Study of ink release from Gravure cells using neural networks and CFD simulations

In rotogravure printing, engraved cells collect ink from a bath and then release it onto the substrate forming a series of printed dots that comprise the printed image. The purpose of this work was to explore the ink release in the rotogravure process to improve its predictability and scientific understanding. Two complimentary approaches have been investigated, an empirical approach embodied into Artificial Neural Network models and a numerical physical approach based on Computational Fluid Dynamics modelling. The Artificial Neural Network approach was based on a statistical correlation of experimental data on cell geometry with optical properties of the resulting print. The developed A.N.N. models were able to accurately predict the effect of cell geometry on ink release, outperforming traditional modelling techniques such as polynomial regression fitting techniques. The models were found to be practical and suitable to integration into manufacturing environments. The A.N.N. modelling highlighted the need for improved cell geometry data; to facilitate this, new software was developed for the automatic and accurate geometric characterisation of the engraved cells from interferometric profiles. A Computational Fluid Dynamic model of the ink release was successfully developed; the process was described as the evacuation of a Newtonian liquid from an axisymmetric cavity, showing the progressive splitting of the ink and retention of ink in the cell. This model is the first that takes into consideration the dynamic contact angle in the analysis of ink release from a cavity. The reliability of the numerical method and of its dynamic contact angle model was verified by comparing specifically designed models with experimental and literature data. The developed evacuation model shows the importance of the evacuation speed on the dynamics of the process and the critical importance of dynamic contact angle.