Prediction Of Lamination-Induced Colour Shifts For Uv Offset Printings By Using A Heuristic Approach As Well As Machine Learning Techniques

To investigate two approaches for the prediction of spectral colour changes, UV offset-printed test charts were laminated with polybutylene terephthalate films in three thicknesses and measured spectrophotometrically before and after lamination. This enabled the identification of the resulting colour shift for each patch. Mean colour deviations of 3.4 Delta E(00 )were determined while the strength of a colour shift depended on the initial colour patch lightness and the presence of paper white. To predict spectral reflections, a heuristic approach, based on the calculation of wavelength-dependent transmission of the lamination, is presented. A mean accuracy of 1.44 Delta E-00 between predictions and actual measured coated reflections was achieved. The method still shows potential for improvements in the prediction of the paper white after lamination. In a second approach, an artificial neural network (ANN) was applied to evaluate the performance of machine learning on this topic as well. After training and validation, using the ANN for spectral prediction led to a higher precision with a mean Delta E-00 of 0.6. In conclusion, both approaches obtain useful results whereby the ANN predictions are significantly more accurate. The investigation also demonstrates the potential of machine learning in the field of print and media technologies in general and in colour science in partiular.