Identification of oil contaminants on polymer coated beverage cans using Raman spectroscopy

Abstract

Impurities in beverage can production cause serious quality problems. In this paper, an alternative method to the analytical techniques currently used to characterise such impurities is suggested and tested. Rather than relying on conductivity measurements or visual tests, a spectroscopic approach using Raman microprobe equipment has been chosen. This allows for surface scanning with almost immediate chemical analysis. The process has been further combined with computational data analysis to increase detection limits and automate analysis by speeding the analysis process and also to insure repeatability of results. Specifically, application of linear discriminant analysis (LDA) allowed the successful creation of a classification model capable of identifying 14 potential contaminants in addition to the background polymer. Principal components analysis (PCA) was used to reduce the large amount of variables measured to a much smaller set to be used in the discrimination process. The detection limit of the Raman technique was found to be 0.1 μl volume of impurity deposited on the can, and at this volume 53% of all artefacts were classified correctly. With higher volumes deposited (0.4 μl and 2 μl) it was possible to achieve classification accuracies of 80% and 100% respectively.