Hybrid Machine Learning Method to Determine the Optimal Operating Process Window in Aerosol Jet 3D Printing

Area of Science:

• Materials Science and Engineering
• Additive Manufacturing
• Machine Learning Applications

Background:

• Aerosol jet printing (AJP) enables microelectronic device fabrication on flexible substrates but faces challenges in optimizing process parameters for quality.
• The complex interplay between print speed, sheath gas flow rate (SHGFR), and carrier gas flow rate (CGFR) significantly impacts printed line quality.

Purpose of the Study:

• To propose a novel hybrid machine learning method for determining optimal operating process windows in AJP.
• To enhance the efficiency and effectiveness of quality optimization in AJP across various design spaces.
• To develop a data-driven guideline applicable to other 3D printing technologies.

Main Methods:

• Utilized Latin hypercube sampling for experimental design in a 2D space.
• Applied K-means clustering to analyze the influence of SHGFR and CGFR on print quality.
• Employed support vector machines for optimal operating window determination.
• Incorporated transfer learning to efficiently identify windows at different print speeds.
• Used an incremental classification approach to determine a 3D operating process window.

Main Results:

• Successfully identified optimal operating process windows for AJP.
• Demonstrated significant reduction in required line samples for new print speed optimization via transfer learning.
• Established a comprehensive 3D operating process window balancing multiple parameters.
• Validated a data-driven approach superior to traditional experiment-based methods.

Conclusions:

• The proposed hybrid machine learning method effectively optimizes AJP quality by exploring and transferring knowledge across design spaces.
• This approach provides a robust guideline for quality optimization in AJP and potentially other 3D printing technologies.
• The method offers a more efficient and systematic alternative to traditional experiment-based optimization techniques.