Filament Type Recognition for Additive Manufacturing Using a Spectroscopy Sensor and Machine Learning.

Area of Science:

• Additive Manufacturing
• Spectroscopy
• Machine Learning

Background:

• Fused Filament Fabrication (FFF) enables multi-material 3D printing.
• Accurate filament identification is crucial for optimizing print parameters and ensuring successful multi-material prints.
• Current methods for filament recognition can be complex or lack the necessary precision.

Purpose of the Study:

• To develop and validate a novel, compact, and high-accuracy filament recognition module for FFF processes.
• To integrate multi-spectral spectroscopy with machine learning for robust material identification.
• To enhance the autonomy and versatility of multi-material 3D printing systems.

Main Methods:

• A multi-spectral spectroscopy sensor module measuring 18 wavelengths (visible to near-infrared) was utilized.
• Filament samples included PLA, TPU, TPC, carbon fibre, ABS, and carbon fibre-blended ABS.
• Data were collected using the Triad Spectroscopy module AS7265x at distances of 12 mm, 16 mm, and 20 mm.
• Machine learning models (kNN, Logistic Regression, SVM, MLP) were trained and optimized with hyperparameter tuning.

Main Results:

• The Support Vector Machine (SVM) model achieved the highest classification accuracy of 98.95%.
• Optimal performance was observed using data from the AS72651 sensor at a 20 mm measurement distance.
• The developed module demonstrated high accuracy in distinguishing between various filament types.

Conclusions:

• A compact, high-accuracy filament recognition module for FFF was successfully developed.
• The integration of multi-spectral spectroscopy and machine learning offers a powerful solution for automated material identification.
• This technology can significantly improve the autonomy of multi-material 3D printing, enabling dynamic filament switching and parameter optimization.