Dynamic Modeling of Intrinsic Self-Healing Polymers Using Deep Learning.

Area of Science:

• Materials Science
• Polymer Science
• Computational Mechanics

Background:

• Traditional characterization of self-healing polymers relies on destructive testing, limiting dynamic property analysis.
• Understanding the mechanics of polymer self-healing often requires complex theoretical derivations or simulations post-experiment.

Purpose of the Study:

• To propose a novel model, the self-healing property evolution using energy functional dynamical (SPEED) model, for predicting polymer self-healing mechanics.
• To enable the prediction of macroscopic property evolution using dynamic imaging and machine learning, avoiding destructive testing.

Main Methods:

• Developed an energy functional minimization (EFM) model to extract dynamical systems from time-series 2D cut images of self-healing polymers.
• Integrated the EFM model with a static property prediction model to create the SPEED model.
• Trained the machine learning (ML) model using a large dataset of healing polymer images (over 100,000 frames).

Main Results:

• The SPEED model successfully predicted the temporal evolution of material properties, specifically toughness, in a self-healing conductive polymer.
• Demonstrated that the model can capture the underlying physics of self-healing dynamics, including potential and interface energies.
• Validated the model's ability to predict macroscopic property evolution with minimal training data.

Conclusions:

• The SPEED model offers a non-destructive method to predict the dynamic evolution of self-healing polymer properties.
• This approach significantly advances the understanding and characterization of self-healing materials by providing temporal insights.
• The methodology is applicable to various self-healing polymers, facilitating efficient material property prediction.