Machine Learning Enables Process Optimization of Aerosol Jet 3D Printing Based on the Droplet Morphology

Area of Science:

• Materials Science
• Engineering
• Computer Science

Background:

• Aerosol jet printing (AJP) offers high-resolution fabrication of flexible electronics but faces challenges in printing quality affecting device performance.
• Improving printing quality is crucial for advancing AJP technology in microelectronic device fabrication.

Purpose of the Study:

• To develop a hybrid machine learning method for analyzing and optimizing the AJP process based on deposited droplet morphology.
• To fundamentally improve printing quality by understanding the formation mechanisms of printed line characteristics.

Main Methods:

• Utilized space-filling experimental design (Latin hypercube sampling) to explore the 2D design space.
• Employed K-means clustering to link droplet morphology with printed line characteristics.
• Applied support vector machines for optimal operating window identification and Gaussian process regression for droplet property modeling and multi-objective optimization.

Main Results:

• Identified an optimal operating window for AJP based on deposited droplet morphology, ensuring improved printing quality.
• Successfully optimized droplet morphology for high controllability and desired thickness, addressing conflicting objectives of droplet diameter and thickness.
• Established a systemic investigation into the formation mechanisms of printed line characteristics.

Conclusions:

• The proposed hybrid machine learning method effectively optimizes AJP printing quality by focusing on deposited droplet morphology.
• This data-driven approach provides a guideline for enhancing printing quality in other noncontact direct ink writing technologies.
• The study significantly advances the potential of AJP for fabricating high-performance flexible electronic devices.