Related Experiment Video
Updated: Sep 3, 2025
06:16
Scalable Stamp Printing and Fabrication of Hemiwicking Surfaces
Published on: December 18, 2018
7.4K
Investigating Machine Learning Techniques for Predicting the Process Characteristics of Stencil Printing
Péter Martinek1, Balázs Illés1, Norocel Codreanu2
1Department of Electronics Technology, Faculty of Electrical Engineering and Informatics, Budapest University of Technology and Economics, Műegyetem rkp. 3., H-1111 Budapest, Hungary.
Materials (Basel, Switzerland)
|July 27, 2022
Summary
Machine learning models accurately predict solder paste deposition in stencil printing. Regression trees outperformed other methods, offering a robust approach for electronics manufacturing optimization.
Area of Science:
- Materials Science and Engineering
- Manufacturing Processes
- Artificial Intelligence in Engineering
Background:
- Stencil printing is critical for mass assembly of electronic circuits via reflow soldering.
- Predicting solder paste deposition characteristics (area, thickness, volume) is essential for process control.
- Existing methods may lack the precision required for optimizing complex printing parameters.
Purpose of the Study:
- To investigate and compare various machine learning (ML) methods for predicting stencil printing outcomes.
- To identify the most effective ML model for accurately forecasting solder paste deposition.
- To establish a robust methodology for ML model optimization and validation in electronics manufacturing.
Main Methods:
- Experimental data generation by varying printing speed, stencil aperture size, and solder paste particle size.
- Assessment of multiple ML algorithms including Adaptive Neuro-Fuzzy Inference Systems (ANFIS), Artificial Neural Networks (ANN), boosted trees, regression trees, and Support Vector Machines (SVM).
- Hyperparameter optimization and cross-validation were employed to fine-tune models and prevent overfitting.
Main Results:
- Regression trees demonstrated the highest performance in modeling stencil printing characteristics.
- Artificial Neural Networks (ANN) with Bayesian regularization learning showed comparable, slightly lower performance.
- The study successfully optimized and compared diverse ML approaches for predicting solder paste deposition.
Conclusions:
- Machine learning, particularly regression trees, offers a powerful tool for predicting and optimizing stencil printing processes.
- The developed methodology for ML model fine-tuning and comparison is adaptable to various electronics manufacturing applications.
- Accurate prediction of solder paste deposition can enhance the reliability and efficiency of electronic circuit assembly.
