Spatial strain correlations, machine learning, and deformation history in crystal plasticity

Area of Science:

• Materials Science
• Condensed Matter Physics
• Computational Mechanics

Background:

• Systems far from equilibrium exhibit history-dependent behavior, making response prediction challenging.
• Crystal plasticity involves complex microstructural features and strain gradients due to processing history.
• Spatial strain correlations offer potential predictive information beyond current system properties.

Purpose of the Study:

• To demonstrate the use of spatial strain correlations for inferring and classifying prior deformation history in crystalline materials.
• To apply machine learning techniques for statistical inference of deformation history.
• To investigate the impact of nanoscale size effects and data uncertainty on classification accuracy.

Main Methods:

• Utilizing 2D discrete dislocation plasticity simulations to generate data for uniaxially compressed crystalline thin films.
• Employing machine learning algorithms to analyze spatial strain correlations.
• Evaluating the influence of size effects and structural uncertainty on the classification of plasticity regimes.

Main Results:

• Spatial strain correlations were successfully used to statistically infer and classify prior deformation history.
• Machine learning techniques effectively classified different plasticity regimes based on strain correlations.
• Nanoscale size effects and structural uncertainty presented challenges but did not prevent classification.

Conclusions:

• Spatial strain correlations are valuable for understanding and predicting the deformation history of crystalline materials.
• Machine learning provides a powerful tool for inferring material history from microstructural data.
• Further research is needed to fully address the impact of size effects and uncertainty in nanoscale plasticity analysis.