Analysis on Microstructure-Property Linkages of Filled Rubber Using Machine Learning and Molecular Dynamics...

Area of Science:

• Materials Science
• Computational Science

Background:

• Understanding microstructure-property relationships is crucial for material design.
• Extracting key filler aggregates from complex morphologies is challenging but essential for optimizing rubber mechanical properties.

Purpose of the Study:

• To develop and compare machine learning techniques for extracting stress-contributing filler aggregates in filled rubber.
• To validate the identified aggregates' impact on mechanical properties.

Main Methods:

• Quantifying filler morphology using persistent homology and persistence images.
• Developing binary classification models with logistic regression and convolutional neural networks.
• Validating extracted aggregates using coarse-grained molecular dynamics simulations.

Main Results:

• Both logistic regression and convolutional neural networks successfully extracted filler aggregates contributing to high stress.
• Comparison of extracted aggregates revealed insights into shapes and distributions that enhance stress generation.
• Coarse-grained molecular dynamics simulations confirmed the significant effect of extracted aggregates on mechanical properties.

Conclusions:

• The proposed machine learning methods effectively identify critical filler aggregates that determine rubber mechanical properties.
• This approach provides a robust framework for optimizing material performance through microstructure analysis.