Development of automatic charging control optimization technique for continuous-type inkjet printers using charged-droplet flight simulation and Bayesian optimization

Abstract

Conventionally, repeated experiments by trial and error have been used to improve the printing quality of continuous-type inkjet printers. This is because the trajectories of the ink droplets are affected by external forces such as air drag and Coulomb force and thus their analysis is difficult. In this study, a simulation technique to predict the trajectories of ink droplets, the Kriging model, and a multi-objective genetic algorithm were combined to develop an optimization system which determines a droplet charging pattern with better printing results. To enable parallel evaluation of the printing results with an arbitrary charging pattern, the simulation technique was developed based on OpenFOAM, which is an open-source software for computational fluid dynamics. A multi-objective optimization problem was defined by design variables which control the ink droplet charging pattern and by two objective functions which quantitatively evaluate printing quality from simulation results. The Kriging model was used to calculate the expected improvement of the two objective functions and to maximize this improvement by the multi-objective genetic algorithm (MOGA). The developed optimization system was applied to the digit “9” expressed by a 5×5 dot matrix. As a result, the expected improvement-based (EI-based) Bayesian optimization could find the solution which dominates most of the solutions obtained by the MOGA. Additionally, reaching this solution required only 1/7 of the number of simulations required by the MOGA optimization. The obtained solution was used for an experiment on an actual continuous-type inkjet printer and the quality of the printing results was evaluated. It was confirmed that the printing quality provided by the solution obtained by EI-based Bayesian optimization was sufficiently high, and the usefulness of the developed optimization system was demonstrated.

Keywords

• bayesian optimization
• continuous-type inkjet printer
• gaussian process regression
• kriging model
• multi-objective genetic algorithm
• numerical simulation
• openfoam
• printing quality control