Publication Date

8-2022

Date of Final Oral Examination (Defense)

6-27-2022

Type of Culminating Activity

Thesis

Degree Title

Master of Science in Electrical and Computer Engineering

Department

Electrical and Computer Engineering

Major Advisor

Harish Subbaraman, Ph.D.

Advisor

David Estrada, Ph.D.

Advisor

Kurtis Cantley, Ph.D.

Abstract

Printed electronics are emerging technologies that can potentially revolutionize the manufacturing of electronic devices. One promising technology for printed electronics is inkjet printing. Inkjet printing offers both low-cost processing and high resolution. Being a subset of additive manufacturing, inkjet printing minimizes waste and is compatible with a wide range of inks. However, inkjet printing of electronic devices is still in its infancy. One major challenge for inkjet printing is the complexity of the process optimization and uncertain high throughput production. To achieve a high-quality print, there is a complex parameter space of materials and processing parameters that needs to be optimized. To address this challenge, in this thesis work, we develop a machine learning algorithm to connect the processing parameters to print morphology for inkjet processes. To achieve this goal, we developed more than 200 experimental samples and processed the print images automatically with OpenCV-based codes. Finally, we correlated the morphology specifications, i.e., print line width, overspray, and roughness to the processing parameters, i.e., cartridge height, nozzle voltage, and drop spacing, via a neural network model. The order of machine learning model accuracy from high to low is for line width, roughness, and overspray, respectively. The model's low predictability of overspray can be attributed to our limited dataset, Dimatix unreliable performance, or the low dependency of overspray on the processing parameters of this study.

DOI

https://doi.org/10.18122/td.2007.boisestate

Download

Share

COinS