잉크젯 프린팅에서 기계학습을 이용한 액적 토출 웨이브폼 최적화

Abstract

Inkjet printing has involved a challenge to find proper jetting behavior. The drop can be ideally ejected by optimizing waveform and fluid modification. Although many pieces of research have studied the effects of them on drop ejection, the different results have been reported due to the complicated interrelation between them. We propose the machine learning approach for finding waveform which generates optimal jetting behavior at the specific ink. Machine learning can recognize the pattern in a complicated relationship. The characteristics of jetting behavior were extracted such as the drop velocity and the number of drops as outputs. The elements of waveform and Z number which represents fluid physical properties were selected as inputs. The data of drop ejection at various model inks collected from a high-speed imaging system. The predictive model of jetting behavior was built from a machine learning model that has an accuracy of more than 90%. The optimal waveform can be obtained through a feedback algorithm based on the predictive model. We confirmed that our proposed method found optimal waveform successfully at the unknown ink.