About this Abstract |
Meeting |
2025 TMS Annual Meeting & Exhibition
|
Symposium
|
Additive Manufacturing Modeling, Simulation and Machine Learning
|
Presentation Title |
Machine Learning Guided Prediction of Printability During Additive Manufacturing |
Author(s) |
Stanford White, Yu Zhang, Samrat Choudhury, Mo Maniruzzaman |
On-Site Speaker (Planned) |
Stanford White |
Abstract Scope |
The quality of an Additive Manufacturing (AM) process is governed by a large set of processing parameters and environmental conditions. Determining the conditions for printability among billions of combinations of processing and environmental parameters is expensive and highly time-consuming. This study introduces a machine learning guided approach using Light Gradient Boosted Machines to optimize print quality in a piezoelectric inkjet printer for pharmaceutical drugs. An initial dataset of prints, rated from 0 to 4 with increasing printability was used to train the model. Differential evolution optimization was used to generate potential printing conditions maximizing the probability of class 4. Machine learning guided printing conditions were tested experimentally, added to the dataset, and the model was retrained iteratively until convergence. The insights gained in this work provide efficient methodologies for optimizing the quality of AM processes with only a subset of data generated as our machine learning iteratively guided subsequent experiments. |
Proceedings Inclusion? |
Planned: |
Keywords |
Additive Manufacturing, Machine Learning, Other |