Physics-informed machine learning enabled virtual experimentation for 3D printed thermoplastic.

Area of Science:

• Materials Science and Engineering
• Computational Materials Design
• Polymer Science

Background:

• The performance of 3D printed thermoplastics is highly sensitive to ink formulation.
• The vast chemical space of monomers makes identifying optimal formulations challenging.
• Experimental data scarcity hinders efficient material property prediction.

Purpose of the Study:

• To develop a virtual experimentation platform for predicting 3D printed thermoplastic performance.
• To establish a correlation between ink composition and material properties using physics-informed machine learning.

Main Methods:

• A multilayer perceptron (MLP) model was trained using physics-informed machine learning.
• Dimensionality reduction of stress-strain curves to principal components (PCs) addressed data scarcity.
• Physics-informed descriptors were integrated into the model's input dataset.

Main Results:

• The model achieved high prediction accuracy for fracture strength (R²=0.97) and toughness (R²=0.95).
• Virtual experimentation predicted 100,000 ink formulations, generating corresponding stress-strain curves.
• Experimental validation confirmed a strong agreement between predicted and actual material performance.

Conclusions:

• The physics-informed machine learning platform enables efficient virtual experimentation for materials discovery.
• This methodology provides a generalizable approach to correlate complex input variables with material performance metrics.
• The developed platform accelerates the identification of optimal 3D printing materials.