Session: 10-01 Interactive Presentations

Paper Number: 98765

98765 - Print Parameter Prediction Using Deep Learning Technique to Achieve Realized Electrical Performance and Geometry on Ink-Jet Platform 

In recent years printed electronics are rapidly gaining attention and are replacing the traditional manufacturing techniques, especially in consumer electronics. A closed-loop deep learning approach for correlation of the print parameters with realized electrical performance and geometry estimations on an ink-jet platform was attempted. For reliable printing with fine conductive traces, a prediction model of the changes in the print parameters and the realized print dimension is necessary. The inks used in this study were silver ink (particle and particle-free) and involved a comparative study of the same. A closed-loop control algorithm was used to attain the desired electrical and geometrical values by changing the print parameters without any user intervention. This was achieved by an automatic print parameter sensing system using a camera that captures the print to identify the geometry and dimension of the same. Once the realized print parameters are identified, a deep learning neural network regression model based on these parameters can be used to predict the desired input print parameters and is applied to achieve the desired geometry and dimension of the print. These new parameter values can be modified to attain the desired geometry and printing characteristics. This closed-loop system presents print characteristics sensing system using a camera as well as a deep learning neural network regression model to predict the new print parameters and an automatic update system for changing the values in the printing software. These combinations of the system are used to correlate the print parameters with the realized electrical performance and geometry of the print on an ink-jet printing platform. 

Presenting Author: Shriram Kulkarni Auburn University