of Applied Sciences, Germany)

This paper investigates clustering techniques for revealing quality-related structures in industrial coloring process data. The multivariate dataset contains process data of paint mixtures labeled as either successful or unsuccessful based on quality inspection results. The proposed approach combines dimensionality reduction and statistical analysis with clustering techniques to explore relationships between mixture compositions and quality outcomes. Different clustering methods are compared in order to identify structural patterns in the ingredient composition of the mixtures. The results suggest that clustering methods capture structural patterns in ingredient usage.