Applied Machine Learning to Predict Stress Hotspots I: Face Centered Cubic Materials

TL;DR

This paper combines crystal plasticity simulations and machine learning to predict stress hotspots in face-centered cubic copper alloys, achieving a ROC-AUC of 0.74, revealing insights into microstructural influences on stress localization.

Contribution

It introduces a data-driven approach integrating full-field deformation models with machine learning to predict stress hotspots in FCC materials, demonstrating the method's effectiveness and limitations.

Findings

Achieved ROC-AUC of 0.74 in hotspot prediction.

Identified microstructural features influencing stress localization.

Showed machine learning's potential and constraints in materials microstructure analysis.

Abstract

We investigate the formation of stress hotspots in polycrystalline materials under uniaxial tensile deformation by integrating full field crystal plasticity based deformation models and machine learning techniques to gain data driven insights about microstructural properties. Synthetic 3D microstructures are created representing single phase equiaxed microstructures for generic copper alloys. Uniaxial tensile deformation is simulated using a 3-D full-field, image-based Fast Fourier Transform (FFT) technique with rate-sensitive crystal plasticity, to get local micro- mechanical fields (stress and strain rates). Stress hotspots are defined as the grains having stress values above the 90th percentile of the stress distribution. Hotspot neighborhoods are then characterized using metrics that reflect local crystallography, geometry, and connectivity. This data is used to create input feature vectors to train a random forest learning algorithm, which predicts the grains that will become stress hotspots. We are able to achieve an area under the receiving operating characteristic curve (ROC-AUC) of 0.74 for face centered cubic materials modeled on generic copper alloys. The results show the power and the limitations of the machine learning approach applied to the polycrystalline grain networks.