Evaluation of neural networks for modelling ink transfer in the gravure printing process

Modelling of the ink transfer in gravure printing presents significant difficulties due to the complexity of the process. This paper outlines the development and evaluation of Artificial Neural Networks for the modelling of the ink transfer. On the basis of experimental data, models were created for the estimation of the optical density of the print as result of engraved cell geometric characteristics (specifically screen ruling, screen angle, volume). The developed models have been shown to be accurate and reliable, being able to estimate the density within the normal human range of sensitivity to colours and outperforming more traditional modelling techniques such as polynomial regression fitting techniques.