Reinforcement Learning-Based Dynamic Optimization of Driving Waveforms for Inkjet Printing of Viscoelastic Fluids

In digital printing, the design and optimization of a driving waveform for piezoelectric printheads are critical for the precise patterning of functional materials. This study introduces an approach using a deep reinforcement learning (DRL) algorithm to dynamically control the waveform for functional inks, which vary in properties with environmental conditions. We developed a prediction model using a multilayer perceptron algorithm that accurately forecasts drop velocity and jetting morphology based on the ink's rheological properties and waveform parameters. Integrating this model into a DRL framework, we achieved precise control over the waveform, attaining a target drop velocity of 3 ms^-1 for quantum dot ink within 20 steps. Further, we implemented the trained DRL agent into a drop-watching system, enabling real-time waveform adjustment to maintain optimal jetting despite changes in ink properties due to temperature variations. Our results demonstrate the significant potential of machine learning for improving precision and adaptability of industrial inkjet printing processes.