Related Experiment Video
Updated: Sep 29, 2025
14:16
Fluorescence detection methods for microfluidic droplet platforms
Published on: December 10, 2011
22.4K
Deep-Learning-Based Microfluidic Droplet Classification for Multijet Monitoring
Eunsik Choi1, Kunsik An2, Kyung-Tae Kang1
1Digital Transformation R&D Department, Korea Institute of Industrial Technology (KITECH), Ansan 15588, Republic of Korea.
ACS Applied Materials & Interfaces
|March 22, 2022
Summary
This study introduces a deep learning method using a convolutional neural network (CNN) to monitor inkjet printing droplet jetting. The system accurately evaluates jetting conditions in single-jet and multijet processes.
Area of Science:
- Manufacturing Engineering
- Materials Science
- Artificial Intelligence
Background:
- Inkjet printing is versatile and scalable for mass production.
- Monitoring nozzle performance is crucial for consistent droplet jetting.
- Variations in ink, nozzle conditions, and manufacturing affect jetting.
Purpose of the Study:
- To develop a deep-learning-based method for identifying droplet jetting status.
- To enable accurate and continuous monitoring of inkjet printing processes.
- To improve the reliability and efficiency of multijet printing.
Main Methods:
- Utilized a convolutional neural network (CNN) based on the MobileNetV2 model.
- Employed optimized hyperparameters for image classification of inkjet frames.
- Captured images using a CCD camera for analysis.
Main Results:
- Achieved high accuracy in evaluating droplet jetting conditions.
- Successfully demonstrated the method in a multijet printing process.
- Realized a test time of less than one second per image for evaluation.
Conclusions:
- The deep-learning approach provides an effective solution for monitoring inkjet printing.
- This method enhances the quality control and scalability of inkjet manufacturing.
- Automated jetting status identification improves process reliability.
