Autonomous Control of a Roll-to-Roll Printing Device

Roll-to-roll (R2R) slot-die coating is widely used in industry in the manufacture of a diverse range of products; e.g. lithium-ion batteries, solar cells and optical films. The printing of such films is dependent on the precise control of the printing parameters to control coating properties such as film thickness and width. During the coating process, an ink is pumped into the slot-die and exits through a narrow opening onto a moving substrate (web). The optimisation of roll-to-roll printing parameters is commonly achieved through a process of trial and error, relying on the skill and experience of the operator. When ink formulations are changed due to material supply issues or advancements in the materials used, a change in the printing parameters is required. Furthermore, disturbances to the optimised process conditions can arise during a coating run from changes in pump pressure, changes in web velocity and changes in the gap between the slot-die and the substrate.<br/><br/>Here, we will present details of a lab scale R2R printing rig that we have constructed. This rig features various additional sensors not found on industrial R2R rigs, including cameras mounted side-on and face-on to the slot-die print head. We have written a set of LabView VIs both to operate the rig and to process and analyse the data from the sensors. Autonomous control of the printing parameters is provided by a machine learning (ML) function.<br/><br/>Our use of cameras focussed on the slot-die head provides real-time in-situ data. This enables the ML function to respond in real-time during a coating run to disturbances in the parameters of the printed film. Changes to the slot-die gap, web speed and ink flow rate can be made by the ML function to modify and maintain the film width and thickness. We also demonstrate that a third camera positioned a short distance from the slot-die and above the web, can be used to monitor and provide feedback to the ML function on the cross-profile of a clear film using a structured light setup.<br/><br/>In conclusion, we will show that our use of sensors provides real-time monitoring of the printed film and that this stream of data can be applied to a ML function for automatic optimisation and responsive control of the printing parameters. Our rig is able to automatically respond to disturbances in the printing parameters that can arise from, for example, changes in the gap due to asymmetry of the pull-on roller or changes in pump pressure. This represents a significant advance over current industrial methods where the printing parameters are fixed at the start of a printing run.