A Comparative Evaluation of Predictive Models for Optimized Droplet Classification in Inkjet Printing

Inkjet printing technology is emerging as a promising alternative to traditional silicon-based semiconductor fabrication, offering advantages such as a simplified manufacturing process, cost efficiency, environmental sustainability, while maintaining competitive performance in sensing and computing applications. However, the effective printing relies on an iterative, time-consuming trial-and-error approach to optimize key parameters for precise droplet formation. To overcome this limitation, this study develops a machine learning-based classification model to enhance efficiency and accuracy in droplet formation along with ink flow. Multiple machine learning algorithms, including Linear Discriminant Analysis, Support Vector Machine, Random Forest, Decision Tree, and K-Nearest Neighbors, are implemented using voltage, frequency, and meniscus as input features. After hyperparameter tuning, model performance is ranged from 73% to 98%, with the Decision Tree model achieving the highest accuracy of 98.46%. Based on this result, the optimized parameters from the Decision Tree model are applied to classify droplets, which ensures better quality inkjet printing.