The design of an inkjet drive waveform using machine learning.

Area of Science:

• Materials Science
• Fluid Dynamics
• Computational Science

Background:

• Optimizing drive waveforms is crucial for reliable inkjet printing, but current methods rely on time-consuming manual parameter adjustments.
• Ink properties significantly influence waveform design, necessitating tailored approaches for different inks.

Purpose of the Study:

• To develop a closed-loop machine learning (ML) approach for designing optimal drive waveforms for satellite-free inkjet printing at a target velocity.
• To automate and accelerate the optimization of inkjet printing parameters, overcoming the limitations of manual methods.

Main Methods:

• Collected high-speed imaging data of jetting behavior for 11 model inks subjected to 1100 distinct waveform designs.
• Extracted large datasets on drop formation and velocity from images to train and compare five ML models.
• Utilized a Multi-layer Perceptron (MLP) model for its superior prediction accuracy in characterizing jetting behavior.

Main Results:

• The ML models accurately predicted inkjet jetting behavior based on waveform parameters and ink properties.
• The Multi-layer Perceptron demonstrated the highest prediction accuracy among the evaluated models.
• A closed-loop algorithm was successfully established, using the MLP to determine optimal waveform parameters for satellite-free drop formation at a target velocity.

Conclusions:

• The proposed closed-loop ML approach effectively designs optimal drive waveforms for satellite-free inkjet printing.
• This method significantly enhances the efficiency and reliability of inkjet printing processes by automating waveform optimization.
• The approach was validated by successfully printing an unknown ink using the recommended waveform parameters.