Machine learning (ML) and three-dimensional (3D) printing are rapidly advancing fields that offer significant potential for modern manufacturing. ML enables computers to autonomously learn from data, while 3D printing allows for creating intricate, multi-material structures with minimal manufacturing expertise. However, determining the optimal printing parameters remains challenging, often leading to increased pre-printing time and material waste.

In this study, we apply ML techniques to understand the relationship between processing parameters and mechanical properties in 3D-printed parts, aiming to optimize these parameters. Our project also focuses on involving undergraduate students in using basic ML techniques and affordable 3D printers to explore parameter settings, process control, and experimental design in Additive Manufacturing (AM). Through this approach, we extract meaningful patterns from the data, reducing the need for extensive experimentation while achieving comparable results. This educational approach seeks to standardize and generalize the learning experience in ML-based quality control of AM.

We use a MakerBot Replicator+ to print a daily-use part, varying process parameters such as extruder temperature, printing speed, infill density, infill pattern, layer height, and wall layers. Our analysis identifies infill density as the most significant factor influencing tensile strength, with other parameters like printing speed, infill pattern, and wall layers also contributing. Conversely, extruder temperature and layer height have minimal impact. Our findings highlight the potential of ML to optimize 3D printing processes while providing a consistent, validated, and adaptable educational framework for students.

Matthew Percival

Affiliation

Refine Results

Machine Learning-Based Quality Control for 3D-Printed Parts